Electromagnética Compatibility

   EMBED

Share

  • Rating

  • Date

    December 1969
  • Size

    75.7MB
  • Views

    1,871
  • Categories

Preview only show first 6 pages with water mark for full document please download

Transcript

ALL RIGHTS RESERVED 0 l%l~ Applications for reproduction of this book or parts thereof, should be directed to: EMC Proceedings Editor, ETH Zentrum - IKT, 8092 Zurich, Switzerland. EMC Symposium & Exhibition, Zurich 7985 Honorary Patron: Mr. F. Locher, Berne Under the auspices of: Mr. R. Trachsel, Director-General of the Swiss PTT, Berne Sponsor: Association of Swiss Electrotechnicians (SEVIASE) Organized by: Institute for Communication Technology of the Swiss Federal Institute of Technology Zurich Cooperating: International Union of Radio Science (URSI), Convention of the National Societies of Electrical Engineers of W. Europe (EUREL), international Radio Consultative Committee (CCIR), IEEE Electromagnetic Compatibility Society, IEEE Switzerland Section, Association of Polish Electrical Engineers (SEP), Committee AE-4 on Electromagnetic Compatibility of the Society of Automotive Engineers (SAE), Information Technology Society of the SEV (ITG) Organizing Commlttee: Prof. Dr. P. Leuthold, Zurich (Symposium President); E. Diinner, Zurich (Vice-President); Prof. Dr. F. L. Stumpers, Eindhoven (Vice-President); Dr. T. DvoNk, Zurich (Organizing Chairman); Prof. Dr. R. M. Showers, Philadelphia (Technical Program Chairman); H. K. Mertel, San Diego (Workshops Program Chairman); U. Welte, Zurich (Exhibition Chairman); B. Szentkuti, Berne (Publicity Chairman); Dr. M. lanovici, Lausanne (Joint Events Chairman); R. Bandle, Zurich; R. Danieli, Zug; G. Meyer, Stafa; J. @rum, Zurich (Chairpersons, Local Arrangements); G. Georg, Allenwinden (Treasurer); Mrs. E. Danieli, Zug; Mrs. V. Szentkuti, Berne (Ladies Program). Technical Program Committee: Chairman: Prof. Dr. R. M. Showers Prof. Dr. P. Degauque, Villeneuve-d’Ascq; Dr. T. Dvorak, Zurich (Pro ceedings Editor); Prof. Dr. C. Egidi, Turin; Dr. J. J. Goedbloed, Eindhoven; Prof. Dr. S. Lundquist, Uppsala; Dr. A. D. Spaulding, Boulder; Dr. R. Sturm, Munster; Dr. A. Whitehouse, London; Prof. Dr. F. Zach. Wien. Advisory CommIttee: H. Bachmann, Noordwijk; Prof. Dr. F. E. Gardiol, Lausanne (Swiss National Committee of the URSI); Ft. Gressmann, Bruxelles (EBU); J. Hamelin, Lannion; J. S. Hill, Springfield (IEEE EMCS); G. A. Jackson, Leatherhead; R. C. Kirby, Geneva (CCIR); J. L. Moe, Fort Worth (SAE AE-4); Prof. Dr. J. J. Morf, Lausanne; W. Moron, Wroclaw (SEP); Prof. Dr. J. Neirynck, Lausanne (IEEE Switzerland Section); Prof. Dr. R. Sato, Sendai; Ch. Scherrer, Berne (BAUEM); Prof. Dr. Ft. Struzak, Wroclaw; Prof. Dr. A. Wedam, Ljubljana; Prof. Dr. R. Zwicky, Zurich. 79754985: Ten years of EMC Symposia Symposium Patrons 1975.1995: F. Lecher Drs. Ph. Leenman R. Trachsel Symposium Chairman: Prof. Dr. F. E. Borgnis (19751979) Sympostum President: Prof. Dr. P. E. Leuthold (1981-1985) Secretary Generallorganising Chairman: Dr. T. Dvorak (19751985) Program Chairman: Prof. Dr. F. L. Stumpers (1975-1981) Prof. Dr. Ft. M. Showers (1983-1985) Workshops Program Chalrman: H. K. Mertel (19751985) Sponsoring organlsatfons 19751965: Montreux Tourist Office, Netherlands National Committee of the IEC, Swiss Electrotechnlcal Association Organfslng tnstitutlons 1975.1965: Montreux Tourist Office, Netherlands National Committee of the IEC in cooperation with the Institute of High Frequency Electronics of the Federal Institute of Technology Zurich, Institute for Communication Technology of the Federal Institute of Technology Zurich Cooperating organlsations 19751985: International Union of Radio Science (URSI), Convention of the National Societies of Electrical Engineers of Western Europe (EUREL), International Radio Consultative Committee (CCIR), International Special Committee on Radio Interference (CISPR), Region 8 of the IEEE, IEEE Switzerland Section, IEEE Electromagnetic Compatibility Society, Association of Polish Electrical Engineers (SEP), Committee AE-4 of the Society of Automotive Engineers (SAE), Nachrichtentechnische Gesellschaft im Verband Deutscher Elektrotechniker (NTGIVDE), Information Technology Society of the SEV (ITG) Certificates of Acknowledgement: (for outstanding support of the Symposium) J. S. Hill (1977), H. K. Mertel (1977), J. C. Toler (1977). Prof. Dr. F. L. Stumpers (1983) Some data on past svmposia I Year Attendance Papers in Record Exhibitors Techn. excursions Workshops *15 summaries of the Session on Sequency 1975 1977 1979 1961 1963 396 108’ 18 444 106 19 507 107 23 529 102 25 829 103 29 -4 41 : : Techniques ” not included Prize Award Papers Honor Roil: Montreux 1971: 1. (ex aequo, in alphabetical order of the first author): lR. W. p. King, G. S. Smith: “Electrical field probes and their application in EMC” *V. P. Pevnitsky, L. V. Tigin: “A stochastic model of a cumulative. process of man-made radio interference and objective evaluation of srgnal distortions produced by these interferences” 2. D. A. Bull, G. A. Jackson: “Interference survey in military transport aircraft” 3. Ft. Struzak: “Electromagnetic compatibility: Urban electromagnetic environment - Facts models, trends” 4. R. Cortina, F. Demjchelis, W. Serravalli: “Anew type of 500kHzmeasuring instrument for long-term recording of radiointerferencefrom power lines” Montreux 1977: 1. (ex aequo, in alphabetical order of the authors): “A. P. Kalmakov: “Analysis of statistical characteristics of click voltages measured with a CISPR measuring set” *A. D. Spaulding: “Optimum reception in the presence of impulsive noise” (ex aequo, in alphabetical order of the authors): lR. J. Hasler: “The measurement of external immunity of domestic receivers-some problems and their solution” ‘R. G. Struzak: “CISPR auasi-oeak measurino channel with extended . . dynamic range” P. Groenveld, A. de Jong: “A simple r.f. immunity test setup” P. G. Galliano: irlmpulsive disturbances on car electric circuitry” Rotterdam 1979: 1. “D. Middleton: “Canonical non-Gaussian noise models: Their implications for measurements and for prediction of receiver performance” 2. (ex aequo, in alphabetical order of the first author): *I. L. Gallon: “EMP coupling to long cables” “J. Hamelin, B. Djebari, R. Barreau, J. Fontaine: ‘Electromagnetic field resulting from a lightning discharge, surges induced on overhead lines, mathematical model” “J. G. Tront, J. J. Whalen: “Computer-aided analysis of RF effects in operational amplifiers” 3. (ex aequo, in alphabetical order of the first author): W. Hadrian: Reduction of electromagnetic disturbances ‘in buildings caused by lightning using conductive facades” A. P. Kalmakov: “Possibilities of reduction of volume of measurements when checking the sources of clicks for compliance with CISPR limits” T. Takagi, t-t. Echigo, R. Sato: “Some characteristics of electric discharge as a noise source in EMC problems-recent studies in Japan” Zunbh 1981: 1. (ex aequo, in alphabetical order of the first author): ‘C. R. Paul: “Adequacy of low-frequency crosstalk prediction models” *F. M. Tesche, T. K. Liu: “Recent developments in electromagnetic field couolina to transmission lines” 2. lR. ‘Bersier: “Measurement of the immunity of TV receivers to AM RF fields in the 3 to 30 MHz range, including the influence of connected cables” 4. M. L. Crawford: “Options to open-field and shielded enclosure elec tromagnetic compatibility measurements” 5. M. Borsero, E. Nano: “Comparison between calculated and measured attenuation of the site recommended by IEC for radiation measurements” 6. B. Demoulin, P. Degauque, M. Cauterman: “Shielding effectiveness of braids with high optical coverage” Zurich 1983: 1. “J. J. Goedbloed, K. Riemens, A. J. Stienstra: “Increasing the RFI immunity of*amplifiers with negative feedback” 2. ‘T. G. Dalby: “Linear antenna near-field decoupling using a radial transmission line” 3. lB. Demoulin, P. Duvinage, P. Comic, P. Degauque: “Penetration through an interruption of the shield of a coaxial cable” 4. K. Bullough, A. Cotterill: “Ariel 4 observations of power-line harmonic radiation over North America and its effect on the magnetosphere” 5. L. E. Varakin: “Electromagnetic compatibility of cellular mobile communication systems with pseudo-noise signals” 6. J. J. Max, A. V. Shah: “Distributed lowpass filters for EMI filtering” * recipients of monetaryawards ___.___ .~ - Table of Contents A, AutomatedEMC measurements 1Al E.L.Bronaugh, P.A.Sikora, Electra-Metrics, Amsterdam, NY: Automated EMC measurements: An overview. 2A2D.N.Heirman, AT&T Laboratories, Holmdel, NJ: Automated immunity measurements. 3A3 J.C.van Essen, ESA-ESTEC, Noordwijk, Netherof an automated EMC lands: Instrumentation test facility for spacecraft. Issy-Les-Moulineaux, 4A4 G.Eumurian, Thomson-CSF, France: Computer-assisted control of EMP measurements on major systems. B, ESD techniques 5Bl P.Richman, A.Tasker, KeyTek Instrument Corp., Burlington, MA: ESD testing: The interface between simulator and equipment under test. 6B2 M.Mardiguian, D.R.J.White, Don White Consultants, Inc., Gainesville, VA: Recent develop: ments in the understanding of coupling paths of ESD through a metallic cabinet. 7B3 L.Inzoli, Honeywell ISI, Milano, Italy: -ESD susceptibility and radiated emissions of EDP peripheral printers. 8B4 B.Daout, H.Ryser, Hasler Ltd., Berne, Switzerland: Fast discharge mode in ESD-testing. C, TriggeredlightningEMP WI I@2 UC3 DC4 H.Kikuchi, Nihon University, Tokyo, Japan: A new model of triggered lightning. The St.Privat d'Allier Research Group, France: Applications of triggered lightning in France: Possibilitiesand limitations. A.S.Podgorski, NRC, Ottawa, Canada; J.A.Landt, Los Alamos National Laboratory, NM: Numerical analysis of the lightning - CN tower interaction. T.Takeuti, M.Nakano, Z.-I.Kawasaki, N.Takagi, Nagoya University, Toyokawa, Japan: Electromagnetic fields on the ground due to lightning strokes triggered with rockets and a tall chimney. D, EMC measurements UDl J.D.Gavenda, University of Texas; J.H.Davis, IBM Corp., Austin, TX: Electromagnetic wave propagation in a semi-anechoic chamber. 14D2M.Kanda, NBS, Boulder, CO: A methodology for evaluating microwave anechoic chamber measurements. 15D3 S.C.Kashyap, NRC, Ottawa, Canada: Field distortions in a TEM cell. 16D4J.H.Davis, W.C.Cockerill, IBM Corp., Austin, TX: Chamber quality assessment. 17D5S.Linkwitz, Hewlett-Packard Co., Santa Rosa, CA: Discriminating between narrowband and broadband EM1 using a spectrum analyser. 18W U.Raicu, G.U.Sorger, Eaton Corp., Sunnyvale, CA: Broadband YIG-tuned preselector for VHF and UHF. 1gD7 G.K.Boronichev, LONIIR, Leningrad, USSR: Measurement of the immunity of broadcast receivers according to the CISPR method and the difficulties encountered. Technical University of Wroc2oD8 T.W.Wigckowski, law, Poland: On the measurement of EM power density using a double loaded loop antenna. E. Printedcircuit board EMC ZEl C.R.Paul, University of Kentucky, Lexington, KY: Printed circuit board EMC. 2232B.Danker, N.V.Philips, Eindhoven, Netherlands: New measures to decrease radiation from printed circuit boards. 23~3D.R.Bush, IBM Corp., Lexington, KY: Radiated emissions of printed circuit board clock circuits. 24~4H.W.Ott, AT&T Bell Laboratories, Whippany, NJ: Controlling EM1 by proper printed wiring board layout. 25~5R.F.German, IBM Corp., Boulder, CO: Use of a ground grid to reduce printed circuit board radiation. 26~6J.W.E.Jones, Portsmouth Polytechnic, England: Achieving compatibility in inter-unit wiring. 2737J.P.Charles, CNET, Issy-Les-Moulineaux, France: Electromagnetic interference control in logic circuits. F, Lightningelectromagnetic pulse 289 C.D.Weidman, E.P.Krider, University of Arizona, Tucson, AZ: Lightning radiation fields. 2%'2F.Heidler, Hochschule der Bundeswehr Muenthen, Neubiberg, GFR: Traveling current source model for LEMP calculation. 30-3C.Weidman, J.Hamelin, M.Le Boulch, CNET, Lannion, France: Radiation characteristics, emission mechanisms and phenomenology of lightning. 31F4M.W.Wik, Defence Material Administration, Stockholm, Sweden: Double exponential pulse models for comparison of lightning, nuclear and electrostatic discharge spectra. 32~5R.L.Gardner, L.Baker, MRC, Albuquerque; C.E. Baum, D.J.Andersh, Kirtland AFB, NM: ~Comparisen of lightning with public domain HEMP waveforms on the surface of an aircraft. 33F6D.Jaeger, R.Rode, MBB GmbH, Ottobrunn, GFR: NEMP and lightning protection requirements for modern aircraft equipment. 34F7F.Pigler, Siemens AG, Erlangen; P.Kronauer, BBC, Mannheim; R.Terzer, KWU, Erlangen, GFR: Prediction of lightning-induced interference voltages on the basis of measurements taken in similar installations. 35F8H.Schiippler, D.Ristau, University of Transport, Dresden; H.Lorke, IPF, Berlin, GDR: Impulse current and voltage propagation in underground telecommunication cables. G, EM wave interactionwith biological systems 3&l Q.Chen, R.C.Huang, B.C.Pan, CARIS, Beijing, China: The hazard of electromagnetic radiation and discussion of safety thresholds. 37~2 T.S.Tenforde, C.T.Gaffey, M.S.Raybourn, University of California, Berkeley, CA: _... Influ_-. ence of stationary magnetic fields on ionic conduction processes in biological systems. Centre, N.Dekleva, D.Vujnovid, Clin.Hospital Zemun; B.Beleslin, Medical Faculty; V.Majid, Electrotechnical Faculty, Belgrade, YUgOSlavia: Magnetostimulation - A method for reestablishment of antibiotic bactericidal action. CNRS, Thiais, France: Specific 3964 A.J.Berteaud, mechanisms of microwave power dissipation in living tissues. 40G5 R.G.Olsen, Naval Aerospace Medical Res.Laboratory, Pensacola, FL: Measurement of specific absorption rate in a full-size man model near a 10.67-m monopole antenna/ground plane system at 2.101 MHz. 41G6G.d'Ambrosio, A.Scaglione, F.De Martino, R. Pennarola, University of Naples, Italy: Ku_ band radiation effects on the eye. 42~7 D.W.Griffin, N.Davias, University of Adelaide, Australia: Wideband evaluation of microwave intensity near the eyes with scattering structures present such as safety spectacles. 3803 H, Statisticalaspectsof noise and limits 43HlA.de Jong, Dr.Neher Laboratories PTT, Leidschendam, Netherlands: Statistical aspects of noise and limits. Department of Trade and InA.C.D.Whitehouse, dustry, London, England: Radio interference - The probability problem. 45H3B.Audone, R.Cazzola, G.Barale, Aeritalia, Torino, Italy: Statistical evaluation of the EMC safety margin at system level. 46H4R.Bersier, Swiss PTT, Berne, Switzerland: -The state of art of TV receiver immunity and recommendations for appropriate construction deduced from test statistics. 47H5A.P.Kalmakov, LONIIR, Leningrad, USSR: Probability distributions of effective voltages of man-made radio interference and their use for the calculation of limits. 48H6Q.Chen, Y.C.Zhu, CARIS, Beijing, China: The application and development of EMC in China. I, EM Phenomenain Power transmissionand distribution 4911 H.-J.Haubrich, VEW AG, Dortmund, GFR: New ways for interference computation and MonteCarlo-optimization to guarantee the compatibility of inductively coupled line systems. 5012 W.MachczyAski, Polytechnic of Poznan, Poland: Potentials and currents along an earthed buried cable exposed to electromagnetic effects of a power line under fault condition. 5113J.L.ter Haseborq, H.Trinks, Technical University Hamburg-Harburg; R.Sturm, NBC Defence Research and Development Institute, Munster, GFR: Coupling and propagation of transient currents on multiconductor transmission lines. 5214F.Paladian, J.P.Plumey, D.Roubertou, J.Fontaine, University of Clermont-Ferrand, France: Response of a single-conductor overhead wire illuminated by an inhomogeneous plane wave. 5315F.Maumy, B.Jecko, O.Dafif, University of Limoges, France: Time domain scattering by thin wire structures above a homogeneous ground. 916 %I1 SchwaA.Strnad, H.RGhsler, Energie-Versorgung ben, Stuttgart, GFR: Noise sources and interference values in high voltage substations. T.Yoshino, I.Tomizawa, University Of Electrocommunication, Tokyo, Japan: Balloon and satellite observation of power line radiation over northern Europe. J, Computerprogramsfor the EMC engineer SJI J.K.Breakall, G.J.Burke, E.K.Miller, Lawrence Livermore National Laboratory, Livermore, CA: The numerical electromagnetic code (NEC). 5752 D.J.Bem, J.Janiszewski, R.Zielidski, Technical University of Wroclaw, Poland: Computer. aided analysis of electromagnetic compatibllity in VHF-FM broadcasting networks. 58J.3 A.Farrar, NTIA, Annapolis, MD: Computer models for determination of satellite powerflux-density limits. University of Tsukuba, Ibaraki, 5qJ4 K.Hirasawa, -- Japan: Computer programs for calculating bounds of interference between arbitrarily shaped wire antennas. aJ5 G.Azrak, Merlin-Gerin, Grenoble; Ph.Auriol, Ecole Centrale de Lyon, Ecully, France: Numecompati~rical simulation of electromagnetic bility in time domain. 61~6W.Krzysztofik, Technical University of Wroclaw, Poland: Electromagnetic wire scattering of thin cylindrical antennas loaded by nonlinear impedances. K, EMI in microelectronics 62~1J.J.Whalen, SUNY at Buffalo, Amherst, NY: Determining EM1 in microelectronics - A review of the past decade. 63~2J.G.Tront, Virginia Polytechnic and State University, Blacksburg, VA: Comparison of the RF1 susceptibility of several typical IC pin drivers/receivers. SUNY at Buffalo, Am64~3 Y.-H.Sutu, J.J.Whalen, herst, NY: Demodulation RF1 in inverting and non-inverting operational amplifier circuits. L. Nuclearelectromasnetic Pulse imoact 65~1O.Dafif, C.Bardet, E.Jecko, University of Limoqes, France: Transient field distribution in a transmission line simulator. 66~2H.-D.Briins, D.KBniqstein, Hochschule der Bundeswehr, Hamburg, GFR: Calculation and measurements of transient electromagnetic fields in EMP simulators. 67~3T.Karlsson, G.Unden, M.Gylemo, National Defence Research Institute, Linkoeping, Sweden: EMP simulation by pulse injection. 68~4 M.E.Gruchalla, A.J.Bonham, J.Gibson, P.G. Johnson, EG&G WASC Inc., Albuquerque, NM: A portable programmable pulser and high-speed, log-weighted peak-level recorder for direct.___ drive testing. 69L5C.E.Baum, AWFL, Kirtland AFB, NM: Black box bounds. 70L6 P.B.Johns, University of Nottingham; A.Mallik, Kimberley Communications Consultants, Nottingham, England: EMP response of aircraft structures using transmission-line modelling. 71=7I.L.Gallon, AWRE, Aldermaston, England: Radiation damping in finite cylinders. 72L8A.Caron, B.Djebari, A.Zeddam, CNET, Lannion, France; Ph.Blech, Y.Dijamatovic, M.Ianovici, EPFL, Lausanne, Switzerland: Validation of EMP calculation methods using the response of an aerial cable to a lightning stroke. M. Power and data line transients EMI 74~2 75M3 76M4 77M5 & W.T.Rhoades, Xerox Corp., El Segundo, CA: Characteristics of unusual power main transients. F.D.Martzloff, GEC Comp., Schenectady, NY: The development of an IEEE guide on surge testing for equipment connected to low-voltage AC power circuits. P.Richman, KeyTek Instrument Corp., Burlington, MA: Changes to classic surge-test waves required by back-filters used for testing powered equipment. V.Scuka, Uppsala University, Sweden: Performance deterioration of metal oxide varistors by current surges. M.Tetreault, Digital Equipment Corp., Stow, MA; F.D.Martzloff, GEC Comp., Schenectady, NY: Characterization of disturbing transient waveforms on computer data communication lines. Statisticaltheory of EMC 78Nl D.Middleton, New York, NY: Threshold signal and parameter estimation in non-Gaussian EMC environments. 7gN2 A.D.Spaulding, NTIA, Boulder, CO: Locally optimum and sub-optimum detector performance in non-Gaussian "broadband" and "narrowband interference environments. 8fjN3N.N.Buga, V.Y.Kontorovich, Electrotechnical Institute of Communications; Y.V.Polozok, LONIIR, Leningrad, USSR: Electromagnetic environment control on the basis of system models with random structure. 0, Spread spectrumand mom communications 8101 P.M.Hopkins, D.N.Cravey, Lockheed Co., Inc., Houston, TX: Spread spectrum communications - Interference considerations. 8202 K.Dostert, University of Kaiserslautern, GFR: EMC problems in data transmission over indoor power lines using spread spectrum techniques. 8303 L.E.Varakin, All-Union Telecommunication Institute by Correspondence, Moscow, USSR: The efficiency of the cellular spread spectrum radiotelephone. H.Ochsner, Federal Institute of Technology, 8404 Zurich, Switzerland: Comparison of spectrum efficiency of CDMA and FDMA mobile radio systems. G.K.Chan, Department of Communications, Ot8505 tawa, Canada: Interference analysis of a land mobile cellular radio system. 8606 K.Fisher, Department of Trade and Industry, London, England: Planning of television band III for use by mobile services. 8707 B.BeriE, Federal Radiocommunication Direction, Belgrade, Yugoslavia: Comparison of field strength measurements and computer prediction in land mobile service. 8808 A.Golas, Telecommunication Research Centre, New Delhi, India: Compatibility of TV and UHF communications antennas mounted on the same tower. 899 S.Satyamurthy, Combat Vehicles R&D Establishment, Madras, India: Design of compatible equipment for land mobile vehicles. P, Shieldingand cable coupling WPl S.R.Ramasamy, Defence Electronics Research Laboratory, Hyderabad; S.Mahapatra, Indian Institute of Technology, Bombay, India: Attenuation of electromagnetic radiation from microwave ovens utilizing corrugated metallic surface combined with magnetic resistive sheets and absorbers. 91P.2W.Hadrian, Technical University of Vienna, Austria: Low-frequency magnetic shielding effectiveness of steel-reinforced concrete platforms. 92P3 B.L.Michielsen, Philips Research Laboratories, Eindhoven, Netherlands: A new approach to electromagnetic shielding. 93p4V.A.Morozov, N.V.Rodionova, USSR Academy of Sciences, Moscow, USSR: Field nonuniformity reduction inside a spherical magnetic shield. 94P5 H.Rahman, St.Louis University, Cahokia, IL; J.Perini, Syracuse University, NY: EMP enclosure penetration and cable coupling. !%P6 B.Demoulin, P.Duvinage, P.Degauque, Lille University, Villeneuve d'Ascq, France: Measurements of transfer parameters of shielded cables at frequencies above 100 MHz. 96P7 K.H.Gonschorek, Siemens AG, Erlangen, GFR: Magnetic stray fields of twisted multicore cables and their coupling to twisted and nontwisted two-wire lines. Q, Power electronics 97Ql M.Di Stefano, Italian Railways, Roma; G.L. Solbiati, SIRTI S.p.A., Milano, Italy: -Project of a railway electrification from the EMC point of view. 98Q2 H.Kunkel, M.Lutz, O.Frey, High Voltage Test Systems, Basel, Switzerland: Coupling and filtering possibilities of transients during EMC tests. ggQ3 F.C.Zach, Technical University of Vienna, Austria: A new pulse width modulation control for.line commutated converters minimizing the mains hacmonics content. 10op4 J.Sack, H.Schmeer, Hochschule der Bundeswehr Muenchen, Neubiberg, GFR: Computer-aided analysis of the RF1 voltage generation by small commutator motors. 101Q5 J.M.Firth, NRC, Ottawa, Canada: Control and reduction of spurious emissions from small DC to DC power converters. 1@Q6 B.Brdndli, J.Bertuchoz, R.Steck, NC-Laboratory, Spies, Switzerland: High current fast pulse measurement with a Rogowski coil. 103Q7 V.Nikiforova, All-Union Research Institute of Energetics, Moscow, USSR: Electromagnetic compatibility of electrical equipment in power and industrial supply systems. S, Systems EMC and protectivemeasures u@l H.Cichofi,H.Trzaska, SARU Region 1 EMC Working Group, Poland: Selective interference in home entertainment electronic devices. US2 I.Oka, K.Ishida, I.Endo, University of Electro-Communications, Tokyo, Japan: Co-channel interference in an on-board processing satellite. R, Key Problemsof spectrumuse 112.53S.Yamazaki, H.Kuronuma, NHK Science & Technil@Rl K.Olms, FTZ, Darmstadt, GFR: Radio frequency cal Laboratories; Y.Noguchi, Nippon Electric spectrum management. Company, Tokyo, Japan: Relation between APD/ l&R2 H.J.Weiss, COMSAT, Washington, DC: The big CRD of automobile ignition noise and resulsqueeze - A selective look at ORB-85/88. tant TV picture degradation. l&R3 A.H.Wojnar, Warsaw Academy of Technology, PoUS4 K.Uchimura, T.Aida, Kumamoto University; T. land: Deformable lattices for efficient freTakagi, Tohoku University, Sendai, Japan: quency management. Electromagnetic radiation caused by silver 107R4 R.Sandell, BBC Research Department, Tadworth, palladium alloy contact switching. England: The prediction of field strength in u4S5 W.van Eck, J.T.A.Neessen, P.J.M.Rijsdijk, Dr. the frequency range 30-1000 MHz and its inNeher Laboratories PTT, Leidschendam, Netherfluence on spectrum management. lands: On the characteristics of the electro1@~5 G.A.De Couvreur, M.C.Delfour, Department of magnetic field generated by video display Communications, Ottawa, Canada: Optimum freunits. quency assignment strategies for radio celluu5S6 W.Biichler,Meteolabor AG, Wetzikon, Switzerlar land: Overvoltage protection circuits. 1@R6 P.Vaccani, Department of Communications, Ot116~7 M.A.Bykhovsky, G.G.Gurianov, Ministry of Tetawa, Canada: A second generation mobile speclecommunications, Moscow, USSR: Iterative trum monitoring system. interference simulator for the division of two FM signals. -c-m- Authors 4 Aida T. Andersh D.J. Audone B. Auriol Ph. Azrak G. 113s4 32F5 4583 6055 6OJ5 B - Baker L. Barale G. Bard& C. Baum C.E. Beleslin B. Bern D.J. BeriB B. Bersier R. Berteaud A.J. Bertuchoz J. Blech Ph. Bonham A.J. Boronichev G.K. Braendli B. Breaknll J.K. Bronaugh E.L. Bruens H.-D. Buechler W. Buga N.N. Burke G.J. Bush D.R. Bykhovsky M.A. 32F5 4583 65Ll 3235, 69L5 3863 57J2 8707 4684 39G4 10296 72L8 68L4 19D7 102Q6 56Jl 1Al 66L2 115S6 8ON3 5651 2333 11657 c Caron A. cazzo1.3 R. Chan G.K. Charles J.P. Chen Q. Cichofi H. Cockerill W.C. Cravey D.N. 72L8 4583 8505 2737 36G1, 4886 llOS1 16D4 8101 D - Dafif 0. d'Ambrosio G. Danker 8. Daout B. Davias N. Davis J.H. De Couvreur G.A. Degauque P. de Jong A. Dekleva N. Delfour M.C. De Martin0 F. Demoulin B. Dijamatovic Y. Djebari B. Di Stefano M. Dostert K. Duvinage P. 5315, 65Ll 41G6 22E2 SB4 42G7 13D1, 1604 108R5 95P6 43Hl 38G3 108RS 41G6 95P6 72LB 72L8 97Ql 8202 95P6 H Gruchalla M.E. Gurianov G.G. Gylemo M. 68L4 11657 6n3 Hadrian W. Hamelin J. Haubrich H.-J. Heidler F. Heirman D.N. Hirasawa K. Hopkins P.M. Huang R.C. 91P2 30F3 4911 29F2 2A2 59J4 8101 36Gl 1 Ianovici M. Inzoli L. Ishida K. J Jaeger D. Janiszewski J. Jecko B. Johns P.B. Johnson P.G. Jones J.W.E. I(_Kalmakov A.P. Kanda M. Karlsson T. Kashyap S.C. Kawasaki Z.-I. Kikuchi H. Kijnigstein D. Krider E.P. Kontorovich V.Y. Kronauer P. Krzysztofik W. Kunkel H. Kuronuma H. c Landt J.A. Le Boulch M. Linkwitz S. Lorke H. Lutz M. fl Machczydski W. Mahapatra 6. Majid V. Mallik A. Mardiguian M. Martzloff F.D. Maumy F. Michielsen B.L. Middletoh D. Miller E.K. Morozov V.A. E Endo I. Eumurian G. lllS2 4A4 N Nakano M. - Neessen J.T.A. Nikiforova V.N. Noguchi Y. E Farrar A. Firth J.M. Fisher K. Fontaine J. Frey 0. 58~3 lOlQ5 8606 5214 9BQ2 0 Ochsner H. _. Oka I. 0lms K. Olsen R.G. Ott H.W. 5 Gaffey C.T. Gallon I.L. Gardner R.L. Gavenda J.D. German R.F. Gibson J. Galas A. Gonschorek K.H. Griffin D.W. 37G2 71L-l 32F5 13Dl 2535 68L4 8808 96P7 42G7 Index fl Paladian F. Pan B.C. Paul C.R. Pennarola R. Perini J. Pigler F. Plumey J.P. Polozok Y.V. Podgorski A.S. 72L8 783 11152 33F6 57J2 5315, 65Ll 7OL6 68L4 26~6 47H5 14D2 6n3 15D3 12C4 9Cl 66L2 28Fl 8ON3 34F7 6156 98Q2 112S3 llC3 30F3 17D5 3538 98Q2 5012 9OPl 3aG3 7OL6 6~2 74M2, 77M5 5315 92P3 78Nl 56Jl 93P4 12C4 11455 103Q7 112S3 8404 lllS2 104Rl 40G5 2434 5214 36Gl 21El 41G6 94P5 34F7 5214 8ON3 llC3 B R&man H. Raicu D. Pamasamy S.R. Raybourn M.S. Rhoades W.T. Richman P. Rijsdijk P.J.M. Ristau D. Rode R. Rodionova N.V. Roehsler H. Roubertou D. Ryser Ii. 94P5. 18D6 9OPl 3702 73Ml 5B1, 75M3 11455 35FS 33F6 93P4 5416 5214 a34 5 Sack J. Sandell R. Satyamurthy 8. Scaylione A. Scbmeer H. Schiippler H. Scuka V. Sikora P.A. Solbiati G.L. Sager G.U. Spaulding A.D. Steck R. St.Privat d'Allier Research Group Strnad A. Sturm R. Sutu Y.-H. lOOQ4 107R4 8909 4166 lOOQ4 35FS 76M4 1Al 97Ql 18D6 79N2 102Q6 1 Takagi N. Takagi T. Takeuti T. Taker A. Tenforde T.S. ter Haseborg J.L. Terzer R. Tetreault M. Tomizawa I. Trinks H. Tront J.G. Trzaska H. 12C4 11354 12C4 5Bl 37G2 5113 34F7 77M5 5517 5113 6332 llOS1 u Uchimura K. Und&n G. 113s4 6n3 V Vaccani P. van Eck W. van Essen J.C. Varakin L.E. Vujnovid D. 109R6 11455 3A3 8303 38G3 w Weidman C.D. Weiss H.J. Whalen J.J. White D.R.J. Whitehouse A.C.D. Wi?ckowski T.W. Wik M.W. Wojnar A.H. v Yamazaki S. Yoshino T. z Zach F.C. Zeddam A. Zhu Y.C. Zielihski R. loC2 5416 5113 64K3 28F1, 3OF3 105R2 62KI, 64K3 6~2 44H.2 20D8 31F4 106R3 11253 5517 9993 72L8 48H6 57J2 Scientific Contributions - 1 Al 1 - AUTOMATED EMC MEASUREMENTS: AN OVERVIEW Edwin L. Bronaugh and Paul A. Sikora Electra-Metrics 100 Church Street Amsterdam, New York 12010 USA Abstract This paper looks at the history of automated EMC measurements and the current technology. It discusses the scope of this session. A philosophical discussion is included to lead to understanding the strengths and weaknesses of current technology and needs for future development. A present-day computer-controlled interference emissions measuring system is described. Introduction Background The desire for automated EMC measurements found its inception in the decade of the 1950's with the greatest incentive arising, perhaps, from the plethora of measurements mandated by military EMC standards on military communications and electronics equipment. The problem most pressing at the time could be summarized as too many measurements to be made resulting in too much data to analyze all in too short a time. From this apparent need arose mechanical attachments for the manual radio noise meters of the day to tune them automatically over their available tuning ranges while driving the X-axis of an X-Y plotter with a voltage proportional to the position of the mechanical tuning mechanism and the Y-axis with the envelope voltage from the indicating instrument drive circuitry. Although these "automatic" instruments were crude and frequently inaccurate, they provided the data much faster, more reliably, and in a more usable graphical form than could be provided by a human operator tuning, reading an indicator, and writing down the data on a pointby-point basis. For several years, EMC instrument manufacturers worked to improve upon this early swept tuned instrument by providing electronic tuning, more responsive detector functions, large dynamic measurement range by use of AGC or logarithmic amplifiers, and untuned wide band antennas and transducers. To this day such instruments are still widely used to make measurements in accordance with MIL-STD-461/ 462 [4, 51 and other military standards, During this same time, spectrum analyzers or panoramic receivers were being developed for somewhat different purposes, but would eventually come to be used for some EMC measure- ments. Then came the era of the computer, and EMC engineers and instrument manufacturers saw advantages to the use of computers to control the EMC test instruments. The computer could operate the test instruments; record data; apply antenna, transducer, cable loss, instrument calibration, and other correction and conversion factors to the data; and plot this reduced data on multi-decade plots for ease in comparing the performance of equipment under test with the limits in the technical standards. Many such systems for measuring interference emissions are in use today. While much automation has been achievedwith interference emissions tests, automation of interference immunity (susceptibility) tests has lagged far behind. One of the many reasons for this has been the more complicated nature of immunity tests. Purposes and Objectives This paper has two purposes. One is to introduce this session on automated EMC measurements by giving an overview of the session, and by discussing the philosophy of automated EMC measurements. The other purpose is to present some details on an automated radio noise (EMI) measuring system incorporating both self- and computer-controlled test capabilities. The objectives are to bring out some of the strengths and weaknesses of automated EMC tests, and to stimulate thinking towards continuing improvement in EMC measurements. The scope of this session is to address the issues associated with the use of computercontrolled or self-controlled automatic and semi-automatic test equipment and techniques to make EMC measurements* Both halves of the EMC test question will be addressed, i.e., both emissions (interference) and immunity (susceptibility) measurements. Some of the automated EMC measurement issues to be raised and discussed are: 1. Emissions and immunity testing for regulatory compliance versus testing for engineering and development; 2. The effects of the test equipment scan rate, the statistics of the radio noise or disturbance being measured or simulated, a mixture of signals and noise, and the characteristics of sources on EMC measurements; 3. Automation of EMC measurements as a tool to - correct deficiencies in present manualmeasurement techniques; 4. EMC measurements for meeting military requirements contrasted with those for commercial requirements (import licensing, type approval, etc.); and 5. Special characteristics of test instrumentation for automated EMC measurements versus that traditionally used in manual EMC measurements. Automated EMC Measurements Before we can improve upon the automatedEMC measurement instrument and extend its use to non-military EMC measurements, we must ask and answer a number of questions. Some of these questions are: 1. Why make automated EMC measurements? 2. What needs to be measured? 3. What are the strengths and weaknesses of the present technology? 4. What goals should be set for advancingautomated EMC measurement technology? To answer some of these questions and see the need to ask others, we must understand the sometimes conflicting requirements of EMC measurement standards and regulations. What do the technical standards of such bodies as IEC, CISPR, ANSI, ISO, VDE, FCC, CSA, JASO, SEV, and many others, and the military establishments of several countries have in common and where do they differ? Can an automated system be made to adequately deal with the differences? Should tasks that a thinking human operator can do easily be automated for an unthinking computer controller to do poorly, e.g., click measurements [a]? Why Make Automated EMC Measurements? This question was basically answered in the introduction, and the reasons are yet with us these days. Even though we have achieved some degree of automation in EMC measurements, the growth of the use of electronics with its concomitant growth of EMI causes the problem mentioned earlier of "too many measurements, too much data, and too little time," to be with us continually. This places us in a dilemma of needing either more and better automation of EMC measurements or fewer EMC measurements to make. The latter choice would tend to imply less or poorer control of radio noise andelectromagnetic interference, and, thus, a worsening lack of compatibility among our uses of the electromagnetic spectrum and electrical and electronic appliances and equipment. Even now, some regulatory agencies are reducing the amount of testing for compliance with their regulations in attempts to achieve a more acceptable balance between the amount of testing needed and the degree of EMC obtained,with the economics of both issues being a major consideration, What Needs to be Measured? A complete answer to this question is beyond the scope of this paper; however, an outlineof the answer would obviously include both interference emissions from equipment and interference immunity (susceptibility) of equipment. Also, both conducted and radiated interference emissions and immunity measurements would be included. To get into a more detailed analysis of what to measure, one must study in detail 2 - the various EMC measurement standards and regulations along with the current and predicted EMC problems in the geographical area of interest. The fact that geographical areas are an important factor is obvious if the interference regulations of high population density, high technology areas are compared with those of low population density, high technology areas. Strengths and Weaknesses of Automated EMC Measurements The present technology for automated EMC measurements uses desk-top calculators and small computers to operate the test instruments, record and reduce data, provide calibration, plot results, and even write test reports. At its present state, the technology has both advantages and disadvantages. Some of the advantages are summarized by Mr. D.N. Heirman of AT&T in his paper [l] in this session: "Automation of EMC Testing should be viewed as an engineering tool and not as a replacement for the engineer who must determine compliance with either regulatory or corporate EMC criteria. As a tool, automation if implemented properly decreases the likelihood of measurement error due to operator inattention, test instrumentation misadjustments, and inability to recreate all the test conditions on a repeatable basis. The cost of automation must also be weighed against the increased test time normally associated with manual operations." The advantages of these automated test systems seem to be abundant, but they also have serious disadvantages. Automation of EMC tests now often provides us with much more incorrect data faster. It is difficult to make the computer think and understand what is being measured, while the human operator who can think and understand can't make measurements as fast and tends to record data incorrectly or lose it entirely. However, a major reason for the incorrectness of the automated data lies in the typical understanding of the state-of-the-art many years ago. Many an EMC engineer is so happy to have the measurements done quickly with less labor that he or she has forgotten that many measurement errorproducing compromises were the state-of-theart years ago and are still present andaffect the correctness of the data taken by modern instruments. An example of this is in field strength measurements. Most EMC field strength measurements are made under conditions in which the electromagnetic fields to be measured are not homogeneous, but the antennas used to make the measurements are calibrated for and operate properly only in homogeneous, planewave fields which might be found in free space many wavelengths from their sources. Mixed signals and noise can pose particularly difficult problems. An example of this may be seen in testing a vehicle for ignition noise emanations. The vehicle is to be usedin the vicinity of sensitive receivers for long wave, medium wave, short wave, etc. communications, broadcasting, and navigation; thus, its ignition noise must meet stringent limits from 10 kHz to 1 GHz. The vehicle is large and the testing organization has no large shielded chamber in which to test it, so it must be tested outside. The outside environment - contains much noise and many narrowband signals throughout the required test frequency range, and most of these signals are so large that they produce indications in the EMI analyzer or radio noise meter far above the limit specified for the ignition noise emanations from the vehicle under test! Current EMC Instrumentation Technology The radio noise meter characteristics [2, 31 are the primary factor that determines if the ignition noise in the above example can be measured throughout the range of frequencies from 10 kHz to 1000 MHz. The regulatory requirements [4, 5, 6, 71 are secondary factors in the accurate and successful measurement of EMI in such a non-ideal real-world measurement situation. A typical simulation of the array of input signals and noise which the radio noise meter must resolve in the example above may consist of: 1) The signal from an impulse generator set to produce a level of 52 dB(uV/MHz) which simulates the vehicle ignition noise; 2) A pulse generator operating at 50 kHz producing a pulse amplitude of 0.0025 ~VS [68 dB(uV/MBz)] which simulates low frequency industrial noise in the vicinity of the test site; and 3) Two cw signal generators set to produce signals at 22 kHz and 8 MHz at levels of 64 dB(pV) and 49 dB(llV), respectively, to simulate two of the many narrowband communications and broadcast signals also in the ambient of the test site. A thinking, well-trained and experienced human operator using an ordinary radio noise mater would have an exceedingly difficult time resolving this spectrum of noise to determine correctly the level of the impulse generator, but this appears to be an almost impossibleto-solve problem using a computer-controlled radio noise meter unless it and the control computer software have capabilities that exceed those usually found in "standard" EMI analyzers or radio noise meters and controllers. In the 10 kBz to 150 kHz frequency range, the standard CISPR radio noise meter [2] has a bandwidth of 200 Hz and the standard ANSI radio noise meter [3] has bandwidths of 200 Hz, 1 kHz, and 10 kHz. In a 200 Hz bandwidth the impulse generator produces a level of -22 dB(!.N), the pulse generator produces a level of -6 dB(pV), and the 22 kHz cw generator produces a level of 64 dB(UV). In a 1 kHz bandwidth these levels become -8 dB(lN), +8 dB(!Jv) , and 64 dB(pV), respectively. In a 10 kHz bandwidth, the levels become +12 dB(uV), +28 dB()N), and 64 dB(BV), respectively. The simulated ignition noise (the impulse generator) which must be measured is far below the interfering signals and may be below the impulse sensitivity of the radio noise meter in a 200 Hz bandwidth. As can be seen from the above data, when the bandwidth of the radio noise meter is made larger to bring the simulated ignition noise UP to a level where it can be easily measured, the bandwidth is so wide that the 22 kHz narrowband signal begins to override the simulated ignition noise in the skirts of the radio noise meter selectivity characteristic. This effectively prevents the detector in the radio noise meter from properly responding to the simulated ignition noise, It may be seen that theproblem of relative noise levels continues on above 50 kHz, 3 - 1 Ad A well trained, experienced human operator using visual techniques with an oscilloscoPe on the radio noise meter output may be able to make some satisfactory measurements, given enough time. A computer-controlled analyzer would need to be extremely sophisticated to do as well as the human operator. perhaps the best that could be done by an automated system, would be to determine that no satisfactory broadband noise measurement could be made in this frequency range under these conditions, and so inform the operator. Because of problems such as this, the US Air Force has seen fit to issue an application note [61 recommending that "official" measurements be made in one bandwidth and compared against one limit no matter what the nature of the EMI, broadband or narrowband. The United Kingdom is in the process of issuing regulations to this effect [7]. Both of these documents assume that measurements can always be made in a low ambient noise environment, such as a shielded enclosure, although this is often not possible. In the current technology, CISPR and ANSI instruments are specified in such a manner as to imply that manual EM1 measurements are to be made. At the same time, the military presumes 153 that some form of automated measurements will be made, and test laboratories performing EMI measurements to comply with military standards are generally making automated measurements. Also, automation has begun to pervade EMI measurements made to comply with standards and regulations, such as those of the VDE [8] and FCC [9], covering consumer electronics equipment. The above discussion applies only to the measurement of EMI, but similar instrumentation problems exist in making interference immunity (susceptibility) measurements. From one viewpoint, the worldwide community of EMC scientists and engineers is better off with respect to making immunity measurements since few regulations exist covering these measurements. This allows those who wish to make immunity measurements much freedom to develop instrumentation and methods that are timely and appropriate. Mr. Heirman demonstrates this in his Paper CU. This does not mean that automated immunity measurements are intrinsically any easier to make or more reliable than automated emissions measurements. Immunity measurements will be addressed by other papers in this session. An Automated Measurement and Analysis System The problem posed above wherein several different signals and noises are superposed was investigated further with the objective of finding a way to automate the measurements and yet obtain valid results. First, the needed attributes of the system are discussed, then ways one might manually measure the various signals and noises are investigated, and finally a method combining hardware and software is realized. The discussion applies to the 10 kBz to 150 kHz frequency range, but similar problems exist in, and similar techniques can be applied to, other frequency ranges. In order to insure that impulsive and cw signals are properly measured, the automated system must be able to make several decisions without manual intervention by the operator. First the system must be able to identify all -4- cw signals. This can be done with the use of a discriminator which recognizes these signals when they are encountered. This is relatively easy to do by monitoring the FM video output of the receiver for a D.C. shift in the output level. Once the presence of the cw signal is determined, a more difficult decision must be made: Is there a significant impulse level superposed on the cw signal? Since all NIL-STD type measurements must be made with a peak detector, the level of the cw signal read includes any additive impulse level; the problem is to isolate the impulsive signal and measure its level. Because of the logarithmic scaling of the radio noise meter output level, a high level cw signal can almost entirely mask an impulsive signal. By way of example, consider a situation where one finds a narrowband signal present at 15 kHz at a relatively high level, 60 dB(uV). A typical source of such a narrowband signal would be switching regulated power supply operating at a switching frequency of 15 kHz. The narrowband signal is one line, the fundamental, of the spectrum of many harmonics created by the rectangular switching waveform, and appears in the 4 kHz bandwidth of the radio noise meter as a cw signal. In addition to this signal, there is an impulsive signal at a level of 50 dB(uV/MHz). Since the radio noise meter impulse bandwidth (6.31kHz) is a relatively large fraction of the tuned frequency (15 kHz), the narrowband signal will appear to have a very wide response envelope, and a considerable signal level will be present at the start of the frequency range (10 kHz). In addition, less than four bandwidths away resides the high level second harmonic of the switching power supply frequency. Contrasting the above narrowband signal is the 50 dE(uV/MHz) impulsive signal. Due tothe fact that at 15 kHz we still have a 6,31 kHz impulse bandwidth, now relatively narrow compared to the reference bandwidth of 1 MHz, we can see only a very small component of this impulsive signal. The actual voltage level will be approximately 6 dH(pV), as shown by equation (1). The change in impulsive voltage level due to bandwidth difference from the 1 MHz impulse reference bandwidth can be calculated as follows: actual BW in MHz ALE = 20 log ( (1) 1MHz AdB = 20 log(O.O0631/1) A~B = -44 do Vi = 50- 44 = 6 dB().lV) One can then change both levels to voltage, add them algebraically, then reconvert their sum back to a level in dH(uV), to find the difference in meter indication that is caused by the presence of both signals simultaneously. These calculations are shown in equations (2.1) and (2.2): x dB(PV) = 20 (2.1) and manipulating to equation (2.2) y )lv= log-l(~) (2.2) First, considering the narrowband signal of 60 dB(uV) using eq. (2.2), y = log-1(60/20), we find 60 dB(HV) = 1000 uV. Next let us consider the 50 dB(uV/MHz) impulsive signal level, which we have already calculated to be ~6 dB(UV) in a 6.31 kHz impulse bandwidth, Using equation (2.2) again, y I.~V = log-1[6 dH(uV)/20], we find this level is =2 I-IV. We now can algebraically add the two volttage levels for a combined signal level of 1002 uV. The next step will be to convert back to a decibel scale to find the meter reading of the combined signals. Using equation (2.1), x dB(uV) = 20 log(1002), we find a level of ~60.017 dB(uV), showing that the 50 dB(uV/MHz) impulsive signal adds a meagre 0.017 dB to the level measured with the narrowband signalonly. d%iV) 605040302010O-lO-2oI FREQUENCY IN KHz Fig.1: Measurement in Wideband with Detector in Peak Position Herein lies a large part of our problem. With the accuracy and precision of most radio noise meters being such that a difference of 0.017 dB is insignificant, and probably unmeasurable, how do we measure a not so insignificant level of 50 dB(uV/MHz)? Let us first consider the options we would have in performing these measurements manually, then we will try to develop an automated method. The major problem interfering with our ability to arrive at a correct impulse level measurement is the presence of a narrowband signal and its associated harmonics. An obvious solution, therefore, would be to eliminate the presence of the narrowband signal, with the use of a sharply tunable notch filter, thus removing the narrowband signal from the spectrum viewed by the EMI analyzer. In the same fashion, using additional filters, one can remove the associated harmonics. Operating one frequency range at a time, being careful not to take measurements on the "skirt" of the filter characteristic, the operator could obtain valid readings on the impulse level present. This method, however, will be cumbersome and may not yield valid results if the encountered narrowband signals are spaced too closely together in frequency. Generally, however, this method can be used to arrive at reliable, valid results, If the operator does not have access to a series of tunable notch rejection filters, another method must be attempted,,The first step would be to decrease the I.F. bandwidth to decrease the frequency range masked by the narrowband signal on the skirts of the radio noise meter selectivity characteristic. The operator must note, however, that by changing the width of the I.F. bandwidth, he is sacrificing some of the impulse sensitivity of the 1 Ad -5- Receiver: NARROWBAND, CARRIER Recorded Level: -10 dB(1_lV) radio noise meter as is shown in Table I. Sensitivity Impulse B.W. 34 dB(HV/MHZ) 6.31 kHz 1.26 kHz 38 dB(FIv/MHz) 97 Hz 50 dB(lJV/MHz) Table I. Typical Radio Noise Meter Specifications Frequency 15 kHz 15 kHz 15 kHz Changing to a 1.26 kHz impulse bandwidth sacrifices approximately 4 dB of impulse sensitivity, but changing to a 97 Hz impulse bandwidth sacrifices 16 dB of sensitivity -obviously too much. From this observation we see that we cannot decrease the impulse bandwidth to less than 1 kHz and still get reasonable impulsive sensitivity. The operator must note, however, that when using a narrower bandwidth, he must apply the appropriately increased bandwidth correction factor to reference to a 1 MHz bandwidth. The bandwidth correction factor calculation is shown in equation (3). where x = correction factor and y = bandwidth used (in MHz) x = 20 log(1 MHz/y) (3) Now that the operator has narrowed the frequency range affected by the skirt of the narrowband signal, he can tune to a point where the narrowband level is a significantly lower portion of the total signal level measured. (The operator must note that the impulse level will have also dropped, probably by about 14 dB, but the narrowband signal level Will have generally dropped significantly more.) The next step that must be performed by the operator is to identify the impulse and narrowband portions of the signal. To do this the operator should change the radio noise meter detector function to a carrier or average detector, tune the receiver to the lowest possible amplitude point on the narrowband signal, and take an amplitude reading in d.B(HV). dB&') FREQUENCY IN KHz Fig.2: Measurement in Narrowband with Detector in Carrier Position The operator must then change the detector function back to peak and take another reading in dB(pV) (See Figure 3). The readings can then be converted back to voltage levels, algebraically subtracted, and reconverted into dB(pV) and dB(pV/MHz) levels respectively. An example of these calculations is as follows: Receiver: NARROWBAND, PEAK Recorded Level: -3 aB(uv) where x = level in lJsr and y = level in dB(HV) x = log'l(y/ZO) For Peak reading converting to HV: x = log'l(-3/20) x = 0.7079 yv For Carrier reading converting to HV: x = log'l(-lO/ZO) x = 0.3162 HV The difference is 0.7079 HV - 0.3162 I_IV = 0.3917 nV. Converting the difference to dB(HV): y = 20 log(o.3917/1) y = -8.1 dB(HV) Calculating bandwidth correction factor using equation (3), x = 20 log(1/0.00126) x = 58 dB Impulse level in dB(yV/MHz) = level in dB(l.lV) + bandwidth correction factor, or -8.1 + 58 y 50 dB(HV/MHZ). dG!J) 6050403020toO-IO- I -2o-, IO I :5 nw&i3 I I I I IN25~H~ :o :i Fig.3: Measurement in Narrowband with Detector in Peak Position Now that we have discovered a viable method of performing these measurements in a manual mode, we must explore a means to arrive at the same results using an automated system, Through experimentation it has been found that the second manual method lends itself very well to automation. The automated system can identify the frequencies of the narrowband signals by monitoring the FM video output. The system can tune through a frequency segment, identify, locate and measure all narrowband signals, storing the data as it goes. Once all signals areidentified in a particular segment, the computer accesses the data it has stored, and analyzes the signal pattern it has encountered. During this analysis the computer locates the best possible frequencies to attempt to gather valid broadband readings. Once this procedure has been completed, the computer then tunes the receiver to the first selected position, and a reading is attempted. A narrow bandwidth is selected, and data is taken first with the detector in the peak function, and then with the carrier function. When data acquisition is completed at this frequency, the computer checks the collected data for a measurable difference in the peak and carrier levels to determine if it is feasible to arrive at an impulsive signal amplitude. Should the analysis show that it is indeed possible to arrive at a -6- valid reading, the computer calculates the impulsive signal level in the manner previously described, stores the calculated data,proceeds to the next previously selected frequency point, and repeats this procedure until all such points are completed in the frequency segment. A problem arises, however, when the computer analysis determines that a valid impulse level cannot be arrived at by the previously described algorithm, At this point the computer notifies the operator that it cannotproceed with calculations at this frequency point and that further manual investigation is necessary, Once the message has been noted, the computer proceeds to the next point to be analyzed. After all data collection and analysis have been completed, the computer adds tranducer factors and other correction or calibration factors, if any, and plots the data against the desired specification limit or reference. u r’ z References [l] Heirman, D.N., "Automated Immunity Measurements," EMC Syqosium & Exhibition, &ich, March 5-7, 1985, Session A [2] CISPR Publication 16 (1977) and Amendment 1 (1980), "C.I.S.P.R. Specification for Radio Interference Measuring Apparatus and Measurement Methods" [3] ANSI C63.2 (1980), "American National Standard Specifications for Electromagnetic Noise and Field Strength Instrumentation, 10 kHz to 1 GHz" [4] NIL-STD-461B (1980), "Military Standard, Electromagnetic Emission and Susceptibility Requirements for the Control of Electromagnetic Interference" [5]MIL-STD-462 (1980), "Military Standard Electromagnetic Interference Characteristics, Measurement of" [6] MIL-STD-462-Application Note (1980), "Identification of Broadband and Narrowband Emissions," Aeronautical Systems Division, Electromagnetic andInterference Compatibility Branch, (ASD/ENAMA), Wright-Patterson AFB, Ohio 45433 [7] united Kingdom Def. std. 59/41 [8]6DE 0871/6.78, "VDE Specification, Radio Frequency Interference Suppression of Radio Frequency Equipment for Industrial, Scientific, and Medical (IsM) and Similar Purposes" [9] Rules & Regulations of the FederalCommunications Commission (FCC), Part 15, Subpart J, "Computing Devices," and FCC/OST MP-4 (1983), "FCC Methods of Measurement of Radio Noise Emissions from Computing Devices" [lo] van Essen, J.C., "Instrumentation of Automated Electromagnetic-Compatibility TestFacility for Space-Craft," EMC Symposium & Exhibition, !&rich, March 5-7, 1985, Session A @.l]Eumurian, KG., "Computer-Aided Control of EMP Measurement on Large Scale Sys_. terns,"EMC Symposium & Exhibition, Ziirich, March 5-7, 1985, Session A . -E-z3 .-L____-_______J: --c_-_______ t: still further, but the number of operator interventions can be greatly decxeased by using this automated procedure. The technique described above is in opposition to that suggested by the ASD application note ES], but it provides the correct data under non-ideal measurement conditions. The approach suggested by the application note cannot provide the correct data under similar non-ideal measurement conditions. Also, we have not addressed the proper application of transducers, so if the mandatory measurement method requires a theoretically unsound use of a transducer such as an antenna, we may still be collecting much incorrect data. : . _-__-_______ ----_I-_____ -30 IO 20 FREQUENCY 30 IN 30 20 IO FREQUENCY IN 35 KHz 35 KHz Fig.4: Automated Systemp;;;dband/Narrowband Proceeding in this fashion, the computer (controlled radio noise meter) can perform a complete EMI emissions test, collecting large quantities of data and undertaking an immense number of mathematical calculations, in a relatively short period of time. Unfortunately, there will be situations encountered that require the judgment of the human operator, showing that inquiry and development must proceed 2A2 -l- AUTOMATED IMMUNITY MEASUREMENTS Donald N. Heirman AT&T Information Systems Holmdel, New Jersey 07733 USA Automation of EMC Testing should be viewed as an engineering tool and not as a replacement for the engineer who must determine compliance with either regulatory or corporate EMC criteria. As a tool, automation if implemented properly decreases the likelihood of measurement error due to operator inattention, test instrumentation misadjustments, and inability to recreate all the test conditions on a repeatable basis. The cost of automation must also be weighed against the increased test time normally associated with manual operations. This paper will address the proper use of automation in immunity testing. The areas where automation is most useful are shown by describing a typical immunity test using a transverse electromagnetic (TEM) cell. Introduction In recent years, the proliferation of a wide range of RF noise sources from commercial broadcast stations to microprocessor-controlled appliances have increased concern for product susceptibility. Of course, in military systems, the need for product immunity (the positive view of susceptibility) is vital for strategic and tactical systems. On the other hand, consumer product immunity is generally designed to respond to pressures of the market place. A too sensitive product to the RF ambient would cause customer complaints and lead them to purchase a competitor’s product. The sheer magnitude of immunity testing has created much automation in testing in an attempt to meet production schedules and to ensure that the product has been made immune to all sorts of RF environments. The advantages and in some case the disadvantages of automation of susceptibility tests are presented from the viewpoint of the test engineer involved with consumer products, Automating Engineering Evaluation Stage Even before a product is well along towards prototype or preproduction models, testing can be used to assess the relative immunity of the product during the development cycle. At this time, it is more important to get sufficient data in a short period of time to evaluate immunity progress. This phase is generally called engineering evaluation. Here automation can provide a quick view of product immunity. All the test parameters can be held constant from test-to-test, especially when instrumentation is computer-controlled and the product response automatica(\y recorded. During engineering evaluation with incomplete or laboratory models, it is generally more important to see if there is any immunity response at all with the minimum test time. Typically, levels higher than the anticipated RF ambient are applied with a frequency scan rate faster than that for a full response of the equipment under test (EUT). The higher field, faster scan is traded for a lower (and perhaps closer to the design immunity limit) level and slower scan for full response. Here automation is a requirement since an operator may not be able to keep up with all the necessary instrumentation settings and EUT monitoring. Response algorithms based on scan rate and frequency response can be written to guide the chosen scan speed. These algorithms can be used and evaluated to ensure that the final compliance test is truly respresentative of the product immunity. The engineer evaluation period also provides a time to fabricate sensing hardware and to adapt automation software to determine better the actual EUT performance degradation as a function of applied level, scan rate, and type of applied signal (AM, PM, FM, impulse, etc.). The need for automating the remote operation of the EUT is also studied during this time. Such operation might be controlled by remote computer terminals, load simulators, or the instrumentation controller itself. If mechanical operations are needed, pneumatics may be used. Recreation of the Immunity Field The RF environment is a complex one in both time and frequency domain. Electronic products generally respond undesirably to certain frequencies and waveshapes, not the aggregate. This response is documented primarily by studying interference cases or by testing to several representative ambient signals. Unless specifically designed for immunity, commercial electronic product performance can be expected to degrade at some point during the life of the product. The seriousness of the degradation may or may not warrant design or in-the-field changes. Examples of degradation include: 1. 2. 3. 4. 5. 6. 7. Increased bit error rate Erratic Operation Abnormal modes of Operation Audio Rectification Seize up modes M,iscalibration Component failure There is always a problem with the ability of any transducer to recreate the in-use praduct RF envlranment. The literature in the USA shows that the vast majority of RF data taken is associated with commercial broadcasters and other licensed radio services.[l,2,3,4]. Hence most immunity tests attempt to recreate these narrowband radiated fields. There &much less data on RF conductedinto the product via the AC power mains and other signal or interconnecting cabling.[5] There is even less data from impulsive or aperiodic signals produced by switching transients and other localized fields such as that from a cooling fan for a power supply. Of course, special RF surveys can be made to better describe the actual ambient at product locations. This requires considerable time and must be funded for proper instrumentation. As a conse- quence, mOSt manufacturers rely on data already contained in the literature in setting up immunity test levels. System Immunity Test A typical immunity test program would include both radiated and conducted tests to include the following: Radiated 1. 2. 3. 4. Electric Field Magnetic Field Electromagnetic (Plane Wave) Field Impulsive Noise Conducted 1. 2. Immunity Immunity Direct Coupled Near-field coupled The radiated test would expose the entire product to a radiated field. The associated peripherals, l/O cables and other subsystems would also be simultaneously exposed. The conducted test concentrates on powerline and signallcommunications lines. The above list is not exhaustive. It is clear that any automation to help relieve such an extensive test program activity is highly desirable. To focus our attention on one of the most used tests, this paper will concentrate on automation of radiated immunity tests where the radiated filed is a narrowband electric field typical of AM, FM, or TV broadcast fields. The basic instrumentation for creating broadcast fields in a controlled chamber is comprised of an RF generator, modulation source, power amplifier, and transducer (antenna). Since broadcast transmitting antennas are generally far enough removed so that a plane wave is incident on the product, the presence of the product does not cause the transmitting power to increase or decrease. However, the field in close proximity to the product, does differ from that with the product removed. Most RF environment surveys measure the field with antennas removed from any object that would affect the measurement including the ground, i.e., the antenna is located several wavelengths above ground except for AM broadcast. This is close to measuring a free space value of field strength. Hence, we want to recreate that fiefd in a controlled manner. All such RF environment simulations have the potential for errors. Immunity measurements even at open area test sites must contend with and account for the ground reflection. Measurements made in enclosed chambers have even more reflections if the walls are not anechoically treated. That leaves few choices of test facilities that readily approximated free space. One choice is a parallel plate capacitor (stripline antenna or a transverse electromagnetic (TEM cell). Both provide a plane wave for frequencies within the passband of the transducer. RF anechoically treated shielded rooms (all six surfaces) are yet another choice for free space measurements. However, anechoic chambers are usually more expensive. Oncethe type of measurement facility is selected, a method for automating the recreation and the monitoring of the immunity field can be implemented. Generation of the necessary fields are relatively straightforward and will not be discussed further. The monitoring of these fields is not straightforward and great care must be exercised in monitoring the field around the EUT. The most popular monitoring procedures are: 1. 2. 3. Real time leveling using a field probe next to the EUT. Recalling from controller memory signal source drive power based on previous measurements of field strength in the test volume with the EUT removed. Recalling from memory the source drive to set a desired field strength based on the calculated field using antenna gain, radiation pattern, and signal level input. Item 1 has the potential for monitoring a field that is largely affected by the EUT, especially at frequencies where the EUT resonates. Items 2 and 3 are preferred if the EUT does not significantly interact with the transducer to affect the calibration of the applied signal. Both of these latter items to be fully implemented require automation to look up the calibration data and control the signal input into the power amplifier. A Sample Automated Susceptibility Test To further focus on the benefits of automated immunity testing, we describe a typical test using a TEM cell as the radiating transducer for launching an RF narrowband electric field ambient. Figure 1 shows typical instrumentation. TEM cell testing provides a passband of operation from dc to a frequency where the dimensions of the cell are approximately equivalent to a wavelength. For a cell with dimensions as shown, the useful upper frequency is about 165 MHz for EUTs with dimensions of up to 10 by 30 by 30 cm. General test guidelines are contained in Reference 161.We now expand those guides for this example. First, the EUT dimensions should be kept small compared to the dimensions of the cell’s test volume. If not, errors in the applied test field will increase due to the capacitive loading of the EUT. Generally the linear dimensions of the EUT should be kept to no more than about 30 percent of the associated test volume dimensions in either the top or bottom half of the cell. The far field immunity level at the center of the test volume (midway between the center conductor and ground plane) can be calibrated by several methods with the EUT removed: Monitoring input RF voltage using a monitoring Tee for frequencies typically below AM broadcast frequencies. Monitoring net power flow into cell using incident and reflected power and a bidirectional coupler. This technique can be used for all frequencies within the passband of the cell. Monitoring the electric field directly with probes. The first two aproaches are accomplished external to the cell which has distinct advantages since no cables exposed to the high RF field. Automation is virtually a necessity to keep track of these levels and to perform net power flow calculations as well as repetitively calibrate the meters. The last approach requires the most care. In the probe approach an optic link is generally required to not disturb the field or become a radiating or scattering structure. The placement of the probe is also critical since at EUT resonance, for example, the field is most perturbed and a probe in the near or scattered field will indicate fields that are different from the nomimal test level. During the actual immunity test, the fields can be monitored using the above three basic approaches. The levels will differ from nominal due to the loading affect of the EUT. If the dimensions of the EUT are kept to the 30 percent test volume criteria, the level differences from nominal will be in the order of &36 dB under cell multimode frequency. The most useful way to evaluate what is happening to the electric field is by using an electrically short dipole or monopole probe. To avoid the near field scatter problem at EUT resonance, these probes can be placed in the half of the cell not occupied by the EUT and at a point which is the mirror image (about the cell’s center conduct) of the geometrical center of the EUT. Above multimode, placement of the probe becomes much more critical to remotely monitor the field at the EUT. The differences between the nominal immunity level and what is read by the probe significantly increases making this monitoring method less useful. Characterizing the effects of all the monitoring methods is a useful undertaking. For example, one of the benefits of such probing may be to extend the useful upper frequency limit of the cell or to allow larger EUTs. In practice, the TEM cell can 2A2 BIDIRECTIONAL MODULATOR CONTROLLER - APPLIED - TELEMETRY/CONTROL IMMUNITY SIGNAL LEADS LEADS PERFORMANCE DEGRADATION SAMPLE NOTE: ElJTlPROBE MONITORING CABLING VIA TEM CELL FIBER OPTIC OR HIGH IMPEDANCE LINES INSIDE CELL. Figure 1. Typical TEM Cell Immunity be used above its normal upper frequency limit. In this frequency region, the field strength is complicated by the multimoding of the cell. Only through use of automated data gathering techniques can the cell be properly mapped to determine the field throughout the test volume. The mapping would be much too cumbersome using manual techniques for recording the orthogonal (and total) field components. In this case automation is the only practical way to extend the test capabilities of the cell. Performance Degradation Monitoring The next area where automation helps is in recording performance degradation as a function of applied field strength, frequency, modulation, degradation type, EUT response time, etc. Much of this is simple data bookkeeping. However, there still persists those who want to visually determine performance degradation. If degradation monitoring were constantly done by this means, especially by viewing a CRT, errors will soon occur due to the long, repetitious and boring nature of immunity tests. No matter how conscientious the operator, monitoring of anticipated, slow to materialize, visual EUT degradation is prone to errors and lack of repeatability. Typical automation of performance degradation would include monitoring analog signals directly onto the IEEE 488 general purpose bus or digital information on an RS 2326 interfaces DIMENSIONS I = 200 cm w = 95 cm h = 65 cm Instrumentation cable. These methods: signals are routed to the bus by one of 2 a. Direct connection to monitoring points within the EUT via fiber optic or high impedance transmission lines. b. Indirectly via connection to EUT performance monitors, external controllers, external circuitry, simulated loads, or peripherals, all of which are not in the test chamber but are connected to the EUT via cabling. The former method requires several telemetry links not part of the EUT. The latter relies solely on using part of the EUT system that is not exposed to the high fields, except of course for the interface cabling inside the test chamber. Proper filtering of these leads through the TEM cell walls are needed to protect the equipment outside the cell from RF on the cables extending through the walls of the cell. Once the degradation is recognized by the computer, preprogrammed operations can be implemented. Some operations are shutting down the amplifier if a destructive level of degradation is reached, sequencing to other EUT modes of operation, and pausing on particular frequencies to evaluated EUT response time to the applied field. Other instrumentation activities can also be conducted while immunity is being recorded. For example, in the TEM cell, 10 - there are relatively high cell Q’s above its normal upper frequency (multimode) limit, the output of the signal source power amplifier chain should be filtered so that the second and higher harmonics are suppressed by at least 60 dB. This will avoid a false EUT response at the signal source frequency when in fact the response is due to a harmonic of the applied signal (generated by the amplifier) which is coincident with a multimode response. Automatic switching of low pass filters is a necessity since the test engineer is concentrating his attention on the EUT degradation and operation and could easily forget this switching detail. It cannot be overemphasized the importance of spending the extra time to automate the performance degradation monitoring. The test controller can do most of this, especially if all degradation can be sent to the controller using analog (via an A/D converter) or digital (via the IEEE 488 bus or EIA RS-232 telemetry) signals. The test engineer should where at all possible take advantage of performance monitoring by sensing signals on the same leads which remotely operate or communicate with the EUT from outside the test chamber. This will avoid introducing additional cabling which itself might be vulnerable to the applied field causing a false degradation indication. Final Immunity Evaluation Once engineering evaluation of the product immunity is finished and suitable mitigation applied, the final compliance test is performed. This test must be highly repeatable and calibrated to judge compliance. Here automation will significantly increase the test repeatability and ensure that separately derived calibrations are always used. These tests are generally longer in duration since the full range of performance degradation is checked and recorded for the final test report. This phase is particularly methodical and a great deal of degradation bookkeeping is necessary. For example, the frequency scan rate may be varied to ensure that a degradation response is not missed. The affect of the complex field within a TEM cell above multimode has to be accounted for here if used. It may be necessary in the multimode range to move the EUT in the cell to expose various circuitries to the full field gradient caused by the standing wave pattern which can amount to field uniformity errors in the order of 10 dB or more. Even under multimode, there are undesirable TE and TM modes launched that at the very least should be accounted for in the measurement error. All of these factors are best recorded and controlled via a well thought out and planned automation program. Conclusions This paper described the usefulness and precautious of automation of immunity testing. Automation if used properly is a powerful tool that can be used to produce a test with less operational errors. However, automation which is not periodically checked by manually performing a test, tends to lull users into a sense that the results of such tests are irrefutable. Periodically it pays to manually set all instruments and see if the results are the same as that found by automation. The paper has also shown the concern for ensuring that the EMC engineer correctly automates the immunity test to replicate the appropriate immunity field and to monitor the proper perform degradation. References PI D. E. Janes, R. A. Tell, T. W. Athey, and N. N. Hankin, “Nonionizing Radiation Exposure in Urban Areas of the United States,” Proceedings, IVth International Radiation Protection Association, Vol. 2, pp 329 - 332, April 1977. 121 R. A. Tell and N. N. Hankin, “Measurement of Radiofrequency Field Intensity in Buildings with Close Proximity to Broadcast Stations,” U. S. Environmental Protection Agency Report ORPIEAD-78-3, August 1978. 131 D. N. Heirman, “Broadcast Electromagnetic Interference Environment Near Telephone Equipment,” IEEE National Telecommunications Conference Record, Catalogue No. 76 CH 1149 - CSCB. 141 G. Costache et al., “Electromagnetic Field Strength Probability Profiles for Canadian Cities,” International Electrical and Electronic Conference and Exposition, Toronto, Canada, October 1981. [51 FDA Medical Device Standard, “Electromagnetic Compatibility Standard for Medical Devices.” MDS-201-0004, October 1, 1979. PI M. L. Crawford and J. L. Workman, “Using a TEM Cell for EMC Measurements of Electronic Equipment,” U. S. National Bureau of Standards Technical Note 1013, April 1979. 3A3 1. Abstract The purpose of this paper iS t0 give a complete overview of an Automated EMC Test Facility in operation, for Emission-, Susceptibilityand Time domain measurements. The contents include system set-up, specifications and drawings with a description of different test set-ups used for spacecraft, subassembly or unittesting. The narrowband and broadband aspects are highlighted, and a plot of the data output from the system are included. Conclusions are drawn with respect of specifica' tions, test time, accuracy etc.. 2. History Since the time when the basic idea of automating the EMC Test Facility was conceived and initial funding became available, the line of thinking had changed quit a bit. Due to improvement of the test equipment, measuring techniques and the budgetary constraints, the original idea of setting up a separate system for Emission- and Susceptibility- Testing had to be abandoned. Instead, a new design was set up in such a way that all instrumentation performs a multiple function and will be used for Radiated/ Conducted Emission and Radiated/Conducted Susceptibility. The existing test equipment is integrated in this system as well. A block diagram,, Fig. 1, shows the basic set up. Due to the fact that a broadband high power requirement will increase the cost of the radiated susceptibility part of the system by 100% or more, it has been decided to keep the requirements of 30 - 60 V/m and to accommodate projects with the necessary power at a given frequency "narrowbanded", which complies with the experience so far. All specifications for this system are derived from spacecraft requirements existing today and in the near future. The system has been designed to meet these requirements. 3. Introduction The system will be used for the following measurements: A) Radiated Emission Measurements over a frequency range from 20 Hz - 40 GHz, B) Conducted Emission Measurements over the frequency range from 20 Hz - 100 MHz, C) Radiated Susceptibility Measurements over the frequency range from 20 Hz - 40 GHz, D) Conducted Susceptibility Measurements over the frequency range from 20 Hz- 100 MHz, With the possibility of injecting CW and pulsed signals, to test power-, signal- and commande lines. The block diagram of Fig. 2 outlines the complete system set up. FIG.1 x- TEST -FACILITYGENERAL SET-UP All equipment used in this system is operated to IEEE-488 standards or equivalent. The possibility of opto-coupler extention is provided for operation in a radiation-hazardous area. The System Controller is a desktop model with a large screen, so the facility engineer is able to program it and to modify the software during the test. For the sweep section we stay as long as possible co-axial, in order to facilitate the test work. However, a synthesizer is a must, due to the frequency accuracy required. The exact specifications of each subsystem will be discussed separately. For the Radiated Susceptibility testing we have for the Low frequency range, Amplitudemodulation, for the Megahertz range Frequency-modulation and for the Gigahertz range Pulse-modulation. - 2 PRINTER /_ HP-9876 A + P-F9 HP-B112 A w CURRENT-FKE?E SOLAR. 6741 1 0: PULSE GEN. MOO SOURCE I $u-6. N &JNO 4. SCKUM AN4LmR HP-B566A lCQi-4OGHz TWT 1077 H12 17 FIELD MONflOR CONOUCi% D -SUSCEPTIBILITY SETUP FOR RADIATED -I CONDUCTED EMISSION NOTE C.5W - COAXIM RL - REIAY t pp&LJ-:_a\,; L. v+ r---- --7 HP-37203 A I :_______--,‘,‘_______I I 15 EXTENDER ,’ ’ .’ ,’ RADIATED-! CONDUCTED SUSCEPTIBILITY SWITCH ESTEC %3?2iii;,~____; / \,--.’ EMC TEST FACILITY FIG. 2 _ - Amplifiers from 20 Hz - 18 GIIzwere existing in the EMC Test Facility and are integrated into the system. This together with two new amplifiers; one from 18 GHz - 26.5 GHz - "K" band and one from 26.5 GMz - 40 GHz - "KA" band. In this way only one waveguide and standard gain horn will have to be used for each amplifier. All cower for the radiated susceptibility test will be fed through a dual directional coupler to the antennas. In order to control the levels on the antenna a Dual-Sensor-Powermeter is used in order to measure set-power and reflected power. For the radiated- and conducted emission a front-end receiver is used. It has to supply the necessary pre-amplification over a frequency range from a minimum of 100 Hz to 18 GHz, and preselection below 2.4 GHz. The unit will be used with the HP-8566 spectrum analyzer. Blockdiagram Fig. 3. This pre-selection below 2.4 GHz increases vulnerability to overload, especially on broadbandnoise, and maximizes sensitivity, while at the same time maintaining the highest possible instantaneous system bandwidth and dynamic range. In addition to this receiving system a broadband electric field antenna is used, working over a frequency range from 20 Hz - 1 GHz. Apart from the existing spectrum analyzer, a second analyzer has been introduced, partly used for susceptibility testing. This in order to control the injected voltage and current on the line under test. Due to the fact that scientific spacecraft are low-noise, but still produce noise with its own frequency spectrum, the susceptibility levels to be injected are small, depending on the applicable voltage- or currentlimit, which can be in the order of 20 dBuR. In order to link the system together several co-axial switches and relays are used, positions and number are shown in the blockdiagram of Fig. 2. To control system operations and monitor the behaviour of the U.U.T. (unit under test) a digital voltmeter combined with a data acquisition unit is used. The data acquisition unit is important for the system, it includes: channel multiplexers, relay multiplexer, voltage- and current DAconverters and a real time clock. Also, time domain measurements are an integrated part of the test activities. A bus controlled oscilloscope is used for this purpose. And, last but not least, a printer iS intergrated with a dump feature, which gives out the test data, like: a narrowband plot containing the narrowband signals only includiny spec-level and frequency printout with measured levels. An example of a plot is included. The same can be achieved for broadband measurements or conducted- and radiated susceptibility measurements. In fact this system produces a complete testreport with detailed information directly after completion of the test. 4. Description and Specifications In this section the system will be described with reference to blockdiagram Fig. 2, and broken down in subsections in order to have a better overview of the system. 13 3A3 - IF-DISPLAY SECTION RF SECTION I Ym.l.b.___._L__=i._ .I7 I_.__._____-__.-.___._.d.q d1.4 :m-? 521.4 MHz IF IN MHz IF OUT :...._..._.......__..____....................................: FIQURE 3 MOOEL 254EC RECEIVER SYSTEM. BLOCK DIAGRAM 4.1 Control part The "control part" consists of a HP-9836-S calculator and a HP-9876-A printer. The main reason for using a "desktop" calculator is that the controller can be handled and programmed by the facility engineer directly, without having to ask for software support, which means no loss of operating time, maintaining and updating of a more complex system Apart from the above, it:is of utmost importance that the "EMC-Engineer" on the job can translate his EMC problems directly into the machine, without external software support. This in order to avoid unnecessarily complications of already complex problems. The calculator has a 12" CRT and two built-in disc-drive units for 5%" floppy disc. Memory capability expendable up to 2 Mbytes, with Basic,(extentions) Pascal, graphics dump and storage. This system is also fitted with an extra HP-IB and BCD interface. More than 4 bus expanders can be added to provide 16 additional slots for memory- and I/O cards. In addition there is an HP-IB extender with fibre-optic interface. The use of fibre-optic links has a special meaning for this system. Due to the fact that in our case the system isa combined one, and used as well for "Emission Measurements!' as "Susceptibility Measurements". It is very important to separate the transmitting and receiving sections to avoid unnecessarily problems due to small beaks, cable- or small ground loops, which would limit the dynamic range of the equipment. The use of fibre-optic links has imporved EMC measurements quality considerably and is now standard in our facility. The thermal graphics printer can handle a graphics dump from the 9836-s CRT within 10 seconds. In this way, the test data are available seconds after the test, a protection against nailbiting and nose eating customers who are nervously waiting for the test result to be produced, to see if they are in- or out of specifications. 4.2 Receiving part The receiving part consists of two units: A- SRD-2548-C precision wideband front-end receiver. B- HP-8566 spectrum analyzer. A more detailed blockdiagram is given in Fig. 3. The specification is the combination of the two instruments. Up to 2.5 GHz pre-selection is obtained by pre-amplification. Above 2.5 GHz we have Yig-pre-selection. Use of this front-end receiver in combination with the spectrum analyzer has imporved the sensitivity, dynamic range and instantaneous band width of the spectrum analyzer without losing any of its features. The system has been set up to operate over the frequencyrange from 20 Hz - 40 GHz. However, from 20 Hz to 100 Hz, extra care has to be taken due to the fact that we have to work so close to the local oscillator and having to extrapolate the antenna factor. The combination can perform measurements up to 18 GHz with a typical noise figure of 10 dB and a dynamic range from 72 dE to 1 MHz BW. From 18 - 26.5 GHz the analyzer is used with a harmonic mixer type HP11970-K and from 26.5 - 40 GHz with a harmonic mixer type HP-11970-A. With a noise level of approximately -110 dBm by 1 KHz. Bw. For equipment layout see Fig. 4. In addition to Fig. 4 we have Fig. 5. Showing the same set up, but with the HP-11517-A Bias Mixer. This has the disadvantage of 20 dB less sensitivity, plus the fact that each frequency line has to be investigated. Must be manually adjusted. Additional software-driven routines allow for automatic calibration of external sensors, such as antennas and current probes. The calibration routines will accept and store externally derived calibration data. Another important feature included is the overload sensing and warning in all the critical areas of the RF signal path. If a signal overload condition ever occurs, automatically the signal gain will be decreased in the appropriate part of the system. In the normal remot digital control mode the interface connects directly the receiver with the analyzer. In this way the receiver controls the analyzer and the entire receiving system need only appear as one device for purposes of addressing and control. In this role the system is both a listener and a talker. All data transfer functions from the spectrum analyzer display section are retained. Fig. 5 The combination of the calculator, front-end receiver, and spectrum analyzer is the most powerful tool for EMC measurements I have seen sofar. It is able to step from one frequency line to the next and evaluate each data point for narrowband or broadband criteria (according to Mil-STD in our case). If necessary at the same time coherent and incoherent broadband noise can be separated, and narrowband and broadband data can be graphed on separate plots and each individual frequency point can be printed out. For analyzing the test results is this a very important piece of information. In our set up a narrowband emission plot will take about 30 minutes (20 Ilz - 1 GHz) Fig. 6 This in combination with a relatively new type of receiving antenna, type SAS-1D from Antenna Research Associated Inc. (Fig. 6) It is an electric field antenna over the frequency range from 300 Hz - 1 GBz. The low-band circuitry is such that the response rolls off at the rate of approximately 20 dB per decade of frequency below 300 Hz. The system consists of two electrically separate antennas, namely a top-loaded monopole for low-band and a discone for high-band. The yolarisation is vertical and the directivity - 15 is omnidirectional. Overload for 1 dB comperssion: Lo-hand 0.5 V/m, Hi-band, 0.1 V/m. The antenna is pOrtable and especially suited for indoor applications. For all our conductive measurements we use the well known Solar current probe type 6741-1, frequency range 20 Hz - 100 MHz, which has a flat frequency response over the frequency range 10 KHz - 100 MHz. Maximum current: 300 Amperes ac or dc;load Impedance: 50 + j 0. ohms. Direct connection to the conductor is not necessary, since the probe may be opened for insertion of the conductor. 4.3 Sweep Section To cover the frequency range of 20 Hz-40 GHz required for this system, three instruments are used: A- HP--8165-A Frequency Synthesizer and Function Generator. 0.1 Hz-50 MHz. B- BP-8673-D Frequency Synthesizer. 50 MHz - 26.5 GBz. C- WJ-1204-40 Milli!meter-Wave Frequency Extender. 26.5 - 40 GHz. The HP-8165-A programmable signal source is a versatile function generator with good accuracy. Microprocessor control ensures rapid programming amplitude output from 10 mVpp 10 VPP, amplitude- and frequencymodulation. The HP-8673-D synthesized signalgenerator has precise signal simulation capability. The frequencies are derived from a quartz crystal time base, via a direct synthesis technique providing extremely low signal sideband phase noise. Harmonically related spurious C-60 dBc.SSB Phase noise<-80 dBc. 10 KHz offset +6 dBm output level at 26.5 GHz. at 10 GHz Leveled calibrated output to -100 dBm. Amplitude, Depth 0 - 908, pulse on/off ratio: >HO dB and frequency modulation maximum c peak deviation is smaller than 10 MHz or (see data sheet). All functions are programmable, including frequency output and RF level setting (in 0.1 dB steps). The same synthesizer is used to feed the Watkins-Johnson frequency extender WJ-1204-40. The most notable feature of this system is its excellent frequency resolution, accuracy and stability. Input power 0 dBm, output power +3 dBm. As stated in the introduction, one of the aims was to stay as long as possible co-axial, in order to facilitate the test work. However, one has to realize that starting from 12 or 15 GHz and going up, the attenuation is increasing tremendously and special attention has to be payed with respect to the length of cable, connectors etc. For this and several other reasons we kept our equipment as mobile as possible and derived a great benefit from it sofar. 4.4 Modulation part This part consists of two instruments which are used for multiple purposes, such as radiated- and conducted susceptibility testing, for testing as modulation sources and also for conducted spikes and commandline testing. From the HP-8116-A, pulse function generator all functions are bus controlled and provide sinewave, squarewave and pulses over the frequency range 100 mHz to 50 MHz, pulse width: 10 nS - 999 mS. Amplitude 10 mVpp to 16 Vpp. 3 - A3 The second instrument in this section is the HP-8112-A, programmable pulse generator with the following specification: : 950 m sec. - 20 n sec. Pulse period : 950 m sec. - 65 n sec. Pulse delay Double pulse : 950 m sec. - 20 n sec. Source impedance : 50 ohm. Output voltage: +/- 8 v into 50 ohm. 4.5 Power Meter Here we use a dual sensor power meter type HP-438A, with a frequency range from 100 KHz26.5 GHz, using the HP-8485-A Thermocouple power sensor. This has been introduced to ._/ BLOCK DIAGRAM OF THE 438A AND ITS TWO SENSORS Fig. I control the power and reflected power on the transmitting antennas. Measurement modes are A, B, A-B, B-A, A/B and B/A. The power range is sensor dependent, dynamic range 50 dB. The use of the power meter is entirely based on Mil-STD testing, which implies that the electric field is calibrated with the transmitting and receiving antenna one meter apart in an empty room, the empty room being the EMC Test Facility covered with absorbing material, in order to reduce reflection. Power levels are taken and stored in the calculator and called up during the test to set the levels. Diagram on Fig. 7. A software routine is set up to determine if the antenna is radiating, check the level and compare the reflected power etc. The been I_- use of electric field sensors has-completely abandoned, under the assumption that no source will increase the radiation power in order to satisfy the form-factor of a given object. 4.6 Monitor part This part consists of 4 separate units. Blockdiagram Ref.nr. 11, 12, 13 and 14 (Fig. 2). First instrument in sequence is the HP-8566-S, spectrum analyzer with a frequency range from 100 Hz - 22 GHz, using external harmonic mixing with a frequency range up to 40 GHz. Amplitude approximately from -137 dBm to +30 dBm, resolution 0.1 dB. Dynamic range 95 dB. Accuracy 2.2 dB over the frequency range from 100 Hz - 22 GHz. With internal software routines like Peakseards-Signaltrack, Signal identification Marker aided measurements-Max hold and saving of control setting. The above mentioned features are a must if a - 16 - ESTEC E ELECTRIC real EMC-Minded program has to be set up. Ref.nr. 12 represents the Data Acquisition unit HP-3947-A including the extender unit. 5% digit DVM which may be programmed for 300 readings per second (3% digit mode) or 50 readings per second (5% digit). It consists of a 40 channel relay multiplexer with a power rating of 1 VA per channel (170 Vp max), relay contact 1 Ohm, crosstalk -40 dB, 32 channel Mercury wetted relays are added, with 100 VA per channel (100 Vp max). Contact resistance 400...mOhm. Crosstalk -30 dB. Further we have a Dual Current D/A converter with an output from 0 to + 10 Volt. These sources can be used to provide a programmable test stimulus or to control voltage programmed devices like power supplies and VCO. And last but not least, this "Data Acquisition Unit" contains a real time clock to support all data output. With Ref.nr. 13 we have our HP-3437-A, High Speed System Voltmeter, 3% digit, voltage range 0.1 - 10 Volt with more than 5000 readings/set. in peak- or RMS value. One of the more important functions of this monitor sub-section is to control susceptibility testing. Mainly, for conducted susceptibility, we have full control on injected voltage and current and are able to plotbZ-. To be in full control means y ou have your ~__ hand on the problems. All this automated equipment -seems to be a major investment, which indeed it is, but it pays back in quality, quantity and time, highly rewarding I would say. 4.7 Amplifiers/Antennas All amplifiers are well defined on the blockdiagram Fig. 2. The measuring set up for Rad.susc. is shown in Fig. 8. We work over the frequency range from 20 HZ 40 GHz. From 20 Hz - 250 MHz : 100 Watt power. From 250 MHz - 18 GHz: 10 Watt power. 1 Watt power. From 18 GHz - 40 GHz : Electric field level is shown on Fig. 9. From the diagram in Fig. 8 we can see that the EMC Community is in great need of a much more -. effective antenna between 30 - 100 MHz. Nowever, one has to consider that this is the standard available power. But from project to project it will be investigated if an effort will be made to increase the power and field level. M C Test facilities. FIELD-STRENGTH GENERRTION Fig. 8 HP-8165 HP-8673 SYNTH /GEN -0 20Hz.-50MHz SOMHz.-26GHz HP-3497-A DATA-ACQ.UISITIOA UNIT RADIATED SUSCEPTIBILITY TEST SET-UP C _?.mnll L Gk12 ?._..I LB GHZ Fig. 9 The Blockdiagram indicates also frequency range and antennas used. For more detailed information, please refer to the data sheet. Generally speaking we can say that amplifiers, couplers and antennas are harmonized to the maximum extent possible as regards frequency range and power. 5. Remarks on Measurements 5.1 Radiated Emission I would like to highlight briefly our measurement criteria. The narrowband signals are measured in 6 scans from 20 Hz - 22 GHz, in standard spectrum-analyzer settings, with respect to "span" etc. Each band is scanned with 2 different band- HPIR SYNTHJGEN -- HP-438 -A DUAL-SENSOR POWER-METER HP-9836.. S SYSTEM CONTROLLER .FREQ.RANGE-IOKHz-40GHz 3 - 17 - A3 Conducted measurements are also carried out in the time-domain. Although it seems that this measurement is not EMC related, it is very important to be able to correlate your Frequency Domain measurement with your TimeDomain measurements. The instruments used are arbitrary; in our case we use standard equipment together with a digital storage oscilloscope: Type TEK-468. BW-ZOOMHZ, buscontrolled. The use of a digital storage oscilloscope has the advantage of good triggering possibilities and thanks to the bus you are able to dump the picture on the printer, including all information. 10 MHz Fig. 11 widths a factor 10 apart, and amplitudes from the first and second measurement are compared with a 3 dB criterion in our case. Correction factors are added before printing the signal output. Fig. 10 shows a narrowband plot. Broadband noise is measured in 4 scans from 10 KHz - 22 GHz. Scans are made in the Peakhold mode with a relatively wide bandwidth. The time set to fill each "Bit!'is equal to the data from an impulse signal with a separation of 50 Hz. Each data point contains half impulse bandwidth, and stored in a temperaly file, is compared with 6 data points before and after; if 4 3 dB (in our case) it will be processed. Fig. 11 shows a broadband plot. 5.2 Conducted Emission For conducted emission in the frequency domain we opt for the same criteria of signal processing as we did for radiated emission. Measurements are carried out on power-and synchronization lines in differential- and commonmode in voltage and current. Data- and command I lines are tested in bundles of wires, separated like; all digital input lines. all digital output lines. all serial digital data. all analogue lines. The signal ground is always tested separately, due to the importance with respect to the quality of collected- or transmitted data. Further more I would like to draw your attention to what we call a "Structure noise Test". Fig. 12. This test supplies us with important information, like leak resistance and capacitance, loops etc. This is important when making the final analyses about the unit under test. Fig. 12 CONOUCiED EM STRUCTURE -- NOISE CURRENT 5.3 Conducted Susceptibility _-.-__ -____ Conducted susceptibility testing is carried out on Power- and Sync. lines both for CW and Spike signals. Voltage and Current are controlled and checked against the limits. Voltage and Current injected are plotted together with the Impedance. If the unit under test is susceptibleted, impedance is an important piece of information to analyze the problem. Fig. I3 shows a typical output plot from a conducted emission test. GIOTTO dBuR/dBuV ESTEC EMC TEST FflCILITY CONDUCTED SUSCEPTIBILITY JPR DPERRTfoNRL MODE :EXP ON DRTE 21-09-1984 TEST ;:;; : 28 VOLT Fig. 13 Particularely signal- and commandlines are subject to conducted susceptibility testing. Due to their important function in a "spacecraft" a lot of our attention is devoted to test those lines. A special test box has been developed for this purpose. Fig. 14 shows the test set up. The command signal is fed through the box and the disturbance is injected via an opto-coupler, in order not to load the circuit. In this - 18 - .way negative pulses are injected into a "1" and positive pulses into a "0". Together, this enables us to determine the safety margin of the circuit. A typical advantage of this kind of equipment set up is the ability to supply the customer with a small box in order to control the test. This box provides him with the possibility of decreasing the signal level until the susceptibility has stopped, and to increase the signal when the susceptibility has passed. At the same time the print-out will indicate the frequency range and the threshold level of the susceptibility. 5.4 Radiated Susceptibility Maybe this part of EMC testing has benefited the most of being automated. We have seen that in the case of manual operation an operator has to control the frequency with one hand, amplitude with the other, check the modulation, overload of the amplifiers, correct T.A.F., cable losses and check the susceptibility criteria of the UUT. All this control.and check functions are now taken over by the system controller. This provides an accurate and fully corrected, measurement calibrated on the spot of each frequency step. Above all one has the ability to repeat this measurement any number of times without the slightest diviation. Also here the susceptibility control box can be introduced (as discussed under cond.susc. testing). An important aspect is that the UUT is not unnecessarily overtested. With manual operation we have seen errors up to 20 dB or more due either to human errors or mismatch. Not only do we avoid overtesting but we have also the possibility of investigating the so called "Window" effect by automatically increasing and decreasing the power level. Sofor we have not seen this "Window" effect in our facility. The fact that we are working in a shielded room is the cause of other problems like non-uniformity and antenna position this can be the subject of a calibration routine. Reflections can be controlled through the installation of reasonably sized anechoic material. But this will be only effective from a few hundred megahertz. Near field problems are even more difficult. Remember that most of the requirements call for starting at 10 KHz. Building a facility where 10 KHz is in the far field is sheer UTOPIA. And near f!ieldmodeling is a sophisticated guess. Therefore the only solution to the near field problem is to move away from Low-Frequency Radiated Susceptibility Testing. Fig. 8 shows the test set up used for radiated susceptibility testing. The use of the system interrupt box and dual sensor power meter has already been explained. 5.5 Susceptibility to E.S.D. In our facility we have extremely good use for the schaffer-NSG-430 discharge gun apart from normal applications like testing of "Spacelab" equipment which is subject to discharges from 10 m-joule. Discharges which are produced by astronauts during Spacelab missions. The same test can be made on large ground stations or computers to detect bonding faulr ts and ground loops. It cuts back expensive facility time considerably. 6. Conclusion An attempt has been made to set up a system using commercially available equipment (micro processor only). Starting on automating, one is afraid to face high cost implementations. The cost can be rather limited on the basis of "growth system". It is possible to achieve a very sophisticated system by planning the cost over a long period of time. The new system should be developed in tight cooperation with an EMC engineer, and aim for achieving flexibility and growth capability. Automatic measuring techniques has brought us many additional advantages. Like the possibility of compensating for measurements errors, direct comparison of test results and use the same results for prediction. And it protects the test-facility against poor preformance. It is not possible to go deeply into the theoretical background of the topics discussed. And a routine engineer can ask many awkward questions. However, it is hoped that successful innovating, will achieve what we call "EMC". 4A4 19 - - COMPUTER-ASSISTED CONTROL OF EMP MEASUREMENT ON MAJOR SYSTEMS Gregoire EUMURIAN THOMSON-CSF ISSY-LES-MOULINEAUX, FRANCE 1. - INTRODUCTION EMP experimentation on major systems, for reasons of cost and efficiency, require accurate prior preparation of the tests. The time available is limited and most often it is impossible to resume the tests if later analysis shows inconsistencies. It is thus necessary to have an automatic system to avoid handling errors and a control assisting system which enables refining the results and adapting the theoretical experimental program to the actual situation. It enables measuring low excitation fields (100 V/m to 1 kV/m) . The passband extends from 70 Hz to 150 MHz. The sensor takes the form of a sphere with a diameter of 106 mm. All these performances (high sensitivity, wide passband to low frequencies and small dimensions) were obtained by implementing the active sensor concept. The EMP data acquisition system consists of 5 main elements : sensors, optical transmission lines, optical link processors, digitizers and a data processing system (fig.17). Principle 2. - SENSORS To obtain a minimum volume, optimize the sensitivity, dynamic range and passband of the magnetic and electrical sensors, coil-type highimpedance sensors have been used along with built-in amplifiers and correctors and also electronically matched high-impedance electrical sensors. The total dynamic range of the sensors is extended by a sensitivity change remote-control system (0 to 70 dB 10 dB steps). Figurel8indicates the main sensors as well as their dynamic ranges (between noise and maximum saturation 1evel)passband external dimensions. The sensor takes the form of an internal sphexe (comprising the ground frame) surrounded by two half-spheres forming the positive and negative frames (Fig.)). The two antennas are connected to a differential amplifier with unit gain, via two highimpedance 40 dB attenuators. Att~ts-2M6 A,%.&lad6 Figure 1 - Principle and geometrical characteristics of free-space sensor (E30) 2.1 - EXAMPLE OF A SENSOR 2.1.1. Free space electric field sensor. This type of sensor is used to determine the incident field and the field within cavities. It has a full-scale sensitivity from 2 1 V/m to 3.16 kV/m with a total dynamic range of 120 dB (50 dB of instantaneous dynamic range and 70 dB by switching). The differential stage is followed by a 20 dB low-impedance attenuator, an amplifier (gain 46 dB)and a second low-impedance attenuator. The attenuators can be remote-controlled (via the optical link) and their sensitivity is switch-selectable over a range of 70 dB by 10 dB steps. - 20 - i.e. an effective height of : _-PARTICULARITIES OF THE SENSOR - Geometry Figure 1 indicates the geometrical characteristics of the sensor. The spherical shape . 'was selected to optimize the effective height of the antenna. Figure 2 indicates the distribution of the 'electric field over the surface of the sensor. The field, perpendicular at all points to the surface of the sensor, depends on the incident field to infinity (Ei) and its angle with the sensor axis (0). h 6IIrg EO (6) eff = ch+ce with the values indicated in figure 1. The effective height is : h 6Ilx (0,053)2 eff = = 6.68 mm (7) 3611x 10g x 70 x lo-l2 -Connection of the sensor to a high load The utilization of attenuators and high impedance differential amplifiers (Re = 22 MO) enables reducing the lower cutoff frequency to less than 100 Hz. F B - 1 2 II (Ch + Ce ) Re - = 69 Hz (8) An output voltage directly proportional to the incident field is thus obtained (non-derivative response). Figure 2 - Electric field on the surface of a metal hemisphere The value of the field is given by equation (1) : Er = 3 Ei cos g It is drawn up from two limit conditions : -zero tangential field on the surface, -field indentical to the incident field to infinity. . Figure 3 gives the equivalent diagram of each half-sensor. Ch represents the capacitance between the hemisphere and the internalsphere and Ce, Re, the capacitance and input 'resistance of the amplifier. .Figure 3 - Equivalent diagram of a half-sensor _ (free space electrical sensor) The incident field, Ei, induces a charge Q (2) within each hemisphere of radius ro : -Dynamic range For a given value of the incident field, the dynamic range depends on the effective height of the antenna as well as the noise level (N) at the input : S -= N Signal p-p = 2 Ei . heff Noise rms N (9) With a noise level of 50pVrms (100 Hz to 100 MHz band) and an effective height of 6.68mtn the dynamic range is 48.5 dB for an incident field of 1 V/m. 2,1,2. Free Space Magnetic Field Sensor The magnetic field sensors are based on a coil with one or several turns. To obtain satisfactory sensitivity to low frequencies, the section and number of turns j has to be increased. Conversely, to increase response to high frequencies, a low capacity coil (reduced number of turns) is required and sensor size should be small with respect to the wave length Moreover, to obtain satisfactory rejection of the electrical field, the coil is shielded which in turns limits frequency response via a stray capacity, The first type of sensor consists of a flat coil shielded with a split shield. Q =j!S EO Er. ds (2) where s is the surface area of the hemisphere. Taking into account the value of Er (l), the induced charge is : Q = ~01.f~Er ds = soi/3Ei ds cos 8 = 311r02EoEi (3) It can be observed that this charge is three times that of a disk with a radius ro. The voltage Ve at the input of the ampli-, fier depends on this charge, Q, as well as on the capacitance between the hemisphere and the sphere and the input capacitance : Ve = Q ChtCc: 31Ir20scEi -= Ch+Ce (4) Under these conditions, the effective height par hemisphere is : Ve heff l/2= Ei 3dso = Ch+Ce (5) ThistYpe of device can be obtimized to obtain satisfactory sensitivity to low frequencies (large diameter, large number of turns') or response to high frequencies (small diameter, small number of turns). The second type of sensor is a Moebiusloop formed by two half loops made of coaxial cable 4 A4 Since the coaxial structure is adapted at output, the limitation at high freauencies does not show up as long as the wavelength remains significantly longer than the diameter. Sensitivity to low frequencies corresponds to that of a frame of the same dimensions with two turns, therefore poor. The approach adopted consists in making the coil by using a coaxial line. The outer conductor (shielding) is discontinued at each turn to prevent formation of short circuit loops. The effect of the break in impedance due to the splits in the outer conductor is compensated for by a corresponding adaptation which is obtained by connecting a network, corresponding to the characteristic impedance of the line used for the coil. To increase this device's sensitivity, while maintaining compactness, a magnetic core is place inside the coil. 4. - PARTICULARITIES OF THE TRANSMISSION CHANNEL. 4.1 - CONNECTORS The required passband of 100 MHz for lengths up to 300 m lead to usinggrad indexe fibers (100/140 urn). Although actual technology enables manUfacturing connectors with low optical lOSS (1 to 1.5 dB) the use of optical connectors on the worksite requires particular care at each connection-disconnection operation to avoid optical attenuation with the resulting decrease in cfynamic range. To prevent this derating and obtain a constant transfer function, the optical elements and their associated electronic circuits are enclosed in the electrical connectors. All the connection-disconnection operations are thus performed electrically. This dual sub-channel device with electrical input-output comprises a bidirectional optical modem. 4.2 -TRANSMISSION This type of sensor provides a maximum sensitivity of + 1 v/m ( S/N >40 dB ) for a bandwidth of TkHz to 150 MHz. 3.- OPTICAL TRANSMISSION PRINCIPLE The transmission line provides two channels : -a telecommand channel routing the various commands to the sensors (switching on, range switching, calibration, etc..), -a signal channel transmitting information from the sensors as well as service signals (calibration, battery, condition, etc..). The telecommand channel has a restricted passband (50 kHz) and makes use of frequency coding of commands as well as analog storage of signals (memory capacity) in order to obtain maximum immunity (operation in ambient electric fields up to severa hundred kV/m). The signal channel should have : -a transfer function independent of temperature, ageing, condition of the connectors and the optical fiber, -a wide passband (100 Hz - 100 MHz), -satisfactory linearity (2%), -a wide dynamic range (50 dB). Electrical/optical conversion can be performed by a laser diode or a LED. The laser diode has a passband and a power input about ten times higher than that of the LED (500 MHz against 50 MHz and 2 mW against 200 uW) *However, it is not very stable in temperature and requires thermalservo-control of the transmitted power). The LED was finally selected since its association with a compensation circuit enables considerable extension of its passband. Frequency compensation is obtained by progressively increasing the gain at higher frequencies. However, this operation requires : -a sufficiently fast diode, -a slow decreasing slope, -a passband independent of the level and temperature up to at least 150 MHz. These conditions are fulfilled by a LED specially developed for this application 50 pm chip and micro-lens). Figure 7 below indicates the frequency response for this type of diode before and after frequency compensation. The passband at -3 dB elec. is increased from 50 MHz to 150 MHz. ? the LED used in the optical line -._-.__ before compensation _------ after compensation. - 22 Figure 8 - Shows'the compensation.circuit diagram.. Figure 8 - Principal of the LED frequency compensation circuit. 4.3 - RECEPTION The optical electrical conversion can be performed by an avalanche orPIN photodiode. The avalanche photodiode has a gain of 10 to 100 times that of the PIN photodiode. However, it is very sensitive to the temperature and requires a high supply voltage (100 to 300 V). The thermal instability of the avalanche photodiode can be corrected by a servocontrol device.Its high gain enables building sensitive receivers with a low input impedance However, the high dependence between the junction temperature of avalanche diodes and their gain results in a non-linear thermal capacity effect deforming the low frequency signals (100 kHz). This deformation shows as a non-linear differentiation effect (fig.9). Figure 9 - Deformation of an AF signal produced by the thermal effect of the junction of an avalanche diode For these reasons , preference was given to a PIN photodiode with a low junction capacitance associated with an impedance matching amplifier with a very high negative feedback resistor (20 kQ )(figure 10). - This device applies a reference voltage to the input of the optical line, detects the level received by the optical receiver and, after A/D conversion and linear/log transcoding, commands a set of attenuators. It has the advantage of an extended passband, which is independent of the attenuator position, and a total absence of distortion. 6.- DIGITIZING The digitizing system developed for acquisition of EMP data enables : -high-speed sampling (2 ns) -a large number of samples (5000) -8 bits precision. 6.1 - OPERATING PRINCIPLE Actual technology provides no simple means of sampling followed by digitizing at a speed of 2 ns. This digitizing is obtained by using 50 samplinq channels which are scanned sequen- Figure 111- Block diagram of the high-speed digitizer Each sampling head thus works at a lower speed of 50 x 2 = 100 ns (figure 12). loons 100 s Figure 10 - Prinzple of the optical receiver This very high negative-feedback resistor, after solving certain technical difficulties (necessity of an unwanted capacitance 0.50 pF across the resistor) enables obtaining a dynamic range of 50 dB (optical power 30 VW) while preserving a passband of 150 MHz. The transmitter-receiver assembly associated with a grad indexe fiber of 100/140 urnenables building a 100 Hz - 100 MHz/50 dB optical line over 300 m. 5. - OPTICAL LINK PROCESSOR The optical link processor enables telecommand of the sensors via the optical link and pre-processing of the signal before digitizing. This pre-processing consists of : -calibration of the optical link with automatic adjustment of the transfer function, -filtering of the signal as a function of the required passband and the sampling rate speed, -matching the output level to the digitize input level in order to optimize the dynamic range. The constant transfer function is obtained, by an AGC device using a calibration signal. Figure 12 - Distribution in time of the instants at which the various channels are w. After sampling, the signal is stored in analog memory (100 memory elements per channel) The process is continuous and after 50 x 100 samples (time = 100 ns x 10 = 10~s ) a part of the old information is lost and is replaced by new information (drum memory) (figure 13). When the equipment is triggered, the sampling procedure is stopped, the 50 analog memories are read and the A/D conversion is performed at a slow rate. The continuous sampling before triggering enables providing information of events before as well as after the instant of triggering. The digitizer is controlled by a microprocessor which, in particular, performs the corrections required by the inaccuracy of the sampling heads and the analog memories. - 23 6.2 - PARTICULARITIES OF THE DIGITIZER The diqitizer has three main properties adapted to EMP meaSUrement : -drum memory associated with a System of Pre and post triggering, -very high digitizing rate (2 ns) I -large number of points (5000). The drum memory and the pre and post triggering system enable digitizing signals whose exact shape and polarity are unknown.Figuras 13 & 14 givean example of a signal with a positive and then a negative half-cycle, produced by equipment adjusted for negative synchronization but with pre-triggering. The high sampling rate associated with a large number of points enables digitizing signals which simultaneously have high speed and long duration (signals occuring on long transmission lines with protecting elements limiting the signal) (figure 15 and 16). For this type of signal, the digitizer provides a 150 MHz analysis passband for 5000 x 2 ns = 10 us. FinalLy, the large analysis window enables acceptence of signals occurring at different moments without using delay lines. 7. - DATA PROCESSING SYSTEM The coreof the data processing system is the IBM-PC, version AT, with disk drive (20 Mbytes) and floppy disk drive (1.2 Mbytes), color monitor, graphic printer and color x, y plotter. The software consists of 6 modules : -device management -results filing (raw or after pre-processing) -pre-processing of results -results display (raw or pre-processed) -firing aid -mathematical processing. 7.1 - Device Management This module ensures remote control of the digitizers and sensors and setting of the optical processors connected to each optical line. It performs all the channel calibration operations and monitors system operation (battery status, internal tests, etc). 7.2 - Results Filing This module enables two results filing modes : on disk (temporary) and on Flo.ppy disk (permanent filing). It also enables accurate description of the test conditions : date, time, composition of each system, position and direction of each sensor. 7.3 - Results Pre-processing Pre-processing of results enables shifting the signals to a common time origin and reduction, with operator consent, of the number of points for each measurement, if the signal width does notwarrant filing of all the points. 7.4 - Results display This module enables display of one or several signals ; this display facilitates correlations. In particular, simultaneous display of the incident fields and inducted currents and voltages enablesa betterundertanding of the induction processes. - 4A4 In the same vein, simultaneous display from a given of the fields obtainedviewed point and with incident fields of different values enables clear identification of the system's nonlinearities. Finally, comparison of the fields obtained from neighboring sensors enables detection of operating incidents. 1.5 - Firing Aid This module assists the operator in setting the various parameters in the acquisition systems : -sensitivity (sensor) -filtering frequency (optical processor) -sampling rate ) ) digitizer -off-set -pre or post triggering ) -etc. One to three fixings are necessary to reach an Optimal setting. This relatively rapid setting procedure is possible due to the sensor's high in&antaneous dynamic range 1>50 dB). The dynamic range, along with the switched dynamic range (70 dB) allows measurement Over a dynamic range of 120 dB. Example : Electrical field sensor (ranges : + 3kV/m, + lkV/m, 5 38@J/m, + lOOV/m, t 3OV/m, T lOV/m, 5 3V/m, $- 1 V/m). For the first firing, the SenSOl iS Set On a lesssensitive range (3 kV/m). If after digitizing, signal amplitude is greater than 3kV/m the program provides the option to the operator to change sensors or to decrease the incident field. If the field is included between 3kV/m and 30 V/m, the program suggests setting the sensor to the nearest range. If the field is less than 30 V/m, the sensor is set at this level of sensitivity and another firing is started. The new value obtained is used for resetting the sensor. If the range selected is the most sensitive and the signal amplitude is still not adequate, the program suggests either to increase the incident field (if the other sensors or installation allows it)or to improve the dynamic range by using an 80, 40, 20 OX 10 MHz filter (gain of 3 to 12 dB). Hence, with a 20 MHz filter is it possible to detect signals starting at 5 mV/m. If the computer controlling data acquisition is linked to the pulse generator delivering the signal, setting the range in function of generator voltage is easier. In the same manner the operator can set the limits for the ranges by restricting the measuring dynamic range. A second important setting is selection of the filtering frequency and sampling rate best suited to the measurement. These selections are made after determining, fwom the fastest transition, the passband of the signal under analysis. Based on this value, the filter which is best adapted to the measurement (maximum of dynamic range) and corresponding sampling rate is used. This selection is particularly of interest when examination of long events (low frequency resonance or computer sequences) require a large analysis window and slow sampling rates, - 24 - hence, a risk of undersampling in the absence of an adequate filter. 7.6 - Mathematical Processing This model enables calculation of a certain number of mathematical functions (FFT, correlation, powers, etc.). When voltage sensors are used and if circuit impedance is knowen,the module allows going from voltage to current via a Laplace transformation. 8. - CONCLUSIONS The simultaneous utilization of active sensors with high sensitivity and a large dynamic range, an optical transmission line with a wide band and stable transfer function, and solid state high-speed digitizers with a large measuring time range is perfectly suited to the measurements required by the EMP simulator. The associated date processing system also enables maximum automatizing of the fire control system while leaving the operator the final decision, but after having provided him with the elements for that decision. 1~ 5!LA I Figure 16. Digitizing in the presence of fast signals with a long the duration (2) Figure 13 -Figure 17, EMP data acquisition system block diagram. Figure 14.Principle of the drum memory. (2) Figure 18. Main Sensors - - BETWEEN 25 5 - THE INTERFACE ESD TESTING: SIMULATOR AND EQUIPMENT UNDER P. Richman KeyTek Corporation Massachusetts, Crucial issues regarding the interface between the ESD simulator and the Equipment Under Test (EUT) include discharge current peaks that are vastly different from simply-calculated Values, and failures of the EUT at both low and high, but not intermediate voltage These phenomena can be exlevels. plained and mathematically modeled in terms of circuit inductance and freeThe more inclusive space capacitance. circuit model that results, gives significantly improved agreement between calculated and experimental electrostatic-discharge current waves. TEST and A. Tasker Instrument Burlington, Bd U.S.A. a hand-held metal object key, bracelet or ring. like a tool, Values called for by various Standards and used by individual organizations range from 60 to 300 pfd for C, and from 10 to 10,000 ohms for R (l-5). ESD ,- EUT Introduction The peak current that flows during an electrostatic discharge (ESD) from an ESD simulator can be vastly different from the value intuition might lead one to expect. It can be at least as low as one-tenth, or at least as high as ten times, the value computed by dividing stored -- or test -- voltage by the simulator's nominal internal resistance In addition, the discharge current waveform in both simulator and actual human-body discharges often bears little relation to the simple, single R-C equivalent circuit in widespread use (l-5). GROUND Fig. 1: Conventional, for Personnel Discharge PLANE Single R-C Model Electrostatic Fig. 2 shows discharge current due to a typical human-body discharge from a hand-held metal object. Instrumentation for converting the discharge current into a voltage suitable for oscilloscope monitoring was built as per reference (1); oscilloscope bandwidth was 400MHz. Even though the circuit The two factors most responsible for these often huge discrepancies are circuit inductance and capacitance to free space. The conventional model for personnel electrostatic discharge is given in Fig. 1. It consists of a simple capacitor C, charged to voltage V, and discharging into the victim equipment -the EUT or Equipment Under Test -through resistor R. The "low" end of C is most often connected to a ground plane or to a point on the EUT, or both. A discharge tip, connected to the resistor R, is advanced toward the EUT until an arc occurs, simulating the spark that leaps from a finger or from Fig. 2: Typical @SD Current Discharge Wave from a Hand-Held Metal Object. (Steep-rise edges retouched for readability) 5kV Initial Charge Level 2.5A/half cm, 2ns/half cm - model of Fig. 1 is in common use, there is simply no way in which it can begin to account for the real-world Fig. 2 waveform, specifically for the sharp, high-amplitude initial spike. (Others have also reported initial spikes (6).) The single R-C model of Fig. 1 is inadequate in that it ignores: 1. The human body and/or ESD simulator circuit inductance, which ranges from 0.5 to 2 uH. 2. The 3 to 10 pfd, almost inductancefree capacitance to free space of the human hand. 3. The typically 5 to 20 pfd, almost inductance-free capacitance to free space of the victim EUT itself. Circuit Inductance Reference (l), an IEC draft ESD standard for Process Control, specifies a one-meter long ground return of 20 mm width. However for calibration purposes, the same draft standard calls for a discharge circuit, including the ground connection, that is "as short as possible". Calculations, confirmed by tests, give total circuit inductance including internal simulator circuitry as well as the ground return itself, of about 1.7 yH for the one-meter ground return. Similarly, a figure of 0.7 uH results for a typical R-C network with a calibration-length ground, with a length on the order of 30 to 40 cm. 26 - Fig. 3 shows the addition of total circuit inductance L to the simpler circuit of Fig. 1. Table 1 shows the large effect that L can have on network "efficiency" v\ , defined as the ratio of peak current Ip during discharge, to the "intuitive" peak of V/R; multiplied by 100, to obtain per cent. (Efficiency without L must be lOO%.) Calculated values of I were computer-derived from appropria! e solutions to the series R-L-C circuit of Fig. 3, and were spot-checked via experiment. I was calculated for a stored voltage o!? 5kV, but for different voltages the values of Ip can be scaled proportionately; ignoring preionization and other effects. R and C values come from representative ESD test Standards, as listed in Table .te;fy v - TIP 1 RETURN, 1 GROUND PLANE Fig. 3: ESD Model Modified to Include Total Circuit Inductance, L Table 1 R and C Values for the R-L-C Equivalent Circuit of Fig. 3, with Efficiency hgiven for Realistic Ground Return Inductance, 1.7 uH Ip is peak current for a stored voltage of 5kV tp is time of occurrence of Ip Standard Organiza- or Draft tion (l-5) Standard - LP Q. =lOO,lR (p:d) (ohms) --- (%) 150 64 Ip for 5kV (A) tP ns 21 18 1. IEC(1) 65 (Seer) 80 (Draft) 150 2. MIL(2) 883 B 100 1,500 97 3.2 5.6 3. NEMA(3) Part DC33 (Draft) 100 1,500 97 3.2 5.6 4A. EIA(4) PN-1361 (Draft) 100 500 87 9 4B. EIA(4) 1, 60 10,000 100 .5 1.4 5. SAE(5) 51211 300 5,000 100 1.0 2.9 60 10 6 6. Cart simulation 28 10 1.6 561 27 - Inductance L was taken as 1.7 uH, 1. representative of a typical simulator including a 20 mm wide ground return of about one meter, the length recommended by the IEC draft standard. Note the vast differences, particularly for the IEC 150pf/150fi network, between V/R (33A for 5kV) and the Calculated value, 21A, for peak current Ip for 1.7 PH circuit inductance. For the "calibration" circuit inductance of 0.7 uH, the same 150pf/150n IEC network gives a calculated peak current of 25A. Typically, arc and corona effects reduce this still further, by as much as 20 to 30%. Thus a 5kV stored voltage with a 150 ohm resistor will typically result in a peak current of only 16 to 18A. "Calibration" in test laboratories may report defective simulators, with only one-half required output! (The IEC specifies a peak current of 50% to 90% of stored voltage divided by resistance, thereby covering the situation quite completely. Unfortunately many calibration laboratories simply calculate V/R, and either ignore the IEC specification or neglect to calculate the effects of even the 0.7 uH "calibration" inductance.) Neither peak current Ip nor peak time tP respond to network differences in a simply proportional way. The effect of changing from 150 pfd/l50 ohms to 60 pfd/lO ohms, for example, is rather small; current peak increases from 21 to 28~, time to peak decreases from 1B to 16 ns. Yet the nominal "efficiencies" of the two networks differ by over an order of magnitude. The explanation is that inductance is the controlling factor. Until the simulation circuit resistance gets large -- 500 to 1500 ohms -- or more accurately until network efficiency exceeds go or g5%, circuit inductance dominates performance. Capacitance to Free Space; Interaction with Inductance Every object has capacitance to free space -- or to the walls, floor and ceiling of the room in which it is located. For a spherical object of diameter dl and a room (also taken as spherical, for simplicity) of diameter d29 the capacitance is given by reference (7) as: C=O.556 x Kdld2/(d2 - dl) in which dland d2 are in cm, and Km for air. (1) 1 For a human hand or arm in a room of typical dimensions, the term d2/(d2-dl) approaches unity, so that c- 0.556 dl (2) For a hand of approximately 9 cm "diameter", capacitance is thus on the order of 5 pfd. Note that this capacitance is almost inductance-free. The inductance of a finger, hand and/or forearm may be calculated from reference (7) as: L=O.O021 [2.303 loglO(4j/d-11 uH (3) in whichQ and d are length and diameter, respectively, of the finger, hand or forearm; again in cm. Table 2 gives results, along with approximate values of capacitance to free space, for all body segments involved. Use of a hand with key has been assumed, as this is rapidly becoming a de-facto standard for worst-case ESD simulation. It represents an ESD event involving a handheld metal object such as a tool, ring, bracelet, or indeed an actual key. Table 2 Approximate Dimensions, Estimated Capacitance and Estimated Inductance for Various Sections of the Human Body (d = diameter,&= length) d cm -- Q.c L cm pf' --- PH Fingers holding key 2 6 2 .02 Entire hand holding key (to wrist) 7.5 12.5 5 .02 Forearm (wrist to elbow) 9 30 10 .l Full arm (wrist to shoulder) 9 60 20 .27 Torso (shoulder to waist) 30 60 20 .13 Whole body (torso plus lower body) 30 120 40 .43 Computer solutions are given in Table 3 for peak current I and peak time tp, from the differentia P equations that describe performance of the circuit of Fig. 3. Solutions are given for values representative of appropriate combinations of the hand and arm from Table 2, using a compromise inductance value of 0.1 PH. Solutions are also given in Table 3 for the R-C values specified in various standards as set forth in Table 1, for both "calibration" (0.7 uH) and normal l-meter (1.7 uH) inductances. A resistance of 200 ohms is used for the small capacitance val- - 28 - Table 3 Computed Values of Peak Current Ip and Peak Time tp Ip computed for 5kV; simply scale for other voltages (For virtually all parameter combinations except R=lOK, risetime,T, lies between 25% and 65% of peak time tp.) C (Pfri) R (ohms) c& I, (amperes) for bandwidth= Infi400 60 100 MHz nite MHz MHz 50 50 200 :1 .1 10 19 16 6 15 2.4 14 ;:", 10 200 200 1 :1 18 18 16 17 60 10,000 .5 ; 7.5 100 100 150 300 500 1,500 150 5,000 1 .7 1.7 .5 .5 ; .7 1.7 ;:: .7 1.7 .7 1.7 25 21 1.0 1.0 1.0 1.0 ues that simulate the hand and arm, but is not a major determinant in Ip or tp, over a wide range of resistance values. It is assumed that the victim EUT has significant capacitance to free space in the surface area immediately adjacent to the point of ESD application; i.e., it is another, larger, and also virtually inductance-free capacitance. From reference (7), for example, the capacitance of a 30 to 40 cm diameter disc to free space can be calculated as 12 to 15 pfd. This might represent that portion of a victim EUT panel or keyboard at whose center the ESD was applied. Without such a "ground plane", free-space capacitance effects due to finger, hand and arm will be very much lower, due to the lower total circuit capacitance that will result. (EUT "ground plane" capacitance is effectively in series with hand capacitance.) In this connection it is worth noting that the IEC-designed coaxial "target" (1) performs far better in making current-spike measurements when it is mounted to a ground plane on the order of 40 x 40 cm. In add tion to the "infinite bandwidth" theoretical values given in 1: ;:: 25 24 21 21 1.0 1.0 1.0 1.0 for bandwidth= 400 loo 60 MHz _- MHz - MHz - .4 .6 .7 .7 :; 1.0 1.2 1.2 1.8 1.3 2.0 1.0 1.1 1.4 1.5 2.1 2.4 2.4 1.4 ;:: z-2 . :: 5.2 10.1 5.7 10.6 ::: 2:: 5:; . 10 11 .7 1.7 tn_(ns) Infinite .6 10 18 1.3 2.9 10 18 3.5 4.0 8 12 7.9 9.1 2.8 10 14 11 12 12 20 13 21 11 17 11 17 Table 3 for Ip and tp, computer solutions are also included in the table for the same waveforms viewed with oscilloscopes of finite bandwidths: specifically 400 MHz, 100 MHz and 60 MHz. Data in Table 3 go a long way towards explaining differences between measurements made by different investigators. Simulations with 60 pfd and lOK, for example, will be vastly different depending on the simulation capacitor's physical size, and on whether the simulation resistor is 12 cm long -hence not simulating a finger/hand combination -- or short, and contained within a metal enclosure to which it might have, for example, .5 ofd stray capacitance. For 6Opfd/lOK, Ip is .5A at 5kV. But if stray capacitance -- or capacitance of the simulating 60 pfd to free space -- is considered and a 400 MHz scope used, then from Table 3, I will be 6A for an arc resistance of !ZO n; and 10A with infinite oscilloscope bandwidth. Yet the value shown with 60 to 100 MHz instrumentation will range from only 1.5 to 2.4A. And after all, the 0.5 pfd stray is only 10% of the 5 pfd representative of the human hand -- which at 400 MHz gives 16A for 2OOn, as shown in the table. - Thus all of the simulation circuits in references (1) through (5) miss the the hand/forearm combination point: has a 5 to I.5 pfd capacitance to free space, and it is coupled to the discharge arc with only 0.05 to 0.1 uH. The result is a super-fast edged, short-duration (1 to 4 ns) spike of 15 to 30A, for a stored voltage of Only 5kV. Experience shows this spike can be crucial in causing EUT malfunction, but it is neglected by all existing standards. And if it is accidentally viewed on an oscilloscope, its amplitude is typically underestimated by a factor between 2 and 4 by the 60 to 100 MHz instrumentation in common use. Any saving grace that a simple R-C ESD simulator may have is that the simulation capacitor can itself have a capacitance to free space! This accounts for the sharp wave-start so frequently seen in simulator current waves. Fig. 4 shows a typical case, for the EIA values of 100 pfd and 500 ohms (4). But this is a far cry from the sharp, 1 to 4 ns spike of 15 to 30A peak (for a stored voltage of 5kV) that is generated by a hand-held key, as shown in Fig. 2. ESD simulator's discharge current output, the amplitude of the sharp initial edge generated by the simulator capacitor's own capacitance to free space, is usually less than peak current due to the simulator's basic R-C. For this reason it has been seen as merely an unpleasant anomaly in the wave, due to "parasitics". In point of fact, it provides whatever inadequate sharp-risetime "punch" such simulator waves do have. 29 - 5Bl Corona effects at higher voltages -above 3 to 6 kV, depending on discharge tip geometry -- reduce the sharpness of the initial spike or step. This effect most likely accounts for the fact that many equipments that can pass ESD tests at 10 kV, say, at which level corona has seriously reduced risetime, will fail at only 5kV, due to the steep initial rise of the spike or step. At voltages of 15 to 20 kV, failures may start again, as the sheer magnitude of the di/dt, even with heavy corona, once again becomes high enough to cause equipment malfunctions. Fig. 6 shows computer-generated current discharge waves for the Dual R-L-C circuit of Fig. 5, both without (6a) and with (6b) simulated arc oscillations. Hand-simulation values are 7.5 pf, 200 n and 0.1 yH. Body-simulation values are 100 pfd, 5OOfi and 0.7 PH. Fig. 6b corresponds well with the human-discharge current of Fig. 2; i.e., the Dual RLC model works. In the typical The New, Fig. 4: Typical ESD Current Wave from a Single R-C ESD Simulator (100 pfd, 50051) (4). Sharp WaveStart is due to the Simulator Capacitor's own Capacitance to Free Space 5kV Initial Charge Voltage 2.5A/half cm; 2ns/half cm Dual RLC Circuit Model Fig. 5 shows the new, Dual RLC Circuit Model that seems to best replicate the effects of circuit inductance, plus hand capacitance, in an electrostatic discharge. RB 150-1500 LB .5-z DISCHARGE Two parallel R-L-C paths are provided; one for the body (CB, RB, LB), the other for the hand (CR, RR, LR). Since the impedance from Discharge Tip to ground during the discharge is generally low -- due to the victim EUT's own capacitance to free space -- the two R-L-C paths function as almost independent current sources. Thus the waves they generate are superimposed. CR/RR/ LH generates either the steep-rise initial spike, or for higher voltages at which corona effects cause pre-ionization, a large initial step. CB/RB/LB then generates the longer wave that follows, carrying the often less-damaging energy stored on whole-body capacitance. Fig. -- 5: The Dual RLC Circuit Model for ESD; Incorporating Separate, Parallel Paths for Body and Hand Discharge _ 30 - Fig. 7 shows a typical current-discharge wave from a practical ESD simu. later that was designed to reproduce the Dual R-L-C model of Fig. 5. It agrees well with both Figs. 2 and 6b. 0 Fig. 6: 4 8 1211s 0 4 8 12 ns Computer Solution for Discharge Current from the Dual R-L-C Circuit of Fig. 5; (a) Without and (b) With Superimposed, Simulated Arc Oscillations cH=7.5 pfd RH=2OOn LH=O.l uH CB=lOO pfd RB=5OGn LB'1.7 I-tH 5kV Initial Charge Level 5A/division, 4ns/division probably still causes a large proportion of ESD-simulation failures. 4. A parallel RLC/RLC circuit model (the "Dual RLC" model) gives excellent general agreement with initial edge and initial spike experimental data. It is quite practical to simulate 5. the Dual RLC model with physical components, while nevertheless retaining the convenient one-meter ground return. Data from such simulators agree well with both calculations and data from actual personnel electrostatic discharge. The Dual RLC model represents the situation well: CH discharges through a low inductance to give the initial spike simulating the human hand; CB discharges through a higher inductance, to simulate the longer wave that conveys energy stored on the entire body. 6. Simulators not incorporating the CHRH-LH hand-spike simulation path may well not be able to induce failures at 3 to 6 kV in the same way that actual personnel discharges can do. Thus ESDtesting with such instrumentation may not represent the reality the equipment under test will face when placed in service. References Cl1 International Electrotechnical Commission IEC 65(Secr)80 Draft Standard: Electrostatic Discharge (for Industrial Process Control). i-21 MIL STD 883B, Test Methods and Proceedures for Micro Electronics. Fig. 7: Actual Discharge Current from a Practical ESD Simulator Embodying the Dual R-L-C Circuit of (Steep-rise edges reFig. 5. touched for readability) 5kV Initial Charge Level 2.5A/half cm, 2ns/half cm c31 NEMA, Residential Controls, Envi- ronmental Testing for Electronic Controls, Part DC33, Proposed June 24-25, 1982. c41 EIA (Electronic Industries Associ- ation) Standards Project ~~-1361, Environmental and Safety Considerations for Voice Telephone Terminals, Draft 5, Nov. 24, 1981. Conclusions 1. Any ESD circuit model that doesn't include inductance can't simulate reality well enough for test purposes. c51 2. Capacitance of the hand to free space causes a spike at voltages to 5kV, and an initial fast edge at higher voltages. Both are grossly under-estimated by 60 to 100 MHz instrumentation, while measured adequately with instrumentation of 400 MHz and above. C61 King, W.M. and Reynolds, D., Per- The inevitable 0.25 to 0.5 pfd ca3. pacitance to free space of the simulator's internal capacitor is the only tie to fast-edge reality that many simulators have; and it is too small by a factor of at least ten. Nevertheless it SAE Standard Recommended Practice Information Report J-1211, June 1978, P 20.99. sonnel Electrostatic Discharge: Impulse Waveforms Resulting from ESD of Humans Directly and Through Small Hand-Held Metallic Objects Intervening in the Discharge Path, Proc. IEEE Int'l Symposium on EMC, Aug. 18-20, 1981, pp. 577-590. II71 Terman, F.E., Radio Engineers' Handbook, McGraw-Hill, 1943, pp 48, 113. 6 - 31 - RECENT OF COUPLING DEVELOPMENTS IN THE PATHS OF ESD THROUGH Michel UNDERSTANDING A METALLIC CABINET Mardiguian and Donald R.J. Don White Consultants, Inc. Gainesville, Virginia, USA Abstract White This also serves as a reference to further shielding effectiveness assessment. One must remember that the ESD, although quoted Based on recent studies made on ESD event statistics and ESD modelization for furniture and human discharge onto a computer frame, in-depth analysis is made of ESD current routes on a metallic frame and mechanisms of re-radiation inside equipment. Although a metal box should behave as an efficient shield, it k shown that this does not happen because high-frequency spectrum (up to the Gigahertz region) of ESD excites all existing seams and slot leakages. Measured values of E and H fields (inside the cabinet) near the discharge area and some peculiar aspects, like the effect of a discharge on a screw protruding significantly inside, arc discussed. The Electric and Magnetic field amplitudes lead to some discussion on the wave impedance of the ESD reradiation inside the box and its near field/far field transition. This, in turn, allows a better prediction of the noise voltages induced in nearby PCB traces or flat cables. Grounded Metal Plate Electric 01 Magnelic Field Probe - Ground Plane E \ \ Spectrum Background _____- Figure In the past few years, significant progress has been accomplished in understanding the electrostatic build-up and discharge mechanisms, their simulation and the ESD hardening of integrated circuits. Several computer manufacturers and independent experts have disclosed the results of their studies. However, in contrast with the large amount of data on human body and furniture voltages, capacitances and resistances, discharge rise times and waveforms, etc., relatively few quantitative studies have been done on field amplitudes around an ESD discharge [ 1,2,3]. This paper is a follow-up of a study of Ref. 4 which examined sequentially the mechanisms of the ESD coupling via equipment cabinets and external cables up to the distributed victim circuit. Here now is an attempt to quantitatively evaluate the Electric and Magnetic fields near an ESD discharge path, and the factor influencing field re-radiation inside a typical electronic enclosure. In these tests, the electrostatic discharge was simulated using a Schaffner NSG 430 (150 Q, 150 pf Network). The E and H fields were measured by miniature monopoles, short balanced dipole and magnetic loop (electrically shielded). The probes were connected to an Electra-Metrics ESA 1000 Spectrum Analyzer. A slow scan speed and sufficient RF attenuation were chosen to avoid Spectrum Analyzer error due to the broadband nature of the measurement. E and H Fields Values and Polarization, Discharge to a Vertical Structure, B2 from Over a Conductive an ESD Ground First, the test set-up of Fig. 1 has been arranged to measure the field amplitude facing to an ESD “Zap,” in the absence of any protective shield. This would be the case of a discharge to a metal object near an equipment having only a plastic enclosure. l-Experimental Analyzer set-up for ESD field measurement “static,” is certainly anything but a static phenomena: Within few nanoseconds, a localized electric field of several kilovolts per cm (corresponding to several hundred kilovolts/meter) collapses to zero while in the same time, a localized magnetic induction raises up to several Gauss! With a scaling factor (a few amperes discharge instead of ten kilo amperes, and a spectrum of few hundred MHz instead of few hundred kilohertz) the ESD is in fact a miniature version of a lightning stroke. Few documents have stressed this fact [ 1 & 41 and others have reported field strength values. These values were generally measured at one meter. Although the non-uniformity of the field makes closer measurements less accurate, during our study E and H field magnitudes have been measured at 10 cm, 30 cm and 1 meter. The vertical structure was a 50 cm by 6 cm aluminum plate, firmly bonded to the copper ground plane. The ESD gun was set to 10 kV and an arc discharge, with a slow repetition rate was made on the upper tip of the plate. The ground return for the ESD gun was a flat strap about 30 cm long to avoid the possible influence of both inductance and location of the return conductor. For the same repeatability reason, the orientation of this strap was kept always in the vertical plane formed by the gun and the structure which was discharged upon. Figs. 2a, 2b, and 2c show the results of electric and magnetic fields, after bandwidth and antenna factor correction. A few remarks are in order: a) The 1 meter results correlate within few other reported measurements, discharge simulators about done + 15dB with with similar b) Compared to the I meter results, the 30cm and 1Ocm results seem to show a (distance)W”2 dependancy instead - of a (distance)-2 or (distance)’ as one would least in the induction (near-field) region. expect, 32 - at structure, the ESD generates a predominately magnetic (low impedance) field in the induction region, tending to a 120 x ohms wave impedance in the far field zone. Since the change over of near to far field is wavelength dependant, the transition occurs at different frequencies for the various distances of the experiment; the change is very pronounced for the IOcm case. In this experience, it must be reminded that the radiator is the whole circuit formed by the simulator and Frequency 1 3 10 I” MHz 30 100 300 3MH7 1 30MH/ 1 IOOMHI 1 300MH/ 1 5OOMHr ~~ m men a ion u Table l-Average .m 1 3 10 30 Ftequency Figure 2a-ESD Field at 1 meter 10 300 Free Field Radiation Frequency 3 100 in MHz 30 in MHz 100 300 120 120 sz 100 100 z. 9 % 80 5 80 g D s N : 60 60 g 40 40 g d 20 20 m’ Q N N 3 10 30 Frequency Figure 2b-ESD 100 P m I” Wave Impedances of ESD Radiated Fleids The statement that the field is predominately magnetic near the discharge path may be surprising. There is a common belief that ESD, “being electrostatic, has to be an electric field.” A close look at the discharge network can clarify this: Simple Field theory says that in near field region, low impedance (<337B) sources will radiate predominately magnetic fields, while high impedance (>377Q) sources radiate predominately electric fields. The ESD simulator used follows the IEC-65 recommendation, with an internal resistance of 15OQ. Therefore,it behaves more as a magnetic source in near field. Will “real-life” electro-static discharges really appear like this? In the opinion of the author, the 15OQ value is a fair compromise, but it has the drawbacks of every compromise. Actual furniture-type discharges 161 from large metal objects, carts, chairs, etc. may exhibit dynamic impedances IO times smaller or even less, creating more magnetic field in the near region. While human body resistance, being at least 10 times higher, will create less magnetic field. The IEC-65 somewhat makes up for this discrepancy by recommending such voltages (8Kv and 15Kv) that they force a current about similar to a furniture discharge current. The possible effect of wave impedance on victim circuits exposed to the ESD radiated field will be discussed in the next section. 300 in MHz Fields at 30 cm Frequency 10 30 I” MHz 100 d) A rough integration of the electric field spectrum over the frequency domain gives the following approximation for its time-domain peak value: at 1 meter = 70 volts/meter at 30 cm = 120 volts/meter at 10 cm = 220 volts/meter 300 100 N : 80 3 Voltages ._____~ Induced in Nearby Printed Circuit -___ ___-. and Other Small Circuits Boards -.__ .._ m” 60 D 10 Figure 30 100 Frequency ,n MHz In the second part of the experiment, the antennas were replaced by a PCB having a unique trace representing a loop of 1Ocm x 10cm. This trace was alternatively terminated into 1 kilohm, open-ended, then terminated into a short. The voltage picked-up was read on the spectrum analyzer, with all precautions to prevent possible pick-up by the coaxial cable. 300 2c-ESD Field at IO cm the vertical discharging structure. Seen from an antenna located at 3OOB above 300MHz --@ 1Ocm from about 1OD around 1OMHz to 500 above 300MHz. Fig. 3 shows the induced voltages being oriented tangent to wave front. that for actual arcing on a metallic the PCB The effect of varying the far end terminating resistances is interesting in the prospective of understanding which of H field or E field coupling predominates. Fig. 4 shows the traditional model for a small rectangular circuit illuminated by an electromagnetic field. The E field creates a transverse voltage V2, which appears as a higb impedance source (current source) with an open circuit voltage: V2 = E x 2L x h x This indicates in dBpV/MHz, T cos 0 cos o! x (1) - 110 1 I Frequency in MIir 10 30 100 3 33 300 90 80 80 70 70 60 60 like the coupling l/F as evident coefficient on Fig. 2, while 10 MHz, in Eq. (3) starts to create series insertion 150-200 MHz, available ing the available voltage By comparison, the wir- loss, caus- at the 5OQ end. Finally, the H field spectrum above itself decreases like I/F’, to collapse B2 at the same time increases like F. Above ing less and less voltage 100 90 decreases ing impedance 110 100 6 - even more caus- rapidly. the values for magnetically induced voltage using Eq. (2) and (3), or the graphical method of Ref. (5) are shown also on Fig. 3. They are in fair agreement with the measured data. Curves Frequency I” MHz ! 1 3 10 30 100 I 300 A and B of Fig. 3 correspond on far end, which minimize left is the electrical contribution. is one order thermore C, Calculaird 0 Calciilated voltage based on ti Field coupling voltage based on E-Field couplmg the only only of magnitude supports “standard” Figure 3-Broadband voltage induced rn a 100 cm:‘PCB run located 10 cm from the ESD path, parallel to wave front (PCB not oriented for maxlmum H-Field interception). less then the magnetic is, Then, the PCB was rotated wave front, such as to intercept magnetic contribution termination contribution. 0 = angle between the E field plane of the loop and the direction trum the SOQ impedance analyzer) a high impedance will double the available will it. nullify On the other VI appearing and the direction of the “victim” receptor end the H field as a low impedance voltage, while 1kQ a shorted creates a longitudinal source of of that in field, near one. This fur- the radiation of predominantly 90” to be perpendicular the maximum magnetic case. In this set-up is so pronounced IOOOQ (curve A) to the flux. The in exactly the however, the that even with a far end it overrides the electric (voltage 60 of Frequency A B C end (spec- on the far end, like transverse is in a small circuit u = angle between propagation. tiiven statement on Fig. 5 and can be interpreted same way as for the previous Where What It is clear that this contribution the previous discharge results are shown induced contribution. magnetic. c Figure 4-TradItional model for voltages illuminated by an EM field. to a high impedance the magnetic or in all cases, wlh Ziar end = m (wen) with Zlai end”0 (short) z,,,, end (Receptor)=50 I1 I m end in MHz with Zlar end = 1 kll Figure B-Broadband voltage Induced In a 100 cm;‘PCB run located 10 cm from the ESD path perpindicular to wave front (PCB Intercepting maximum magnetic field). voltage source) with a value: In fact the difference MHz correspond between approximately curves (A) and (C) below to the ratio of IO lOOOn to 5OQ termination. Interestingly there the influence voltage vx = of varying impedances Vx across the receptor VI,,, is totally different. The end is: corresponds Effect Z, ,,,,(,,I,,, il) -~ ~--.--------~----1, t & + z,,..,,,,, (3) -= 5On Z ,,,,,,,,.=O.l A high magnetically impedance induced Z, voltage. n + .iw x 0.4~H on the A shor/ far end on /he fur nullify the Metallic by the normal Cabinet shielding on ESD Radiation ___ ____~_ the cabinet metallized) of curve (C) 15 volts. housing the ESD field the electronic should effect of the material. for the external has been discussed in a former pick-up paper by I/O well protected. real life enclosures are full of slots, scams, apertures, disrupt integrity. the shield cables, which [4], it seems that boards and wiring should be fairly be at- Therefore, internal However, etc., which end will max. iUli,lcJ il. Every shield discontinuity across the ESD current path will “shine” inside, with an efficiency proportional to its length com- On Fig. shorted. (or the spectrum of about and not accounting circuit will of a Typical is metallic tenuated Z, integrating If, instead of being plastic, circuit in our experiment, enough, to a peak voltage 3 the curve Therefore, (C), corresponds the electrical contribution to the far end being is minimum while the magnetic contribution is enhanced. The flat portion of the voltage spectrum corresponds to the domain where the H field pared to the half to and above wave length. 500 MHz, hibit significant leakage. mechanism, the “witness” The ESD spectrum any slot longer than extending up a few cm wilt ex- To show this effect on the ESD PCB was placed inside in lmm thick - aluminum rack; to calibrate the experiment, all mating surfaces have been thoroughly brushed and screwed and the ESD gun, set to IOkVolts was discharged on all sides and especially in the seam areas. No value exceeded the sensitivity level of the test set up. Then, several typical shield imperfections have been introduced by removing some of the top cover screws and inserting Imm cardboard liners under the seams to simulate an ungasketed cover with ordinary manufacturing tolerances. Fig. 6 shows the results. The induced voltage could be read up to 100 MHz, for the enhanced magnetic coupling (far end termination shorted) which demonstrates two points: a) the thin seam unequivocally spoils the protection by the box to ESD coupling b) the re-radiated magnetic. field inside is, once again, offered predominantly An interesting effect was also simulated: One of the threaded holes used to attach the top cover has been painted and the cover was mounted using a long screw, protuding about 2.5cm (I”) inside the box. The results are also shown on Fig. 6. It seems that the screw generates a secondary arc inside, between the fillets and the inner box surface. Frequency I” MHz 100 90 : E 90 N 80 : 80 s 70 7ov 60 60 m" u - Discussion -_____ m" It has been shown that a circuit illuminated by a typical IOkVolts ESD discharge is exposed to field values in excess of 200 volts/meter. With the 150fi/150pF of the standard ESD simulator, the field in the near region is predominantly magnetic. Because of this, the orientation of the victim circuit versus the potential ESD areas, and its source/load impedances arc determinant: A typical logic circuit wilt consist of a “victim” end being a logic gate input with input impedances of few kfi (TTL, Schottky) or more (CMOS), while the far end will be a logic output with source impedance ranging from 3013/15OQ (TTL, Schottky) to 3OOQ (CMOS). This arrangement makes the magnetic contribution worse. Decoupling the signal line at victim’s end wilt always be beneficial and matching the termination at receptor end has the same advantage. Otherwise, decoupling anywhere on the circuit may have no result, or even (I detrimentul one. Also, a fcrrite bead, though providing series insertion toss, may be disappointing because the bead should exhibit more added resistance than the gate input resistance, which is impossible to achieve with small size beads. To the contrary, if this is the power supply bus which has to be protected from ESD induced spikes, ferrite beads can be very efficient since the impedances are tow on both sides. A future study wilt investigate more deeply some aspects of the ESD re-radiation through seams and the induced voltages on ribbon cables running along those seams. References 1. Michael King: CORNELL Frequency I” MHz 1kll A ESD on protruding screw head, PCB terminated lilt0 B ESD on protrudlng strew into a short C ESD on vertical or horizontal seam with a forced PCB being 5 cm behlnd discharge point Figure 6-Voltage in an aluminum induced on the head, PCB terminated 10 cm x and Conclusions Finally, it is important that designers pay attention to the integrity of the metal housings, even for equipment which are neither RF equipment or highly sophisticated gear, but simply have to be ESD immune. 100 N 34 1mmgap, IO cm PCB trace housed 2. 3. 4. rack. 5. 6. DUBILIER Report on EM1 Susceptibility (December 1973). Peter Richman (Keytek): ESD Protection/Test Handbook. Michel Aguet: Perturbations dues aux decharges statiques. (COMPATIBILITE ELECTROMAGNETIQUE Ecole Polytechnique de Lausanne-1983). Michel Mardiguian: DESIGN FOR ESD IMMUNITY RATHER THAN RETROFIT--IEEE/EMC Symposium, Washington 1983. Donald R.J. White: EM1 Control METHODOLOGY and PROCEDURES (DWCI, Gainesville, VA 22065 USA). Michael King: Impulse Waveforms from ESD-IEEE/EMC Symposium. September 1982. - 35 7B3 - ESD Susceptibilityand Radiated Emissions of EDP Peripheral Printers Luciano Honeywell 201000 The E. S. D. emission the most stems We susceptibility (R. E.) and particularly ply with standard ship different between E.S. the relation- if described: the spark of mechanical gap sam- interferences electrically between floating electronic parts, (Printed differences only apparently xample, the E.S. a printer is then Last item GND bonding radiated D. and Board). mechani- grounded,are then such is outlined. As susceptibility level (inside the influence the printers) from EDP TEST Real-world, E.S. ted out in ref. H- field of effect pulse That’s and voltage is, drops, respectively, depend disrupted shielded peripheral the following D.air five routes: generated, ge induction H-field, from ground mechanical struc- effect, discharge current as a matter in two differents on E.U.T. as poinvia injection. of fact, items: predischar3. E-field, The last can be se5a. E-field gap, I. ESD loop; to only experience influence ex- discharge plate, internal enhance and the effects we’ve evaluation on nomena. For both lowing effect IEC laboratory l., and in the bus our 5a, methods 64 standard) and field diagnostic first on signal focused oriented, the printers voltage structure, discharge the out, in transient on mechanical Therefore, pointed on peripheral to be identified secondary 1)A effects: air generation generation spark as expeto enhance effect. main The of five direct or on E-field H. I, S. drops to the host by each gap for or on H-field has 1. predischarge, 2. spark methods, , can be used I. gene rated ternal on the ca- discharge in H. I.S. last of equipments radiation; E-field; 5. noise date discharge, disrupt different of METHODOLOGY (11, Two rienced As with emissions E. S. D. parated 1 current or strong. shielded computer ESD weak e- predicted. concerns connecting Circuit to understand phenomena on parts, between and a method troublesome the E.S.D. structural P. C. B. voltage analyzed, 4. tests. tire. The corona 5b. injected impedence ples; bles, 1‘ FIG. generators then, between coating .penerator. of E.U.T. effect, are, interferences unit of ES.! . L Test set&p for table-top9nounbd equlpmeht, field printer status :on IihC with compubr. test). items and surface cal Italy prin- and to com- D. structure under specific Pulse - many parts, no malfunction we analyse mechanical (equipment with regulations. first - Milan sy- E. D. P. environment In this paper The E.D.P. with to exibit and office - for concern be designed Two - now perhaps are topics peripheral devices, and electromechanical in home and the Milanese Italy peripherals. such ters; mechanical must Systems and radiated problems important Information Pregnana - Inzoli lines. analysis 5b, and 0 f ESD we defined two test with phc_ (folset-up, different purpose. one reference a common is characterised plane, GND star under point by: the E.U.T., for as the dischar- - ging system and E. U. T. 2) An electrical sible connection, between inside the plane the EUT, in ground data The power cord path of ESD The second 1) On line one The path power first then, cords set-up re- without (fig. of EUT, 1) by: frequency pulse, invol- purposes chanical equipotentiality structure 2) Voltage drops pressure are, printer level GND second star set-up systems surface (screws, of EUT as function purpose of are as following: l)Noise conversion unbalanced tive signal cable their transfer “Direct gap”: return to avoid its flat ground impedence “direct external spark discharge” H-f;eldESD and EUT cable .) effect in the case (n. 5b) and ca- and to arranged with mo re, “S”; internal Division a as shown C=150pF) oriented current spark with injection” gap “Sl”. it can be a T. D. R. Reflectometer, ARC stributed (R= lOOn, gap Further discharge, or lumped (Ti- R=50R ), to measure electrical The mechanical side only, has rent ways: Ni-coating;ZnCr04, of sample, been coating di- parame- grounded coated one in two diffechromate on electro mechanical and arc -deposited parts es can where and current ESD generator c ircuit V The equ iva lent with: delay-lines wave reference floating tact as parallel output I’ are inj%%d model guide finger and sample, to-end capa city. by IESD representative sample representative of the sample, res ist ence, excited current between CP re- voltage can be built-up (DL), plane, side ce coating be verified. is behaving circuit, E.M. can be evalua- discharge-surfa sample sonant T. “Interna&ark method, a flat impedence; discharge” oriented inte rferenc on or unshielded etc.. emission cable, generator can be: ted, of e- generated connections, can be: shielded 1. ESD printer struc- between radiated has 2. path, characteristic meters The (shielded oriented, in fig. as possible, cable, the sample Zinc (yellow and colourless conversion). In such a way, the oscillation modes of capaci- busses or fault connections comzison prevalent __I__ (i.e. mechanical or ground noise computer bles, to dif- frequency circuits). 2) Interference host of current between and power lectronic common due to high path coupling tures from mode, shown much conversion ferential structural ters. point. diagnostic invol- Therefore current as two mecha- method laboratory as short, me coatings). se t-up ESD the EUT analyzed. of a significant without on bonding 3) Susceptibility of me- of EUT. contacts, be first test The between and the test verified, The and cables. diagnostic must in fig. with connections. current ved control as following: 1) High The of EUT, condition of ESD in the connections. cables structure been characterised signal 2) Return ving is interferences on the or of pulse. condition cables running external filter The nical behavior connector. current signal point identified is not involved running external as pos- and the GND on line cable turn 3)Self-test short alternatively connection external as 36 between C, of RSURF ESD of and con- generator T,~return cable FIG. Test end- 2 set-up: Mechanical sample surface in g. with coat- - The results 1. can be Surface tor sure is face impedence 2OOgr. No bution is tact pressure lb, open are effect is The pres- flat has strongly 4. and screw exposure to Na Cl discharge nerator As point 4b.If corrosion voltage measures ESD gen. of smoothed input impedence point on a mechanical ranging then, if Cp=@, that’s sonance). The generator input equivalent loons). No electrical resonator is pointed return (std. pulse is used, and sheilded the current pulse. the 5,contact current shaand- pressure The very dangerous oscillations are spark ESD due to ge- gap S, series to disrupt related to the cable method) is is in fig. are circuited waveguides as fig.1 injected 6;the “derivated” cy oscillations generator arranged the typical shown shape and the ESD are triangular and excited set- current high by CRF clas- frequeninfluences. re depends- wn to lo- along structures if E-field, the If = 10; ESD or if its E-field 3 compling inthe in the case effect to significative rne- out. FIG. excited a mechanical floating coatings, discharge dependent. one Sl. If the sic fo=55MHz to surface from rela- to surface short resistence, up indicated, parasytic (2: “Sl” in fig. coating TIZD as it can be seen is deplaced to is on discharge classic of a second “S” related This is used, dependent. V parameter point correlation 4c is “S” frequency first : fo= 150 MHz coustant resistence, the to discharge resonance; dumping and no dependencies thod from a lumped impedence, the discharge related to 0.4KV. (i. e. a A/4 that’s is specific depends parameters cal high the peak structures: 1. 2KV coupling if Cp=80pF, out, oscillation at the from frequency, from ge- neration pointed cable gen. pe is shown (1 OKV) is used. value is dependent.ESD is an arc is a tringular coating VESD, parts. no correlation return shape test. 2. the and if ESD 04. at min., This is sample, as dis charge method IESD’ generator ( 10KV) is used. 4a.If holes. colourless, R,,,__=1;2Ofi(ty2;. 8fi&%% points). pically), Rsupr=* Ni: RSURF= + . 1; RSURF=. 05;. 5flafter 16h, point of mechanical coating. as RsURF=10;400~, edges parts. ESD is ranging discharge it is on flat is, dependent, is used. of oscillation ted phenomena; con- if coating, value on edge 0. 5KV contri- sharpness ( IOKV) 1. 2KV, fig.4,is exactly resi- edge peak from used. near 7B3 3 the sur reflections RSURF coloured, circuits lb. Zn Cr skin dependent, 04, contact a contact wave out. genera- in fig. merely to travelling pointed Zn Cr . shown significant been la and lc As - as 3* generator ‘ESD as following: TDR R SURF’ and a typical is needed, stence. la. summarized resistence 31 part (N. mechanical _-of prevalent 5-a) of a printer structure grounding’impedence (high frequency) of ESD effect is is rised values, the is the most FIG. 5 TDR 1 OOA/div pulse 20 ns /div. FIG. 4 Peak value ZOOV/div. 10 ns/div. FIG. ESD 6 current 1 OOA/div 50 ns /div. 38 This is a common coating with internal or grounding screws To test tion, are printed wire ference 1) with (as one mechanical capacitive than 1 pF. The results representative faced to printed is very surface input sharp mV). 1 limitations, Th: logitudinal the total V A VG zvIN is neglibible, more than ESD input be very &V will The to ESD is a real complex analysis matrix mechanical methodology has printer, with structure, been set-up a segment The H.I.S.I. significant parts, the following be measured 2a. electrical different with Characteristics tion times, delling such /REF. resonancies, must electrical a T. D. R. impedences of the equivalent bet- be pointed parameters As must lines trum mo- how resonators. GND. accurately pointed ILRNiZ repreand as noise can be easely is solved, on all by E.S. D. in Ref.[Z, value is given e- integration the spectrum. of the spec_ . The 31 by asymin dB, for each of time peak on line derived. value by the integration, in above in the pulse, program final of the referenced real will of phase one. the range information Many excess value redu- algorithm, in the significant error of tranbe a ve- papers,we or of the integration representation. mated system, current as approximated then due to the loss and propaga- pulse is shown approximation shown in the spectral a simple its ced the excess’err set-up: delay is obtained ry good out. the and the equation predicted sient, ween 2. All g scale in drops involved modulus log-lo an be deribed. domain for segments; transform the broad-band integration, ptotic set- 6. Its spec- in fig. circuits, ranges spectrum method be test approxima- Fourier of voltage P.C.B. A flow-chart, be or the preof 2a impedences. provided frequency will shown from connecting value ones, must domain voltage in the frequency Y vESD time of the injected for of 2 if are current, as can then peak choice by fig. if laboratory is be calculated source by mea- The by comparison pulse pulse significant approximated t zo) suscptibility susceptibility T. investigation 1. All representation, trum E. S. D. relatively of the tral sharp. poor! If the E.U. tion stimated, zo/(50 ESD a good of the inductance 2. results. up is used, sentation suggested step the prevalent indicated parts, connected”. parameters Typically, can band- value. = 2vESD. are inje cted devices suggested measured. 4. All or current no significative will 20, resonator The voltage and the printer very large. with (L is as ones,is the fi- noise is floating, can be very IESD, model, valent resona- high are of the 7 if distributed 3. The if different mechanical range values results diffe- devices. step by different parameter modulus, impedence part important by fig. Zn is cting or electromechanical and 2c measurement Only the can be very (AVClOO (Zo= and capaci- tonne electromechanical lurrped parameters of the R of Ni or Though impedence discharged guide and resi- inductances lines, “low-frequency surement less characte- (BW> A from the source so far-field approximation is the that the generalize Later we will valid. to include the near fields as results usually are well, since measurements at a range of 3 m, which is only made 0.3X at 30 MHz. Our model semi-anechoic chamber is with a perfectly-conductrectangular partially-reflecting floor and ing The chamber has and ceiling. walls The length L, width W, and height H, antennas transmitting and receiving respecand h are at heights h, The Fioor. above the chamber tively, transmitting antenna is located on the axis of the chamber at a discentral - The rewall. tance D from the back ceiving antenna is located at distanwall and R from the ces D from a side transgitting antenna. plane conduc ti ng single While a only one image, a source beproduces tween a pair of conducting planes will have an infinite set of images in each reflecmu1 tiple plane, representing of planes one pair let tions. We with represent the floor and ceili.ng, transmitting antenna h, above the the A second pair of planes reprefloor. the front and back walls of the sents with the source a distance D chamber, of the chamber. wall the back from represents planes The third pair of the source at with side walls, the w/2. 66 - Here equations. suppressed in these efthe = FR(Oi)FT(O )Ri(Oi) is Ki surf vity fective reflect of the ith account also the into taking faces, transml. tthe angular dependences of receiving antenna patterns, and ti ng phase net The and FR(Oi)* R (Oi) an image is computed by s X ift @L for each of summing the phase shifts for that produce which reflections the includes that Eq. (3) image. Note only the far-field radiation term. The E = E the floor, The reflectivities of front wall, and walls, side ceiling, back wall are assumed to .be fR f y. “Z: TRY, -RF, and -RR, respective verttcally-polarupper signs are for while the lower signs signals, ized are for horizontal polari. zati on. Its by ’ (2rL f r,s,t where corresponding are R rst + D>i (2tH f (2sW f2, (1) . . . for +The ‘rst =: (-RF)f(r)(R,,RF)IrI(~Rw)f(S) x Rw21sI(*Rf) f(t)RCitl, when the upper Eq. (I), and lower signs. (2) signs f(i) We can now calculate the electric field E at any point in the chamber by generalizing method of Reference the 1. We sum the contributions from the source and the various ima es, noting that the signal from the i!!_ P; image is given by KiEO = ()exe[j(Bd di i + $i >I, (3) where di is the distance from the image, g = 2x/h, and i represents a particular combination of .r, s, and t. E. is the electric field a unit distance from the transmitting antenna. The time dependence exp(-jwt) is squared = R02J$$ N -CO6 Ki Kk [ c i>k=ldidk Comparison + W/2)? + h&, = 0, fl, reflecttvities where f(i) = 0 are chosen in = sgn(i) for the Ei +2 9 = f from exe[j(Pdi magnitude [El2 for imA general position vector ages representing all possible combfnfrom the reflections Of ati ons different surfaces is given by zrst si.gnal combined N images iS + (R*g) Ki (4) @,)I. is given 2 g(di-dk)+‘$i-‘$kl with 1 l (5) Measurements receiving antenna at a We locate x = x The distance z = hr. Y = y,, the (r,s,t> image to the refE:rn di ceiving antenna is found from di 2 = b$st - x,2 - y,; - h,x(* 2 9 (6) If we take the antennas to be point FT(O )FR(Oi) = Sin2(OI). dipoles, near-field Note that ef I ect the of terms be $nc$uycl, b:) addi;: can (3 cos20i - 1)(1/g di when computing the Ki for use In this Kq. (4) [21. The calculated chamber response is sensitive to the phase $i as well as the magnitude Ri of the reflectivities of the various surfaces. If there is a frequency-dependent absorpti on in the wall surfaces, there must also be a frequency-dependent phase shift to satisfy the Kramers-Kronig dispersion relations. Al though some measurements of R are available, there are none of 4. Consequently, it has been necessary to model an absorbercoated wall as a partially-conducting surface with some assumed conductivity and dielectric constant. For these preliminary calculations we have chosen a very simple model the wall surfaces, and we ignore effects of non-normal incidence of for the the 1301 - 67 - A proper model should also signals. the from scattering include diffuse uneven surfaces resulting from the use The cones. absorbing pyramidal of reflectivity of a normal-incidence with a surface partially-conducting constant E = 1 is given by dielectric t31: A - 1 1/2 R=(-) 9 where A = -+{[I + ($2]1’2 + 1}1’2, (8) 2 reflecand f is the frequency of the The phase shift of a signal ted wave. reflected from such a surface is 4 = tan_1(1/S), (9) where l3 = 2& {[l + ($)2P2 - 1}1'2. (10) We can choose the conductivity a as for fitting measured reparameter a flectivity to that calculated from Eq. simplest way to do that is The (7). to replace 2a/f by fl/f, where f1 is However, this fitting parameter. the falls which reflectivity loads to a as l/f at high frequencies, while off reflecthe reported measurements of tivity fall off approximately as l/f 2 . replacWe can model this Qehavior by ing 2a/f by (f2/f) , and using f2 as a fitting parameter. signal Figures 1 and 2 compare the amplitude calculated for 258 different measurechamber with actual paths chamber Is 18.3 m long, The ments. Wall and 12.2 m wide, and 6.1 m high. ceili.ng surfaces are covered with pyrCone lengths amidal absorbing cones. 1.22 m, and 0.66m, with are 1.83 m, the the longer cones in regions where occur. Transreflections principal antennas were receiving mitting and 94455-l biconicals. Model Ailtech deterwas Measured signal amplitude subtracting antenna factors mined by from measured data, so it is sensitive to errors in the antenna factors. to emphasize the effects In order fixed chosen of reflections, we have which highlight the antenna heights actual resonances in the chamber. In the attenuation measurements, site antenna is the receiving height of 4 meters and the from 1 to scanned maximum signal at a given frequency is This reduces the effects of recorded. measureresonances on the chamber Some of the discrepancies bements. signal measured tween calculated and amplitudes may be caused by failure to size of take into account the finite biconical diFinite antennas. the response angular poles not only have differ from that of a which patterns driving their point dipole, but also will be affected by nearby impedances conducting surfaces. that It has been found empirically some of the effects of reflections can be reduced by choosing a direction of is not parallel to which propagation verticallySince the chamber walls. polarized antennas remain either parachamber all lel or perpendicular to the equations derived above surfaces, are valid for this case. horizontally-polarized anHowever, neither parallel nor perare tennas arbitrary pendicular to the walls for The equaof propagation. directions separately to applied tions must be different sets of images: one for two current antenna the component of the parallel to the walls, and one for the walls. component perpendicular to the net signal is obtained by summing The the two results before squaring to get the amplitude. Summary and Conclusions for signal model A theoretical chamtransmission in a semi-anechoic developed which incorber been has porates multiple-reflection paths in a Current limitations systematic way. of the model include: Lack of measured data on the phase 1. amplitude of the reflectivity and of absorber-coated walls. Assumption of point-dipole, in2. stead of finite-length, antennas. the crudeness of the In spite of used in the simplifying assumptions calculations reported here, the generdata are al features of the measured The next step will be to reproduced. adjust the various parameters in order to improve the ability of the model to reproduce actual measurements. it has already been possiHowever, draw some general conclusions ble to regarding the effects of the different Here we chambers. surfaces of the situation where the consider the and the receiving antennas are source (the nearer to one end of the chamber other (the than to the back wall) front wall). dominate Floor-ceiling reflections for horizontal polarization, producing the large oscillations at the low freend of the curves in Figure 1. quency and The relatively small side, front, wall reflections can be reduced back propagating at an even further by angle to the walls, rather than paralThe resulting image anlel to them. further away from then point tennas the receiver. 200 30 30 FREQUENCY (MHz) Fig. 1: 200 FREQUENCY (MHz) Measured (left) and calculated (right) sIgna transmission between horizontally-polarized antennas located one meter above the floor of a semi-anechoic chamber. The distance between the antennas -Ls three comparison, the signal calculated for an ideal open site meters. For (perfectly-conducting ground plane, far field terms on1 > is shown by 1 /E02). the dashed Line. Signal amplitude S = 10 log10(49.2)E( __.___-. ____ _____.P\ ._.__,_....-- _.._..__.-- - ~~~~~~~ 200 30 30 FREQUENCY (MHZ) Fig. 2: FREQUENCY [MHZ) Measured (left) and calculated (right) signal transmission between floor of a vertically-polarized antennas located one meter above the the same as for Fig. 1, chamber. The conditions are semi-anechoic signal is except for the polarization. ideal open site Again, the shown by a dashed line. back For vertical polarization the causes a long-period oscillation wall function of in signal amplitude as a distant more while the frequency, but front wall causes a short-period, When the receivoscillation. weaker, ing antenna is on the central axis of reflections from the side the chamber, signals other walls combine with the relatively sharp dip the produce to When the shown in Fig. 2 near 35 MHz. off the moved antenna Fs receiving from the centraL axis, the reflections walls travel different dissLde two other. tances and tend to cancel each This greatly diminishes the dip. have not yet had time Although we to find the optimum parameters for an is already clear chamber, it actual only that model calculations will not existing lead to improvements of provide also chambers, but they will future guidance for designers of chambers. REFERENCES [II Burrows, 12, J. C.R.: Bell System 45-75 (1937) The 121 Ring, R.W.P.: Harvard Antennas, 700 (1956) t31 Good, R.H., Jr., Classical Theory Magnetic Fields, 383-388 (1971) Theory Univ. Tech. of Linear Press, Nelson: and T.J. Electric and of Academic Press, - 69 1402 - A METHODOLOGY FOR EVALUATING MICROWAVE ANECHOIC CHAMBER MEASUREMENTS Motohisa Kanda Electromagnetic Fields Division National Bureau of Standards Boulder, Colorado 80303 U.S.A. field, and the methodology for evaluating the corresponding errors associated with antenna, EM1 and EMC measurements in an anechoic chamber. The anechoic chamber measurement is evaluated in terms of the net power delivered to a transmitting antenna, the near-zone gains of open-ended rectangular waveguides and horns, pyramidal rectangular and The on-axis reflections from chamber walls. field intensity of the standard transmitting horn in an anechoic chamber is calculated in terms of the net power delivered to the transmitting antenna. The resulting data can the overall estimating used for be chamber the anechoic uncertainty in measurements. The statistical control of the of transfer process by use measurement the will monitor antennas standard This uncertainties. measurement paper for evaluating methodology discusses a anechoic chamber measurements. I. The Measurement of the Net Power II. to Open-Ended Waveguide and Pyramidal Horn Antennas Open-ended waveguide and pyramidal horn antennas are used to establish standard electromagnetic (EM) fields in an anechoic chamber. Part of the uncertainty in our knowledge of the EM fields arises from the uncertainty in the net power delivered to the horn. In turn, this uncertainty reflects our lack of knowledge of the amplitudes and reflection and phases of the various transmission coefficients in the power delivery system, as well as the uncertainty in measurements of the power incident upon and reflected from the horn. Introduction anechoic chambers are Microwave currently in use for a variety of indoor electromagnetic measurements, antenna and electromagnetic interference (EMI) The prime compatibility (EMC) measurements. requirement is that an appropriate transmitting antenna at one location within the plane-wave field generates a chamber throughout another volume of the chamber of dimensions sufficient to perform EM1 and EMC measurements. This volume is frequently referred to as a quiet zone and its 'quietness", or reflectivity level, will determine the performance of an anechoic chamber. The National Bureau of Standards (NBS) anechoic chamber is shown in a side view in figure 1. Pyramidal horns or open-end waveguide (OEG) antennas are used as sources of chamber illumination, positioning them in the access doorway with their apertures inside the plane of the absorber points on the chamber wall. A cart on precision tracks located under the measurement axis can be moved horizontally through a distance of 5 m by a stepping-motor drive system. There are gaps in the absorber on the floor to accommodate each rail. The purpose of this paper is to discuss how to establish a standard electromagnetic In our system for establishing standard electromagnetic fields, we compute the net delivered power to a standard transmitting ___ DIRECT PATH L____ DE0 OR HORN _-_ LABORATORY.. ----$/A?3 . PROBE REFLECTBN FRO,” REAR WALL c---.-a-___ t 40 ml 1 rcrx&~, FWY CA”, : ’ 3’ - 6.63 m Fig. 1: A side view of the NBS anechoic chamber U.S. Government work not protected by copyright. - antenna from measurements of incident and reflected power obtained with a dual directional coupler. Our power delivery and measurement system can be represented by a four port black box as shown in figure 2. The port terminations and numbering are: (1) power meter to monitor the forward (throughput) power, (2) power meter to monitor reflected from the transmitting antenna, (3) source of the cw rf power, (4) transmitting antenna. the power standard 70 - In an ideal coupler, i.e., zero reflection coefficient for all coupler input ports and infinite directivity (Sll = S22 = s33 = s44 = S14 = S41 = S23 = S32 = 0), and for a matched power meter at port 1, i.e., Cll. rl = 0, it g(S,r) = h(S,r) can be shown Cl] that The terms, = 1. g(S,r) and h(S,r) involve the products of complex values of the system S-parameters and the reflection coefficients. For this reason, unless the phases of the system magnitude and S-parameters and the reflection coefficients are well determined, g(S,r) and h(S,r) are not calculable. The extent of deviation from unity is, therefore, taken to be an error contribution in the determination of the net power delivered to the standard antenna. Although the degree of deviation from unity is a function of the system S parameters and the reflection coefficients r, it is found to be in general, less than 1% Cl]. To determine the net power delivered to a transmitting antenna, the terms S34/S13 and Although the l/S24 need to be determined. magnitudes of S13, S24, and S34 could be measured with a network analyzer, the system establishing implemented here for being standard electromagnetic fields is a selfcalibrating system which utilizes a standard flat-plate short and a matched termination. When a short (r4 = 1) is placed at port 4, the ratio of power measurements P2 and P1 gives Fig. 2: Measurement of the net power to a standard gain antenna The net power delivered to the transmitting antenna is the difference between incident and reflected power and is given by Cl] !L I I '24 '34 3 s13 p1 (2) P inc - 'refl net = Pa 2 p1 II s34 = (1 - (rJ2) p2 s13 lg(S,r)12 1 lh(S,r)12 (1) where Al(S,r) is a complex quantity much less The second step in than unity Cl]. evaluating the net power delivered to a transmitting antenna, Pnet, is to replace the short at port 4 with a well matched power two power ratio of the meter. The measurements PI and P4 is l (1 - lr212) lS2412 The symbols PI and P2 are, respectively, power meter readings at ports 1 and 2. rl and r2 represent the corresponding reflection coefficients observed looking into power meters 1 and 2. Sij is the scattering parameter defined as the ratio of the complex wave amplitude emerging from port i to that incident in port j. g(S,r) and h(S,r) are functions of the system S-parameter and the reflection coefficients of ports 1, 2, and 4 p1= s13 p4 s34 21 - lr1j2 2 (1 + b2121. 1 - Ir,l (3) where ~~ is a complex quantity much less than From (2) and (3), a value for unity [ll. is obtained. In summary, we perform ll/S2412 - 71 - two power ratio measurements with a standard short and a matched termination in order to determine /S34/S131* and (S2412. Al and A2 involve the products The terms, of complex the system S parameters and values of Since the coefficients r. reflection of these complex phases and magnitudes quantities cannot be easily determined, the extent of deviation from zero is, therefore, taken to be an error contribution in the determination of 14~2 for the TEIO mode equals JlT , with k being the free-space wave number, and r is the reflection coefficient of the TEIO mode from the end of the waveguide. The constant AE, which is related to the amplitude of the incident TEIO mode, will be defined later. Z IS34/S1312 and IS241** Moreover, the uncertainty in the power ratio and the measurements, and Pl/P4, P2/Pl / uncertainty in the reflection measurements r1, r*, and r4 also contribute an error in I I I I the determination of IS34/S131 and IS241. The I detailed discussion on this topic will be given elsewhere Cl]. The net power delivered to a transmitting antenna is then determined from two absolute power measurements, PI and P2, using (1). I I -Y I I r: Near-Zone Gain of Open-Ended Rectangular Waveguide III. The electromagnetic field measurements in an anechoic chamber are usually performed in the near-field region of the transmitting standard antenna, and the approach used to establish the standard field is to calculate the radiated field intensity in the nearfield region of the transmitting antenna. These antennas consist of a series of openended waveguides below 500 MHz and a series of rectangular pyramidal horns above 500 MHz. The near-zone gain of an open-ended, unflanged rectangular waveguide is calculated from forward near-field power patterns, which is determined from theoretically predicted far-field power patterns by use of the plane wave scattering theorem [2]. The open-ended geometry of the rectangular waveguide is shown in figure 3. The E-plane pattern, EE(e), is predicted quite accurately by inserting the E and H fields of the propagating TEIO mode into the Stratton-Chu formula and integrating over the aperture of the open-ended waveguide [3]. Thus, for EF(n), EE(o) = AE . X Fig 3: Geometry of open-ended rectangular waveguide In the case of the H-plane fields, the aperture Stratton-Chu integration of the formulas with the electric and magnetic f-i;lcIcIt;fa;c!;T~~e~~$ neglects the fringe produces much too broad an H-piane pattern: Using an accurate estimate of the fringe currents on the x = f a/2 sides of the rectangular waveguide from a numerical solution to the electric field integral equation applied to the openended rectangular waveguide [4], we obtain for H-plane pattern (case t EH(e) = AH[ f) + r(cose (Jj)*- (% - f) sine)* + Co1 [l + a case + r(1 - # cose)l [1 +a + r(l - 0 cos(% f-11 sine) . (5) sin(J$ sine) !$ sine (4) where the normalized propagative constant R/k The constant AH is related to AE by AE = AH{(i)* ((1 + f) + r(l - I)] + Co} . (6) - 72 - Ihe constant Co is calculated by equating the radiated power determined from the far-field to the total power input power determined from the TEIO mode field. Once the far-field power patterns of a open-ended rectangular waveguide are determined, the plane wave scattering theory enables us to predict its near-field power patterns [5]. The near-zone gain of an openended rectangular waveguide is then determined by integrals of its near-field power pattern. The evaluation of the uncertainty of the near-zone gain of an openended rectangular waveguide will be performed by comparing the theoretical near-zone gain with the experimental results using the plane wave scattering theorem and will be reported elsewhere 161. IV. where C and S are the Fresnel integers defined as C(w) - jS(w) = jw exp(- jft2) dt (9) 0 and their arguments u and v are defined as ;' = f (*ha;+/2 +&) % l/2 ’ f-1 H AH=-------. r + RH (10) Near-Zone Gain Calculations of Rectangular Pyramidal Horns The approach used at NBS to establish a standard field at frequencies above 500 MHz involves the use of a series of rectangular pyramidal horns. In deriving the near-zone gain of a pyramidal horn by the Kirchhoff method, Schelkunoff accounted for the effect of the horn flare by introducing a quadratic phase error in the dominant mode field along the aperture coordinates [7]. Geometrical optics and single diffraction by the aperture edges yields essentially the Kirchhoff results. The proximity effect in the Fresnel zone can also be approximated by a quadratic phase error in the aperture field. To improve Schelkunoff's equation by taking into account the reflection of the diffracted fields from the horn interior and double diffraction at the aperture, the concepts of the geometrical theory of diffraction are used to determine the on-axis near-zone E-plane pyramidal gain of an Taking into account the preceding horn. considerations, the improved near-zone gain of a pyramidal horn is given by [Sl G _ 32 ab - --2-- RE RH , 7lh Fig 4: Pyramidal horn dimensions The E-plane factor RE is given by (7) RE = 1 + cos@o exp(- jkREcos@,) 4w2 (7) + 2 v (Xi, n: - 4,) + Z's2 where RE and RH are the gain reduction factors due to E-plane and H-plane flares, respectively. The pertinent horn dimensions used in (7) are shown in figure 4. The factor 32 ab/(rh2) is the gain of an in-phase distribution uniform across field one dimension of a rectangular aperture and cosinusoidal across the other. The H-plane flare of the horn is given by (7) where b w= RH = % (11) 12 2 {C(u) - C(v)\2 + {S(u) - S(v)12 , (8) 2 (2hR~)I'2cos~o/2 ’ (u - VI Xi=-. rRE r + RH 3 1402 - 73 - = - exp(kx 2COSa) {1 v(a,a) 4ka [C ((-$ l I/2 4kRo I/2 -jS((--$ _ (1 + The failure of the chamber to provide a truly free-space test environment affects the anechoic the accuracies in measurement The performance of a chamber measurements. can be chamber anechoic rectangular rf checked by measuring the relative insertion loss versus separation distance between a source antenna and a receiving antenna [91. j) cos :, (12) cosF)l} . The factor S2 is defined as m v(R~, $ - ibo) f(di, Jl s2 = Reflections from Anechoic Chamber Walls V. The factor v(ro, a) is given by 71 - $,, T - i@,) (13) Insertion loss is the ratio of power received by a receiving antenna or probe for the initial test position to that received It is assumed for different test positions. that the source and probe input impedances the that power constant and remain transmitted by the source remains constant. If the anechoic chamber is a perfect freespace simulator, the relative insertion loss varies with distance according to the freespace transmission loss formula given by where f (d, oo) = 0, v(d, 8 - e,) + v(d, e t no) (15) Pr/Pt = gsgp(h/4nd)2 , (14) di = 2~ sin(ig,) is the ray-path length between single and double diffraction, and m is the largest integer less than ~/2@~. The near-zone gain of a pyramidal horn is used to calculate the radiated field intensity in the near zone of a gain antenna. The typical gain reduction factors HE and RH expressed in decibels are shown in figure 5. The evaluation of the uncertainty of the gain reduction factors for pyramidal horns will be performed by comparing the theoretical gain reduction factors with the experimental results using the plane wave scattering theorem C2] and will be reported later C61. 0 r -l.O- I r.o- 2.0 , I 3.0 I I , 4.0 I , 6.0 Range (meters) Fig. 5: Near-zone gain-reduction factors RE and RH of a pyramidal horn (a = 0.828 m, a' = 0.248 m, R = 0.612 m, RE = 0.812 m, RH = 0.943 m, @o = 0.386 rad) at 1000 MHz ( 8.0 where = net power transmitted source antenna, = power received antenna = near-zone antenna = near-zone probe d = antenna separation distance, m h = wavelength, m. pt pr gs gP by gain gain the of of by the probe source receiving Measured data are compared to free space transmission loss calculated using the appropriate near-zone transmitting antenna gains. Disagreement between the measured insertion loss and calculated transmission loss is a measure of reflections from chamber surfaces, the assuming near-zone gain calculations are exact for the separation distance considered. The measured relative insertion loss versus antenna separation distance provides voltage standing wave ratio (VSWR) data by means of a longitudinal probe scan. Rear-wall reflections and source-toprobe interactions are often resolvable at all frequencies, but reflections from the ceiling, side walls and floor are difficult to identify at frequencies below 500 MHz because the VSWR period is so long. Figure 6 shows an example of measured relative insertion loss with calculated free-space transmission loss along the on-axis of the horn antenna. standard antennas will measurement uncertainties. monitor the References 0 I * I J Cl1 Orr, R. David and Kanda, "Evaluation of microwave measurements." chamber published. La Kerns, David M. Plane-wave scatteringmatrix theory of antennas and antennaantenna interactions. Nat, Bur. Stand. (U.S.) Monograph 162; 1981 June. 162 P* c31 Stratton, Julius A. Electromagnetic theory. New York: McGraw-Hill; 1941. 615 p. c41 Yaghjian, Arthur D. Approximate formulas for the far fields and gain of open-ended rectangular waveguide. Nat. Bur. Stand. (U.S.) NBSIR 83-1689; 1983 May. 34 p. CSI Arthur D. Efficient Yaghjian, computation of antenna coupling and fields within the near-field region; IEEE Trans. Antennas and Propag., Vol. AP-30, 133-138; 1982 Jan. II61 Motohisa, Near-zone gain Kanda, calculations of open-ended waveguides and rectangular pyramidal horns for anechoic chamber measurements. To be published. c71 Schelkunoff, S. A. York: New waves. Reinhold; 1943. 530 p. C81 Errors in the Edward V. Jull, predicted gain of pyramidal horns. IEEE Trans. Antennas and Propag. Vol. AP-21, 25-31; 1973 Jan. c91 FitzGerrell, Richard G. Using freespace transmission loss for evaluating IEEE chamber performance. anechoic Trans. Electromagnetic Compatibility, Vol. EMC-24, No. 3, 1982 Aug. SEPARATION, i Fig. 6: Relative insertion loss between horn antenna and probe along the on-axis horn of antenna with free-space transmission loss curve fitted at 1 m VI. Concluding Remarks A methodology is presented for evaluating an anechoic chamber measurement in terms of the net power delivered to a transmitting antenna, the near-zone gains of rectangular open-ended waveguides and rectangular pyramidal horns, and reflections from chambers. The measurements of net power delivered to a transmitting antenna are determined by using a short and a matched termination in a self-calibrating system. gains of an open-ended The near-zone waveguide and rectangular rectangular pyramidal horns are calculated from forward which -e patterns, power near-field determined from theoretically predicted fi;field patterns by use of the plane wave scattering theorem Lll. rectangular performance of a The anechoic chamber is evaluated by measuring the relative insertion loss versus separation chamber the source of between distance illumination and a receiving antenna. A lack of fit between the measured insertion loss and calculated transmission loss is a measure The of reflections from the chamber walls. resulting data can be used for estimating the overall uncertainty in the anechoic chamber measurements. The statistical control of the process by use of transfer measurement Motohisa, anechoic To be Elc tromagnetic Nostrand Van - 75 - 15D3 FIELD DISTORTIONS IN A TEM CELL S. KASHYAP DIVISION OF ELECTRICAL ENGINEERING NATIONAL RESEARCH COUNCIL OTTAWA, ONTARIO CANADA KlA OR8 ABSTRACT The distortion effects due to an equipment under test in a TEM cell are calculated assuming static conditions. It is shown that the size as well as grounding conditions determine the field in the TEM cell and must be taken into account for electromagnetic compatability measurements. INTRODUCTION This paper concerns the calculation and measurement of fields in a TEM cell in the presence of an equipment under test. Figure 1 shows a typical TEM cell. It consists of a rectangular or a square coaxial transmission line. The ends of the line are tapered to allow coupling to an ordinary coaxial line. The TEM cell provides a shielded test environment with an essentially uniform and linearly polarized TEM field. Various uses of the TEM electromagnetic cell include compatibility measurements [1,2], study of biological effects [31, and calibration of electromagnetic field probes [4,5]. It has also been shown [6] that the TEM cell can be used for accurate antenna factor measurements even if the dimensions of the antennas are larger than the recommended one-third of the distance between the center conductor and the cell wall. CONNECTORS CENTRE ACCESS TEM DISTORTION EFFECTS In this paper, the distortion effects due to an equipment under test in a TEM cell are calculated. Both grounded and ungrounded equipment are treated and results of some measurements are also reported. Static approximation is used which implies that the dimensions of the equipment under test are much smaller than the wavelength of the incident wave. Electric fields are first computed for a cross-section of the TEM cell using an iterative procedure [7]. The effect of the tapered ends is excluded in this calculation. Figure 2 indicates how the iterative technique is applied in finding fields in the TEM cell. It shows a cross-section of a square TEM cell. Only one quarter of the TEM cell is considered because of the symmetry. In the method of iteration, the voltage at each point marked * (2.74 CONDUCTOR DOOR X 2.74 CELL X 5.48m) Figure 1 - ITERATIVE TECHNIQUE FOR FINDING IN A TEM CELL 76 - ELECTRIC FIELDS FOR FIELD ENHANCEMENT A METAL CYLINDER IN A TEM CELL ,’ LA-UL 8 h 7 CELL 1 x NBS MEASUREMENTS -*- G. MEYER - PRESENT WORK WALLS Figure 2 0.6 0.4 0.8 1.0 h/b Figure 4 ELECTRIC FIELD VARIATION IN A TEM CELL x TIPPET & CHANG ITERATION - METHOD 2 ? 2 5: 5 5: wm i.oiiD "2 z" I- !A!0.9 0 Li: W I 0.8 9 F I: & 0 I I I I 0.2 0.4 0.6 0.8 x/b Figure 3 I 1.0 is first assumed to be a certain value (say half of the center conductor voltage). The voltage at any point is calculated by averaging the voltages around the point. This is repeated until the voltage at any point does not change. Figure 3 shows some of the results obtained by this method. It shows the variation of the electric fields in the TEM cells with square or rectangular cross-sections. The results obtained by the iterative method agree quite well with those obtained by Tippet and Chang [B] using the method of conformal transformation. The fields in the presence of a metallic rectangular cylinder in a TEM cell are calculated in similar way. The fields are then compared with the fields at the same point in the TEM cell in the absence of the rectangular cylinder. Figure 4 shows some of the results obtained. For comparison the field enhancement reported by Meyer [q] and experimental results of Kanda [lo] are also plotted. Figure 5 shows the field enhancement for a grounded metal cylinder as a function of its height from the TEM cell wall. It shcws that the field distortion increases with the distance from the cell wall. Figure 6 shows the field enhancement for an ungrounded metal cylinder. Comparison with Figure 5 shaws that the field distortion is less severe when the metal cylinder is not grounded. These results suggest that both the size of the equipment and the grounding conditions nest be taken into consideration when a TEM cell is used for electromagnetic compatability measurements. 15D3 ELECTRIC ELECTRIC FIELD METAL ENHANCEMENT CYLINDER FOR A IN A TEM FIELD GROUNDED ENHANCEMENT IN A TEM BY A DIPOLE CELL CELL 4.0 r 20 &_- b/a ------t = 0.6 MEASUREMENTS NBS h/b Figure 7 x/b -se UNGROUNDED - Figure 5 GROUNDED - 15 20 25 JO FREQUENCY ELECTRIC FIELD h/b-0.210 w/a=0.175 0 0 I 02 0.3 0.4 0.5 R 40 45 M Wiz) ENHANCEMENT IN A TEM 35 EQUIP. EQUIP. BY AN EQUIPMENT CELL Figure 8 0.6 x/b ELECTRIC FIELD ENHANCEMENT EQUIPMENT IN A TEM Figure 6 BY AN UNGROUNDED CELL TEM cells are often used for calibration of probes [5]. However, the field disturbance caused by the probes is not very well known. Figure 7 shows the field enhancement due to a dipole in a TEM cell at a point directly above the dipole. Figure 8 shows some experimental results obtained for an equipment in a TEM cell. It shows that the field enchancement is much higher in the case of a grounded equipment. The location and height of the peak in the grounded case is dependent on the size and length of the ground cable. The effect of the ground cable on field enhancement will be discussed in detail in a future paper. - 78 CONCLUSIONS Field enhancement due to an equipment under test in a TEM cell has been calculated assuming static conditions. Both grounded and It has ungrounded cases have been treated. been shwn that both the size of the equipment and the grounding conditions determine the field in the TEM celt and must be taken into account for electromagnetic compatability measurements. - (41 M.L. Crawford, "Generation of Standard EM fields using TEM transmission cells", IEEE Trans. Electromagnetic Compat. Vol 16, pp. 189-195, 1974. [51 E.B. Larsen, "Techniques for Producing Standard EM Fields from 10 KHz - 10 GHz Monitors," Evaluating Radiation for Symposium on Proceedings of the 1978 in Biological Electromagnetic Fields Systems, Ottawa, Canada, June 28-30, 1978, pp. 96-112. [61 S.C. Kashyap, "Measurement of Antenna Factors with a TEM Cell," Proceedings of IEER 1984 National Symposium on Electromagnetic Compatibility, San Antonio, Texas, April 24-26, 1984, pp. 9-11. [71 W.H. Hayt, "Engineering Electromagnetics," McGraw-Hill Book Company, 1967, pp. 167177. [al J.C. Tippet and D.C. Chang, "Radiation Characteristics of Small Devices in a TEM IEEE Trans. ElectroTransmission Cell," magnetic Compatibility, Vol. 18, pp. 134140, 1976. [91 G. Meyer, "The TEM Measuring Ltne - A Critical Review" Proceedings of 1981 EMC Symposium, Zurich, pp. 407-412, 1981. t101 M. Kanda, "Electromagnetic-Field Distortion Due to a Conducting Rectangular Cylinder in a Transverse Electromagnetic Cell" IEEE Trans. Electromagnetic Compatibility" Vol. 24, pp. 294-301, 1982. REFERENCES [II r21 [31 M.L. Crawford, "Generation of Standard EM Fields using TEM Transmission Cells," IEEE Trans. Electromagnetic Compat., Vol. 15, PP. 189-195, 1974. I. Sreenivasiah, D.C. Chang and M.T. Ma, "A Method of Determining the Emission and Susceptibility Levels of Electricaly Small Objects using TEM Cells," NBS Tech. Note 1040, National Bureau of Standards, Boulder, Colorado, April 1981. W.T. Joines, C.F. Blackman and M.A. Hollis, "Broadening of the RF PowerDensity Window for Calcium ion Efflux from Brain Tissue," IEEE Trans. Biomed. Eng. Vol. 28, pp. 568-573, 1981. - 79 16~4 - CHAMBER QUALITY ASSESSMENT J. H. Davis and W. C. Cockerill International Business Machines Austin, Texas The IBM Austin 045 semi-anechoic chamber compares favorably to an open field site when measured against two models of site attenuation currently being used to measure site performance. The IBM Austin chamber's site attenuation measured to within 2.7 db of the OST 55 theoretical model of an open field site and, compared to the ANSI model, the Austin chamber shows a maximum positive deviation of 3.5 db. CHAMBER DESCRIPTION This test facility is a semi-anechoic chamber with all surfaces covered with pyramidal absorber materials except for the floor which is steel with a covering of vinyl tile. The steel enclosure is 60 feet long, 40 feet wide, and 30 feet high. The primary anechoic cone material is 6 feet in length. For some secondary reflection paths smaller cone material is used. The free area in the chamber is 48 feet long, 28 feet wide, and 14 feet high. both horizontal and vertical polarizaion, 2) over the entire frequency range of 30 MHZ to 1000 MHz, and 3) taking into account product volume. The model as described in the proposed revision to ANSI C63.4 is: A = 279.1 AFRAFT F Max MHzED Where: = frequency in MHz FMHz AFR = antenna factor of receiving antenna AFT = antenna factor of transmitting antenna EMax - maximum electric field in D receiving antenna scan range (See Reference [l] for equation) OST 55 Model To show consistency with other data taken for open field registration, we used a tuned dipole to measure the horizontal attenuation of our chamber against the OST-55 site attenuation model [3]. We made the measurements with a set of Roberts dipole antennas. i,,f; logloD+20 kk'lOFM -Gs -G -27.6-R+B R 81 D = 3 meters FM = frequency in MHz G = antenna gain = 2.15 db R = 4.2 for 3 meters reflection factor B = 2x.5 db balum loss m V rtical--Theoretical ANSI Theoretical Model To show chamber quality, we made site attenuation measurements using a biconical antenna for the 30-200 MHz band and a log periodic antenna for the 200-1000 MHz band. The results were compared to the model in the proposed ANSI C63.4 [2], to answer questions of chamber performance 1) for Fig. 1: Theoretical 3 Meter Site Attenuation, 10 db/div, 30+200 MHz - Method of Measurements The technique we employed to make all site attenuation measurements is described in the proposed revision to ANSI C63.4, Reference [2], where site attenuation equals the two antenna factors plus space loss. We made all open field measurements at Southwest Research Institute in San Antonio, Texas. This 30 meter site is a registered class B site. All measurements (unless noted) are with a source height of 1 meter and the receive antenna scanned 1-4 meters. Site Attenuation Model Verification To verify the ANSI model and check antenna factors, we made measurements in the open field at 10 meters and 3 meters. 10 Meters: The site attenuation for horizontal polarization at 10 meters dis tance was compared to the theoretical site attenuation showing the delta between the theoretical and actual. The difference is within t2.5 db. The maximum difference for vertical antenna polarization on the open field at 10 meters and the model was +2, -3 db. 3 Meters: Because our class B chamber measurements are at 3 meters, we also made a set of measurements on the open field at 3 meters to compare to the model. The results (Figure 2) show that the model describes an open field at 3 meters to within +2 db, -3 db. Because both vertical and horizontal polarizations are within +2, -3 db, we can say the model is an excellent 80 - representation of a good open field. These measurements also verified our antenna factors since a significant error in the antenna factors would produce a common offset in these results. Antenna Factors To make all site attenuation measurements, we used a biconical antenna (AIL TECH 94455-l) from 30 MHz to 200 MHz and a log periodic antenna (AEL-APN 113C) from 200 MHz to 1 Ghz. Using the standard antenna method, we measured the biconical dipole antenna factors from a set of dipole antennae calibrated by the National Bureau of Standards. The log periodic antennas were calibrated using the standard site method with the source antenna at 4 meters high, distance of 10 meters, and the receive antenna scanned 1-4 meters on the open field. Since the horizontal and vertical antenna factors are almost identical, we used the average antenna factor. _ 111 boritortal-A?tennz Facto’ -_ -33 .--i --3: i_. _ _.--.-_- a -- 11,11,1IN 15 ___._.__-.__--- Y---t-t----t----t-f--l * * --1 Vetical: Measured-Theor tical ---.---- g 18 --- / -L r: .___. .i I-, Fig. 2: Open Field Minus Model At 3 Meters, 30 MHz to 200 MHz, 2 db/div Fig. 3: ._.-i.._ ,.._._,_____ ...i_.._ ..-.._. .,~.. .._ _. .f ...i.-._ -c _.__.. _ _ -.__ __-_--__I_ I ~._. .j_..._ __.. .-^__i__..._ t ..-.. -...._ _ . __.__. ._. . ____.._._.._ . .___.. _.._.._ _._..__ Measured Antenna Factor For Log Periodic Using Standard Site Method,10 db/div - 81 - FREQUENCY (IN MHz) 30 40 50 60 70 80 90 100 120 150 175 200 225 250 275 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000 FIG. 4: ATTENUATION (MEASURED) 13 13 12 13 13 15 16 17 17.5 18 18 19 21 22.5 23 23.5 24.5 27 26 27 29 29.5 30 30.5 31.5 33 34 33.5 34 33 DIFFERENCE (MEAS-CALC) 2 2 1 2 1.8 2.6 2.6 2.7 1.6 .1 -1.2 -1.4 -.4 .2 -. z ::7 .6 -1.4 -1.3 1 ::4 -.6 -.7 -.3 .6 1.1 .l .l -1.3 Chamber MeasurementsVs. The FCC OST-55 Model 16~4 Chamber Quality Data The IBM Austin chamber's site attenuation measured to within 2.7 db of the OST 55 theoreticalmodel of an open field site and compared to the proposed ANSI model, the Austin chamber shows a maximum positive deviation of 3.5 db. Chamber Quality Vs. OST 55 Model Figure 4 shows the chamber results when compared against this FCC model of an open field. Chamber Quality Vs. ANSI Model To give a more complete picture of the chamber quality, we compared chamber meas urements to the proposed ANSI model [2]. The maximum positive differencebetween the chamber data with the source at the center of the turntable and the theoret ical model is +2.5 db and the maximum negative error is -4 db. Above 80 MHz, the error is within +2 db, -3 db for both polarizations. Horizontal: Figure 5a shows the meas ured attenuationdata for the chamber with the antenna in the horizontalpolarizations. Figure 5b shows the differencebetween actual data and the theoreticalopen field model for horizontal polarizations. The horizontal results are within +2.5 db. The positive errors will result in the apparent EM1 from a product to be lower than it should be on an ideal open field. L-_._I.____L__/&_-._I lDDDHl Fig. 5a: Chamber Site Attenuation,Horizontal, 30 MHz To 1000 MHz,10 db/div .,l *I +B .l .I Fig. 5b: Chamber Minus ModelJHorizontal,30 MHz to 1000 MHz, 2 db/div - 82 - Vertical: Figures 6a, 6b and 6c show similar data for the vertical polarization. Figure 6a shows the chamber site attenuation. Figure 6b shows the theoretical open field. Figure 6c shows the difference in 6a and 6b. The site attenuation vertical error (Figure 6c) is within +2.5 and -3 db of the theoretical. Fig. 6a: Chamber Site Attenuation, Vertical, 30 MHz To 1000 Mhz, 10 db/div a?---- --- ---- --.-----.-~. -- . . . .--.- -.... -.- .._. __ ._ ._._ . .._.. _ . _.__... --r-. ..-. .. -... ----.---..-____ .-. _.__ J- __._.__.___.____ High Band Fig. 6b: Fig. 6c: Theoretical Open Field Site Attenuation, 10 db/div Chamber Minus Model, Verti cal, 30 MHz To 1000 Mhz, 2 db/div __..__ ,... I_ - Chamber Quality With Source Position Variation To assure that these 3 meter measure ments are representativeof data which would be obtained from the volume occupied by a product, a series of tests were made in which the position of the source antenna was varied. For both horizontal and vertical polarizationswe used four source positions. Since all sources in a 1 meter square product rotate to within the distance of 3 to 3.5 meters of the receive antenna when the turntable is scanned, measurementswere taken at both 3 and 3.5 meters. The maximum positive deviations over this volume in our chamber measurements,which would cause a tested unit to read low, are within 1.5 db of the open field results. Therefore,we expect similar results as measuring in a a3 - good open field. The maximum negative deviations are also within 1.5 db of the open field results. Horizontal: Figure 8 shows the source antenna positions for the horizontal polarizations. Figures 9 and 10 show the most positive and most negative deviations from the ANSI model for the four horizontal source positions. The most positive maximum deviations in the chamber is 3.0 db above the theoreticalmodel. As seen in Figure 2, on the open field, we measured deviations of 2 db above the model. Therefore,we expect chamber horizontalpolarizationmeasurements to be essentiallyequal to those on a good open field. RECEIVE SCANNED l-4M 1M \1 Fig. 7: Horizontal Source Position Variations MOST POSITIVE/NEGATIVE Fig. a: DEVIATIONS FROM THE THEORETICAL Chamber, Max. Deviations From Model, 30 MHz To 200 MHz Horizontal, 2 db/div Fig. 9: FOR SOURCE POSITIONS SHOWN Chamber, Max. Deviations From Model, 200 MHz To 1000 MHz Horizontal, 2 db/div - 84 - Vertical: Figures 12 and 13 show the most positive and most negative deviations from the ANSI model for the four vertical source positions. The maximum positive deviations is 3.5 db above the theoretical, which is within 1.5 db of the results we saw in the open field test. The most RECEIVE negative deviations are also within 1.5 db of the 3 meter open field results. Therefore, we expect chamber vertical polarization measurements to be essentially the same as those on a good open field. SOURCE Vertical Source Positions Conclusion 111 For the total frequency spectrum, horizontal and vertical polarizations, and positional variations of the source, the IBM Austin 045 semi-anechoic chamber compares very closely to the theoretical models of an ideal open field and should give results essentially the same as a good open field. Q .a *. .z I -2 -4 Acknowledgements d AA -II r Fig 2: 1 Chamber, Max. Deviations From Model, 30 MHz To 200 MHz Vertical 2 db/div II The authors wish to thank Albert A. Smith, Jr. of the IBM Poughkeepsie EMC Lab for his advice in preparing this paper. We would also like to express our appreciation to the IBM Boulder, Boca Raton, and Lexington EMC personnel who participated in the review effort. to .8 References 111 A. A. Smith, Jr., R. F. German, J. B. Pate: IEEE Transactions on EMC Vol. 24, No. 3, August, 1983, 260-265 [21 "Open Area Test Sites," draft No. 5 (April 1982) addition to American National Standard C63.4 - II a*a Fig. 12: Chamber, Max. Deviations From Model, 200 MHz To 1000 MHz Vertical 2 db/div [31 "Characteristics of Open Field Test Sites," FCC Bulletin OST 55 - 85 1705 - DISCRIMINATING BETWEEN NARROWBAND AND BROADBAND EM1 USING A SPECTRUM ANALYZER ______ ___----__--____- __--__--____________________-Siegfried Linkwitz Hewlett Packard Co. Santa Rosa, California, USA Abstract The discrimination between narrowband (NB) and broadband (BB) interference is important for commercial and military EM1 tests because different interference limits apply. The measurement of NB signals which are mixed in with BB signals can be difficult using commercial test methods. In MIL-STD tests the selection of bandwidth and their corresponding correction factors can lead.to errors. These issues are investigated and alternate test methods presented. Since the bandwidth is not specified for MIL regulations, a test has to be performed to determine whether the EM1 is NB or BB. For BB signals a bandwidth and signal type dependent normalizing factor has to be applied to compare the measured voltage to the BB limit. This approach can lead to misinterpretations of signal types and errors in determining the appropriate normalizing factors. (1) Introduction Electra-magnetic interference is classified as NB if its spectrum is contained within the bandwidth of the measuring EM1 receiver and BB if the spectrum is wider than the receiver bandwidth. In CISPR type, commercial tests, the bandwidth of the EM1 measuring receiver is specified. In addition, quasi-peak detection is specified in order to properly evaluate the annoyance level of BB interference. It has been found recently, that a low level NB signal which is mixed in with a larger amplitude BB signal can be more annoying to TV viewing than the BB interference by itself. The quasi-peak level, though, might give little indication of the presence of the NB interference. MKRD.B HZ fP RE IB dW Fig.2: Measured signal amplitude depends on receiver bandwidth. The normalized BB signal strength in dBuV/MHz is constant only for a few bandwidth settings. To properly measure the interference potential of different signal types it becomes important to understand the characteristcs of the measuring instrument, how they affect the measured signal amplitude and how they can be used to distinguish between signal types. Spectrum Analyzers The correct evaluation and measurement of EM1 is simplified by observing its spectrum on a spectrum analyzer display with varying bandwidth, frequency span, sweeptime or video bandwidth. tENTER18.9822 HHZ RESBW 18 kH7. Fig.1: VW 108kHZ SPAN588.8kHZ SW 38 al*00 A lower amplitude NB signal is masked by quasi-peak detection. Fig.3: Spectrum analyzer block diagram. - 86 - The absence of preselection below 2000 MHz spectrum analyzers may lead to overload of IEe input mixer and a false spectrum display. analyzer, a dB change in input attenuation will give an identical dB change in the amplitude of the displayed spectrum. This can easily be detected by changing the input attenuation. In linear operation of the Four basic methods are available to distinguish between NB and BB signals. (3) METHODS TUNINGTEST "TUNE" A BWi PRFTEST A SWEEPTIME NB A AMPL ) 3dB NO A SPACING (LINE PEAKVS.AVG.DET CRTRESPONSE BB MODE) A AMPL < 3dB SPACING (PULSE MODE) NO A AMPL A AMPL NO A AMPL A AMPL AVIDEOBW BANDWIDTHTEST A RESOLUTIONBW Fig.4: Methods for NB and BB analysis. In the Tuning Test, the variation of the spectral amplitude over the bandwidth of the analyzer is the criterium for classifying signals. Detector Parameters 1. 2. The PRF Test looks for changes in spectral line spacing as function of the analyzer sweep time. True spectral lines are unaffected by sweep time but impulsive signals with PRF's less than the analyzer bandwidth will change their line spacing and therefore be recognized as BB. CISPR EM1 receivers employ quasi-peak detection (QPD) for anannoyance weighted signal indication, while spectrum analyzers use peak detection to determine the absolute signal amplitude. As a result, a BB signal such as a constant amplitude pulse train with a PRF less than the analyzer bandwidth, i.e. with several spectral lines within the analyzer bandwidth, will be displayed as constant amplitude time domain pulses. The QPD indication, though, will increase with PRF but always be lower than the peak indication. 110 3. In the Peak vs. Average Detection test, a reduction in the post detection video bandwidth reduces the amplitude of BB signals (smoothing) without affecting NB signals. 4. The Bandwidth Test looks for spectral amplitude changes with varying analyzer bandwidth. A signal whose spectrum is contained within the analyzer bandwidth will display constant amplitude and is NB. With an increase in bandwidth by a factor of 10 the signal display will increase 20dB for a coherent BB signal and 1OdB for a random BB signal. A measurement ambiguity arises when a signal behaves as BB only over a certain range of bandwidths. Fig.2 100 90 Fig.5: BAND B 10.1540 vpz= l/Z MHz, x 0.316p”* = 9 kHZ x 1.05 = 9.5 kHz Peak, quasi-peak and average detection of pulses according to CISPR Publ.16 - 31 The average detected signal indication increases linearly with PRF and coincides with the QPD and peak detector readings when the PRF exceeds the analyzer bandwidth, i.e. when individual spectral lines are measured. For B8 signals the average detected signal amplitude is lower than either QPD or peak detected indication. This property can be used to determine the amplitude of a NB signal which is mixed in with the BB signal. (2) Average detection with a spectrum analyzer is obtained by reducing the video bandwidth to less than the resolution bandwidth. The amplitude display has to be in Linear mode. In the Log amplitude mode, video filtering smoothes the logarithmically distorted detector output signal. For BB impulsive signals the smoothed indication is considerably lower than the average value of the impulses. 20 - RESOLUTION BANDWlDTkl10 “KWD BANDWIDTH lOOkHl ,PEAKI 3 Hz IAVG & SMOOTHED, 17D5 - Impulse Bandwidth The absolute measurement of BB signals requires knowledge of the effective analyzer bandwidth in order to normalize the measured amplitude to that of a reference bandwidth as is necessary for MIL-STD tests. For coherent BB signals the impulse bandwith is determining, for random BB signals the noise bandwidth has to be known. The impulse bandwidth relates the peak output pulse voltage from a filter to the spectral intensity of the pulse at its input. A short duration, large amplitude pulse at the input to a bandpass filter results in a reduced amplitude but longer duration pulse at its output. Consecutive filtering stages in a receiver decrease the pulse amplitude, each contributing to the impulse bandwith. kHI NONE Fig.6: Post-detection video filtering of impulsive signals. This smoothing effect in Log mode allows a more accurate measurement of the NB component in a mixed NB/BB spectrum than average detection in Linear mode. Furthermore, the measurement dynamic range is larger so that even low level NB signals in the presence of larger amplitude BB signals can be measured. The video bandwidth needs to be reduced only to the point where the rapid fluctuations of the BB signal are smoothed. Further reduction will not change the measured value but will increase the required settling, analysis and measurement times. Using a spectrum analyzer, mixed NB/BB signals can often be measured without the help of video filtering since the NB signal is clearly visible on the display, especially in Log amplitude mode. 1. AVERAGE DETECTION 2. DISPLAY SMOOTHING IN LINEAR AMPLITUDE IN LOG AMPLITUDE Fig.8: The impulse bandwidth is often assumed to equal the 6dB bandwidth. A relatively simple procedure allows the accurate determination of impulse bandwidth. By measuring the response to a pulse train of variable PRF where the first measurement is taken at a PRF significantly below the receiver bandwidth and a second measurement at a PRF at least three times higher than the bandwidth, the impulse bandwidth can be calculated. MODE MODE Fig.9: Fig.7: Pulse amplitude reduction from bandpass filtering in a receiving chain. Mixed NB/BB signals in Lin and Log amplitude modes and with video filtering. Determination of impulse bandwith from measurements of a variable PRF pulse train. The only assumptions for this test are a pulse shape which is constant and independent of the PRF and a pulse spectrum envelope of constant amplitude over the resolution bandwidth. The latter condition can be verified from the spectrum analyzer display or by detuning the receiver. - Every bandwidth limiting element in the receive chain affects the impulse bandwidth. In particular, the video bandwidth has to be three to ten times wider than the resolution bandwidth to minimize its influence and for the impulse bandwith not to vary with pulse width. 120 c “BW 100 Hz -2, lMP”LSE BANDWDTH M ‘100 80 - BW3ds=100Hz 60 - lOO&lSEC 10 PSEC 1000 @SEC 10.000 @EC PULSE WIDTH TEFF Fig.10: Variation of impulse bandwidth with pulse width for different post-detection video bandwidths. aa For commercial EM1 measurements, low level NB signals with serious interference potential can be "masked" by impulsive signals when quasi-peak detection is used. Peak detection allows both NB and BB signals to be observed on the spectrum analyzer display. NB signals can be further "enhanced" on the spectrum analyzer display by using video filtering, especially in the Logamplitude display mode. Military EM1 measurements which require the discrimination between NB and BB signals seldom specify bandwiths for these measurements. Often, the normalized BB measurement results are constant only over a specific bandwidth range. Furthermore, the actual impulse bandwidth should be determined before normalization. The spectrum analyzer's display, range of bandwidths, frequency span rates, peak and average detection are powerful1 aids for the analysis and measurement of NB/BB signals. The change of impulse bandwidth with pulse width is caused by the non-linear relationship between the peak voltage at the video filter output and the width of the pulse at its input. It becomes significant when the video bandwidth is less than the resolution bandwidth. Once the impulse bandwidth is known, a measured BB signal voltage in that bandwidth can be normalized to a reference impulse bandwidth, i.e. 1 MHz for MIL-STD tests. The normalized voltage is then compared to the appropriate BB limit level. COMMERCIAL LOW LEVEL NB AND BE SIGNAL . SA DISPLAY USING PEAK DET . VIDEO FILTERING FOR AVERAGE . LOG AMPLITUDE FOR INCREASED To overcome the additional testing required and the ambiguity in the choice of bandwith for the measurement of a signal, a new bandwidth specification is being considered such that a BB signal measuring at the NB level would normalize to the BB limit level.(l) 120 DET DYNAMIC RANGE r OBWi = NB LIMIT-BBLIMITdB -20 NB/BB - . WIDE . FOUR 4’01 ’ 10 ldiz Fig.12: ““0” 100 ktlz ’ “““” 1 MHZ ’ 1”““’ 10 MHZ ’ “““” 100 MHZ ’ ’ “““1 1 GHZ ’ “‘UJll 10 GHZ FREOUENCY Fig.11: NB/BB limits for MIL-461 REOZ and impulse bandwith required to match BB with NB limits. In practice, the continuously changing bandwidth will have to be approximated by discrete settings which in turn requires a single limit level without consideration of NB or BB siqnals. Conclusion Performing commercial or military emission tests can result in unique measurement difficulties. SIGNAL RANGE NBIBB DISCRIMINATION OF BW’SAVAILABLE DISCRIMINATION METHODS Difficult EM1 measurements simplified by the versatility of spectrum analyzers. References (1) R.B. Cowdell, An analysis of MIL-STD-462 Application Note: Identification of BB and NB Emissions, IEEE Symp. 1983, CH 1838-2/83/0000-0038. (2) FCC Rules & Regulations, Vol. II, July 1981, Part 15, Radio Frequency Devices, Subpart J, Appendix A, Section 4.2.2. (3) MIL-STD-462 Application Note: Identification of BB and NB Emissions, 5/1980, Electromagnetic and Interference Compatibility Branch, Wright-Patterson Air Force Base, Ohio 45433. - BROADBAND YIG-TUNED 89 18 - PRESELECTOR D. Raicu, R&D Center California The dynamic range of a spectrum analyzer extends between its noise level and the highest signal level for which the nonlinearity remains tolerable. The measurement range is, however, broader, since the use of input RF attenuators makes it possible to handle input levels up to the available power of the signal source. A low noise preamplifier improves the overall noise figure, extending the measurement range of the spectrum anaMeanwhile, the dynamic range lyzer. decreases, since the input nonlinearity threshold is affected stronger than the noise level. Broadband preamplifiers bring about an additional problem. When different signals with widely varying levels are present at the input, they may overload the first stages and distort the spectrum analyzer indication. It is recommended therefore to use at the input a very linear tracking filter,‘ which eliminates the undesired frequencies and their overloading effect. Widely used in the microwave range are the magnetically tunable YIG filAt frequencies in the VHF and ters. UHF range (below 1 GHz), regular YIG filters up to now were not available. The reason is that achieving ferrimagnetic resonance without early limiting below 1 GHz requires a totally new YIG filter design. Low frequency limitations in the range of YIG filters Two factors are mainly responsible for the low frequency limitations of the YIG filters as presently built. U.S.A. field H, the internal magnetic Hi is field = H - Nz 41rMs Hi where N is the axial demagnetizing factor 2nd 4nMs the saturation magnetization. The resonance frequency spheroid is given by w 0 Y "-0 The first is related to the effect of the demagnetization. For a ferrimagnetic spheroid immersed in an axial FOR VHF AND UHF G. U. Sorger Eaton Corporation, Sunnyvale, ~6 = y, of the YIG [II - (N 2 -ii T) LITMUS= Hi+NT 4nMs 1 (2) in which y stands for the gyromagnetic ratio a8d N is the transverse demagnetizing T actor. It appears that even neglecting the internal field required for saturation, there is a lower limit of the resonance frequency l*) 0 = yoNT 4nMs (3) min For spheres of pure YIG, this gives a limitation at about 1.6 GHz, and YIG microwave filters operate mostly above this frequency. By using oblate spheroids and doping the crystal, both NT and 4nM can be decreased by an order of magn?tude. Under such conditions, however, a second factor takes over. It is related to the anisotropy field. The resonance frequency of an anisotropic material depends on the orientation of the field with respect to the crystallographic axes as well as on the magnitude of this field. For negative anisotropy crystals magnetized along the easy axis, for instance, the minimum resonance frequency is w 0 2 =--y 0 Ha min 3 - where H is the anisotropy field. For pure YI8, at room temperatures, this means about 150 MHz. taking into account onTheoretically, ly the first order anisotropy, there exist certain directions of the magnetization vector for which the resonance frequency can reach zero for a certain field magnitude. Our goal was to determine the locus of these directions and to select one of them as the spheroid axis, along which the magnetic field is directed. As mentioned before, an oblate spheroid should be used rather than a sphere, and the crystal should be doped for low saturation magnetization. This complete set of conditions should make possible to obtain a sharp resonance down to significantly lower frequencies. We were interested to cover a range starting at 100 MHz, and extending up to 2000 MHz. LOW frequency ferrimagnetic in an anisotropicaterial without affecting the generality of In a spherical system of the results. coordinates, this becomes, for example cos$* Tf, f RAO 1 (7) vu u* M 0 (8) From (7) and (10) we obtain Hij=H 2K1 K1 3 .+-- oi- 21\11 U. oJ MO J 0 Normalizing and denoting spherical h ix ferrimagnetics 2 2 (11) HM field h=L 21K11 we obtain, in the magnetic K1 s =-, /K1/ coordinates = cos+sinO[ho+s(l-cos2$sin20)] 11. = sin$sin8[ho+s(l-sin2$sin28)] lY = cos?[ho+s sin201 (12) The spherical coordinates $. and oi of the vector E should satisf$, in turn, a relation similar to (6), which implies a minimum magnitude for the d.c. field, ho, which can be obtained For posialong the direction (S,$). tive anisotropy s = 1 and the calculation yields ho> -cos20+ For negative obtain sin2$sin$cos8 anisotropy (13) s = -1 and we (14) We normalize the relation Q= the frequencies 2 2 2 VA = K,, (ix1 ~1~ + o/22c13+ a3 ~1~ ) (9) by using lL IKll STY 0 and we introduce the variable ho + s x=------3 Then, the normalized cy is given by in which VclUA is the gradient of the energy with respect to anisotropy the components of the unit vector z parallel to the magnetization (see, for example, [ll). For cubic (10) 0 ho& sin2 0 where G?. is the internal field and E the d.c? part of the anisotropy fiel 4: defined as TI, = - HA= - (6) In a saturated crystal, the magnetization vector is produced by an equivalent d.c. field parallel to it, whose expression is = where ~1. are the projections of z on the axe&, and K the first order conConsequently stant of anisot&opy. hiz (5) HO - resonance Consider a cubic single crystal in a coordinate systern with axes parallel to the cube edges. A given direction is characterized by the angles $1, $2, Due to the symmetry, $3 to the axes. we can confine the study to a domain within which the angles JI. satisfy a J relation of the type sin 0+sin @ 2 sine. >cose>o 90 resonance frequen- n2 = x2-2sx(sin28sin2$cos2+cos28) + 3 sin40 cos28sin2$cos9 (15) Now, the conditions for the lowest possible resonance frequencies can be found by looking for the roots of the equation (15) and imposing for them the requirement to be consistent with 91 the conditions tively. (13) or (14), respec- For negative anisotropy, the calculations show that this is possible only for fJ = n/3?. Extending this condition from the elementary domain considered to the complete range of spherical coordinates, it defines the orientations of the magnetization lying in crystallographic planes (100). For positive anisotropy, one of the roots of Equation (15) is acceptable for $ = r/4. Again, extending this condition over the complete unit sphere it defines orientations of the magnetization lying in (110) planes but only within the angles between pairs of adjacent 11111 axes separated by one LllOl axis. 18~6 - polygones whose contours constitute the sets from Firr. 1. The dashed lines represent the maximum tolerable departure from the contours, defined by the requirement of reaching a given WO=&, small but still different from zero. It appears that the accuracy requirements are less tight in the neighborhood of the polygone vertex (that is, in both cases, around the axes of hard magnetization), and most critical around the axes of the [110-J type. The second root of (15) is consistent with the condition (13) for orientations of the magnetization covering a narrow two dimensional domain around the tllll axis - which is the hard axis of magnetization for positive anisotropy crystals. In order for the magnetization to be that close to the hard axis, the applied field should be even closer. In fact,the tolerable departure from the [llllis so small, that we can assume that the condition is met only along the CL111 axes, so that we have no practical extension of the one-dimensional domain obtained from discussing the first root of Equation (15), and we can restrict our considerations to the former. Fig. I. shows the loci of the intersections of the unit sphere of coordinates with the orientations of the magnetization vector which make possible a low frequency resonance. For a negative anisotropy, the sphere is separated into eight spheric equilateral triangles and for a positive anisotropy, it is separated into six spheric squares. In Fig. 2, a detail of the locus for positive anisotropy close to one [ill] axis is represented, showing the small two-dimensional domains previously discussed, mentioned for the sake of theoretical completion. It should be kept in mind that orienting the magnetization vector along an axis belonging to the loci in Fig. 1 is just a necessary condition for achieving a minimum (theoretically zero) resonance frequency. A further requirement is for the equivalent field to have the precise magnitude obtained from Equation (15). Any orientation belonging to the locus from Fig. 1 can be selected, but from a technical viewpoint, it is important to find one which is less sensitive to unavoidable slight misorientations of the single crystal disk. Fig. 3 and 4 show separately two of the spherical Fig. 1: Magnetization orientations compatible with a zero frequency ferrimagnetic resonance This is a reason for considering the hard axes of magnetization as the optimum choice. An additional benefit comes from the fact that only along the main axes are the internal magnetic field and the resulting magnetization parallel, which makes the design simpler and potentially more accurate. As shown before, the demagnetization effects should be also considered. They account for a difference between the applied field and the internal field. A flat shape, normal to the direction of the applied field, is required for having a possible low resonance frequency. When the flat rotation ellipsoid is obtained from a single crystal so that its rotation aXiS is parallel to the hard axis of magnetization, all the conclusions derived before, by considering the effects of anisotropy, remain valid if a corrected value of the affective field is used (Bo+N 471-MSinstead of Y Ho). l - 92 - Pure YIG has a negative anisotropy. The most common way to decrease its saturation magnetization is to dope it with Gallium, which does not change the character of the anisotropy. The calculations presented above covered the general case because for other compositions the ferrimagnetic material can have a positive anisotropy. Actually, it could be beneficial to dope the crystal with ions whose effect would be to reduce the negative anisotropy, to compensate it completely or even to cause a positive resulting anisotropy, if Kl remains small. A reduction in the magnitude of the first order anisotropy constant, no matter what its sign is, makes the relative orientation of the ellipsoid rotation axis and the crystallographic axes less -_)....--I 1 c1 critica1 . mL.z,.-..7 2 J.t.dU 7 -,r-lto slrrlpu_II11,122c;"UIU cations in the technology of the ferrimagnetic body used as a resonator and in the construction of the filter. Design Fig. 2: The vicinity of a rlll] axis where the magnetization vector can satisfy (15) and (13) for positive anisotropy. Fig. 3: Tolerable misalignment (dashed line) of the magnetization from the optimum orientation (full line) for negative anisotropy. Considerations A one stage filter was designed based It is on the YIG resonator described. desirable to use the uniform precession mode only, avoiding the effect of other modes. This is not easy to achieve, particularly considering the fact that for a disk shaped resonator, the density of the magnetostatic modes is substantially higher than for a spheric resonator [?,I. It is important to have Fig. 4: Tolerable misalignment (dashed line) of the magnetization from the optimum orientation (full line) for positive anisotropy. - 95 - 19D7 MEASUREMENT OF THE IMMUNITY OF BROADCAST RECEIVERS ACCORDING TO THE CISPR METHOD AND THE DIFFICULTIES ENCOUNTERED G.K.Boronichev The USSR Ministry of Telecommunications Moscow, USSR Difficulties are conaidered,which arise when controlling the immunity of broadcast receivers by a CISPR method. It is shown that a response of an AM receiver to a test interfering-signal (a test stimulus) is critical not only to the parameters of an interference but to a greater extent depends on a difference between the frequencies of a wanted signal and interference, on the relationship between their amplitudes, on the frequency stability of the measuring generators which are used, etc. A thesis is questioned, that it is no use controlling the immunity of USW FM broadcast receivers in that case when an interference affects an IF stage. Data of experimental investigations are given, which confirm the results of the performed analysis. A conclusion is drawn that the CISPR methods of controlling the immunity need further improvement. l.Introduction Solution of EMC problems by guaranteeing a required immunity of radio facilities to external interference is very promising. This stimulates development of corresponding investigations. In this connection, attention should be drawn to a CISPR document [I] where an attempt is made to work out methods of measurement of immunity of TV and broadcast receivers to external electromagnetic fields as well as to interference currents and voltages Induced by these fields in connecting cables and other elements of the receptor circuit. These methods were tested on TV receivers but they seem not to be polished enough when applied to broadcast receivers. The present Report is devoted to analysis of reasons which allow to draw such a conclusion. 2,Rffect of an interference on the RF and IF stages of an AM receiver 2.1. Theoretical considerations. In accordance with 117 . the immunity of IF and LF stab& is controlled in AM receivers. The immunity of RF .*+#%nnr. .8nIL"" n** rllralaur=iu.. mn~‘-..rn~a o"oe~'u .bL9 A test interfering-signal (a test stimulus) is introduced into the radio receiver by means of a TEM cell which simulates an interference field, or by means of special devices which inject interference currents and voltages into different points of the receptor. AM oscillations are accepted as a test interfering-signal, and during the. tests, at the input of the radio re&iver, It is shown in [2, 3] that currents and voltages which arise in the most sensitive points of the receptor under the influence of the electromagnetic field of an interference source, are a basic mechanism of the effect of interference on a radio receiver. The RF and IF stages of a radio receiver should be assigned to such sensitive points. Thus, the problem resolves itself into a detern,4nn+ir\n nC a response of a receptor III.l.IIcI"I"~I V.I. to an effect of an AM interference in the presence of a wanted unmodulated signal. This response should be determined at the output of a low-pass filter which has a cutoff frequency f > 1000 HI,and which is connected t8 the output of the radio receiver. As a first approximation, this problem can be solved on the basis of analysia of the voltage at the output of the detector to the input AllI mirm.a, ..-A --lr\..‘...* of *“h-i-h en a&I%4 a ElL11UrY”Idal"'~~~$"a~~ ~~~~?~d [4] : - frequency part of the range and causes a slight increase in the insertion loss at the higher frequencies. The input l-dB compression point of the filter was higher than+10 dJ3m over most of the range. It decreased at the lower frequencies rather sharply, but at 100 MHz remained higher than 0 dBm. In order to keep the filter response in the passband spur free, the measures described before for suppressing spurious magnetostatic modes had to be complemented by a controlled nonuniformity in the DC magnetic field, obtained by a slight departure from the parallelism of the two planes limiting the This proved to electromagnetic gap. be effective, especially in the higher part of the frequency range, where the problem of mode purity was more critical. A low noise amplifier was built to work Its in conjunction with the filter. frequency characteristic was designed so as to complement the insertion loss characteristic of the filter throughout Since the increase in the the range. measurement range brought by the combination is accompanied necessarily by a decrease in the dynamic range, the gain of the amplifier was selected SO as to maximize the sum of these two It was thus possible to imvalues. prove the overall measurement range by 17 dH, with a reduction of only 3 dB The gain of the in the dynamic range, cascaded amplifier and filter was flat in the frequency range, and the decrease in the filter response at the frequencies close to the lower limit remained manifest only in the corresponding increase of the overall noise figure at these frequencies. Conclusion A variety of technical problems impedes the operation of YIG filters at frequencies as low as 100 MHz, but the most important of them is the theoretical limitation imposed by the anisotropy of the single crystal resonators. Our analysis produced the conclusion that it is possible to substantially expand downwards the operating frequency range of such resonators by properly orienting their crystallographic axes with respect to the applied DC field. On the other hand, in order for a low frequency resonance to be compatible with the requirement of magnetic saturation of the material, the resonator has to be shaped like an oblate ellipsoid (or disk), rather than a sphere, with the DC field applied along its Both conditions can be rotation axis. met simultaneously only with the resonator axes and the crystallographic axes accurately positioned with respect to one another at the time when the resonator is manufactured from the 94 - single crystal. This problem does not a.r:lse,obviously, for spheric resonators which have no preferred axis, but is essential for achieving the ferrimagnetic resonance at relatively low frequencies. Once the resonator as described is available, a whole series of measures (related to the coupling circuits, to the magnetic circuits, to the details of the resonator shape etc.) become effective in improving the filter performance up to levels which make possible its application as a preselector in spectrum analyzers. Together with a low noise amplifier, it brings a very significant improvement in the instrument performance, in a frequency range where the advantages of preselection with YIG filters seemed to be out of reach. References Ill Sodha, M.S. and Srivastava, N.C.: Microwave Propagation in Ferrimagnetics. Plenum Press, New York, 1981 Magnetostatic ISI Walker, L.R.: in Ferromagnetic Resonance. Rev. 105, p. 390, (1957) Modes Phys. Magnetostatic I31 Dillon, Jr., J.F.: Modes in Discs and Rods. J. Appl. Phys., 31, po 1605, (1960) - 97 - 19D7 receivers requires additional confirmation because it could happen that the test data were affected by the measurement errors which were left out of account. At the same time, a comparison of data of Tables 1 and 2 shows that the immunity parameters of receivers at the frequencies of the second channel of reception are lower than at the intermediate freu quency. to some degree. Toward this end, having calibrated a receiver by means of a wanted signal whose modulation factor remained at the previous level, a test interfering-signal was generated, which had a form of unmodulated harmonic oscillations. l3y tuning the generator that produced these oscillations, it was achieved that zero beats appeared at the output of the receiver under test. Then, while removing the modulation from a wanted signal and modulating the interference oscillations, the interference level was varied so as to obtain, at the receiver output, an interference voltage which corresponds to the receiver output power that is by 20 dB lower than its standard output-power. Thus, conditions were created, for which expression (3) is valid. It turned out that, with a suitable stability of HP generators being used, the tests of radio receivers made by such a modified technique lead to obtaining stable results. Tables 1 and 2 give averaged data of experimental investigations of immunity parameters. Table 1 pertains to that case when the frequency of a test interfering -signal was equal to the frequency of the tuning of the radio receiver to the second channel of reception. It follows from Table 1 that broadcast receivers are sensitive to effect of interference at the frequencies cf the second channel of reception, the receiver immunity decreasing with increasing tuning-frequency. 3. Effect of an interference on the Kr' and JP staQes of an FM receiver 3.1. Theoretical considerations. It is indicated in tll that when an electromagnetic field affects the IF stages of USW FM broadcast receivers, the control of immunity of such receivers is not of great importance. This is accounted for by the presence of an amplitude limiter before a frequency-sensitive detector. It can be shown that this assertion is not quite true. Let us suppose that an interference affects an RF or IF stage of a receiver. As a result, at the input of a frequency-sensitive detector, there will be an interfeaddition to a wanted ;;;;;lUflFI in Let us assume for simplicity tha?*an interference has a form of unmodulated harmonic oscillations. Suppose we have that f is a wantedsignal carrier-frequency, f PI is an interference frequency 4 4dif 111 =/fs-f& K>1 l Table 1 Type of a receiver and the frequency of its tuning, ft Method of introduction of interference energy An open TEM cell, V/m Type I, ft = 0.25 rm~ 357 w3 !Cype1, ft - 1 MHz 221 10'3 3 Type 2, ft I 0.25 MHz Type 2, it zc 1 MHz 1.8 Table 2 pertains to that case when the frequency of interference is equal to the intermediate frequency of the receiver. Data of Table 2 evidence that broadcast receivers respond also to interference whose frequency coincides with the intermediate frequency of the receptor. It turned out that for type 1 radio receivers there is no dependence of the measured immunityparameters on the tuning frequency. The fact of such dependence for type 2 Into the antenna input, V 1.6 Into the supply mains, V 0.5 lO-3 10-3 45 10'3 29 10'~ 29 10-3 0.4 50 10’3 Then (according to [4])a signal frequency proves to be modulated by a difference frequency and its harmonics. In this case, Afdhffrequency deviation Af depends on a diffeT;ic;:frequency Aidif and on a va- 1 + mosd Af=Afdif I+k2+2kCOSd ' (4) 96 US USm +xCOS Ad + COS(Sl+AO)t + E +~KhCOS(J2-A@t + .a, ] (I) , where Kd is a detection coefficient; U is a signal amplitude; URFI mSand 9, are, respectively, amplitude, modulation factor and modulation frequency of the interference carrier* ’ K ,#I$ AOis a difference between a wanted . _ eignal frequency and an interference fre uency. After oscillations ?I) have passed through the amplifier and low-pass filter, the voltage measured at the output of the radio receiver becomes equal to (2) where r is a gain factor of the amplifier. It is seen from (2) that at the output of a radio receiver, besides the oscillations of frequency SL , in the case of AO,CB the oscillations of frequencies AO, SL+AO and P-00 can be present whose amplitudes are appoximately 36 times as high (Kc30) as the signal of frequency B , i.e. the signal whose value is used for controlling the immunity of the radio receiver. The amplitude of oscillations of frequency A0 is the highest and is not dependent on the modulation factor m. It follows that the error of measurement of immunity of RF and IF stages of a receiver may prove so high that it becomes of no use to carry out the above-mentioned operation. In this case, the variation of the modulation factor of the interference carrier is of no importance both for a receiver response being measured and for a decrease of a measurement error. Moreover, under the influence of frequency instability of generators which are used to simulate a wanted signal and interference, the response of a radio receiver to a test interfering-signal is instable in time. This fact makes the control operation difficult and leads to an additional measurement error. The situation can be improved if a generator which causes an interference at the input of the detector, is tuned in such a way that AO=O, and if this condition is not broken under the influence of frequency instability of RF generators used in measurements. In this case, the receptor response being measured will correspond to the expession 2.2. Experimental investigations. Two types of broadcast receivers (three receivers of each tvne4 were taken for measurements. Typ& 1 was an all-wave receiver, type 2 was an all-wave receiver with a builtin tape recorder. Measurements were made by means of an open TEM cell (Fig.1 in CIS'PRdocument ['1]) as well as according to a circuit of Fig.16 of the same document (by injecting the interference currents into the antenna input of the radio receiver and injecting the interference voltage into the mains input). Each radio receiver was tested at its tuning frequencies of 0.25 and 1 MHz. Because (as it was indicated in Clause 1.1). there were reasons to suppose that an interference can affect both an RF stage and an IF stage of the radio receiver, as a test interfering-signal such A&Ioscillations were used, whose carrier frequency corresponded to a secondchannel frequency or to an intermediate frequency of the radio receiver. All the requirements of document [I] were fulfilled in the measurements. The experimental investigations showed that their results are considerably affected by frequency instability of generators which simulate wanted signal and interfesence, A variation of a modulation factor of an interference practically does not influence a measured value of a receiver response. Thus, the conclusions drawn in Clause 2.1 were fully confirmed, Moreover, it was found out that it is also difficult to make measurements of a receiver response which in accordance with IN should be by 30 dl3lower than a standard output-signal, because this response is at the same level as the internal receiver-noise. In order to overcome the abovementioned difficulties it was decided to change a measurement technique - RP generators is sufficiently high and that a measuring procedure and a criterion of evaluation are changed. It should be noted that the CISPR methods do not cover a control of immunity of radio receivers to interferences which affect the receivers at such frequencies as the frequencies of the second channel of reception, intermediate frequencies (for FM receivers), etc though the corresponding tests may be of great importance. Thljs, the methods of control of immunity of broadcast receivers, which were set forth in /'I/, need further improvement. 99 - 19D7 References 1 2 3 ;;;II/E(Secretariat)27,August . M.Borsero, E.Nano Some considerations about interference vol.tag@ measurements and relevant Limits, EMC, 1983, Zurich CbSCr~P/E/Pr%(6oySoH~r.reB,l~aCMAbeB, $~pMUHOl3/ccb,;P) J. ,HO.W$ b !-9:x) 0 CkUl3M N!ieXJ.Jy p%3JIWJH1_rMI4 IiO3@$l/lqMe~~aMMI'lOMexO3a~~eHHOCTM TeneBM3MomHx l7pMeMHBKOkl. L3:d .H.liono~ ~~:ELTeMaTMYeCKMifj aHtUIM3 GuemM. SocaIie)rovra~aT,ElocEtsa~lewmrpa~~, I95 E,T. - 98 - Table Type of a receiver and the frequency of its tuning, ft Method of introduction of interference energy An open TEN Ctjll. Type 1, ft = 0.25 !Pype1, Tt I 1 MHz 2 MHz IO 35 70-3 8 IO 28 1O-3 7 5 150 10'3 2.1 0.4 7 10-3 4 Type 2, ft = 0.25 MHz Type 2, ft SI1 MHz 'V/n1 Into the anten- Into the input, supply mains, V V na where d 23dA fdif t. It can'found from (4) that the frequency-deviationamplitude will be equal to (5) Let us consider the simplest cxample. Suppose we have that the upper frequency of an LF amplifier of a receiver is equal to 15 kHz, and when testing a radio receiver its output is not terminated by a low-pass filter. If during the setting a standard output power of a radio receiver an FM signal with a deviation of 22.5 kHz was used and II15kHz, 'fd&d the then an interference level radio receiver output wi_llbe by 30 dB lower than a reference level in that case when gf = 750 kHz, to which KI 40 corresponds. It the modulation of interference oscillations had been taken into account, the calculated value of K would have been even higher. This, an USW FM receiver proves to be sensitive to an AM interference which affects its RF or IF stage. At the same time it should be noted that the abovementioned effect will also take place when a lowpass filter with a cutoff frequency f >I000 Hz is connected to the radi.8receivex. But this will happen io:yK;; the case of Afdif. If the antenna diameter is several times smaller than the distance to the nearest NM field source, the field acting in the vicinity of the antenna consista of the superposition of fields generated by elementary eleotria dipoles which have been placed uniformly along the radiant sources [5,7] l H’(p) =$H:(Phi’krq where: - % ,thecomplex amplitude of magnetic component :2X4 field at the centre of ,the antenna produced by q elementary electric dipole, 3 the distance from Taint P = 0 to point which is orthogonal projection of point p on q direction. 102 - Eiq(0)e complex amplitude of the even electric field component at the centre of the antenna; this component depends on odd magnetic field component [5,7], dependent from W,cPI = fwzction the magnetic component of the EM field and the anten*a parameters [L5,7] , X,(@ = function dependent from the electric component of the EM field and the antenna parameters [1,5,7] , "1 = impedance loaded circular loop antenna. If the size of the antenna is smaller than the wavelength, ]kbl</l For small p/l the asymptotic result. 27~7 quation (3) yields On each PC board, decoupling capacitors were used to avoid mdesirable EMI on lxy+xxtraces : the prpose was to control as much as possible the piths of the logic signals, and to reduce the pllution of the other bxrds through the power leads. The choice of the type of capacitor must be made according to the wide frequency range of the harmxk content. chly ceramic oapacitors have a sufficient frequency range to be efficient above one megahertz. The problem no is to choose the right value of this capacitor ; we have used different values of ceramic capacitors O,l, 1, 10, 100 nF (fig. 4, 5, 6 and 7). The radiofreguency levels decay, k&-ken the value of the capacitor increase and we find a slop of 20 dB by decade for the first harmonics. Dut above 10 MHz, the increasing of the capacitor frcm 10 to 100 nF has no more effect on the level of con.ducted.interference. This phenanenon cones frcxn the resonance of the cqpacitor with its leads and the inductance of traces. The total self-inductance Cl], L of a straight conductor is given by . (4) Fig. 4 : Conducted emission on Vcc with c = 0,l nF (74 LS 00) 1 : length of conductor (in meter) r : radius of conductor (in mater) where NC%?, let us consider a capacitor which has two leads (fig. 3). r ---- I I N 1 1 Fig. 5 : Conducted emission on Vcc with c = 1 X-G (74 LS 00) !p! -I------ Fig. 3 : Schematic diagram of a ceramic capacitor ~-II_ - Fig. 6 : Conducted emission on Vcc with c = 10 nF (74 Ls 00) The self inductance Lc of the leads is given by : IQ = 2(L - II) here (2) L : self inductanceof a cxoductor (length 1) M : mutual inductance of the two conductors c21 where 1 : length (in mn) D : distance between the two leads (in mn) r : r&ius (in nm) Fig. 7 : Conducted emission on Vcc with c = lOO.nF - (74 LS 00) 148 Tee chip has its cm3 self irductance I-p, manufacturer. tihich is given by the capacitor we can canpute the total Usirq the formula (3), self inductance T_T of the different capacitors we have us& and their resonance frequency fr. The results are presental in table 1 ati 2 dlere : (5) % = L, + I+, - @z see tile tiu~~rtance of the leri~thof lea&i this is tiiy for in the resonance frep~lcy atr_l scmre difficult cases, it is recurmeticd to use and ceramic chip capacitors, for tiiicll L, = U Lr, is smiler. The biyger a cerakc capacitor is, tilesb.Xter the leak must ke to have &feetive filtering lXoprt_ies. The position of the capacitor on tilePC Wrd is also imlmrtant : it must k as close as possible frcm the integrated circuit to miriirkize the self irtiuctance of the Z'C peer traces, iutiiz~avoid k&111xAlution of tile other traces. ? Ground planes 1V.J. 1,65 390 Iiyround lkne is rlathirqmore tim a sheet Oi: that is placed close to interconnectihy traces ami is tied to the per supply return. It is certainly the kst kay to reduce radiateo interference with fast logic circuits. tiroun~ planes cdyik_e&siyned in a nurker of different ways. In this article, we will see two rrainmys in l.xALdinga PC kmard witAld ground plane. mztxd 1,45 132 1,65 39 2,15 2,6 10,O 3 Lath my (fig. cj. ‘i’able 1 : Calculationof the total self inductanceversus the capacitancevalue (1 = I,6 ma) c (a?) LP (nM LT (nti) a) fr (me) O,l 0,7 9,4 164 1 0,5 9,2 52 10 0,7 9,4 16 100 181 11,c 4,6 *1 000 1,9 11,4 1,413 -- cxmes frun transnlissionline Uieory h Table 2 : Calculation of the total self inductanceversus the capacitancevalue (I = 10 mn) * not used h in our tests The dimensions of the capacitors we used are given in table 3 : a b 1 10 100 1 000 Fig. 8 : Ground plane Structures : symetrical ccplanar waveq.zidestructure microstrip line structure The gecmetric dimensions of these guidirly structures give us characteristic impedances of the associated transmissionline [2], L3]. kcrostrip lines radiate less energy than simple structures without ground planes (see fig. 9 ati 10 with a [a] structure) kecause tileelectrmaynetic field is cc&in&i in the hiyhest prmettivity dielectric. Table 3 : Dimensions of the ceramic capacitors To go further in reducing radiated emissions, we have to caubine a txansmission line structure with ceramic capacitorson per traces. - 149 27~7 - I --- h Fig. 11 : Cross-sectionaldimensions : nomenclature The maxim..xn coupling factor is given by : with : zo2 = Zoo zoe (8) zo Fig. 10 : E field radiation with a groundplane characteristicimpedance Zoo odd mzde characteristicimpedance Zoe even rrodecharacteristicimpedance and frun these equations, the impedances rquir& are : Discontinuities, SUCh ZlS abruptly qencircuit micro&rip (i.e. open ends), stqs and bends will all radiate to a certain extent. James and Henderson [4] slow that at frguencies where the surface wave is highly trapped in the substrate and with h//i, , wEft/Ao LL 1, the radiation conductance G, is approxinetelby : where : WeCC zo is the effective microstripwidth [3] is the characteristicimpedance of the microstrip line h thickness of the board 6,ecf effective permittivityof the dielectric wavelength ( A, - C/$ ) A0 G, increases with frequency and that is why any time a transmission line is longer than the signal wave length, it is necessary to ntltchthe line. A middle way must be found between 1~ impedance lines with 1~ propagation delay but high p.cwerdissipation in parallel terminations, and high impedance lines with high prcpagation delay but 1~ power dissipation.A ground plane is also a low inductance return path for hantonits of clock rate and the effect of decoupling capacitors -willbe reinforced. Transmission line theory gives us a solution for problems of crosstalk in digital systems. Suppose we have a coupled microstrip structure (fig. 11). The designer knm the maximum capling factor he can tolerate, and he will canpute Zoo and Zoe; with these two impedances,curves (or soft ware) will give him the minimum s/w he can choose for its design [53 arA [3]. R practical rule is s/w ) 4 to avoid all problems of undesirable crosstalk between parallel microstrip traces. V - CONCLUSION The designer of logic circuits must be aware that if the fondamentalfrequencyof its clock is lc~, the harmonic content is particularly wide and is the source of broadband radiofreguency interferences.The control of RF1 begins at the design of the PC boards : drawing of traces to avoid problems of crcestalk, ground plane or stripline structure, ceramic capacitors on per leads. It is only the sum of this different elements that will give a lc~ EM1 level PC board. All these cares in PC boards design are cunplementary ard lead to a ICW level of md1 (fig. 12 and 13) : sane PC board witl-outany care, and with a ground - 150 plane structure a, and 10 nF ceramic capacitor). 'rJhen the frequency increases,ground planes structures bxune absolutely necessary in order to avoid an eledranagnetic pollution. With EXX, logic family,microstrip structureswith terminations at each end are used to reach high speed clock rate witbut extreme cnrershootard ringing: and of course, pcwer leads must bz filtered. - For the ELcctrmalFietic cuxnyutibi1it.y pint of view, high speed d.i:jital circuit are certaitlly the w.xse interferincjcircuits. Even high order I1arxDni.c (speciallyin the rh According to equation (5) the refere%e of current is given by: tm time = t - r/co+ z/co. (7) For z = h the actual reference time of the current &urce is The current at the point of strike is known. =,@Z+~. t C m/h 0 The electromagnetic field caused by a tran-, sient current in a vertical ideal conductor over a perfectly conducting plane can be divided into an electrostatic field term (EQ)an induction field term (EirHi) and a distant field term (Edi, It is produced by the sum of infiniEdi)[Sl* tesimal vertical dipoles of the length dz with a' current i(z,t), as shown in Fig. 4: dE(r,8,t)=dEQ(r,8,t)+dEi(r,8,t)+dE di(r,8,t), (3a) dB(r,B,t)= dMi(r,8,t)+ dHdi(r,8,t) (4a) with % dz 2-3sin20 dEQ(r,B,t)=- -2Trreo r3 & , (3b) (3c) 0 dz dEdi(r,@,t)- -2aso ai(z,tx) at I (3d) I (4b) 0 and dz sine dBi(r,6,tJ=--i;; 2 i(z,txl r ai(z,tx) dz sin0 dBdi(r,e,t)=- 2a car at (4c) The electromagnetic field retardation is given by tx = t - r/co. 3.1 Electromagnetic field components for Z-Z-h, The current and the current derivative (in point X) is di (z,tx) dt (ga) dio(tm) (gb) The current which flowed during the time internal t-m- tm/z at the altitude z yields the dissipated charge t t X m i;i(t,-r)d~ = (SC) By integration the field components from equations (3,4) yield EQ(t)= - hi S 1 3sin2e-2 2TE0 r3 r m (It,,zio(T)dr)dz, 0 hX 1 s Ei(t)= 271EO EX Edi( X’ With respect to the time delay of the current in the lightning channel, z/c,,a visitor in point X sees the current flowing at the altitude z: , =-Z----. (5) For a visitor in point X the current source seems to travel with a velocity vx= dhx/dt at an altitude h (8) 0 Since the local current distribution has a discontinuity for z = hx the electromagnetic field components are calculated separately for 2 < h, and z = h,. (For z > h, are no field components!). i(z,tx) = io(tm) i!z,T)dt, dz 2-3sin28 dEi(r,O,t)=- 2aso c r2 i(z,tx) sin28 c2r C and 1 %. 3sin20-2 io(tm)dz, 2 cOr 2 1 sin 6 -co2r aio(tm) at dz (10) - 160 (11) o 3.2 o Electromagnetic field components for 2=h - 11 For the distant field terms according to equations (3.4) is: is the altitude hx in front h resp. h ag&+behind t&a-local current. At the altitude hx the current derivation ai/at(z=hx, tx )causes a local current derivation ai/az with the velocity az/at along the infinitesimal way dz: ai(z,tx) dz = g $$lz = vx $3.~ = x Total electric and magnetic field The total electromagnetic field corresponds to the part caused by the current distribution along the lightning channel (lO,ll), and the part taking into consideration the traveling current source (14): E(t) = Edict)+ Ei(t)+ SQ(t). Comparison between the Traveling Current Source and the Transmission Line model 4.1 Analytical expression of current io at the point of strike /i,dt, i,, di,/dt and d2io/dt2 must be known for the.calcuiation of E and H fields respectively E and H signals. Therefore i, must be able to be differentiated twice for t 2 0. An analytical expression of i, therefore is desirable,whereby d2io/dt2 = 0 and (i,),,,, (dio/dt)max and the decay time can be chosen independently. The most in high voltage technics used double exponential current function is not suitable for the present just as the current functions given in [lO,ll]. Therefore the following formula was chosen: i k" io(t)= 5 -e, kn+l -t/r (17) To enable a fundamental comparison between the TCS and the Transmission Line model the LEMP calculations are done with a constant return stroke velocity v and T = 100 1~s.The point of strike is assumed to be in a distance of s = 1 km. For both models the radiated LEMP is calculated for a first and a subsequent stroke. The calculation show, that in th: case ?f the TCS model the peak values of the E and H signals are nearly 5 times higher than in the case of the Transmission Line model. +d{itidnally in the case of the TCS model E and H signals show a bipolar shape leading to a initial peak in the shape of E and H field. 4.2.1 LEMP during the front of the first stroke current For the current of a first stroke the current parameters of equation (17) were chosen to n = 10, z = 30 kA, tl = 3 ~.ls(corresponding to (di/dtjmax = 26 kA/vs). The return stroke velocity is assumed to be v = co/3 = 100 m/us. Fig. 5 shows the results of the LEMP calculation. h with (16) with k = $-. n is a correction factor for i. 1 4.2 LEMP calculations vxdi .vx*io(tm,h). 3.3 1 ' Spew ai (t - li;;co (,2+h2) x o m/h)' x h Xsine io(tm)dz. Hi(t)=-&;T r 4. For the electrostatic and the induction field terms according to equations (3,4) is E = E.= H. = 0. Q - xsin2e aio(tm) dz + Edict)= -!- / 2a~~ o c2r at 0 (15) 3sin20-2 io(tm)dz, Ei(t)= 2 cr t h, O2 3sin 8-2 tJm . 1 r3 tm,zloWdr)da; EQ(t)= 21~~~ ; and H(t) = Hdi(t) + Hi(t) with 1. n Hdi(t)=-& lXSineO 0 OO+ a) ai (tm) cOr at dz - 8 '----c(rs) k l+(LIs)di,/dt d2i 0 /dt2 - 161 0 29 - F2 __c_.__---_( 4 t-*(us) b) * H (1) ci(t) cl E (t) (t) E full line TCS model: Transmission Line model: Fig. 6: dashed line LEMP of a subsequent stroke at a distance of 1 km 4.2.3 LEMP during the decay time of the first stroke current In fig. I the current and field shapes are presented for a longer duration than in fig. 5 and 6. In opposite to the Transmission Line model in the TCS model the electrostatic field change caused by the lightning discharge process is taken into consideration. 0 t -“---* E c) (ps)a i (1) TCS model: full line Transmission Line model: Fig. 5: rt! dashed line (kAJ3'TI\ LEMP of a first stroke at a distance of s = 1 km 15 t. 4.2.2 LEMP during the front of the subsequent stroke current I\ 01 A subsequent stroke current-is simulated by equation (171, whereby n = 8, io= 12 kA, tl = 0,5 us (corresponding to (di/dt),ax= 61 kA/ps). The calculations were done for v = 2/3 co= 200 m/ps. Fig. 6 shol ; the LEMP results.- a) 0 \ 250 t -(PI) 560 i 50 200 0 b) H TCS model: full line Transmission Line model: 0 0 E(t) 6 t-&m) Fig. 7: (t) dashed line LEMP of a first stroke at a distance of s = 1 km 4.3 Fundamental physical differences between the TCS and the Transmission Line model The fundamental physical differences between the TCS and the Transmission Line model result from the different kind of transient current distribution along the lightning channel. In table 1 the TCS model is opposed to the Transmission Line model. b) f (t) 4.4 Possibilities of modification of the TCS model The TCS model can be varied very easily and applied to special problems. In its fundamental idea the TCS model bases on the hypothesis of a traveling current source which feeds its current into any complex network. Simple equi- - 162 - TCS model Transmission Line model Traveling current source Traveling current wave From earth to lightning channel Lightning current supply From lightning channel to earth Current source traveling with velocity v from earth to the cloud Speed of propagation along the lightning channel co in opposite direction of v Velocity of increasing lightninq channel Any velocity is possible Indentical to the speed of propagation along the lightning channel Relation between i,(t) and i(z,t) i(z,t)= i,(t+z/c,) i(z,t)= i,(t-z/v); v = constant Altitude of lightning channel Resulting from velocity v and current duration Indefinite Physical model of current distribution - Equivalent Circuit Current flow -- -.- Stationary current source at the point of strike V Electrostatic field change No consideration Table 1: Fundamental differences between TCS and Transmission Line model valent circuits can be dirived. ;5 .Summary In the TCS model the interference source propagation is completely seperated from the interference propagation. Therefore a traveling current wave can move with light velocity along the lightning channel while the current source can move with any suitable velocity. For present LEMP calculations some simplifications were made concerning the return stroke velocity v and the current lo; nevertheless the calculated LBMPs show a very good agreement with the measurements. 1~ comp?rsion with the Transmission Line model E and H signals are of higher frequency and have much higher peak values. E and H fields show a characteristic initial peak. The changed electrostatic field component caused by the lightning discharge process is taken into consideration. The TCS model in principle allows a lot of variations and extensions (e.g. the consideration of the transient earth resistance). Elektra 69(198O),S.65-102. [41 Trapp, N.: Erfahrungsbericht iiber die erste MeRperiode in der BlitzmeBstation auf dem Peiflenberg. ICLP(1983), The Hague, S.23-30. I51 Feuerer, R.: Zeitliche Anderung der Magnetischen Induktion bei negativen Erdblitzen. Ph.D. Thesis HSBw Mtinchen, 1983. 161 Djebari,B.,*Hamelin,J.;Leteinturier,C.;Fontaine,J.: Comparison between experimental measurements of the electromagnetic field emitted by lightning and different theoretical models-influence of the upward velocity of the return stroke.EMC(198l)Zurich, 5.511-516. II71 Weidman, C.D.; Krider, E-P.: The submicrosecond structure of the electromagnetic fields radiated by lightning. ICLP(1983) The Hague, S.65-74. [81 Uman, M.A.;McLain, D.K.;Krider, E.P.: The electromagnetic radiation from a finite antenna. AJP(1975),vo1.43,S.33-38. 191 Uman,M.A.;McLain, D.K.: Magnetic field of lightning return stroke. J.Geophys.Res. (1969),vo1.74,No.28, 6899-6910. References [II Berger, K.: Methoden und Resultate der Blitzforschung auf dem Monte San Salvatore bei Lugano in den Jahren 1963-1971. Bull.SEV, 63(1972)24,S.l403-1422. [21 Garbagnati, E.; LoPiparo, G.B.: Parameter von Blitzstrbmen. etz 102(1982)2, 5.61-65 [31 Anderson, R.B.; Eriksson, A-J.: Lightning parameters for engineering application. [lO]Rajici,D.: Beeinflussung einer Darstellungsweise der atmosphdrischen Entladung auf den maximalen Spannungswerten in den einzelnen Punkten der einfacheren Blitzschutzinstallationen. ICLP(l973) Portoroz,R-2.12. [ll]Jones,R.D.: On the use of tailored returnstroke current representations to simplify the analysis of lightning effects on Systems. IEEE Transactions on EMC,May 1977,S.97-98. - 163 RADIATION CHARACTERISTICS, EMISSION MEWANISMSAND PHENOMENOLOGY C. Weidman, Centre 30 - National J. Hamelin F3 OF LIGHTNING and M. Le Boulch d'Etudes des Telecommunications R.P. 40 22301 Lannion Cedex FRANCE Measurements of the radiation emitted by naturally occurring lightning discharges at four frequencies in the VHF and UHF bands are presented and compared with simultaneous recordings of magnetic (B) fields and photoelecThe RF signals, which are impultric data. sive and often similar at the different frequencies studied, are complex; shape parameters such as amplitude, pulse width, and interval time may vary over several orders of magnitude Large amplitude B field during a discharge. variations below 1 MHz often accompany imporand, thus, stepped leaders, tant RF emissions, return strokes and certain intracloud discharge processes are shown to radiate strongly at VHF and UHF. 1. INTRODUCTION Much of what we have learned about lightning is based on measurements of electric and magnetic fields at frequencies below a few or a few tens of megahertz /l, 2/. There is renewed interest, at present, in lightning The phyemissions at VHF and UHF frequencies. sical processes responsible are not well understood, and because these emissions may occur during times when there is little or no low frequency signal information, there is much that we might learn about lightning discharge Lightning RF emissions are currently physics. being used in atmospheric electricity research, for example, to locate discharge sources and follow channel developement in time within the cloud and to relate this to thunderstorm dynamics and structure /3-11/. Lightning RF emissions are impulsive and often have large amplitudes, and, thus, present a potential hazard to any system which is sensitive to transient fields. Past measurements, often made with narrow bandwidth receivers which do not resolve the fastest variations, are often not suitable for a proper evaluation of this risk. With these two objectives in mind, that is to better understand the lightning discharge processes important at VHF and UHF frequencies and to better characterize the lightning caused electromagnetic environment between a few tens of megahertz and a few gigahertz, the French National Telecommunications Research Center (CNET) has developed instrumentation and has been making measurements at six VHF and UHF frequencies for the past several years. 350 kHz bandwidth receivers were used and signals were recorded in precise time correlation with other lower frequency measurements so that the emissions associated with different lightning processes could be determined with microsecond In this report we would like time resolution. to present and discuss examples of data obtained during the 1983 experiment at the St. Privat d'Allier station in south central France. 2. EXPERIMENT Lightning RF emissions at 60, 100, 175, 300, 500 and 900 MHz were measured using six tuned, vertically polarized, dipole antennas mounted on a 9m metal mast. Antenna spacing was such that effects due to mutual coupling between adjacent antennas and due to reflections from the ground were minimized. Antennas were connected to receivers, located 10m away in a metal trailer, using semi-rigid cable. Receiver characteristics at each of the six frequencies are summarized in Table 1. Logarithmic amplifiers were used to amplify and detect the intermediate frequency signal giving a receiver dynamic range of about 80 dB. In addition to the RF signals, horizontal magnetic fields were measured using two orthogonally mounted, wideband (150 Hz - 20 MHz) antennas /12/, and measurements of return stroke luminous emissions at elevation angles of l", 5", 10" and 20" were made using four photoelectric detectors /13/. Each detector had an angular resolution of about l", a risetime of less than one microsecond and a nearly 360" horizontal field of view. Continuous recordings of the RF signals, the four optical signals and one component of the magnetic field were made using a 400 kHz bandwidth instrumentation tape recorder (FM mode). Time synchronization pulses were superimposed on all channels at the recorder input to permit precise correlation in time, Magnetic fields were also recorded, redundantly, on modified video tape recorders /14/ with 3 MHz bandwidth. The measuring station was located 3 km away from and with a direct line of sight view of the principal lightning triggering station, A number of decisions such as antenna placement and orientation, the choice to record only one component of the magnetic field, and amplifier - 164 gains were made with the assumption that most data recording would be for triggered dischargUnfortunately, due to meteorological conthe trigz:tions and some technical problems, gered lightning data from 1983 is very limited and not representative of data from past years. We are thus only able to present data from naturally occurring activity. TABLE 1 RF Receiver Characteristics Sensitivity Bandwidth Frequency C/N;) W'o) ("$) 90 30 50 90 600 337 348 360 350 385 100 175 300 500 900 Time Scale Thunder was audible, howges are not known. that the Storms ever, and thus we estimate were between 5 and 20 kms from the recording station. Perhaps the only feature common to each of’ these records is the abrupt onset of the emisThis is true of essentially all events sions. and as we shall see, large we have studied, occur siamplitude magnetic fields often The initial few hundred millimultaneously. seconds of activity in each discharge then on this time scale, to consist of imappears, pulses closely spaced in time. The rate of The offset occurrence can be quite variable. from zero between 85 and 1IDms in Fig. l(a) is produced by impulses which occurred at interval times less than the 3 rs time constant and were thus effectively integrated by the receiver. The initial activity abruptly, as between 3. DATA 3.1 Millisecond - Structures Examples of 175 MHz radiation produced during four lightning discharges are shown in Figure 1. These records reproduce accurately Signal variations as fast as 100 kHz, and were obtained by replaying the analog data tapes at reduced speed and rerecording the data on a second recorder. These rerecordings were then displayed, again after reducing the playback speed, on a strip chart recorder. The signals produced by these discharges at the other frequencies had a very similar overall appearance and are not shown. Distances to these dischar- may also stop and restart 40 and 55 ms in Fig. l(b). Late in the discharges, emissions occur in Some of more isolated, often intense bursts. these are probably the "Q noise", the "solitary pulses" and the "fast bursts" described by /I7/, /6/ and /8/, respectively. The sequence of closely spaced impulses followed by the isolated bursts may repeat itself as in Fig. I(b). We nOtf? finally, that it is not possible, on the basis of these signatures at 175 MHz, to identify the different types of discharge processes. The large amplitude signals at points 4 and 8 in Figs. l(c) and l(d) were produced by Fig. 1. 175 MHz radiation produced by 4 different lightning discharges. Increasing signal amplitude is in a downward direction. The amplitude calibration at left applies to all 4 signals. Event (b) continued for an additional 120 ms. k 0 100 200 300 400 508 ms 165 a s3 CmV/m 800 ps 1000 600 400 200 0 b CmV/m 1 .. -1 0 Fig. 2 1 I 200 I I 400 I I 600 I I 800 ps 1000 175 MHz radiation and simultaneous magnetic (B) fields. The signals in (a) and (b) correspond with points 1 and 3, respectively, in Fig. 1. first return strokes and thus these are apparently cloud-to-ground discharges. RF signals at 60, 100, 175 and 300 MHz emitted at the "beginning" of discharge (d) in Fig.1 (point 7) are shown in Figure 3. Note that the 175 MHz record is inverted with respect to the others. Emissions at 500 MHz were not reproduced due to a recorder malfunction, and no simultaneous activity was observed at 900 MHz. Interval times between the RF bursts at 60 and 175 MHz in Fig. 3 are somewhat larger than in Fig. 2, but, otherwise, the overall appearance is quite similar. A fast time resolved view of the magnetic field and 175 MHz activity at point 2 in Fig. 1 is shown in Figure 4. Here, the magnetic field impulses have shorter widths and a higher rate of occurrence than in Figs. 2 and 3. The magnetic field and 175 MHz peaks in Fig. 4 generally coincide in time. 3 0 ,3 t I CLI 0.2 3.2 Microsecond Time Scale Structures 1 In Figure 2, we show on a faster, 100 rs/div, time scale the magnetic field waveforms (B) and the 175 MHz emissions produced at points 1 and 3 in Fig. 1. These records were obtained by digitizing the magnetic tape data using a multi-channel transient waveform recorder. Time synchronization between the two records in Figs. Z(a) and Z(b), and in the figure examples which follow, is better than 10 ps. The bipolar magnetic field waveforms in Fig. 2 are typical of the larger radiation field pulses produced by intracloud lightning and are discussed in detail by /15/ and /16/. The largest RF signals have widths of a few tens of microseconds and occur in time coincidence with the magnetic field pulses. In agreement with z F----mV/m ,, 0.5 0 1 100MHz 300 MHz Yt.. 1 0.2 0 0 0.2 I ’ F3 /17/ there is a tendency for the RF signals to peak during the initial half cycle of the B field waveforms. 0 1 30 - L-L-1 0 Fig. 3 I 200 1 400 I I 600 I I I 1 800 ps 1000 The initial radiation at 100, 300, 60 and I75 MHz (top to bottom) observed during discharge (d) in Fig. 1 (point 7) 0 Fig. 4 200 400 600 800 ps 1000 175 MHz radiation and simultaneous magnetic (B) fields corresponding with point 2 in Fig. 1. First return stroke signals are illustrated in Figure 5. A magnetic field waveform showing the event at point 4 in Fig. 1 and the simultaneous radiation at I75 MHz are shown in Fig. 5 (a). Emissions at four different frequencies produced by the other first return stroke (point 8 in Fig. 1) are presented in Fig. 5(b). The strong 10" optical signal, the third, uppermost trace in Fig. 5(a) was an important factor in identifying this as a return stroke discharge. In Fig. 5(b), the B field component recorded simultaneously on magnetic tape was very weak (indicating a field perpendicular to the preferred direction of the antenna) and identification of this event is based on the shape of the second, orthogonal, B field recorded with a video tape recorder. This latter signal is shown immediately below the RF waveforms. Because the time correlation between the different recorders is only f 500 ps, the horizontal positioning is somewhat arbitrary. Between 0 and 400 s, the 175 MHz signal in Fig. 5(a) consists o I"narrow impulses which occur at intervals of tens of microseconds. These emissions may coincide with individual steps of the stepped leader. The 400 kHz recorder bandwidth may not have been adequate to reproduce leader step B field pulses, which typically have widths of only 1 or 2 rs /18/. A large burst of 175 MHz radiation occurs, between 400 and 500 ps, in coincidence with the first return stroke magnetic field change. This is in contradiction with results given by /19, 20 and 21/, who have reported delays ranging from - 166 - 3 OPTICAL SIGNAL 0 0-3 a 0.2 1 1 I I 200 0 3 I 200 0 A L ,, I I 400 I I 600 I I I I 800 ps 1000 I I 400 7 I I I 600 _1 I 800 pslOOO R 0 b -3 I , I 0 0.2 ! 0 Fig. b 6 t! -_400 600 200 800 ps 1000 175 MHz radiation and simultaneous magnetic (0) fields. Signals (a) and (b) correspond with points 5 and 6, respectively, in Fig. 1. mV/m 100MHz tI.1.1 I -20 0 Fig. 5 130 260 390 520 ps 650 RF radiation and magnetic (B) fields produced by first return strokes. The three signals in (a) show (top to bottom) the light emissions at a 10" elevation angle, the B field and the 175 MHz emissions produced by the event at point 4 in Fig. 1. The return stroke RF radiation and B field in (b) correspond with point 8. of microseconds between the tens to hundreds first return stroke radiation field peak and Because of the the peak in the RF emissions. uncertain time correlation, we cannot determine whether a delay exists between peak RF and B We note that the peaks fields in Fig. 5(b). at each of the four frequencies, though somewhat difficult to define precisely, do not appear to occur simultaneously. As final examples of time resolved data, we show in Figure 6, the magnetic fields and 175 MHz signals which occurred at points 5 and 6 in A similar event occurred at point 9 Fig. 1. and produced the emissions given in Figure 7. These examples are representative of many of the isolated bursts which occur late in a discharge. The magnetic field waveforms in Fig. 6 have amplitudes comparable to that of the first return stroke (point 4) and the impulsions at point 3, both of which occurred earlier in the same discharge, but forms which are distinctly We are not able to identify the different. discharge processes in Fig. 6 on the basis of the magnetic field signatures. 60 MHz ’ + 0 0.2 1 1 0 Fig.7 I I 200 I I 400 I I 600 The RF emissions at 100, 175 MHz which correspond in Fig. 1. I I I I 800 IJs1000 300, with 60 and point 9 4. DISCUSSION The data in this paper complement measurements of the RF emissions produced by triggered lightning discharges, made in New Mexico during the summer of 1982 /22/. Space limitations do not permit more than reference to a detailed study of those results and of some 1983 data given in /23/ and /24/. Here, we have only considered natural lightning and have also limited our examination to discharge processes which emit not only at VHF and UHF, but also produce, simultaneously, lower frequency magnetic fields. This latter choice was made, of course, because we hoped to be able to identify the type of discharge from its radiation field signature, We have seen that many kinds of signals are present in a discharge and often we have not studied enough data to notice any general tendencies. An - 167 exception are the very first emissions in cloud-to-ground discharges, which we would now like to discuss in more detail. 4.1 The Activitureceding Discharges Cloud-to-Ground A detailed study of the electric fields precloud-to-ground lightning discharges has They divide these recently been given bY /*5{i) a ,,pre,iminary fields Into two sections: variation" portion which usually begins 60 to 70 ms before the first return stroke and generally lasts for 0 to 20 ms, and (2) a stepped leader portion which follows and has a mean duration of 27 + I5 ms. Of particular interest is their observation that a sequence of regularly spaced, bipolar impulses often mark the transition between the preliminary variation They call these and the stepped leader phase. large amplitude fields "characteristic pulses." /15/ have observed similar pulse sequences preceding first return strokes and give the following mean signal shape parameters: ceding mean interval time between impulses mean total full pulse sequence width duration mean interval time the pulse sequence and the first return stroke 130 + 50 rs 41 + 13 /Js I- 2ms 53 i- 40 ms The initial polarity of the bipolar field is generally the same as the first return stroke which follows. The bipolar fields discussed by /15/ are probably the characteristic pulses defined by /25/, although we note that the mean stepped leader durations, 53 * 40 ms and 27 + I5 ms, respectively, are appreciably different. We have identified the B fields in Fig. 2(b), and similar signals in five other cloud-toground discharges recorded in 1983, as characteristic pulse (CP) sequences, using the mean parameters given by /15/, Somewhat surprisingly, in Fig, 2(b) and in three of the five remaining cases, the CP sequence occurred in coincidence with the initial abrupt commencement of the RF radiation in the discharge. That is, there were no RF or B field signals preceding the characteristic pulse sequence, and, thus, no evidence of a preliminary variation phase in these four discharges. In the remaining two cases, the characteristic pulse sequence occurred tens to hundreds of ms after the beginning of the VHF emissions. In Fig. l(d), for example, the characteristic pulses occurred at point "CP", 55 ms after the start of the 175 MHz radiation. Measured intervals between the CP sequence and the first return stroke which followed ranged from 9 to 95 ms. The mean value, 52 ? 34 ms, agrees well with the data of /15/, but both are a factor of two larger than the mean stepped leader duration given in /25/. In five out of six cases, the initial bipolar field polarity was the same as that of the first return stroke which followed. The examples in which the CP sequence coincides with the beginning of the RF radiation The large bipolar field ampliare interesting. tudes imply large channel currents which occur without any evidence of "preparatory" activity. 30 - F3 This does not seem reasonable, and we are forced to conclude that, in these cases, the vertical component of the RF E field produced by the preliminary activity is just too weak to be detected at our station. 4.2 The Submicrosecond Structure of RF Radiation In Figs. 8(a) and 8(b) we show, on 10 ps/div time scales, portions of the 175 MHz signals in Fig. 5(a) (between 250 and 350 I_'s)and in Fig. 4 (between 300 and 400 IS), respectively. The 350 kHz bandwidth receivers used in this study have impulsive response times of about 3 ps. The single pulses at the left of the trace in Fig. 8(a) are clearly at this limit. The signals in Fig. 8(b) and in many of the other signals in Figs. 2 through 7 appear to be superpositions of these 3 ps wide waveforms. Thus, even on this time scale, approximately 2500 times faster than in Fig. 1, the RF emisWith 3 ps sions still have an impulsive form. receiver response times, we cannot resolve these fastest structures, which, experimental data indicate, may have durations and interval times less than 1 1s (see /10/,/22/ and paper Fl of this conference). It may be possible to infer some characteristics indirectly. /23/, for example, has measured RF signal amplitudes at the six frequencies given in Section 2 and has found differences in the relative amplitude spectra for different types of discharges. We have also recently begun measuring the fast E field variations which accompany RF emissions. For this we have used a specially designed, 10 MHz - 1 GHz, conical antenna connected to a very fast transient digitizer, We hope to be able to report on this experiment in the very near future. i .mV/m L 175 MHz 250 !-= .~--------__~________---__-________. 8 1 -b mV/m a J 350 0.2.- , 300 Fig. 175 MHz ps 8 I 400 Fast time resolved records of portions of the 175 MHz signals in Fig. 5(a) and Fig. 4 (upper and lower traces, respectively). The single, 3 s wide impulses at the left of trace (a5 are the impulsive response of the 175 MHz RF receiver. - 168 REFERENCES Ill Jman, M.A., and E.P. Krider, "A Review of Natural Lightning: Experimental Data and Yodeling," IEEE Trans. tlectromag. Camp., EMC-24, 79-112, 1982. 121 J. Hamelin, C. Leteinturier, Pulses Emitted L. Nicot, "Electromagnetic by Lightning," presented at the Intl. Aerospace Conf. on Lightning and Static Electricity, Oxford, England, March, 1982. l3l 141 ISI Djebari, B., Proctor, D.E., "A Hyperbolic System for VHF Radio Pictures of Lightning," J. Geophys. Res., 76, 1478-1489, 1971. Obtaining Taylor, W.L., "A VHF Technique for SpaceTime Mapping of Lightning Discharge Processes," J, Geophys. Res., 83, 3575-3583, 1978. Warwick, J.W., C.O. Hayenga and J.W. Brosnahan, "Interferometric Directions of Lightning Sources at 34 MHz," J. Geophys. Res., 84, 2457-2468, 1979. 161 Rustan, P.L., M.A. Uman, D.G. Childers and W.H. Beasley, "Lightning Source Locations from VHF Radiation Data for a Flash at Kennedy Space Center," J. Geophys. Res., 85, 4893-4903, 1980. /7/ Proctor, D.E., "VHF Radio Pictures of Cloud Flashes," J. Geophys. Res., 86, 40414071, 1981. I81 Hayenga, C.O. and J.W. Warwick, "TwoDimensional Interferometric Positions of VHF Lightning Sources," 3. Geophys. Res., 86, 7451-7462, 1981. /9/ Richard, P., A. Delannoy, G. Labaune and P. Laroche, "UHF Interferometric Imaging of Lightning," in Addendum 8th Intl. Aerospace and Ground Conf. on Lightning and Static Electricity, DOT/FAA/CT-83/25(A), Fort Worth, Texas (USA), June, 1978. /lOI Proctor, D.E., "Lightning and Precipitation in a Small Multicellular Thunderstorm," J. Geophys. Res., 88, 5421-5440, 1983. IllI Taylor, W.L., E.A. Brandes, W.D. Rust and D.R. MacGorman, "Lightning Activity and Severe Storm Structure," Geophys. Res. Lett., 11, 545-548, 1984. I121 Hamelin, J,, J. Karczewsky and F. Sene, "Sonde de Mesure du Champ Magnetique dD a une Decharge Orageuse," Annales des Telecomm., 33, 198-205, 1978. /13/ Weidman, C.D. and E.P. Krider, "Time and Height Resolved Photoelectric Measurements of Lightning Return Strokes," Trans. Am. Geophys. Union, 61, 978, 1980. 1141 Hubert, P. and G. Mouget, "Signaux de Foudre Enregistres sur Bande Magnetique Video," Report CEA-R-4818, Commisariat a 1'Energie Atomique, Saclay (France), 1977. 1151 Weidman, C.D. and E.P. Krider, "The Ra- diation Field Wave Forms Produced by Intracloud Lightning Discharge Processes," J. Geophys. Res., 84, 3159-3164, 1979. 1161 Leteinturier, C. and 3. Hamelin, "Analyse Experimentale des Caracteristics Electromagnetiques des D&charges Orageuses dans la Bande 200 Hz - 20 MHz," Annales des Telecomm., 39, 175-184, 1984. 1171 Krider, E.P., C.D. Weidman and D.M. Levine, "The Temporal Structure of the HF and VHF Radiation Produced by Intracloud Lightning Discharges," J. Geophys. Res., 84, 57605762, 1979. 1181 Krider, E.P., C.D. Weidman and R.C. Noggle, "The Electric Fields Produced by Lightning Stepped Leaders," J. Geophys. Res., 82, 951-960, 1977. 1191 Brook, M. and N. Kitagawa, "Radiation from Lightning Discharges in the Frequency Range 400 to 1000 MC/S," J. Geophys. Res., 69, 2431-2434, 1964. /20/ Levine, D. and E.P. Krider, "The Temporal Structure of HF and VHF Radiation During Florida Lightning Return Strokes," Geophys. Res. Lett., 4, 13-16, 1977. I211 Hayenga, C.O., "Characteristics of Lightning VHF Radiation near the time of Return Strokes," J. Geophys. Res., 89, 1403-1410, 1984. 1221 Laroche, P., A. Eybert-Berard, P. Richard, P. Hubert, G. Labaune and L. Barret, "A Contribution to the Analysis of Triggered Lightning: First Results Obtained during the TRIP82 Experiment," in Addendum 8th Intl. Aerospace and Ground Conf. on Lightning and Static Electricity, DOT/FAA/CT-831 25(A), Fort Worth, Texas (USA), June, 1978. I231 Le Boulch, M., "Caracterisation et Mecanismes du Rayonnement VHF-UHF des Decharges Orageuses," Thesis, Universite de Clermont II, Clermont-Ferrand, France, 1984, /24/ Le Boulch, M. and J. Hamelin, "Rayonnement VHF/UHF des Eclairs," to be published in Annales des Telecomm., January, 1985. /25/ Beasley, W., M.A. Uman and P.L. Rustan Jr., "Electric Fields Preceding Cloud-toGround Lightning Flashes," J. Geophys. Res., 87, 4883-4902, 1982. Copies of references 121, 191, 1141, 1221 may be obtained by writing to the authors. and /23/ - 169 31 F4 - EQUBW EXPONDPIAL BlL%MlDELs EORCDMPARISGNOF LIGHTNING, NUCLEAR AND KLRCTRDSTATICDIGCRARGK SRRCIRA MANUELWWIK Defence Materiel Administration ElectronicsDirectorate s-115 838stockholm The stiltaneous protection against all kinds of transient fields is a very Serious pr&lan. Ccanpatiblebroadband interference control calls for integration of protection requirements and the nesd for unified standards. In the ccnning years the EMC cmity must strive toqards integrationof protection against radiated and conducted lightning, nuclear from interference electrostatic ~lses, electranagnetic discharges and unwanted radio emission.This paper gives examples of simple double exponential pulse tiels for ccxnparisonof electrostatic nuclear and lightning, discharge radiated spectra. Background Until recently nuclear EMP (-1 protection has mostly been treated apart from protection against other EM hazards. There are special military reguirements,but the awareness of the high altitude N@P threat for the whole society leads to the need for compatible military and civil especially specifications for telecglmunications and electric Per systems. NENP specialists have previously focused on the threat that the short risetimes represents. For systems employing long electrical conductors studies are no,~ also performed wnceming the extremely 10~ frequency magnetohydrodynamic(MEJD)effects of NEMP. These effects are in scme way c-able to aurora1 gmgnetic storms. Nuclear EM? trends are towards new families of enhanced weapons. Napoleon has said that "when you can use the lightning, it is better than a cannon". Now it seems that modem Napoleons are making sofisticated lightning for military purposes. Recent progress in micrmve technology has resulted in powerful m3dulated wn nuclear EMP devices. Micrcwave pulses can penetrate joints and slits in shields and cause internal induction of the low frequency pulse envelope. Directed energy wwer sources such as pulsed plasma magnetchydrodynamic cartridges can wnvert chemical energy directly into pulsed electrical energy Cl]. Space technology is a major support of more effective nuclear weapon systems. Nuclear weapons are also beconing part of laser and beam weapon technology [2]. Lightning experiments done during the last tl-ree or four years have shown, that the certain types of fields radiated by lightning strokes can have rise times of Whereas lightning several nanosewnds. studies in the past have focused w the high energy lm frequency parts of lightning strokes, especially for the protection of buildings, the interest is rriw directed tmds the relatively low energy la& high frequency parts of lightning and its impact on electronic systems. The rapid increase in deployment of susceptible semiconductor and associated ccmiponentsin a number of new applications throughout society results in a situation where we are all beccming more dependent on reliable electronics than ever before. Digital systems involve large bandwidths. High probability of EM1 of lightning With digital communication syste3ns has been observed C31. These observations indicate that the -cation systems unfortunately also mid be highly susceptible to mclear JZMP. Conducted switching transients have been measured networks interconnecting electric c and electronic loads [4]. SO far mostly transients With relatively long durations have been observed. In addition to these, shorter spikes are able to appear at least locally. These spikes can emanatefran varicus sources such as static electricity and can create broadband interference. High circuit voltage G&j breakers produce very broadband transient fields (up to several hundred MJZ). Transients can result in no visible impairment or in temporary impairmentof a system's -ration or in permanent damage. Even tien no iqairmsnt is &served frcm-ia single transient it can cause an aging prwess leading to a much shortened lifetime. In the future this calls for increased attention to short transients and their lasting effects on eguipnent performance [5]. This is also important ti bear in mind &en designing interferencespecificationsand verification procedures. The development and installation of fiber optical and satellite ocamunications, ccquter networks and electronic processand wntrol eguipnent in industry and mer - 170 exponential pulse rmdels. Neither lightning nor nuclear or electrostatic discharges in real life will look exactly like the node1 pulses. The xcdels have only been created for design pm-poses. The double exponential pulses happen to be very simple to treat. The effect of variation of @se amplitude, rise time and pulse duration can easily be studied [7]. The real pulses have a large variation in appearance which makes it feasible to describe bands within which spectra can fall with various probability. systems increasing. rapidly * specialist?will have failed if they Z interference issues unable to make understandable and incorporated into system initial the from acquisition conceptualizationphase cmward. Performance requirements cannot be levied after-thefact for systems intended to function in stressed environments. In Sweden an EMC working group was famed about ten years ago in order to help incorporating EC and NEMP specifications into systsan acquisition [6]. After a proposal frcnn the group the Swedish Electrotechnical Commission has recently decided to form a working group for nuclear ?ZMPstandardization. The double exponential tima function can bs characterizedas: f(t) = A * (exp(-at)- exp (-bt)) ...(l) Risetime (10 - 90%) = t, = 2,2/b Pulse duration (50 - 50%) = td = 0,69/a Peak amplitude= A*(l+(a/b)*(ln(a/b)-1)) A proposal has also been made to build a separate NEW-protected higher class of teleccmmmication secure network for subscribers needing priority in case of crises or war-time. The network should be built with separate qtical fibres in new cables that the Swedish Telecamnmications Administration will deploy in the future. The separate fibres should be connected to separate cable repeaters, exchanges and subscriber terminals provided with separate pawer supplies and EM protection.The system should also interconnect a NEW protected mobile telephone network. the Spe&nm\ In the frequency damin asymptote is constant and equaltoA/a for 2 Vf. For intermediate Ocwb the spectrum asymptote decreases as A/w2 or 40 dB per decade. Another waveform which is more like natural exponential bshaviour of physical processes is the reciprocal of the sm of two exponentialsgiven by: Double exponentialpulse models f(t)=A*(exp(at)+exp(-bt))-' Spectrum function This paper doss not claim to give accurate lightning, nuclear and electrostaticdischarge spectra. The purpose is rather to draw the attention to ways of carq?aringthe different EM sources in order to work tumrds the integration of protection requirements. In this particular paper examples are given of sms double - -I- Nr [ME tr h iA) a (ns) t T-l d (I-IS) 133 133 67 33 266 486 10 10 5 10 0,173 0,05 0,2 O,O2 0,05 6 800 20 70 7 OS04 127 5 8 9 10 11 12 13 14 15 16 9,8 9,s 4,6 20 20 200 20 20 200 156 156 73 318 318 3183 318 318 3183 L17 60 L16 300 530 530 1200 L600 3200 ...(2) In this case the risetime is equal to 4,4/b or twice the risetime of the previously mentioned caouble exponential pulse model. Pulse duration and peak amplitude are the same for both types ofmdels. At the second break frequency, the latter lrodelisplies a more rapid (exponential) decrease of the spectrumasymptote. a (s-l) AIN FREQU =?qX-dB over (s-l) (A/m) As/m) b A f 2 (MHZ) 4x106 4,76x1: 139 1,38x10? 2,2x10' 174 3,45x106 2,2x10* 73 3,45x10+ 4,4x10g 46 1,38x10' 2,2x108 348 89 98 93 118 92 630 2200 550 5500 2200 76 35 35 70 35 lo4 1,1x10& 800 22 1,6 18 0,03 2,3x10' 4,4x108 160 103 3700 70 22 22 0,4 45 40 40 40 50 50 3,14x104 188x16 158 3,14x10' 3,67x10 157 1,73x106 1,9x103 106 1,53xlOY 7,3x104 323 1,73xlOY 3,49x10' 328 1,73x10V 3,49x10' 3286 1,73x10' 1,83x10 336 1,38x10' 1,38x10' 337 1,38x10' 6,88x10: 3531 46 46 84 34 34 14 34 32 12 5 5 280 2,4 2,8 2,8 2,8 2,2 2,2 3 5,8 3 1,2 0,56 0,56 0,29 0,22 0,ll Fig.l: EXAMPLES OF IXUBLE EXQJENTIAZIPULSES GIVEN BY f(t) = A*(exp(-at) - exp(-bt)) - 171 31 - F4 Exoatrrospheric nuclear W (W) For design rannoses the generalized free field nucl&-I?& fran an exoatmospheric burst is often characterizedas a free field (Z = 377 ohm) plane wave C8l with the magnetic field function according to plse nr 1 in Fig. 1. This is close to a maximum threat. -ever one nust expect that the field waveform and direction varies considerably with lurst location, observation point, type of weapon and other factors. At the location of max peak field pulse nr 2 is assmed, near tangent radius south of the burst pulse nr 3 and near surface zero @se nr 4. The 50 kV/m peak field of pulse nr 2 is doubled in nr 5 for conparison purposes. The spectra are shm in Fig. 2. dB (over 1 As/m) FREQUENCY -40. -. Fig.3: SNEMP SPECTRA, 1 Mt, H = 800 A/m -6O.---- dB (over 1 As/m) O---n-qT~, , ,,,1,,,1 . -20 I LEMP I ( ,(,,, IO4 105 FREQUENCY IO6 IO7 108 109 f -100 (HZ.) FiR.2: HNEMP SPECTRA -120 Field interaction with objects result in modified field functions. The amplitude can be enhanced or decreased. The time function can bs considerably changed. A special case is the danped sinusoidal function. Surface nuclear EMP (SNEMP) The qeneralized surface altitude NIWP waveform- for the azimuthal magnetic field for a dnal 1 Mt explosion is given for a 800 A/m peak field, 20 ns rise tims and 70 p's pulse duration to half value. The model is extracted fron ref [9], see nr 6 in Fig. 1. The field uncertainty is ass& to bs approximately210 dl3,see Fig. 3. Lightning W (LEMP) Suns liqhtninq rulse models are qiven in Fig. 1 and 4. -me peak magnetic -field fi (A/m) at a distance r (m) frcm a,lightning channel with a peak current I (A) is , ,,,,” , ,, IAMP! \ 8 w -*Ot-----I 103 I 13 I -40 t 102 (Hz) I I I I I I l\il I\\ \\ 1 \, \\I\ I 1 i 1 I I I l\Y I 1 5 I I I IN-\1 1 -140t 102 I I I I I I I I I 103 104 IO5 m’D*nrlch,r” 106 ,“li \ 1 1 108 109 I\\ \ I \\ \ IO7 Fig.4: LEINPSPECTRA, 10 m CHANNEL DISTANCE estimated as fi= (f)/(2?Yr).This is nore of an engineering and less of a scientific approach. Calculations are based cn a 10 m distance frcm the lightning channel. At a distance r (m) the field spectrum A/a (dB) is altered a factor 36 - 20 log 2tir (dD). Pulse nr 8 is an approximationof amplitude spectra given by [lo]. Tne lightning pulse rise time is 117 ns. In pulse nr 9 it is - 172 - reduced to 60 ns which has also been measured [ll]. For a linear rise, @se no 8 nds to an approximatevalue of di/dt :"yOTy A/s. Pulse nr 10 is based on average values measured by C123. Pulse nr 11 is extracted frcana specification by Cl31 modified by [ll]. This shows approximately6 earlier than values higher times current 50% maan measurements for derivatives.Pulse nr 12 is based on a node1 by [14] also used as part of a aode described w [15]. The lightning current is 20 kA. In @se nr 13 this is increased to 200 kA, see also Fig. 4. Pulse nr 14 is a simplified version of a pulse given by [13] for specificationpurposes. Pulses nr 15 and 16 are extracted fran AFHP moderate and severe threat definitions given by [16]. Electrostaticdischarge EMP (ESD EMP) Electrostatic discharge currents and associated fields can also be delled as double exponential pulses if interactions are disregarded. Nany examples of different ESD pulses are given in literature, e.g. [17]. In this case the peak magnetic field is calculated as for lightning currents but assuming a 5 cmdistance from awirewith a 40 A peak current discharge having a 5 ns rise time and a 30 ns pulse duration to half value, see Fig. 1 and 5, pulse nr 7. The spectrum uncertainty is assumed ti be +lO dB at ICBJ and +5 dD at high freque&ies. Ho&ever, in tiz near field region within cne wavelength of the wire it is not possible to accurately relate the electric and magnetic fields to each other using simple far field relations. Spectra mnparison Spectra of pulses nr 1, 6, 7, 8 and 13 dB (over 1 As/mb of Fig. 1 are shown together in Fig. 6. For better clarity the regions of uncertainty have been left cut. EJcwever,in a thorough ccmparison they should be included. l'h@ SNENP spectrumnr6 exceeds nest spectra up to about 50 MHz. However, if a severe lightning stroke such as represented by pulse spectrum nr 13 is chosen, it will exceed SNEMp up to about 5 MHz and other spectra up to about 10 MHz. The frequency region frQn 1 to 10 MHz is of particular interest concerning resonance excitationdue to the carmon presence of structures with physical dimensions in the order of a few meters tens of several meters to half the tiich for (frequencies correspondingwavelength equals the physical dimensions). LZNP spectrum nr 8 exceeds HNRW nr 1 up ti almost 1 Mlz and is the same as nrlupto3MHz.HNEMPspectranr3 and4 are lo&r than theLEMP spectrum nr 8 up to about 10 MHz. It should be noted that the HNENP spectrum nr 1 is a worst case while LFMP nr 8 is an estimate based on average measurements. (equal to physical Frun 10-100 Wlz dimensions 15-1,5 m) the SN!BP spectrum nr 6 strongest, follmed by almost equal Gtlues of HNEW (nr 1) and ESD (nr 7) spectra while LIMP (nr 8 and 13) are lower. At frequencies above 100 MHz WJM?, HNEMP and ESD spectra are similar in Fig. 6. In a ttore detailed comparison the accumulated energy -tents of the spectra as a function of frequency should also ba ccmpared. The HNJM? (nr 1) crmtains about 10% of the total energy Bela 100 kHz, about 60% between 100 kHz and 1 MHz and almost 30% betwefm 1 and 10 MHz. dB (over 1 As/m) 0 ,( I I!(,W I I1,1111, ,rp-TTqnm--l-rr,,,r . LEMP I _20_SNEMP .-. I 6\ Y\ -40 t LEWd -1401 FREQUENCY (Hz) Fig.5: ESD SPEK!TRA,5 cm DISTANCE ,- 13 ‘\ 8 I _ \ I I I I+A-f! I I I I I lyQ$J 102 103 104 IO5 FREQUENCY 106 107 IO8 109 (Hz) Fin.6: COMPARISON OF HNEMP, SNEIP, LEMP AND ESD SPECTRA - 173 _ _ - -: Flrsl ReturnStrokes -175 IO’ 105 Frequency Id Id IO’ klz) Fig.7: COMPARISON OF LIGBTWING MODEL SPECTRUM No. 8 WITH MEAN AMPLITDDE SPECTRDM OF 24 FIRST RETURN STROKES GIVEN BY Ref. [IO]. ACCORDING TO cl01 CURVES (2) AND (3) ARE BEST APPROXIMATION'IOTHE TRUE SOURCE AT HIGHER FREQUENCIES.SPmRAL AI"lPLITUDE VALUES ARE FOR E-FIELD AT 50 km DISTANCE. In Fig. 7 it is shov~n that lightning spectrum KU-. 8 agrees well with mean amplitude spectrum of first return strokes according to ClOl. Under certain circumstances and in certain regions the spectra from the different sources can agree. This fact encourages integration of protection against radiated broadband interferencefrcm various sources. of one only considers spectra (according to Fig. 6) and mt coverage area the conclusion is, that if cne has ti protect against FSD at high frequencies and lightning at mr frequenciescne has also roughly covered the HNEMP spectrum. The ESD spectrum average area is however small (calculations were made at 5 cm distance). Therefore only fbr dimensions objects with small very protection against radiatedESD interference will also be of value against radiated HNEMP. However, good protection against ESD of equipment oabinets and associated cables incorporates careful shielding and bonding, utilizing topological benefits as well as other measures to filter out broadband radiated and conducted interference. Until recently such disciplines have many times been overlooked leading to a variety of unexplained functional disturbances, degradation and failures. Now these protection methods are a mst fbr most electronic e&-t. This also leads to better resistance against radiated LIMP and NEMP fields. LISP F4 The difference in E8D EMP, m and NEMP coverage areas is of vital importance fbr interferenceand susceptibilitystudies. ‘Ihe risetime and duration of the m field differ significantlyfrcm tkboseof lightning and power transients. Thus, protection against these effects will ~~JIz necessarily protect against l%P Clgl. Daublc -55 r Id 31 - spectra only compares within 10 m distance of the lightning channel. The dcaninating threat frcnn m (and to a certain extent also franSN$M?) is the very large average area. Telephone and per lines will be exposed to radiated fields along their whole length and this can result in large currents [18]. This secondary conducted interference poses the largest threat ti electric and electronic systems. double exponential time Although the functions given lq the difference of two exponentials are camrw3nlyused. and often cited they are not truly representing physical processes C20, 211. The functions start with a discontinuity and a finite slope at t=O, reach a peak value after a short time and then decay slowly to zero. The waveform consistingof the reciprocalof the sum of two exponentials is rather to be recmded from physical point of view. Nevertheless, the first model gives a rough upper limit of the spectrum frequency content at frequenciesnot too far above the second break frequency. Double exponential@se mxlels are actually just models and comparison of field spectra should be made with caution. Fields fran lightning strokes are rmch more oaaplicated than w plane wave free fields (without interaction). Several interesting lightning observations and studies have been made recently and should be continued in order to improve protection of especially new generations of electric and electronic eguipnents and systems. First lightning strokes can start with a slaw front rising for several ,us to about half of the peak. This can be followed by a fast transition to peak in the order of several nanoseconds r-111. Because of near field phenomena present only at short distances from the lightning channel measurements at short (meters)and 1ong (kilometers)distances are not expected to correspond fully. The lightning current @se &es not propagate at a constant speed and the fields are cylindrical in compilrisonwith the HNElMp plane wave field. In a sore complete ccmparison field polarisation and direction should also be taken into account. Conclusions Sinple double exponential pulse rrPdels for lightning, nuclear and electrostatic disharge radiated spectra can be used for discussion of broadband radiated interference. Recent measurements of lightning spectra have shm that fields radiated by certain types of lightning strokes can have rise times of several nanoseconds correspondingto very broad&nd interfe;I';;;Ze Ccmparison of spectra shm that w compares to severe lightning at short distances up to some M~IZ. Lightning and high altitudeNE?@ spectra can be qrable around 1 MHz depending on assumptions.At frequenciesabove 10 MHz ESD can be ccanparableto m. Hawever the difference in coverage areas is of vital importance for radiated interference and susceptibility studies and accentuates the m threat against telemunications and electric power systems. - References Cl1 Gill, Steven P.: Directed Energy Power Source Could Generate EW Technology Revolution. Defense Electronics,April 1984, 116-120 c21 Beam Defense. Bibliography1983. Aero Publ, Inc, 329 West Aviation Road, Fallbrcok, CA 92028 USA c31 Scuka, Viktor: EM1 of Lightning with ication Systems. Int. Digital Ccrnnun Wroclaw Syq on EM-Z84. Institute of High Voltage Research, Uppsala University, S-755 90 Uppsala, Sweden [4] Wemstrijm, H&an et al: Transient Overvoltages on AC Power Supply Systems in Swedish Industry. EDA Report E 30002-E2,April 1984. National Defence Research Institute, S-581 11 Linkaping, Sweden [51 Scuka, Viktor et al: Lasting Effects of Transients on Eguipnent Performance.Workshop organized by URSI Ccmnission E during the 1983 EMC Svmp in Ziirich.UURIE: 147-83, Uppsala 1983 [63 Practical Methods for Electromagnetic InterferenceControl. Ericsson Network Dept., S-126 25 Stockholm, *eden [7] Wik, Manuel W: Double Exponential Pulse Models for Canparison of Spectra fran Lightning, Nuclear and Electrostatic Discharge Sources. EMC 84, Tokyo [8] EMP Engineering arrdDesign Principles, Bell Telephone Laboratories,Technical Publ, Dept., Whippany, New Jersey, USA, 1975 C9] Longmire, C.L.: The History and 'Physicsof EMP. Mission Research Corp., Santa Barbara, Calif 93102, USA. Paper given at NEM 84, Baltimore 1984 [lOI Rider, E.P., and Weidman, C.D. The Amplitude Spectra of Lightning Radiation Fields in the Interval fra 1 tx 20 MHz. 1984 IntemationalAerospace and Ground Conference on Light&q and Static Electricity, Florida, USA [ll] Unxan,Martin A.: Application of Advances in Lightning Research to Lightning Protection. 1984 174 - InternationalAerospace and Ground Conference on Lightning and Static Electricity,Florida, USA [12] Kuhhnan, B.P., and Reazer M.J.: Characterizaticnof Fast-risetime ElectromagneticField Pulses Recorded in Airbxne Measurementsduring Florida Thunderstorms.1984 InternationalAerospace and Ground Conference on Lightning and Static Electricity,Florida, USA [13] Fisher, F.A., and Plumer, J.A.: Lightning Protection of Aircraft. NASA Ref. Publ. 1008, 1977 [14] Cianos, N., and E.T. Pierce: A Ground Lightning Envirornnentfor Engineering Usage. Stanford Res. Inst., Project 1834, August 1972. IYTICNo AD 907891 [15] Riley, L.H., and Fdlin, G.R.: Lightning Tests of Pershing II. 1984 InternationalAerospace and Ground Conference on Lightning and Static Electricity, Florida, USA 1161 Lippert, Jack R. et al: Progress of the Atmosperic ElectricityHazards Protection Program. 1984 International Aerospace and Ground Conference on Lightning and Static Electricity, Florida, USA [17] King, W.M., and Reynolds, D.: Personnel ElectrostaticDischarge: Impulse Waveforms Resulting Fram ESD of Humans Directly and through Small Hand-Held Metallic Objects Intervening in the Discharge Path. 1981 IEEE Int. Symp cn EMC, Boulder, alorado, USA Cl81 Wik, Manuel W.: Hardening of Teleccrsnunication Networks against Electrcxnagnetic Pulses. Ericsson Review rxo1, 1984 1191 TJRSIFactual Statement on Nuclear Electxcxtxagnetic Pulse (EMP) and Associated Effects. URSI XXIst General Assembly, Florence 1984 [20] Lee, K.S.H.: I%? Interactia: Principles,Techniques and Reference Data. Al%%-TR-80-402,Air Force Weapons Laboratory, Kirtland AFB, New Mexico, USA [21] Gardner, R.L. et al: CXsnparison of Published HEMP and Natural Lightning an the Surface of an Aircraft. Paper given at NEM 84, Baltimore, USA - 175 32 - F5 COMPARISON OF LIGHTNING WITH PUBLIC DOMAIN HEMP WAVEFORMS ON THE SURFACE OF AN AIRCRAFT* R. L. Gardner and L. Baker Mission Research Corporation Albuquerque, New Mexico (USA) C. E. Baum and D. J. Andersh Air Force Weapons Laboratory Kirtland Air Force Base New Mexico (USA) INTRODUCTION High altitude EMP (HEMP) and its concomitant electromagnetic environment potentially HEMP, a short (-0.1 threaten an aircraft. vs.) pulse of large amplitude ("50 kV/m) arrives at aircraft essentially as a plane wave. Although this pulse generally contains no oscillations (zero crossings), it's Fourier transform shows frequency.content over a wide band, with significant content up to 100 MHz. For present purposes we use a well-known public domain HEMP waveform. Lightning, another potential threat, can interact with an aircraft in two essentially First, for a nearby strike, different ways. the electromagnetic fields generated in and near the stroke channel impinge on the airSecond, for a direct strike on the craft. aircraft, the stroke current actually flows on the conducting structure of the aircraft. The first of these effects may be called field interaction and the second, current injection. It is reasonable to expect that the latter may have larger effects than the former because the strike current path is along the aircraft. Because of the increasing concerns about these two threats, this study assessed the differences between the electromagnetic environment associated with HEMP and that associated with natural lightning, including the manner in which they affect aircraft. The investigation was based on the environments suggested by public domain literature for HEMP and by published data for natural lightning. These environments are described in Section 2 of this paper. The comparison of the two threats was based on the currents and charges on a simple geometry representative of the characteristics of an aircraft that were caused by the two electromagnetic environments. In Section 3 several simple analytical models are presented to relate the currents and charges to the environments. These models are then used to compare the two threats to aircraft in Section 4. In Section 5, operational considerations for the two threats are presented. Lightning is improbable but damaging to aircraft. Exposure to HEMP is essentially certain for aircraft in war. Section 6 presents the conclusions of the study, that below about 1 MHz lightning dominates, above 10 MHz HEMP dominates, and between the two limits the interaction of the environment with the aircraft is sufficiently complex that either may dominate, depending on the details of the aircraft. II. ENVIRONMENTS This section presents the electromagnetic environments produced by HEMP and lightning. The environment for HEMP is that presented in the public domain (Ref. 1). The lightning environment is derived from a number of references which present actual measurements of lightning electrical characteristics. In this paper the mechanism of HEMP and the various arguments used for determining a lightning environment are only summarized. For more detail see reference 2. HEMP Environment The generation of HEMP by a nuclear device is described in detail in an article by Longmire (Ref. 1). Sophisticated codes are used to calculate the field levels for HEMP and these calculations agree well with experimental data. The incident HEMP waveform depends on a number of factors including: height of burst, device type, atmospheric conditions, and distance from the explosion. To avoid this complexity during the system design process a guideline waveshape is used. A waveform presented in reference 1 is: E(t) = -tbf Eo(e -e --) u(t) (1) where E. = 60 kV/m (saturation field) 250 ns is the fall time ,rf = 7 = 2 ns is the rise time constant r and u(t) is unit step function. In this comparison of HEMP and lightning, HEMP will treated as an incident plane wave with the waveform described by Equation 1. With this waveform, the peak amplitude is the predicted saturation value of 60 kV/m. - 176 Lightning Environment In this report, the current, that is the electromagnetic environment for direct and nearby strikes, is assumed to be produced in a return stroke because it typically has the largest currents and rates of rise. Detailed descriptions of the sequence of events in a lightning discharge and relevant definitions are contained in Uman (Ref. 3) and Golde (Ref. 4). To characterize the lightning environment, which is the lightning current, three figures of merit are sufficient to specify the double The exponential waveform of Equation (1). three figures of merit considered here are: (1) (2) (3) Peak Current Peak rate of rise of the current Integral of the pulse A reciprocal double exponential has these three figures of merit as parameters, is easy to work with, and has an analytic Laplace transform. Since the detailed theoretical modeling effort applied to HEMP has not been applied to lightning, it was necessary to use empirical techniques to determine the figures of merit listed above. All of the available measurements from which estimates of lightning current parameters are derived may be divided into three classes: (1) (2) (3) Tower measurements Measurements on aircraft in flight Radiated field inference of current Data from each of these sources were used to estimate the current in the lightning channel. This current within the channel establishes the lightning current waveform that constitutes the threat. Tower Measurements. Tower measurements of lightning currents are made using current sensors installed on metal towers located where there is normally a great deal of lightning activity, generally mountain peaks. Since the tower is part of the lightning discharge circuit, the effect of the tower itself on the measurements must be considered. Useful summaries of lightning currents and rates of rise of the current are given in Uman (Ref. 3), Golde (Ref. 4), and Garbagnati (Ref. 5). Of those summaries, only the data presented by Garbagnati is fast enough to see characteristic times of 100 ns or less, so that data will be shown here. The longer version of this paper (Ref. 2) contains a more complete presentation of the data. The maximum rate of rise reported is less than lOI A/s. Other sets of tower measurements confirm this data. When corrected for ground reflection even the maximum rate of rise observed by Ericsson (Ref. 6) is very near 1011 A/s, as well. At this time, tower measurements constitute the only low altitude, cloud to ground lightning current measurements available. Aircraft Measurements. Electromagnetic measurements made on an aircraft in flight represent another useful data base for determining the electromagnetic Two recent environment caused by lightning. sources provide data on the effects of lightning direct strikes on aircraft. The peak rate of rise measured on the boom in front of the F-106B (Ref. 7) is particularly interesting since it has the largest rate of rise of the current. In spite-of the low (13.9 kA maximum) peak currents, the peak rate of rise found by taking a graphical derivative of the current records was 1.3~10~1 A/s. Significantly, the maximum value closely approximate the 1011 A/s maximum rate of rise seen in the tower measurements. Currents Inferred from Field Measurements. Another method of determining the current in a discharge is to derive the current from distant field measurements. The difficulty with this method is that the current derived from the fields is not unique and unfolding the very complicated early time current evolution of lightning is not a trivial task since no quantitative model exists for the early part of the return stroke. Since the recently published research using this method suggests very fast rates of rise in return strokes the methods used will be more closely examined. Uman, et al. (Ref. 8) derive the relationship between the electric field and the current in the lightning channel under a restrictive set of assumptions. The initiation point must be at the ground and the current waveform must propagate up the channel at uniform velocity and without distortion of the waveshape as it propagates. It is also assumed that the fields are entirely in the radiation zone. Correcting the above modeling to account for the return stroke currents that initiate from a point about 100 m from the ground rather than at the ground reduces the current and derivative values by a factor of two. For subsequent strokes there is no initiation region as described here. However, for subsequent strokes there is a memory of the location of the channel. The breakdown wave is limited in propagation velocity by the velocity of light rather than the velocity of propagation for return strokes. Near the ground, i.e., at early times the appropriate velocity for the Uman model is that of light. Corrections for the data reported in reference 9, which shows the distribution of the derived rates of rise from the fields, brings the data from the fields into much closer agreement with the data reported from tower measurements. During the initial stages of a return stroke, the return stroke channel radius is small; consequently the channel is simultaneously resistive and inductive (Ref. 2). Detailed numerical simulation of the evolution of the channel indicate that the channel diameter is less than about 2 mm. The impedance of the channel for high frequency of the initial - 177 channel does not allow current waveforms with high rates of rise to propagate without distortion along the channel, thus violating the A assumptions of the Uman model (Ref. 2). model consistent with this small channel at early time is that of a local current source rather than a propagating wave. A final item to consider in deriving currents from electromagnetic fields of lightning is the possible branched configuration near the ground. A conceptual picture is shown in Fig. 1 which is supported by photographic eviThe effect of this branched dence (Ref. 2). configuration is that currents derived from the fields of lightning, at early times, are potentially a factor of two to three too high. 21, About 20 US After Closure 32 - F5 maximum current of IO kA. It should be noted that the lower current results in a larger frequency content above a few MHz. 15v1 12 t 2 E 2 I)- ar > .I4J ld > 6- .r b 0 3- 0. -20 I I I I I 0 20 40 60 80 Time (US) Fig. 2. Time derivative of current described by Equation '2. Possible branched configuration showing closure region where the transition from leader to return strokes occurs. Fig. 1. Summary of the Lightning Threat. A review of the available data lightning suggests a particular waveform fOonr the lightning return stroke current. The waveform is similar to Equation (1) for the HEMP fields but has different parameter values. We conclude the lightning current should be represented by: I(t) where kIpk = -(t-t&“, e +e (t-t@‘f (2) k is 1.025 and I = 100 kA pk Tr is the rise time constant = 2.5 x 10m7, chosen to give a maxmum rate of rise lOI1 A/s at t = t0 = 50 ps is the fall time constant Tf to is an offset time Figure 2 is a plot of the current described by Equation 2. Fig, 3 shows the corresponding frequency spectrum for the waveform described in Equation 2. Based on an examination of available data, there is a severe threat described by Equation 2 with a maximum rate of rise of 1011 A/s and a maximum current of 100 kA. Also an expected threat should be represented by the same rate of rise but a Frequency (Hz) Fig. 3. Frequency spectrum of current from a Fourier transform of Equation 2. III. INTERACTION In addition to the differences in the electromagnetic environments produced by lightning and HEMP, these two threats differ in the way they interact with an aircraft. This section describes both qualitatively and quantitatively, using simple analytical models, the interaction of lightning and HEMP with aircraft. - 178 Physics Of Interaction of Lightning and HEMP With Aircraft Direct Strike Lightning. While understanding of the physics of interaction of direct strike lightning with aircraft is still in its early stages, a qualitative description of the interaction may be When an aircraft enters a thundergiven. storm, it is under the influence of an eleCtric field which by polarizing the aircraft induces image changes and causes currents to If a leader flow on the aircraft's skin. streamer approaches the aircraft, those fields become much larger, particularly around sharp As the corners and edges near the streamer. local fields approach a level of about 3 MV/m the air begins to break down in the field enhanced region or forms a corona. This coron a exhibits non-linear characteristics which complicate the electromagnetic behavior of the aircraft under the influence of the nearby streamer. As the leader attaches to the aircraft, coronal activity increases and a channel forms. As current flows through the channel, the air gets hotter and the number of active physical processes increases dramatically, Hydrodynamic expansion, radiative transfer, thermal conduction, joule heating, and field emission from the metal surfaces all play a part in developing the channel that in turn forms the current carrier of the lightning direct strike. The system of clouds, channel, and aircraft should by viewed as an electrical circuit. Initially, because the channel is cool and narrow, it is both resistive and inductive; consequently it limits both the current and its rate of rise. As the lightning channel forms and as the return stroke current flows, the channel becomes hotter and larger in diameter. These increases, in turn, diminish the resistance and inductance of the channel, allowing more current to flow more rapidly. Free Field Interaction. HEMP and nearby lightning both interact with aircraft as a superposition of plane waves. When the electromagnetic wave interacts with the aircraft, it induces currents and charges on the aircraft as if the aircraft were an antenna. While non-linear behavior near sharp points and edges may occur in free field interaction, the effects are not as dominant as they are in the direct strike case since the coupling is not as efficient and there is not the direct charge transfer that exists in the direct strike case. The incident fields penetrate in the aircraft through apertures and along system cables entering the electrical enclosures of the fuselage. Internal system cables may be excited by several apertures. For HEMP excitation, the cable drive from these various apertures will in general be in-phase, such that the various drives from the apertures reinforce. This inphase drive is in contrast to the aperture - drive from direct strike lighting where the phase of the aperture excitation depends Of the velocity of propagation of the direct strike along the aircraft. Basis For The Comparison. Comparison of the effects of HEMP and lightning on aircraft requires that a point in the interaction process be chosen such that That is, like quantities must be compared. incident fields compared with incident fields and surface currents compared with surface currents. In addition, the comparison should take place as close to the outside of the aircraft as possible to make the comparison as simple as possible. Both the surface magnetic field and electric fields should be compared since they drive apertures with approximately equal efficiency and the surface charge is a more effective driver for lightning. There are a number of simple interaction models that may be used for the comparison for simple geometric shapes that may be used to represent some of the important features of aircraft. The ones used here are shown in Table 1 and described in more detail in Ref. 2. TABLE 1. INTERACTION MODELS TO BE PRESENTED. Medium Slab Continuity Ellipsoid Sassman T-Line T-Line In this table, low frequency means subresonant or below about 1 MHz. Medium frequency means in the resonant or l-10 MHz region. Finally, high frequencies are those for which aircraft structural' details become important, or above 10 MHz. IV. COMPARISON In this section the actual comparisons will be presented along with the effect of the corona on the natural frequencies of the aircraft by the surrounding corona and attached streamer. Natural Frequencies and Evidence for Corona. In investigations of the F-106B data (Ref. 10) Trost and Turner have extracted the natural frequencies of the currents on the F-106 and models of it in configurations representative of both nearby and direct strikes. Attachment of wires and the direct 179 strike data show natural frequencies with much higher loss components than those of the bare aircraft described in more detail in Ref. 2. The importance of this shift in natural frequencies may be seen by examining the curves in Fig. 4. This figure shows a hypothetical transfer function from one exterior environment to an interior system. Two possible exterior environments are superimposed on the transfer function. The threat to the SYStern is found by multiplying the transfer funcFor example, tion by the threat environment. suppose environment A is HEMP and environment Because environment A has B is lightning. peaks at the same frequencies as the transfer function, and environment B has peaks at the minima of the the transfer function, environment A results in far higher currents at the For the choice of environinterior system. ments given above, HEMP would be the dominant The opposite conclusion, however, threat. would be drawn if environment A were lightning and environment B, HEMP. Since the a priori knowledge of the identity of the environment is not available the result is a region of uncertainty in the l-10 MHz region when the two threats are compared. -Transfer Function from Exterior to Interior Circuit ........ Exterior Environment A ---Exterior I Environment B I 32 - Fig. 5, it is clear that lightning dominates at low frequencies and HEMP dominates at high At intermediate frequencies frequencies. there is a region of uncertainty. 100 kA Lightning \/- Lightning /HEMP Fig. 5. Spectrum of H for two waveforms for direct strike lightning and for HEMP. The uncertainty envelopes show the effect of the resonant region. For more complex models, such as transmission line models (see Ref. 2 for details) the conclusions are the same for both magnetic field and normal electric fields as shown in Figs. 6 and 7. ;s; “\. . . ‘1.. *..... . . . . ‘A., . ... ... -4,. ‘, :.’ .. . _’ Fig. 4. 1 / / / / / <\ ,‘\?. . . .:.. :.. . . \ \\ :.: Y \ ‘. : \ \ \ Frequency (MHz) Transfer function products. Comparison of Simple Models. The first comparison of HEMP and lightning to be presented uses the slab model for HEMP and the continuity of current model for lightning. In Fig. 5, the two models are used to compare the surface magnetic field, on a 1 m radius cylinder, generated by HEMP and by direct strike lightning. HEMP is calculated from the incident waveform in Equation (2) with a peak field of 60 kV/m, as the saturation field. The surface magnetic field is then doubled to account for reflection. Both reasonable worst case and moderate threat lightning waveforms are used in the comparison. In i Frequency (Hz) ‘..* --.+ F5 Frequency (Hz) Fig. 6. Comparison of magnetic field intensity H for HEMP and lightning using the more sophisticated models. V. OPERATIONAL ENVIRONMENTS A comparison of HEMP and lightning is not complete without considering differences in - the operational environments for aircraft exposed to these two threats. Reported statistics indicate that 37 percent of reported lightning strikes result in some sort of precautionary landing or mission For lightning mishaps, that occur abort. rather infrequently, this level of mission For abort causes little operational impact. HEMP, during wartime, essentially all aircraft are exposed to HEMP for each high altitude If each exposure resulted nuclear explosion. in a 37 percent mission abort rate, as lightning statistics indicate for lightning (Refs. 11, 12), then almost a third of our bomber fleet could be wiped out by a single high altitude nuclear explosion. -I-’ :: .r- r ‘. ‘i ‘, . : ., t 1L R. L. Gardner, et. al., "Comparison of Published HEMP and Natural Lightning on the Surface of an Aircraft," Lightning Phenomenology Note 12, Air Force Weapons Laboratory, Kirtland Air Force Base, NM, 1984. 3. Uman, M. A., Lightning, McGraw-Hill, New York, 1969. 4. Golde, R. H., "Lightning Currents and Their Parameters," in Lightning, Vol. I, Physics of Lightning, R. H. Golde, ed., Academic Press, London, 1977. : : I ., ;; ‘\\ “\1 P :\ ,, :,: y \\ Fig. 7. 5. Garbagnati, E., et al., "Lightning Parameters--Results of 10 Years of Systematic Investigation in Italy" in Proceedings of the International Aerospace Conference on Lightning and Static Electricity, Oxford, England, 23-25 March 1982. 6. Melander, B. G., "Effects of Tower Characteristics on Lightning Arc Measurements," in International Aerospace and Ground Conference on Lightning and Static tlectrlclty, June 26-28, 1984 , 0 rlando, Florida. 7. Trost, T. F. and F. L. Pitts, "Analysis of Electromagnetic Fields on an F-106B Aircraft During Lightning Strikes,"Proceedings of International Aerospace Conference of Lightning and Static Electricity, St. Catherine's College, Oxford, Eii$%%T, 2325 March 1982. 8. Uman, M. A., D. K. McLain and E. P. Krider, "The Electromagnet/c Radiation from a Finite Antenna," --Amer. J. Phys., 2, 33-38, 1975. 9. Krider, E. P. and E. D, Weidman, "The Submicrosecond Structure of Liqhtninq Radiation Fields," in Proceedings-of the 8th International Aerospace and Ground Conference on Lightning and Static Electricity, DOT/FAA/CT-83/25, Federal Aviationdministration, Technical Center, Atlantic City Airport, NJ. 10. Turner, C. D. and T. F. Trost, "Laboratory Modeling of Aircraft-Lightning Interactions, Final Report," NASA Grant, NAG-l-28. 11. Rasch, N. O., M. S. Glynn, and J. A. Plumer, "Lightning Interaction with Commercial Air Carrier Type Aircraft," in International Aerospace and Ground Conference on Lightning and Static Electricity, June 1984. 12. Corbin, J. C., "Lightninq Interaction with Aircraft," in Proceedings of the 8th International Aerospace Conference on ‘Lightning and Static Electricity, DOT/ FAA/CT-83-25, Conference Publication. June 1983. : \ . \ Frequency (Hz) Comparison of the surface charge density for HEMP and lightning. VI. CONCLUSIONS For several methods of comparison of HEMP and lightning the conclusion is the same. Lightning dominates at frequencies below about 1 MHz. HEMP dominates above 10 MHz. Finally, in the region between 1 and 10 MHz the interaction between the two threats and aircraft is so complex that either threat may dominate. In addition, the two threats of direct strike lightning and HEMP interact with aircraft in fundamentally different ways requiring different hardening techniques. Finally, the operational impact of particular vulnerability levels for lightning and HEMP have very different effects on aircraft operations. In particular, only very low levels of failure probability may be tolerated for HEMP since the entire fleet may be exposed to the threat simultaneously. REFERENCES 1. 2. A\\ : : \ : : : 10-l - : . .:. __---_. . _--.a! “-< , * 2 2 180 EMP Interaction Principles: Techniques, and Reference Data, K. S. H. Lee, editor, ArFW Force Weapons Laboratory, Kirtland Air Force Base, NM, 1980. - 181 33~6 - NEMP and Lightning Protection Requirements for Modern Aircraft Equipment D. Jaeger, R. Rode Messerschmitt-Bclkow-Blohm Military Aircraft Division Ottobrunn, W. Germany 1. Introduction The objective of this paper is not to discuss final requirements concerning NEMP-hardening and lightning protection for aircraft equipment (only first drafts are available) but to point out the differences between these requirements. A simple aircraft coupling model is used to demonstrate the interaction with the NEMP- and lightning enviroment. The calculations present an imagination about the interference signals which might be expected and allow a comparison of the different types of threat. Only cable induced signals are taken into consideration here. The result is that important signals might be induced and that generally ligthning represents the higher threat. It will almost include NEMP hardening from the technical point of view if threat cases with lower probability are considered, too. The relation between ligthning and NEMP-requirements will become more adverse if modern a/c materials like Carbon Fiber Composite are used. 2. Threat to be considered 2.1 NEMP The usual NEMP is taken into consideration with an amplitude of 50 KV/m, a rise of about 5 ns (10% - 90%) and a pulse widts of about 600 ns (10% - 10%). It can be expressed by the formula: E(t)= 52.6.,03&*106t _,-wwo*t ) “,,,, The spectrum of the magnetic field is shown in fig. 1. 2.2 Ligthninq Two types of lightning strokes should be considered which differ in function of time, in amplitude, in area of interaction with the a/c and in probability of threat. It is the initial stroke and the return stroke. A) Initial stroke: It enters generally the a/c in zone I A and leaves in zone I B. (fig. 2). - The current is running through the structure. The initial stroke is characterized by a high amplitude but a low rise. Worst case values are in the order of 200 KA with a rise of 2 us. The pulse width can be about 100 us. This pulse can be expressed by the formula: 182 - 3. Simple a/c - coupling model A survey of the NEMP and lightning coupling meachnism down to equipment level is shown in fig. 4. The current density spectrum is presented in fig. 3. THREAT AIRCRAFT While the NEMP interacts with the a/c via field coupling the lightning stroke eEfects the a/c in the conductive way. (Near by lightning is neglected here because it produces generally more tan 30 dB lower results). The interfaces the a/c presents to the threat are the antennas and the structure. B) Return stroke It can follow the initial stroke and may enter the a/c anywhere in zone II A. It has a smaller amplitude but a higher rise than the initial stroke. The probability is lower than for the first stroke. Some worst case levels are in the order of 100 KA (amplitude), 0,5 us (rise time) and about 50 us (width). The pulse can be expressed by: The current density spectrum can be found in fig. 3. In addition another type of threat exists. Some years ago streamers have been detected by the NASA which have an amplitude of some KA and a rise of some 10 ns. They might arise everywhere at the a/c with or without an initial or return strike following. Because the spectrum in the higher frequency range is very similar to the spectrum of the return stroke, the streamer is not considered here. Because only the cable induced signals will be considered here the follwing ways of coupling are of interest: - field penetration via structure material (NEMP only) - field penetration via apertures (NEMP and Lightning) - currents induced outside on the structure (NEMP and Lightning) In the last case two types of currents can be distinguished: - the pulse itself - resonant signals which are exited by the pulse The currents can be coupled in via magnetic fields through the apertures and via current penetration. In the case of a metallic aircraft it can be demonstrated that mainly coupling via aperatures might be of interest. The conductive coupling does not produce a worth mentioning interference signal (fig. 5). - 183 33~6 - LEMP 4.2 Cable induced signals The signals depend on the H-field of the NEMP as well as on the H-fields produced by the current densities outside on the structure (NEMP and lightning). The following types of signals might appear at the interfaces of the equipment (fig. 9): 5. Discussion of some results Some results are presented in fig. 12 and 13 for a cable of 5 m length for the NEMP and the initial lightning stroke entering and leaving the aircraft in zone the pulse of the threat signal if the electrical length of the cable is short compared with the rise of the pulse (NEMP only) the outside structure resonant signals (NEMP and lightning) cable resonant signals if the elctric length of the cable is long compared with the rise of the pulse (NEMP and lightning). The different components can be calculated by using the approximation formulas of fig. 10 and 11. - 184 - of 0,5 uH/m. An effective area for the H-fields of 0,3 x O,l m2 is asumed to be typical. ___-________________.----- 4. Interaction of the aircraft model with the threat The signals induced in the cables depend on the outside H-field (NEMP) as well as on the H-fields produced by the current densities outside on the structure. -140 ‘102 10’ ld 105 106 10' f/HI 4 108 F IG.5 : RELATION BETWEEN CURRENT ON STRUCTURE OUTSIOE AND CABLE INSIDE Therefore the coupling mechanism can be simplified to the model of fig. 6. IIf NEM q LEMP . It can be realized for a typical aircraft by fig. 7 4.1 Currents -p on the structure -_ outside In the case of a NEMP the pulse has a very short rise time. Current densities will appear on the structure outside which correspond to a resonant current of about 2000A. This can be calculated by using approximation methods or formulas of the antenna theory. The resonant frequency will be about 9 MHz in the case of the cylindre, the signal will decrease with a Q of about 4,5. For a real aircraft structure resonant frequencies up to 5 times higher might be of interest. In the case of the initial lightning stroke entering and leaving the aircraft in zone I a current pulse of 200 KA will flow along the structure. Although the pulse is slow compared with the NEMP it might also cause resonant effects. A maximum current of about 300A may appear. It can be calculated with approximation methods on the base of fig. 3. In the case of the return stroke and the streamer calculations are more difficult. Similar results should be expected. The data ob- -I I I D The aircraft shall be symbolized by a metallic cylinder with an aperture. It shall represent a limited cockpit attennuation, a lot of slots or an electrically unsealed access door. There shall not be a typical resonant frequency of this aperture. A cable shall run behind this aperture with a lengths of about 0.1 up to 3 times of the a/c length. It may have an inductance tained for the lightning strokes are first approximations. There are indications and some practical results in literature that resonant frequencies will change to lower frequencies if an aircraft is hit by a lightning stroke. A factor of 2 seems to possible. Although sufficient data are not available, this effect seems to be predominant for the large initial stroke, of smaller interest for the return stroke and almost neglectible for streamers which have a similar spectrum in the upper frequency range. Some typical functions of time are presented for the currents on the structure in fig. 8. - 185 33~6 - While the ligtning stroke produces very high low frequency pulses and only small resonant signals in the case of the NEMP only the resonant signals exist but with higher amplitudes. If other frequency bands (cable lengths) are considered, too, the differences will increase (fig. 14). Modern aircraft get more and more dependent on electronic equipment. A modern flight control system for example requires a reliability of maximum 1 fault per 1.000.000 hours. This means effects like lightning are getting more and more significant and also threat situations have to be considered with lower probability than the initial zone I stroke, that means the return stroke. This ligthning stroke has a lower amplitude but a higher rise and might hit the aircraft anywere in zone IIA, also close to the open area which is responsible for the coupled in signals. If the ligthning stroke hits in 1 m distance resonant currents like shown in fig. 15 might appear (additionally the low frequency pulse!). If the threat case of a hit direct at the open area is considered (lower probability again) the amplitudes of the resonant currents might look similar to fig. 16. Although the ligthning stroke produces a very high low frequency pulse the resonant currents can be higher up to some MHz than the currents caused by the NEMP if cases with smaller probability are considered, too. For frequencies greater than some MHz the NEMP induced resonant currents increase but not significantly. 6. Influence materials- of new aircraft New aircraft are planned to be built more and more of Carbon Fiber Composite (CFC) which is an electrically poor conducting material (about 3000 times less than aluminium). In this case the simple coupling model of fig. 6 cannot be used anymore. In addition conducting coupling has to be considered (fig 5) and in some cases field penetration from the outside, too. The effect is that lightning protection and NEMPhardening problems generally will grow up but in different scales. Because the low pass filter characteristic of the structure ligthning will increase more than the NEMP-effects. - 187 34F7 - PREDICTION OF LIGHTNING-INDUCED INTERFERENCE VOLTAGES ON THE BASIS OF MEASUREMENTS TAKEN IN SIMILAR INSTALLATIONS _ R. Terzer KWU, Erlangen P. Kronauer BBC, Mannheim F. Pigler Siemens, Erlangen Federal Republic of Germany This paper describes the usual procedure in Germany for predicting the expected lightning-induced interference vOltages, based on simulation measurements be suitable for determining the Voltages induced by defined design strikes at specific striking points over given cable routes. in existing installations. 2. 1. Simulation measurements Definition of the problem In some installations with high safety requirements it is often necessary to Measurement of such induced voltages with natural lightning strikes is not economical on account of their rarity; provide evidence, before construction begins, that the safety systems are protected from the effects of lightning even measurements based on triggered lightning are very involved. With triggered lightning the striking point strikes. Nuclear power plants, for example, are such installations, The evidence must be presented with such methods that it can be easily reconstructed and accepted by the experts. could be specified, the problem of conversion to the design strikes would still exist. In 1974 we therefore attempted for the first time to simulate a lightning strike by capacitor discharge, With extensive and complex installations, such as nuclear power plants, a purely mathematical method of determining the electrical stress would be difficult, even with greatly simplified as- On the basis of the model concept, i.e. that after a lightning strike the lightning current propagates progressively, radially and uniformly in the sumptions, and would be controversial, environment and discharges the image To be able to predict with sufficient charges present there as a result of the influence, we set up a spiderlike system of return lines, extending from reliability the expected electrical stress in the instrumentation and control system, caused by lightning strikes, the only method which appears possible to us is that of measuring the induced voltages in similar structures and applying the results to the structure to be assessed. Such a method must the striking point, for the first measurements. This system extends for a sufficient distance over the area to be examined, so that the current distribution in the structures and earthing system approximates that of a real lightning strike (Fig. I). - 188 - tes, a Fourier analysis can be used to some extent for conversion. Clearly, such measurements are useful only if the installation of the external earthing system and of the internal potential equalization system has been completed, and practically all the cables have been laid. At this stage, however, commissioning of the power plant is already fully in progress and, as a result of operational currents and of switching operations,, a considerable level of interference exists, Simulation of lightning strike Lightning simulation measure- at this stage must therefore result in induced voltages considerably in excess of the operational interference level, if the results are to be reli- ment in a nuclear power plant able. The shielding of the structures, 200m Fig. 1 d 148m b which has been increasingly improved, Design lightning strikes are defined for the design of nuclear power plants. A proposal is given in Table 1. If the simulation current form corresponds to that of the design strike, extrapolation becomes simple. If the form devia- Positive lightning (in buildings over IOOm high) has resulted in attenuation values which are SO great that the amount of apparatus required for such a measurement is no longer a practical proposition, Peak current i kA Max.rate-of-rise of current Rise time di/dt kA/ys 15 ps 50 ps 500 i di/dt kA 100 kA/ps 100 *I T2 i di/dt ps ps 5 200 kA 50 200 Decay half-time Negative first Peak current lightning Max.rate-fo-rise of current Rise time Decay half-time Negative secondary lightning Peak current Max.rate-of-rise of current Rise time Decay half-time T1 T2 T1 T2 kA/ps cls ps 500 0,25 150 Table 1 Proposed lightning parameters to be used when designing lightning protection for nuclear power stations. - 189 34F7 - striking point was defined by the ex- A discussion relating to the safety concept for German pressurized-water pert. nuclear power plants led to the conclusion that even with all eXter%al influences, the intact emergency feed water system maintains the reactor in a safe condition. Discussions with the For this reason, we have recently examined only this connecting duct with the connected structures for the measurements, and arranged the injection experts then showed that for the light- point for the surge generator at the ning protection of a power plant, evidence of reliability must chiefly be provided for this system. Measurements taken have shown that most of the voltages induced in the connections between the reactor building and the emergency feed water supply buil- point on the smaller building selected ding are induced in the connecting duct. ding and ducts, efforts must be made On account of the greater current concentration, the induced voltage is higher when the lightning with the same lightning characteristics strikes the smaller building (Fig. 2). by the expert. The length of this connecting duct is approximately 30 m. To be able to obtain with sufficient reliability a measurable induced Voltage, in spite of the usual lightning protection systems installed in the builto achieve a current flow of approximately 1 kA via the duct. With other types of nuclear power plant, there are also a number of systems which are required to maintain the reactor in a safe condition. Duct lengths of up to approximately 100 m are employed in such cases. already mentioned, the intention was that the waveform in the simulaAS tion should approximate that of the design lightning strike as closely as possible. The following holds approximately for the wave front with small supply networks: Where L is the input inductance of the circuit, R is the ohmic input resistance. In order to obtain an adequate meaFig. 2 Worst-case striking point for safety system of pressurized water reactors In German nuclear power plants the sheltering effect to the higher structures has not been taken into account. Although the likelihood of a strike on this smaller building is lower, this surement reaction, spiderlike return lines were no longer used to return the current to the surge system in the latest experiments; instead the current was concentrated on the duct to be measured by using a ribbon line as the return line, This ribbon line consisted of 6 wires spaced at about 1 metre and at a height of 5 metres, which followed the course of the duct, - thus resulting in a uniform field and current distribution on the Surface of the earth, over the entire width and length of the duct. This ribbon return line exhibits a characteristic impedance which is approximately Constant Over the entire length. In order to avoid oscillations, particularly at the wave front, the ribbon line has to be terminated with the characteristic impedance. For the last measurement with a duct length of 100 m, this terminating resistance was 150 ohms. A voltage of 150 kV is developed across this terminting resistance with a measuring current of 1 kA. For these experiments, therefore, we used the largest transportable surge generator available to us? with a charging voltage of 250 kv. (Fig. 3) 190 - During a real lightning strike, the current is distributed radially in all directions from the striking point; only a part of the entire lightning current flows via the duct in which the measured connections are sitUat.ed. In the simulation measurement with the ribbon return line, the entire SimUlation current flows via the duct and some adjacent earthing conductors. TO be able to extrapolate the results of the simulation, therefore, that portion of the current must be determined which flows vie the duct in question during a natural lightning strike. Several methods can be used, some more involved than others, with different physical mode concepts, which nevertheless provide very similar results for our specific case of a nuclear power plant. Comparative measurements with radial return lines from the entire perimeter, and a return line in the form of a ribbon line over the duct, have confirmed these calculation. The voltages to be expected the event of a design strike at the specified striking point are given by: in Fig. 3 Lightning simulation with ribbon return cable Where UK IK The following f'iguresshow oscillograms of the injected surge current, and the uM voltage induced in the cable (Fig. 4). + 143kA is the portion of design, lightning current flowing via the duct, is the voltage measured dur- ing simulation, is the current flowing via 'Sim the duct during simulation. 3. P is the expected voltage, Predicting the induced voltage for other installations L Mith the simulation measurements, the ---I I+-!,!-+ hh=O.Bps Injected surge current Fig. 4 Measured induced voltage Oszillograms from the lightning simulation arranged like Fig.3 correctness of the chosen concept can only be demonstrated on existing installations. A prediction for other power plants can be made if they are similar. - implies not In this context, flsimilarll only the topographic arrangement of the structures, but the similarity in the construction of the buildings in regard to lightning conductors, similarity of the ducts in regard to the connection of lightning current-carrying conductors over the length of the duct and over the isolating gaps, as well as similarity of the duct connections to the lightning conductors of the structures. We have established such a similarity by means of detailed specifications for erecting the structures (Fig. 5). 191 - 34F7 relatively small differences in arrangement between power plants built by ourselves, the error is not great and is on the safe side, We have SPeCified the lower limit of length LK as 30 m. This corresponds to the length for some simulation measurements and is the usual length in the nuclear power plants currently under construction. In the simulation measurement, the voltage is measured in a cable which leads from one structure to the other via the duct. The voltage is not only induced in the duct, but also in both connected buildings. In a conversion proportional to the length, this influence at the ends is also converted correspondingly and results in too low values for shorter lengths. In the case of lengths, which are greater than the measured lengths, the portion of the voltage induced in the buildings is also extrapolated proportionally. In practice, therefore, with Fig. 5 Shielding measures for buildings and connecting ducts For such a similar power plant, the portion of current of the design strike flowing via the duct can be determined in the same way as for the extrapolation of the simulation calculation. The expected voltage UK is then given by: uK q lK 'MxlSimXrG LK Where LK is the length of the duct, LSim is the length of the duct for the simulation. ducts which are both shorter and longer than 30 m the voltage will be lower than that given by the calculation, which complies with the requirement for evidence of safety in plants. power The upper limit of length LK applies where a further lengthening does not result in a voltage increase, For this we have the following physical model concept. The change in field strength during a lightning strike propagates at the velocity of light C. With a mainly inductive coupling, the discharge current induces a square-wave voltage during the linear rise at the wave front. This induced voltage travels in the cable at the velocity of propagation V, which is governed by Here we have assumed that the voltage is proportionally dependent on the length of the duct, for a limited area. This is not quite compatible with the concept of radial distribution of the lightning current. However, with the the dielectric constant of the cable. - Induced current changes and induced voltage travel at different velocities. The maximum voltage is reached when the trailing wave front of the induced current changes reaches the leading wave front of the voltage induced at the striking point. The length Lmax is given by: 192 - carrying capacity for control and instrumentation systems - SO limit applies to this case that this also. In our opinion, the method presented here is the only one capable of providing very reliable predictions relating to the maximum voltages enCOUn- tered during the effects of lightning L max Where C V % = ts is the velocity of light, is the velocity of propagation in the cable, is the virtual duration of in these important, safety-related circuits. Although the proposed design strikes contain characteristic values which, when combined, extend far beyond the strikes observed so far, and although we are on the safe side with the wave front. the specifications presented here for x:&-!?-$ With E = 4 this results in a length of 75 m for a negative secondary lightning strike with a virtual duration of the wave front of 0,25 psec., and a length of 300 m for a negative initial lightning strike with a virtual duration of the wave front of 1 psec. Duct lengths of 300 m are very rarely employed, so that the upper limit for duct connections is insignificant. However, this length can be exceeded by connections using underground cables - which we install with a shield with current- the conversion of the simulation measurement to apply it to a planned installation, a safety factor will be specified in view of the high safety requirements for nuclear power plants. The permissible electrical stress of the electronic components must be above the voltage stress thus determined by the amount of this safety factor. We therefore believe we can state with certainty that the effects of lightning on the safety systems of nuclear power plants can be ruled out in the future, just as in the past, References /I/ Pigler, F., Terzer, R.: Blitzschutz in Kernkraftwerken. Siemens-Energietechnik 3(1981) No. 10, pp. 336-339 /2/ Neuhaus, H., Pigler, F.: Blitzkennwerte als Grundlage der Bemessung von BlitzschutzmaAnahmen etc, Vol. 103(1982) No. 9, pp. 463-467 - IMPULSE CURRENT 193 AND VOLTAGE PROPAGATION TELECOMMUNICATION H. “Friedrich List” Schuppler University GDR - 1. Impulse , D. of 8010 Introduction currents impressed, for upon buriod telecommunicain case of lightning tion cables discharges to ground must be dotermined with regard to the propagation characteristics of a linear buried metallic conductor. up to now impulse current propagation wae determined by means of the system response to a unit step [Z]. By application of the convolution theorem or the Duhamel integral the determination of propagation conditions leads to an extensive numerical integration on the basis of power series expansion. The system response to a current, impact having a waveform of 5/65~s only is available in tabular form in the CCITT Manuel on protection against [4 . In order to lightning discharges 1,y lightning predict effects caused currents flowing along cable sheaths the voltage occuring between cable and the cable sheath must be pairs calculated. The usucl method of calculation to do so uses the sheath current as input parameter. Therefore, the difficulties concerning the calculation of the sheath current as mentioned above are encountered in calculating, the pair-to-sheath voltage, the dopence of which on the distance from the lightning current intake location is of particular interest. Gcncrally two different caees are to be considered : - no breakdown occured between cable pairs and cable sheath - there is a breakdown between cable pairs and cable sheath at the point of lightning current impact. As rcportod earlier [I] the lightning current striking a cable sheath can be described approximately by menns of exponential functions. In moet of the practical cases it can be given in a double exponential expression. In order to develop a comprehensive instanC@, 35~8 - IN UNDERGROUND CABLES H. Ristau, Transport Lorke and Communications Dresden method of calculation for the sheath current and the pair-to-sheath voltage the impulse response of Sunde [Z] has been transformed into the Laplace By application of the tabuladomein. ted complex Batoman function of the = w(x+jy) complex error function w(z) and by observing the Cauchy/Riemann differential equations, length and time dependent solutions could be obtained as shown as follows. 2. 2.1. the propagation Calculatin of impu !! se current Current of time and and voltage location as functions As shown already formerly [ 1 a impulsive current can be descri IYed by a series of exponential functions. In most of the really occuring cases (on condition that Th/ T, > 2.65) it ie sufficient to suppose a double exponential expression. So, as you know, we can describe the lightning current i(t) = I”( ewa’- es”‘) This function has got the advantage of being easy transformed, what is important for the calculation of the pulsed current propagation along a buried cable. For the step response of a linear buried core being known [ 21 , a solution should Duhamel out to of the problem of propagation be formed by means of the integral. However, this turned be practically impossible. For this reason the pulse propagation has been calculated by means of the Laplaca transformation. The method of ;;13;;;ation function i_(t) is as : - CaLculation of the impulse response P(s,t) by the step response E(s,t) bY - 194 - (a) P(s,tl - = Erfc( ConvoLution and exitation of P(s,t) - Laplacs convolution L(P(s, - (2) T$E(s,t) The step response ding to Sunde E(s,t) - = results (4) of (b) In the case occuring at intake the ’ JL {i(t){ (5) = I’R&$ u(s,t) of no broakdown the point of current -R,C,jl Determination of the response function by retransformation = I---’ {F(p)} (6) By applying this method of calculation to a current flow according to equation (l), we get for the current at any point of the cable sheath i(s,t) L”e in = -- ~a[~bhX ,&I - ~&,&=dj (7) - which f3li - series cable PII - resistance circuit Cl - capacity circuit I permeability earth resistance. 9 - specific Correspondingly we can compute the voltages between cable conductors and cable sheath [S] : resistancs sheath of of of the the tha sheath-wire sheath-wire and which (8) P (10) response * i (tl transformation function i (s,t) in I * (3) impulse function at accor- svc) t 14 it)] = L(P(s$l) = F(p) In the case of a breakdown the point of current intake 2.2. = sm (32) Change of waveform Observing a lightning current of the waveform 5/65 US , as it is dona by the authors of the CCITT handbook on Protection against lightning discharges [ 41 results in current paths inside the cable sheath demonatrared In fig. I. Figures 2 and 3 demonstrate the paths of the voltagea between sheath and cable wires. The change of the waveform results in a variation of the frequency spectrum, Figures 4 to 6 illustrate the translation to lower frequencies. - 195 35 - F8 1.0J t f’ t s=Om 0.8 0.Ej- / P++= 1 l?/km ~=I000R m f too i-3 80 6C 0.4 ,- 4c ). 0.; 2( / 1 Fig. t A& I 1 Fig. : Current inside the cable sheath at a waveform of the generating lightning current of 5/65 rs RH=1 R/km L--l s=Om t 3 between sheath and : Voltage wire at a waveform of the generating lightning current of 5 65 pus, in the case of breacdown at the point of f the lightning stroke 4 : Frequency spectra of a waveform 5/65 rs lightning current and the cable sheath current generated by it at distances of 100 and 1000 m :.o 2 lmax 0.6 0.4 Fig. 2 between sheath and : Voltage wire at a waveform of the generating lightning current of 5/65 E/S, without any breakdown at the point of the lightning stroke Fiq. - 196 - 2.3. t GE Waveforms recommended by CCITT For testing telecommunication equipment to be connected to cables CCITT recommends to epply impulse voltages of a waveform lo/700 p s [5] . At figure 2 we can see half-peak times of a few 100 ps occuring along cables hundreds of metres away from the point of the lightning stroke. The longer half-peak times at larger distances are connected to much smaller amplitudes. The decrease of the amplitudes is demonstrated at figure 7. 0.11- 300, I0 k.m -wave form 0.8 5165pS VlkL! 2. 0.1 200 i 16 Fig. 5 : Frequency spectra of the voltages between sheath and by s wavewires , generated form 5/65 ys lightning curwithout any breakdown rent, at rho point of the lightning stroke 100 brcclkdown without breakdown Ic 1 0 Fig. 250 7 560 10b0rll s_ 750 : Amplitude of the voltage appearing in connection with the lightning current between wires and sheath in relation to the distance from the lightning stroke (waveform of the lightning current 5/65 /us) Figure 3 illustrates the impulse voltage between wires end sheath having rise times of a few 10 ,us in the near of the point of the lightning stroke. By application of the method for calculating the propagation of impulse voltages on cables described in [l] , the analysis demonstrated in paper clearly shows that the for testing impulse voltages supposed b the recommendation the CCITT Y 51 has been fixed it imitates divices Fig. 6 : Frequency spectra of the voltages between sheath and generated by a wavewires, form 5/65 ys lightning cur- rent, in down at lighrning the case rho point stroke of break- of the realistic connected conditions to cables. this waveform 10/700 t~s I<17 of well, for for - 3. Critical of the 197 35 - T, appreciation results The illustrated above results are based upon the assumption the lightning current having a waveform of the method of calcu5/65 p s . However, letion used and the computer programs applied for determining the volues of the diagrams permit on the other hand any other waveform to take as a basis which may bo outlined by a double exponential expression (cf. equation (1) ). As a matter of principle, the method of calculation is applicable to any wavef 0 rm . There is some reason to give a hint 60 [6] which at the IEC - Publication defines the waveform by giving parameters on the virtual front time and the virtual time to half value. For practical calculations, the CCITT Manual on protection against lightning discharges [4] defines the waveform stating the parameters true front time and the true time to half value. In this contribution the point of view of the CCITT Manual was applied. That is the reason why the waveform of 5/65 /JS usod in this papor slightly deviates from a waveform of 5/65 ps of the lightning current according to IEC. Figure 8 illustrates the difference which, however, turns out consideto be insignificant to our ration. - virtual front Tz - virtual time T, - true front T/-, - true time 4. Computer BLITZ 3 Calculation current value available (Lightning of the 3) lightning References Beyer,D.; Lorke, H.; Schfppler, H.: Calculation of lightning effects on communication cables. 5th Wroclaw Symposium on EMC, 1980 in Toronto, Beyer, D.; Gotre, S.; Schtippler,H.: Blitzstromausbreitung und resultierende StoBspannungen auf unterirdisch gelegten Fernmeldekabeln. Fernmeldetechnik 22(1982)5 - [41 CCITT : Protection of telecommunication lines and equipment against lightning discharges. Geneva, 1974 151 CCITT : Yellow Book, Vol. IX, against interference, dation of the I< series. Geneva, 1981 - : Defination programs E. D.: Earth conduction effects transmission systems. Norstrand camp. New York, London, 1949 T 8 half BLITZ 4 (Lightning 4) Calculation of the wires-to-sheath voltage at the point of the lightning stroke t i Fig. to BLITZ 2 (Lightning 2) Calculation of the wires-to-sheath impulse voltage under the condition of a breakdown at the point of the lightning stroke 0.Y TM value time [21 Sunde, 7 L half BLITZ 2 (Lightning 1) Calculation of the sheath impulse current and the wire-to-sheath impulse voltage without breakdown ;;,;k; point of the lightning [il 0.5 time to By application of an algorithm for calculating the complex error function by Gautschi are found out the following computer programs by means of which we obtained the values of before: the diagrams * 5. 1 F8 [61 1% Publication 60-2 High voltage test techniques, 2: Test procedures. Geneva, 1978 [71 Gautschi, W, : Algorithmus 363 functions. CACM 12( 1969) z of time used for impulse current the parameter describing an Protection Recommen- - complex Part error - 199 36 - Gl THE HAZARD OF EIJ!XTROMAGNETIC RADIATION AND DISCUSSION OF SAFETY THRESHOLDS Q. Chen R.C. Husng B.C. Pan China Aviation Research Institute for Standardization Beijing, China It is also difficult to unify a quanti- This paper presents fIInOtd.Ond rdatiOnS Of electromagneticradiation pouer density (PD) wwsus LD~ of animals and ED50 by means of analyslng experimental data on the effect8 0. caused by the change in intensity of electromagnetic radiation within UHF band to animals. The Hazard of micro electromagnetic d. The measurement error of the electromagnetic wave is large,when the measurement radiation on personnel is analysed by means of safety large amount of statistics and the threshold for hazard of micro-wave to human body is also discussed. General It is well known that the electromagnetic radiation is harmful to living things. Upon this conclusionmany countrieshave established their own safety standards.Expertsadmit that the harmful effect on living things related to tative thresholdlevel, because differences exist betueen the individualsof living things. is taken in cell shielded by metal nets or plates, its reproducibility is poor. Based on data of PD-LDg,@dpD-Elf;0 eeent endregression equaticnderivmd Mm thesedata,this paper specifies the lower limit of PD, and presents acceptable safety threshold,in which a safety factor of 20 dB is considered. PD-LDw experimentand test results Block diagram of the test set-up the amount of electromagneticpower absorbed, exposure time, wave length, wave form and polarization mode of radiation,though they hold different views on how the electromagneticradiation can be harmful to living things. Up to now the safety values specifiedin the standards established by many nations are not uniform, some exceeds the other by one thousand times. Therefore this question is worth studying.The technloal reasions for the large difference are as the following: a. The limit values of the safety boundary are generally wide, because they are calculated according to the heating effect of electromagnetic radiation and the balance of human body heat exchange. b. The values are somewhat strict, nhen the radiation threshold level uhioh may cause human body with functionaltrouble serves as the safety bounaary. It is also difficult to unify the criteria of the functional trouble of human body and therefore the thresholdlevel. fl8 1 Block Diagram of Tast Sd~up Data of test Test frequency2pulse wave (PW) 3 CHz, oontinuous wave (CW) 2480 MHz Test waveform: pulse wave (TElO) pulse width z = 1.6,~s cycling period T = 2500,~s peak power output 300 KW Test sample: For PW, 250 rats of whioh one half is male, one half famsle. weight: 20&3Og; For CW, 96 rata, taking double standard error (0.3OC) in anal temperatures of 30 sound and adult rats as deoision criterion of EDw. Test results The test results are shown in table 1. DRtA sheet Tab10 1 PoYer 1 N”lTbfJr nu/cn2 --- Hve PD-uJw test Acoumulated number I--- density 1 of dea alive dead total 9 176 7 27 8 2117 3 20 13.1 39oC to 43'C; The anal temperaturesof the dead ranged from 39.5OC to 48Oc. The difference between the two was significant. The dead animals had blood statis, bleed.and edema in lungs, and bleed at endocardiumand epicardium,and digestive tract, and blood statis in meninges. The testis of the alive animals had apparent pathological changes,withnicrosis of genital ephitheliumoccured and seminiferoustubule damaged. Empty bubble can be observed in conical cells,afterNile body of cerebrum was stained. PD-ED~ experimentand test results Experimentation Test frequency and wave form: same as those for PD-LD~o test Test sample: 84 sound rats. The male and famale one half each. weight: lx)-3oOg First were the average ana.3.temperatureof rats under normal condition taken and double standard error computed: 10.0 10.5 11.0 11.5 12.5 total 7.5 8.5 9.0 9.5 10.0 tokl Effective anal temperaturemeasured after test To- Anal temperaturebefore test Decision criteria are: It is effective that anal temperatureis equal or higher than T after radiated by microwave, and ineffectgvethat temperatureis lower than To. where 50 0.00 0.04ca 5.5 16.67 O.OQTI, 6.0 46.67 0.0166 6.5 88.24 0.0062 7.0 IOO.ca o.cab, total 0.0705 3.0 4.34 O.M118 3.25 27.78 0.0112 3.75 61.11 0.0132 4.0 85.71 0.0058 4.5 c0.w tutal 6 *o o ',2 3 -5 o *og o.wOO 0.0319 24.1 0.17 5.9 0.09 The behavior and morbid state of animals under test T - Test results No appreciablechange in behavior of the animals was found,noneof them died throughout the test.Theirweight, blood picture (leucocytes, thrombocytes)did not change.A few ofthem had light blood statis in lung and intestine. Test data are shown in table 3. Tsblc 3 Data ohect of PD-ED5ott?nt s I’:D5() When the animals exposed to the radiation 0.ftest intensity, at the beginning their actions of face-washing increased, then they were agitated and were short of breath, sweating at the 'bellies,four limbs became hectic, finelly they moved restlessly and violently, struggledand fel3.down in a tic and spasm,and died. The weight of all the animals after test differed from that before test. Heating effect was apparent.Thedecreasingrate of weight related to the power density. The higher the radiating power, the faster the decreasing rate of the weight.(SeeTable 2) Table 2 Vartst,on of velght under oxposur* Of different povcr density r , POWW density Exposure time Mean value of weight drop DeCrelLSillg rate of weight aYJe.? (Inin) (9) (R/m) 17.' 22 6.2 0.28 1.8 0.40 0.30--I 173.7 120." 4.5 7 2.1 The temperaturebefore and after testvaried. For example, the temperatures taken in anus erase on an average by 4.4oc.The anal temperatures of the existing alive animals ranged &m (min) 3.9 0.10 8.0 0.11 10.3 0.14 (1. 5 15.2 0.14 0.05 20.1 0.14 PU 0.0?5 CW (min) 14 1.4 ---_ 9.2 20.1 _-- O.Oq 0.10 -I Approximateregressionequation of PD.-LD% Assumption a. When PD is small enough and a specified value, animal and personnel will be safe under radiation of a long period. b. Only the effects of electromagneticenergy and exposure time shall be consid ered without regard to other factors. c. The safety threshold for the animal and personnel shall be proximately the same, comform to the L.O. Hoeft model theory when they are radiated for a long time. - 201 The regressionequation of PbLD50 If the functional relation of the test data is: .. . 1 y;ae+ (a>0 1 y - Power density (PD), in mW/cm2 where x - Exposure time, in min 36 - Return formula 2 to 1: B=blge A=lga a=IOA=l2.34418793 “*= Assuming that: 10.416 1.12.34418 eX l .. 6 The formula 6 describes approximately the functionalcorrelationbetween PD-LD50. Y=lgy Estimationof the interval of test data Therefore: Y=Lga+bI.g X Assuming that: A=lga B=blge Therefore y=AtBX Where B= 10.41633733 b ‘ii: y=ae x= -+ Gl $(X;-X)(yi-y) gxi- X,’ a. Residual standard error =0.20164 ... 2 2s=O.403295 ... 3 A=f?-_Bx The formula 2 is linear regressionequation. The parametersin formula 3,4 are given in tabI. 4, b. The regressive straight Line in interval. with reliabilityof 95.6% Upper limit line y"=AtBXt2S =1.4947574+4.5237578x Lower limit line yl=AtBX-2S =0.6881674t4.5237578X The regressive straight line in interval with reliabilityof 68.3% Upper limit line y"=AtBXtS =1.2931024+4.5237578X Lower limit line y'=AtBX-S =0.8898224t4.5237578x The! regression equation of PGED5C and its linear correlationcheck Assuming that the fuctional relation between the test data is With the equations solved, B=4.523'757816 A=1.091462425 Therefore formula 2 becomes Y=1.091462425+4.523757816 x y=aebX (a>C) where Y_- Power density, in mW/cm2 x-- exposure time, kn min Assumfng that x=x Y=lgy therefore yd.gatblgeX=AtBX where A=lga B=blge With the formula 8 solved ... 7 ... 8 ... 5 Check of the linear correlation significanceof the regression equation Significancecheck shall be done to determine whether it has been correctly equated. e f(Xi-Z,(Yi-7) ;='N &(X-R)’ = -22.3490628 = - 0.140471796 159.1 A= Y-BX = 1.63480769 a= 1OA = 43.1328042 b=B = -0.323448264 Ige _-. __ =- 0153284566 0.027629/50 =0.?22/77673 When the confidencelevel o( = 0.05 y=aebx=43.1328042e~0.323448264X ... 9 The formula 9 describes approximately the functional correlationof PD-ED50. Linear correlationcheck: I-critical=0.878 Y>'Tcritical Formula 5 is meaningful,Y and X are linear correlatedwith a reliabilityof 95% y= +(xi-R.KY-P) J$tXi-%$ &(Yi-yY x -22.3490628 = - 0.979835 22.8090 202 - level when confidence Ycritical D<= 0.05 = 0.878 IYl B r critical Therefore formula 8 is lfnear Estimation of interval: = + correlative, =: 0.12288 ZS= 0.24576 The regressive straight line in interval with reI.iabiI.ity of 95.6% Upper limit yw=A+BX+2S=1.88056769+0.140471796X Lower limit y’=A+Bx-2S=1.38904769+0.140471796x The regressive straight line in interval with reliability of 68.3%: Upper limit y”=A+BX+%=1.757~769+0.140471796x Lower limit y’=A+BX-S=1.51192769t0.14047179611 DiSCUSSiOn 1. Prom formula 6 and Pig. 6.1 we know that a, when x-+00 . Coefficent a characterizes Y the minimumthreshold of LD * 8 PD. According to the safety exposure mode;io developed by L.D. Hoeft,the safety threshold for animal and that for personnel are about the 8ame, provided the exposure time lasts Long enough. Therefore the value of a i8 the approximate safety thre8hoI.d for personnel. Fig. 6.1 2. In consideration of safety, however, internationally 20 dB 8re added a8 safety aLlouance. With the difference between animal and human being and measurement error considered and a safety allowance of 20 dB inaluded, the standard threshold level may be specified as: awnit I & 10 That is to Say, shaIl not be iation RM/C& = 1.23 (mW/cm2> the maximum amount of rab aIlowed tc exceed 1.23 3. The characteristic of PI)-LD varies the form of a exponenti& funat 10 on, When radiated at level. a and taken 20 dB safety factor into aoount the exposure time must be shorcomputed tened. the correspending time can be from formula I (see Mg. 6.1) . Exposure time including 20 dB safety factor: in X” &j=s . = 4.52 (min) i.e. exposure time shall. not exceed 4.52 minutes, when radiated at 1.23 mW/cm2. 4. We can know from PD-LD% test and formula 9 that ED was too long to measure, when PD i8 less t ? an O.O25mW/cm2. It can be seen that the safety threshold of radiating power density shall be 0.025 mW/cm2or less,if the anal temperature was taken a8 dicision criteria. 5. The results of PD-LD50 and PD-ED0 test8 indicate that LD5 end ED5 for the p se wave are one time shor ? er than 8 hose for the J continuous wave with similar poner density. The Hazard of the pulse wave is much severer than that of continuous wave. Patholog%e statistical ena.I-ysis of the microwave radiation hazard General The large amount of statistic analyses on pathological. state of PsrsonneIs who are exposing to microwave fox a long period are meet important for analysing the exisiting hazard of microwave on buman body. Such analyses are very difficult and must be done carefully. The total. study cases are 423, 179 thereof sre male worker8 in field of pulse wave,having a working time of 3 years or more.244 are workers in contact with continuous wave, in which male wrkers account for 60%, having a working time of 9 years or more. The working frequency ranges involved in statistic8 are 2-9 GRz and 140-160 MHz, the power density around the working area is under1QO~W/cm2. The control. group consists of 189 persomels who are not exposing to microwave but they are similar to the uorkers above in respect of wcp king condition, age and sex. The control group for the pu.ISe wave consists of 109 cases, the control group for the continious wave consists of 80 cases. The items of stati.stc8 incIude the nervous system, cardiovascular system, digestive system, crystrilline lens, end hematological examination, Details are described below: The history of exposure to mfcrovave and chief complaints. Reart, lung, liver, spleen. Rest sitting position blood pressures taken on the left and right arms. Etlectro-encephalogram (BEG): In rest sitting position and eyes being closed, the time occupied by the frequencies of each vertexoccipito bi-lead wave of left side and integrated value of amplitude of each wave were recorded by both of B-lead electro-cerebrograph and magnetic recorder. The frequency spectrum of electro-cerebrogxam was anslyaed by computer, CIinicaI electro-cardiogram (ECG);To record rest-sitting position IZ-lead electro-cardiogram* Electra-cardiogram of high rate: Record 12Lead cIinicaI, electro-cardiogram at high rate with a magnetic recorder, then play back at low rate to observe the number of high frequency spike of QRS complex wave on the alectro-card%ogr%Dl. R-R interval record:Connecting cardiotachometer to teletype tc record the intervals between heart beats.It was recorded for 3 minutes each in squating and erect position. The heart rate variability Was tlnalysed by computer. CryataIIine lens checkt After rapid mydriasis by 2.5% neophrine, the opacity of crystalline 36~1 - 203 - lcms of two eyes was checked up with split-lamp. me cases of congenital,traumatic and aged lenticularopaaity was excluded. Routine complete blood count: bloodplatelet, white blood cell, red blood cell, end hemoglobin. The results statistical The results of analyses of path&&$!d. examination are shown in table 5 through table 13. In these tables pulse wave continnouswave as is referred to as PW; cw; coefficient of significance as P; &nifi.caJjce difference as N.S. No .._Il. ...i .._.. ___J L__ Table 11 Tnbl. oh. 6 chid co.pl.int of dlesntivo Incidence I of I I Imtic~rlar 1 opacity system I Note: C-1 - lsnu of both eyes ure transpnrent and clear. (+) - There are D fen point shsged opncitlsa at equator of posterior capsule of one LMS. (+I - There ar. a feu palnt4bap.d both ay.8. Table 12 I Table 8 Duration and amplitude of Q vaw of Hean integratedvalue EM: P of Q "a"e _---Study plFQUP control W'UP 15.44 I'Y -__ ::t"dy CY -group 'antml Rmilp co.05 16.67 15.45 15.81 _ N.!: The changea of Plat*lstaount opaoitis1 in lens of hnqru I UBC count 1 - 204 shows that the characteristic value of autocorrelation function and average heart rate incree sed slightly.It is evident that the pulse wave, as aenviromental load, caused the characterisIncidence opacity or of I , number spiks of iIF , Inoidmaa I IIlorum of P< 0.05 PCO.M)l rtrtAg0 NNO P . 80 min Fig. 1: Orbital temperature. Eye directly exposed to 15 GHz 100 mW/cm2 microwave radiation. Mean values and standard deviations of the orbital temperatures before and immediately after irradiation, in three groups of experimen ts, are reported in table 1. Pi (mW/cm2) At (min) 100 200 200 60 60 120 (“C> 35.6 (.5) 35.2 (.6) 35.3 (.3) Tfl('C) 38.1 (.3) 38.6 (.8) 38.7 (.4) Tf2('C) 36.3 (.4) 37.0 (.8) 36.9 (.6) Ti Table 1: Mean values and standard deviations of orbital temperatures measured du ring three groups of experiments. Pi: incident power density; At: exps sure duration; Ti: initial orbital temperature; Tfl: final temperature of the eye directly exposed; Tf2: fi nal temperature of the other eye. For each treatment the average of four measurements (two points on each eye) has been assu med as the initial temperature; the average of two measurements (on the same eye) gave the fi nal temperature. In the first group (6 experi- 220 - ments at 100 mW/cm', 60min) the mean temperatu re increase was 2.5 'C in the eye directly ex& sed and about 0.7 'C in the other one. With a doubled incident power density (7 experiments at 200 mW/cm2, 60 min) the mean temperature in creases were about 3.4 'C, and 1.8 'C respectx vely. When also the duration was doubled (3 ex periments at 200 mW/cm2, 120 min) no further y~ crease in the final temperatures was observed. In all the exposed animals lachrymation and pho tophobia were always observed, as temporary effects within the first 24 hours after irradiation. More persistent were a pupillary constric tion and altered reflexes to ligth. In the animals exposed 120 min to 200 mW/cmZ (each time) a milky band on the posterior region of the lens was apparent at 24-48 hours after the fix st irradiation. With further identical treatments a well defined opacity took place. In the animals exposed 60 min to 200 mW/cm2 (each time) a small opacity became apparent, in the irradiated eye, during the second week (after two or three treatments), but the following ex tension was less pronounced than in the previ2 us cases. In the animals exposed 120 min to 100 mW/cm' only very weak opacification phenomena occurred. With exposures 60 min long,to 100 mW/cm',no lens modifications were observed. The exam of the fundus did not reveal any alte ration. On the contrary the observation of microscopic capillary vessels of the conjunctiva revealed always clear alterations. About the electroretinographic (E.R.G.) observations, in table 2 some cumulative data (wl_ thout distinction between the three groups of treatments) are collected. Mean values and stan dard errors of tipical parameters of E.R.G. plots (photopic waves amplitudes "a" and "b", "culmination time" of b wave, and the number of Oscillatory Potentials O.P. [5])are reported. Before M.V. S.E. After M.V. S.E. aamp. 13.17 ~I.17 12.5 ztl. 26 b c.t.(ms) 36.67 Lto. 49 39.0 *to.86 b amp.hV> 102.67 h5.41 137.67 29.57 O.P. AO.17 3.17 1.83 20.31 Table 2: Electroretinographic results. Observations before and after exposures. aamp' "au wave amplitude; best.: “CUE mination time" of wave "b"; b amp: "b" wave amplitude; O.P.: number of "Oscil latory Potentials". - 221 From data of table 2 it can be observed that: "arrwave amplitude appear substantially u i) naffected by the microwave exposure; ii) 'b" wave amplitude and latency ("culmination time") both are increased after irra diation; iii) the number of Oscillatory Potentials appe ar reduced. Such funtional alterations are resonably related to the emodynamic changes following the mi crowave induced heating. 41 - G6 References [l]Daily,L., Wavin,K.G., Herrick,J.F., Parkhil1,E.M.: Effects of microwave diathermy on the eye. American Journal Phisiology, vol. 155, 432 (1948). [2]Richardson,A.W., Duane,T.D., Hines,H.M.: Ex perimental lenticular opacities produced by microwave irradiations. Arch. Phys. Med. vol. 29, 765-769 (1948). Conclusions The reported observations, although to be confirmed by further experiments, suggest two major findings: i) Cataractogenic effectiveness of 15 GHz radiation appear quite similar to that well known at lower frequencies (2.45 GHz) from the point of view of both the threshold le vels and the ophthalmoscopic characters. ii) At exposure levels under-cataractogenic, a rather precise estimate of the stress conditions of the eye may be acheived by bio microscopic techniques applied to the microvessels of the conjunctiva, and by func tional test such as electroretinography. - [3]Carpenter,R.L., Van LJmmersen,C.A.: The action of microwave power on the eye. Jour. Microw. Power vol. 3, 3-19 (1968). [4]Cleary,S.F.: Microwave cataractogenesis. Proc. IEEE vol. 68, 49-55 (1980). [S]Brown,K.T.: The electroretinograms, its corn ponents and their origins. Vision Research vol. 8, 633 (1968). Work done with financial support of the Mi nistero della Pubblica Istruzione and of the Consiglio Nazionale delle Ricerche through the Istituto di Ricerche per 1'Elettromagne tismo e i Componenti Elettronici. - WIDEBAND NEAR EVALUATION THE EYES PRESENT WITH SUCH 223 ‘42~7 - OF MICROWAVE SCATTERING AS SAFETY INTENSITY STRUCTURES SPECTACLES D. W.Griffin and N.Davias Department of Electrical and Electronic Engineering The University of Adelaide Abstract A novel monopole antenna method has been applied to a full-size, image-plane type model of a human to determine the microwave field near the eyes and to evaluate the effect of scattering structures, such as metal-framed safety spectacles, on those fields. Investigations conducted in an anechoic chamber yield angles of incidence, test frequencies and wave polarisation values that give details of shielding, enhancement and depolarisation effects due to the scattering structure that appear not to have been previously studied. Microwave cataractogenesis is one of the most extensively investigated exposure effects and yet the influence of such variables as radiation frequency, field polarisation and directions of incidence on the distribution of microwave energy near and in the human eye appears not to have been determined [l]. In the absence of any established method of making the required measurements [3], a novel image plane method has been developed [4] for exploratory evaluation of the effect of the spectacles. Preliminary results [4] for the 2 to 4 GHz frequency range have been extended so that results of investigations over the range 2 to 12 GHt can now be given. I. Introduction The measurement of microwave fields in and around the eyes in models of humans is of prime interest to those concerned with the health implications of exposure to microwave radiation. The principal hazard to the eyes due to microwave exposure is that of cataract formation, and the true measure of the hazard that exists is the energy absorbed by the eye [11. In this study our aim is to measure the effect of scattering structures, specifically metal-framed safety spectacles, placed near the eyes and to determine whether safety margins incorporated in standards for permissible levels of human exposure [z] are compromised by such structures. Significant scattering of incident microwave energy can be expected because the metal frames include closed loop sections to support the lenses and also straight sections extending to the ears. Dimensions suggest fundamental resonances around 1 GHa although these may be heavily damped by the close proximity of the head which exhibits large losses at microwave frequencies. II. Measurement Assembly And Procedure The novel image plane method developed for this investigation and described in detail in a recent paper [4] is illustrated in the measurement assembly of Figure 1 and the standard orientations of the image plane model in Figure 3. A model of the head and chest of a human has a plane of symmetry passing through the centre of the forehead and along the nose and an image plane can be placed in this plane of symmetry creating the entire model so far as electromagnetic behaviour is concerned. The fields near the eyes can be measured by introducing a monopole antenna normal to the ground plane across the top of the nose and past the pupil of the eye. This forms the item under test in the chamber of Figure 1. The response pattern for the monopole by itself can be obtained and used as an overall calibration. Next the model can be placed on the ground plane and a new response pattern obtained. If all other experimental variables - are kept constant then the difference between the two sets of results gives a measure of the effect of the model on the resultant fields ,to which the monopole responds. If then the spectacles are mounted on the ground plane and a new response pattern is obtained then subtraction between the three sets of results gives a measure of the effect of introducing each new item onto the ground plane. r----- _---_-__--- “,crcwave 1 anecbolc chamber I micr~,~a 1 HP 9045 2 System control HP-IO HP-tB cmp"ter - The experimental variables used for this study are as follows. Assuming plane uniform TEM waves incident on the subject, the direction of incidence (0 and C#values in spherical coordinate terms), the polarisation of the incident waves (u@ and u4 components of electric field EJ)and its’ frequency are the variables associated with the source, as shown in Figure 2. Knowledge of the source power is not necessary because if all experimental variables are kept constant between the three sets of measurements, then a direct comparison of the three sets of results will yield the information required. Patterns for the model with and without spectacles which is treated as a receiving antenna are obtained from measurements of the scattering parameter Szl as the model is successively stepped through a sequence of B and 4 orientations under computer control. HPO409A ktmntlc 224 J HP 9072A Olgltal plotter x AY_. L model on plane Y Fig.1. A simplified block diagram of the main functional parts of the automated antenna radiation pattern measurement system based on a computer-controlled network analyser. 270” the origin for the Cartesian and spherical coordinate system used is at the CentFe of the circular ground plane that supports the made1 and the field measuring probe. 4H IF.c----Qp incident TEM wave !A (a) (Hi,) nade -horizontal horizontal poldrlzalion wing (b) (IN) mode -horizontal polarlrrt~on vertical wing T-h L incident wave in the x-a plane so that $=O and 845 (black dots represent 15' steps in 0). Fig.2. Spherical coordinate system used to describe the incident transverse electromagnetic waves and orientation of the model for any set of experimental measurements. (c) (VW mde - vertical polarization horizontal wing (d) (VV) mode - vertical polarlratlon vertical wing Fig.3. Measurements with each of two polarisations ( horizontal, H, or vertical, V) in each of two polar planes ( wings horil;ontal, H, or vertical, V ) gives the measurement modes HH, HV, VH and W as shown above. Measurements have been restricted to two polarisations of the incident TEM wave, two polar planes through the complete three-dimensional response pattern and the frequency - 225 range from 2000 MHz to 12000 MHz in 100 MHz steps. It is possible to identify four distinct measurement modes corresponding to the choice between either horizontal, H, polarisation or vertical, V, polarisation, and the choice between a response pattern in a polar plane parallel to the wing structure of the spectacles, referred to as H, or normal to the wing structure, referred to as V. Figure 3 illustrates the orientations of the polarisation of the incident TEM wave and the phantom relative to the floor of the anechoic chamber. The monopole extends along the a-axis from the origin of the coordinate system, and it remains parallel to the floor of the chamber, the horizontal, H, as does the propagation vector & of the incident TEM wave. III. Results A. HH Mode The response of the field measuring monopole at 5 degree steps in 0 can be plotted from computer files as shown in Figure 4 for a frequency of 6000 MHz. Results obtained with 4 = 0” are presented as negative 8 values, and those for 4 = 180’ are presented as positive 6 values. The total length of the monopole protruding from the ground plane is 35.35mm, corresponding to quarter-wave resonance occuring at a frequency of approximately 2.1 GHz. The ratio .of monopole length, L, to monopole diameter, D, is approximately 30. - 42 ~7 The pattern for the monopole by itself on the circular ground plane exhibits features consistent with diffraction due to the shape and finite size of that plane. These features are part of the reference response against which the effects of introducing the phantom and the spectacles are measured. When the phantom is introduced on to the ground plane, then for negative values of 8, the phantom is placed behind the monopole antenna, when looking from the transmitter, and so should have little effect on the overall shape of the response pattern, but for positive values of 6 the phantom is directly in the path between the transmitter and the monopole antenna, and so would be expected to introduce a significant amount of attenuation to the signals received by the monopole. The finer details of the pattern shown in Figure 4 are typical of the results obtained for the frequency range considered and are due to a number of effects. The phantom is composed of a thin shell filled with a material suitable for simulating the microwave prop erties of the human body. A standing wave pattern will be formed around the phantom and any dips in response could be due to the monopole being placed in the position of a null in the standing-wave pattern. Diffraction effects may also account for some of the details in response pattern as could depolarisation, coupling and scattering effects. When the spectacles are placed on the phantom then, assuming that source power is kept constant and that all other factors concerned with the study are unchanged, a simple comparison of this set of results with the previous one should yield the information we require. Henceforth, the results obtained with the phantom in place will be referred to as case 1 and those for the spectacles in place will be referred to as case 2. Fig.4. Response patterns for the monopole probe at 6000 MHz for the HH mode showing the monopole by itself then with the phantom in place and finally with metal-framed spectacles on the phantom. Consider first the case of negative 8 values. For a frequency of 2000 MHz the maximum in response for case 2, occuring at an angle of about 6 = -65O, is approximately 5 dB down from that occuring for case 1. We find that this remains so for frequencies up to approximately 4000 MHz, where the two maxima then begin to approach each other in level, until at 4800 MHz they have in fact - converged. Above this frequency, the two patterns remain very similar in shape, although there is __~ however a fairly consistent increase of __-_ _ __ a few dB in response for case 2 for the range of negative 0 values. In particular, for frequencies between 4000 MHz and 6000 MHz there is a consistent increase in radiation levels of approximately 5 dB for case 2 in the approximate range of orientations -60’ 5 8 5 -10”. We find that there are dips and peaks occuring for case 2 patterns which are dependent on the frequency and the orientation, however it is of more value to look for consistent changes between the two results rather than isolated cases of dips and peaks which are much more difficult to explain. In the region close to 8 = 0’ it is found that there is a consistent increase in radiation levels with both case 1 and 2 results as compared to those for the monopole by itself. This is to be expected if one remembers that we would expect zero response from the monopole by itself, and any scattering structure introduced near the monopole will increase the component of electric field parallel to the monopole axis. There is also a consistent increase in the level for case 2 results as compared to case 1. Consider now the results for positive 6 values. It is found that there is a fairly consistent increase of approximately 5 dB in the radiation levels for case 2 as compared to case 1, except for orientations close to 8 = O”. This is to be expected since for positive 0 values the phantom causes an obstruction in the path between the transmitter and the monopole and so one would expect very little radiation to reach the monopole. However, introduction of the spectacles, and the subsequent reradiation from them, would increase the fields around the eyes. Cases exist where there is an increase of more than 20 dB and this occurs for orientations where there is a dramatic dip in level for case 1, and subsequent addition of the spectacles completely overcomes this dip. The important result in this study is the comparison of the radiation patterns obtained for the phantom with and without spectacles. Subtraction of the two results will yield the information required, and this is given in Figure 5, where the difference between the two results is plotted against orientation for frequencies of 2000 MHz to 12000 MHz in 2000 MHz steps. Figure 5 illustrates the 22 16 - fact that there is a consistent increase of up to 5 dB over the frequency range 4000 to 10000 MHz for angles of incidence in the range -80” < 8 2 60’. Fig.5. Patterns at 2, 4, 6, 8, 10 and 12 GHz showing the effect of adding metal-framed spectacles to the phantom for the HH mode. This result is consistent with what we expect, since the incident TEM wave has a component of electric field parallel to the wing of the spectacles for all orientations except those close to 8 = f90°, and a component parallel to the rim of the spectacles for all orientations except those close to 0 = 0’. However, the monopole probe is at right angles to the wing structure and also to the equivalent of the loop antenna, the magnetic linear wire antenna. Therefore only a very small amount of re-radiated electric field will be parallel to the monopole probe. But the presence of the phantom and the fact that the monopole probe is in the near-field of the spectacles will increase the parallel component of re-radiated electric field somewhat. B. HV Mode The response of the monopole is given in Figure 6, with results obtained for 4 = 270’ plotted as negative B values, and those for 4 = 90° plotted as positive 0 values. Results for the monopole mounted by itself on the ground plane for this mode are identical to those obtained for the mode discussed in the previous section, and illustrate the precision and repeatability of measurements. - 227 - 42 ~7 relative to the phantom of up to 20 dB for frequencies below 6000 MHz, and for most values of 0. Fig.6. Response patterns for the monopole probe at 6000 MHz for the HV mode. The phantom forms an obstruction between the monopole and the transmitter for all values of 0, except those close to zero, and so would be expected to introduce significant attenuation to the levels reaching the monopole. This is confirmed by the results where we find that the overall shape of the response pattern when the phantom is introduced is similar to that for the monopole by itself, but with a corresponding decrease in levels. This is particularly true for frequencies above approximately 6000 MHz. When the spectacles are introduced onto the ground plane, it is found that the basic shape of the results remains fairly constant for frequencies above approximately 6000 MHz, but accompanied by a shift in level. The basic shape of the radiation pattern obtained is fairly symmetrical around B = O”, and this is consistent with the fact that the orientation of the incident electric field with respect to the spectacles is more or less symmetrical about B = OO. That is, the electric field is in the plane of the rims for all orientations but normal to the wing. Any departure from this symmetry is due to the presence of the phantom. The difference between results obtained for the phantom with and without spectacles are plotted as a function of orientation for frequencies of 2006 to 12006 MHz in 2000 MHz steps and presented in Figure 7. This shows that there is an increase in radiation levels Fig.7. Patterns at 2, 4, 6, 8, 10 and 12 GHz showing the effect of adding metal-framed spectacles to the phantom for the HV mode. The electric field of the incident TEM wave has a component parallel to the rim of the spectacles for all orientations, and this may cause a significant amount of re-radiation to occur, with a component of electric field parallel to the monopole probe. Excitation of the wing structure may also occur due to scattering and depolarisation occuring from the phantom. C. VH and VV Modee Polarisation of the incident electric field is normal to the monopole and so one would expect zero response for all orientations. The absolute level of radiation experienced by the monopole for vertical polarisation of the electric field as compared to horizontal polarisation is reduced by as much as 30 dB which puts all measurements very close to the limits of the dynamic range of the network analyser. When an item is introduced onto the ground plane along with the monopole one would expect there to be an increase in the level of radiation at the monopole due to scattering and subsequent depolarisation occuring from the new item, and this is confirmed by the results. A detailed discussion of results is not possible here, but suffice to say that there is a 228 consistent increase of up to 40 dB in radiation levels at the monopole site for the VH mode, but a consistent decrease for the W mode. IV. Conclusion For measurements with a monopole antenna that extends past the pupil of the eye, introduction of metal-framed spectacles causes an increase in the signal received by up to 5 dB for the HH mode and by up to 20 dB for the HV mode for a wide range of frequencies and for most orientations. Measurements made with vertical polarisation exhibit similar results. Measurements with a shorter monopole that extends only to the corner of the eye and is therefore closer to the bridge over the nose do not exhibit this enhancement effect, indicating that a method of measuring the field distribution in front of the eye is required in place of the monopole that gives, at best, a weighted average of the required quantities. A probe that scans the region in front of the eyes would increase the amount of data to be gathered by a factor equal to the number of measurement positions in front of the eye. A revolutionary method could be adapted to reduce the amount of data to be gathered in the following way. Instead of a large number of tests involving specific field conditions in an anechoic chamber, a paddle-tuned microwave - reverberation chamber might be used for creating a sufficiently comprehensive variety of field conditions for the worst case to be covered. V. References S.F.Cleary “Microwave Cataractogenesis” Proc. IEEE, vol. 68, pp.4955, Jan. 1980 Standards Association of Australia “Draft Australian Standard for Permissible Levels of Human Exposure to Non-ionizing Electromagnetic Energy in the Frequency Range 30 kHa to 300 GHz.” DR 82191, Oct. 1982 The Institute of Electrical and Electronic Engineers “IEEE Standard Test Procedures for Antennas” IEEE Std. 149-1979 D.W.Griffin and N.Davias “The Effect of Metal-framed Spectacles on Microwave Radiation Hazard Near the Eyes” presented at IEEE 1984 National Symposium on Electromagnetic Compatability. VI. Acknowledgements The need for research on this task was drawn to the authors’ attention by Mr. Sasty Sastradipradja of Telecom Research Laboratories, Australia. - 229 43Hl - STATISTICAL ASPECTS OF NOISE AND LIMITS A. de Jong Dr. Neher Laboratories PTT Leidschendam, Netherlands The effect of noise and interference on electromagnetic systems depends on many statistically variable factors. These factors are related to the interference sources, the coupling between source and interfered equipment and to the capability of that equipment to resist interference. Most experience in this field is available in the area of radio interference (RFI) because an open radio transmission system is quite vulnerable to interference. Some considerations are therefore given to the control of interference in radio communication systems by the application of RF1 limits. Introduction Electric and electronic equipment may be disturbed in their proper functions by unwanted electromagnetic signals, often called disturbance, interference or noise. Some of these disturbing signals are useful for other purposes (wanted signals), but disturbing to the victim equipment, a wanted signal for one application may act as an unwanted o.r interfering signal for a different application. Other disturbing signals are not useful anyhow, but are generated unintentionally as unwanted products of electrical or electromagnetic processes. Each electric or electronic equipment is surrounded by a large number of unwanted signal sources, such an overall situation is characterized as: electromagnetic environment. Because of the many variable characteristics of such an electromagnetic environment and because of the variable aspects of the equipment itself a statistical approach is necessary to define the quality or reliability in proper functioning of the equipment in an electromagnetic environment. The proper operation in an e.m. environment is essential for all types of electrical, electronic and electromagnetic (radio) systems and devices. Radio systems, however, are extremely vulnerable to ambient e.m. interference fields because of the open medium used for the transport of information. For this reason the behaviour of radio communication in an e.m. environment has been investigated for many years and a lot of information is available in this area of electromagnetic compatibility (EMC). Moreover, most countries apply regulations to control radio interference (RFI). More recently, growing attention has been paid to non-radio EMC problems. The following paragraphs survey the various aspects which determine the statistical parameters of interference cases whereas more detailed information on theoretical and practical aspects is given in the papers presented in this session. Types of interference signals Output signals (radiated or conducted) of various interference sources can differ considerably from type to type. The following distinction is often made: - narrowband (relative to centre or carrier frequency), - broadband. Such a distinction in the frequency domain is very useful in the case of RF interference (RFI), which considers the interference effect on (narrowband) radio channels. In this case the term narrowband is related to the bandwidth of the operational receiver or the corresponding measuring receiver. In the case of communication in broadband systems, audio and video systems and in the functioning of digital equipment or other electronic applications the signal waveforms in the time domain may be more useful. It must be noted, however, that in complex cases (e.m. environment caused by various sources) the overall interfering signal waveform is not quite relevant for analytical purposes and considerations on the frequency spectrum are more informative. For basic waveforms the transfer from time domain to frequency domain and vice versa by using Fourier rules are well-known. Although simple waveforms seldom occur in practical situations the following basic waveforms are used to simulate actual interference signals. Narrowband signals: - sinewaves as generated by oscillators, transmitters and industrial, scientific and medical (ISM) equipment, - modulated sinewaves (transmitters). Narrowband signals may be generated anywhere in the frequency spectrum without any principal restriction. Broadband signals: - rectangular pulses (digital equipment, radar), - triangular pulses (sweep circuits, electronic circuits), - 230 partial sinewaves (phase controlled thyristor regulators), trapezoidal pulses (a modified approximation of rectangular pulses), exponential rise and decay pulses (transients), white gaussian noise (passive and active components, gas discharges), coloured noise (some active components, program signals). From the broadband signals white gaussian noise is basically flat and the spectrum infinite. In practice parasitic reactances inside the source reduce its level in the higher frequency region. The pulse waveforms generate spectra which are flat up to a certain frequency limit, beyond this limit the level falls off with a 20 dB or 40 dB per decade slope. For this reason in practice broadband man made noise is normally restricted to the frequency range up to 1 GHz. In actual circuits basic pulse waveforms are always distorted by non-linear frequency and phase characteristics and often by non-linear amplitude characteristics of the signal path, so the output waveform is distorted and the resulting frequency spectrum deviates accordingly from the basic spectrum. This is a dominant factor in narrowband transmission channels. The interference potential of a narrowband source differs from that of a broadband source because the frequency of the narrowband source should coincide with the frequency channel of a narrowband (e.g. radio) system whereas the broadband source influences a broad frequency band simultaneously. On the other hand the level of a broadband source is, measured in a narrowband channel, much lower than the level of a narrowband source of comparable interference power. So the statistical interference behaviour of both types of sources is quite different. Some actual interference waveforms have a stochastic appearance (noisy), often in the form of noise bursts, 'quasi impulsive noise', (e.g. sparking contacts, commutator motors, motor car ignition systems). Noise sources in which the noise output is generated directly by the a.c. mains voltage show an amplitude modulation in accordance with the mains waveform envelope (e.g. mains fed commutator motors, thyristor regulators, contacts, gas discharge lighting). This may cause the well-known hum in sound signals and two horizontal bars filled up with noise spikes in TV pictures. Statistical behaviour of an interference source As explained earlier the waveform and the concerning interference level vary from type to type because of the different basic characteristics of the noise source. For a given type of equipment the noise output level of a source varies from sample to sample within a series of the same type of went. This is due to electrical and mechanical tolerances in components and circuit lay-out during the manufacturing process. For this reason in mass produced appliances only statistical testing is economically feasible. This results in a limited confidence in the entire production of a certain type (80%-80% rule of CISPR). - The output level of a single source is often variable in time. Mechanical contacts (switches, commutator motors, ignition systems) and gas discharges (gas discharge lightink, sparks) show short term variations due to mechanical instabilities, temperature, air pressure, etc. This requires observation of the output level for a period of several seconds per spot frequency so that the sweep rate for automatic frequency measurements is reduced to very low values. Long term variations are caused by mechanical wearing, aging, gas diffraction, etc. For this reason measurements on commutator motors and contact devices are carried out after an adequate running-in period. In particular contact burning increases the noise output during lifetime considerably. The output level of a single broadband source is always more or less variable in frequency. In some cases the level variz strongly over the frequency range e.g. in equipment with coils (commutator motors) in which parasitic resonances occur in the RF range. This necessitates measurements in a broad frequency range with adequate frequency resolution. In the case of radiated fields the radiation varies with the direction, particularly at higher frequencies. This requires measurements in several azimuth directions, mostly performed by means of a turn table. Normally, measurements are not carried out in various elevations, but in specific cases (aeronautical interference) it may be also necessary. The final test result is also determined by the testing accuracy. Inaccuracies are introduced by electrical and mechanical tolerances in the measuring equipment, the operating conditions of the unit under test and the test set-up. The lay-out of connecting cables to input and output terminals in the test set-up is very important, so the test set-up should be standardized accurately. This is in particular critical in the case of fieldstrength measurements. Such measurements are also strongly influenced by environmental conditions such as reflections caused by the ground (dependent on weather and season), nearby objects and by the operator. Fieldstrength tests carried out at several test stations on the same signal source have shown differences up to 2 5 dB. For this reason and for other practical drawbacks (open test site, climate) fieldstrength measurements are not very popular. The statistical approach for mass produced equipment is generally acceptable for the interference protection of non-vital services (performance or quality degradation in broadcast and non-vital conuaunication). In this respect the 80%-80% rule (80% of the equipment should fulfill the requirements with 80% confidence) recommended by CISPR has proven to be useful. In the case of vital services (e.g. radio channels for emergency and distress calls, instrument landing systems, vital industrial processing, telecontrolled transportation systems, etc.) more absolute limits are required to control worst case situations. This is, however, only possible by testing all individual sources and conflicting with statistical testing of mass 231 - produced equipment. In this case statistical testing with much tighter limits than the 80X-80% CEPR limits is the only alternative. Coupling between interference source and interfered equipment The noise output signal generated by the source can either be radiated or conducted from the equipment. Radiation depends on the coupling or leakage from the internal source to the outer case or cabinet (external currents) and on the radiation characteristics of the cabinet. Radiation is only possible if the dimensions of the equipment are not very small compared with wavelength, so normal size equipment does only radiate in the MHz range and higher. Noise signals can be conveyed by conduction along cables, in particular the mains cable is an important noise conductor. Part of the noise injected into the mains is conducted by the mains network to other users, part of the noise energy is radiated by the mains itself depending on frequency and mains configuration. At the victim equipment a complementary situation exists. Ambient fields may induce unwanted signals directly in sensitive parts of the victim equipment if this equipment is not properly screened. This is frequently the case in electronic circuits housed in non-metallic cabinets, unscreened and poorly designed electronic circuitry. The induced voltages depend on the dimensions of the sensible circuit loops and increase with frequency, so the effects are mainly important in the MHz range and higher, In addition, ambient fields induce currents (and voltages) in cables connected to the equipment. In this case the noise conducted by the cables is often dominant with respect to the direct pick up of the equipment because of the relatively long length of the cables exposed to the ambient field compared with the dimensions of the equipment. Finally, mains cables may conduct unwanted signals injected from other sources into the mains network. An important factor for the coupling effect is the distance between source and victim equipment. The propagated interference power is inversely proportional with distance according to a 2nd power law for free field radiation, with a 4th to 6th power law for nearby induction fields and between a 2nd and 5th power for RF fields under actual propagation conditions. Attenuation of conducted signals takes place according to an exponential law. Special conditions such as screening, reflection, atmospheric absorption, fading, lay-out of cables, etc. may influence the distance effect considerably. The propagation characteristics of the coupling path are always frequency dependent and occasionally amplitude dependent. For radiated fields the radiation characteristics of the source are the dominant frequency dependent factor, for conducted signals the frequency dependent cable attenuation is the important factor. At the lower end of the frequency spectrum (< 10 MHz) the radiation effect of actual interference sources is relatively small as far as the dimensions of the source (and its cabling) are small compared with wavelength. In this frequency range the attenuation of cables to conducted interference is rather low 43Hl so in practice conduction coupling dominates. In the higher frequency range (> 100 MHZ) radiation is more effective and cable attenuation rather high so in this frequency range radiation coupling dominates. In the transition region (10-100 MHz) the dominant coupling type varies from case to case. From the considerations given earlier it will be clear that wide variations in coupling paths are found from case to case. Some measurements have been carried out to determine the coupling between noise sources connected to the mains and receiving antennas at a specified distance. However, limited data of coupling path statistics are available in literature. Statistical behaviour of interfered equipment Unwanted signals may disturb victim . equipment in various ways. One important aspect is the interfering signal itself as characterized by its time domain (waveform) or frequency domain (spectrum). Another factor is the way through which the unwanted signal penetrates into the victim equipment. Apart from the distinction in radiated and conducted interference it is different whether an unwanted signal enters the equipment through the standard input terminals or via any other path. Unwanted signals entering via the input terminal pass the same circuits as the wanted signals and therefore cannot be easily separated without degradation of the wanted signal (except for more complex signal processing procedures). In telecommunication the protection ratio is defined as the ratio between the wanted input signal level and the unwanted input level which causes a specified degradation of quality or loss of performance. Unwanted signals entering via other paths can be isolated from the wanted signal more simply (screening, filtering, proper circuit lay-out and earthing). In this case the (external) immunity-factor is defined as the ratio between the unwanted interference sianal level and the wanted input level (note thereciprocal quantity compared with the protection ratio) which causes a specified degradation of quality or loss of performance. In both cases the annoyance caused by the interfering signal is of a subjective nature and should be evaluated statistically from subjective observations in order to be translated into objective quantities. The immunity characteristics of an equipment vary from sample to sample due to electrical and mechanical tolerances in circuit lay-out and components just as in the case of interference sources. To determine interference cases the following statistical parameters are to be taken into account: the level of the wanted signal, a statistical quantity which varies strongly in the case of radio signals and generally to a much less extent in the case of cable conducted signals, the overall level of ambient interference signals, which is generally very complex in practice as pointed out earlier, the protection and immunity characteristics of the interfered equipment. - 232 Interference and immunity limits Noise suppression measures are taken in several areas of the EMC field. Standardization of EMC measures is growing in international industrial organizations. The oldest and most widespread area of application is radio communication because of its use of radiated waves in an open transmission medium. To protect radio services limits have been standardized for many years and laid down in legal regulations to restrict the unwanted output of equipment (interference limits). The purpose of these limits is that under average or worst case conditions such a noise source or a combination of noise sources does not cause any unacceptable interference. More recently standards have been established for the required immunity of (radio) equipment against interference. As explained earlier it is evident that the dictation of limits is only a partial answer to the overall problem of RFI, although it covers an important part of actual interference cases. It is not easy to evaluate the effect of new or amended limits in practice. For this reason radio regulating authorities collect statistical information on RF1 complaints. Imperfections of these statistics are: many cases of interference are not reported by the public, so information is limited, normally, no distinction is made in complaints caused by equipment which does or does not comply with the limits, the subdivision in interference sources or interfered victims is not always explicit, amendments in limits are only effective in the long run, because older equipment often has to die out. Nevertheless, these statistics show tendencies in interference situations over years and these statistics are the only practical means for feed back to the authorities responsible for the establishment of limits. Conclusions Abatement of electromagnetic interference and in particular radio interference requires a statistical approach because of the many variable quantities involved with respect to the noise sources, the interfered equipment and the coupling between noise source and victim. Such an approach should also take into account the harmfulness of the interference and consequently the reliability of the measures to be taken. The existing regulatory measures, which prescribe limits for a number of noise sources in the field of RF1 solve the interference problems only partially. More sophisticated models and procedures are necessary to attain optimal suppression measures. The field of application should be broadened to equipment not only used for radio communication. It should be kept in mind, however, that abatement of interference is not only a matter of technical requirements, but of economic and policy considerations as well. - 233 44~2 - PROBLEM RADIO INTERFERENCE- THE PROBABILI'I?! A C D Whitehouse Department of Trade & Industry London,England 1. INTRODUCTION In the management of the radio frequency spectrum the various radio services are planned on the basis that they will suffer some interference. One of the most difficult asoects oE planning is the estimation of the degree of interEerencethat a oarticular service is likely to sufEer. Within the InternationalConsultative Committee on Radio (CCIR), mucn work has been done on estimating the extent of interferencebetween radio services, but comparativelylittle has been done for the interferencewhich is caused by other sources of spurious signals. The TnternationalSpecial Committee on Radio Lnterference(CTSPR) is tasked with developing radio interferencelimits. CIXPR and CCIR are currently reassessingthe limits of spurious signals caused by industrial,scientific and medical.(lSM) equipment. This work includes examining methods of!estimating the probabilityof the occurrence of interference to various radio services. 3 4. THE POSTTLON WITHTN CISPR 'Thequestion of derivinq acceptable limits for ISM equipment is beset on the one hand with the problem of large equipments which are particularlycostly to suppress and on the other with that of safety of life services such as aeronauticalinstrument landing systems (ILS) which require a hiqh degree of protection. CISPR has recommended limits Cl1 but there is substantialevidence that in many parts of the world these limits are exceeded by the majority of ISIMequipments, and yet there are few cases ofinterference [21. 'ChereEore CXSPR in conjunctionwith CCIR is revising its limits to produce values which can be observed in practice but yet afford the required degree of protection to the various radio services. For ISM equipment the CISPR test method is to measure on a test site the field strength of spurious signals at a distance of 30 metres from the equipment. To produce a limit it is necessary to relate the field strength measured at 30 metres on a test site to the maximum permissiblespurious field-strength at the receiving antenna of the radio system to be protected C31. The maximum permissiblespurious signal (Eil when the signal at the receiving antenna is the minimum signal to be protected (Ew) is: Ei = Ew - R dB(uV/m) - (11 Where R is the required signal-to-spurious ratio . It is necessary to determine the average minimum distance (d) of the antenna from the ISM equipment. By using a distance attenuation law (l/d") a maximum permissible spurious signal at the distance of 30 metres from the ISM equipment is calculated. Using this value (E ) as the CISPR limit would ensure a very3Righ degree of Qrotection but the penalty would be to impose high suppression costs on the manufacturersof the ISM equipment - this cost being ultimately borne by the consumers, the general public. In practice buildings provide some attenuation (0) of the spurious signal, while the planned reliabilityof each radio service can have a component (P) to allow for the probabilityof interference. Allowing for these factors the limit (EL) is: EL = E30 + B + P dB(uV/Ml - (21 Although conceptuallya simple model there are significantproblems associatedwith determining the distance attenuation law associatedwith very short range propagation, the attenuationresulting from the building structures and the factor to allow for the probabilityof interferenceoccurring. 3. THEPROBABICITY‘^OF INTERFERENCE Within CCIR the probability that any part of a transmissionis lost as a result of interference is estimated C4] from equation 3. P(I) = CP(B/A) - P(D/A and C)]X Cl-(l-N'TilexQ (-T~N/(~-NT~~)I where P(x) is the probabilityof an event x; P(X/Y) is the probability of x, given y; A is the desired transmitter transmitting; (31 234 B is the wanted signal being satisfactorilyreceived in the absence of unwanted spurii: C is another eguipment producing unwanted signals; D is the wanted signal satisfactorily received in the presence of unwanted interference. N is the average number of interference transmissionsQer unit of time; TW ‘!ci is the length of a wanted transmission; is the length of an interference transmission. Equation 3 is based on the definition in the Radio Regulations C51 of interferenceas "the effect of unwanted energy ..... on reception in the radio communicationssystem, manifested by any performancedegradation,misrepresentation, or loss of informationwhich could be extracted in the absence of such unwanted energy." It follows that there can be no interferenceif the wanted transmitter is not transmitting. Some comments on equation 3 are in order. The factor P(B/A) is just the probability that the wanted signal will be correctly received when there is no spurious signal, in some services this is called the reliability. It is worth noting that if P(B/A) is small - such as may be the case in long distance ionosphericpoint-to-point services or mobile services near the edge of the coverage area - then the probability of interferencewill be small regardless of other factors. P(D/A and C) is the Qrobabil- ; ity that the wanted signal will be correctly received even when spurii are present. For receivers separated from the ISM equipment by distances > d,P(D/A and C) is essentially equal to P(B/A) and hence the probabilityof interferenceextremely small. The following sections consider the Qrobability of interferenceoccurring in various radio services. 4. INI'ERFERENCE IN THE FREQUENCY RANGE150 - 1605 KHz. This section con&iders spurious radiation generated by induction heating equipment and the interferencecaused to aeronauticalnondirectional beacons (NDB) and to LF/MF radio broadcasting C61 For NDBs the aim was to evaluate the probability that an airborne receiver operating on a frequency within the band allocated in the Radio Regulationswill suffer interferenceas a result of spurious signals generated by an induction heatf&r~eguipmentwhich just meets radiation limits specified for a 30 m test site. To evaluate the probabilityit has been necessary to undertake an in-depth survey of conditions existing within the United Kingdom in terms not only of the levels of qurious signals generated by induction heating equipments but also of the operationalcharacteristics of the NDBs. Although many parameters are significant in determining whether interferenceoccurs, only three have been considered. These are the characteristicsof the spurious signal, the frequency of the spurious signal and the separation between the source of the spurious signal and the airborne NDR receiver. 4.1 'Thewanted signal. The minimum field strenath to be wrotected at the edge of service area is 37 dB(uV/m) [71. Within the service area oE an NDR the wanted field strength exceeds this value, but for simplicity it is assumed that the wanted field strength throuqhoutthe service area is 37 dB(uV/m). This assumption errs on the side of safety. The protection ratio for co-channel operation of a service similar to the NDR is 15 dB [81. Interferencemay occur if the level of the spurious signal, ERl, is such that ERl> 37 - 15 dB(uV/m) = 22 dB(uV/m) (4) 4.2 The InterferenceSignal The spurious radiation generated by induction heating eguipment is generally intermittent in nature and the protection ratio required has not been established. 'Theinvestigation indicates that the majority of induction heaters produce bursts of spurious radiation lasting several seconds inter-spacedwith approximatelyequal intervals during which no spurious is generated - a mark-to-spaceratio of approximately1 : 1. 4.3 Frequency The frequency band allocated on a primary basis is 160 - 435 kHz and on a permitted basis is 435 - 526 kHz. The selectivityof a NDB receiver is specified [9l to allow the Qroteqion ratio required to be reduced by 11 dB at - 2 kHz and by 85 dB at f 7 kHz. Thus the 0 dB bandwidth of the NDB receiver is approximately1 kHz. Therefore if an induction heating equipment is generating a spurious signal with a frequency within the 160 - 526 kHz band, the probability that the frequency falls within the bandwidth of the NDB receiver is: Q(f) = = 2.7 x 1O-3 1 526 - 160 - (5) 4.4 Separation Assume a proposed limit for ISM equipment of 90 dB(uV/m) measured at a distance of 30 m on a radiation test site. It is necessary to calculate for an ISM equipment which just meets this limit the distance at which the spurious radiation equals the threshold for interferencelevel of 22 dB(uV/m) (see equation4 1. 44~2 - 235 - 'Thereare two factors taken into account: i) The service volume is Attenuationof the spurious signal as it passed through the structure of the factory; and ii) the propagation factor. lO.?r(RNDBj2km3 Therefore the probability that an aircraft is within the critical volume surrounding the ISM equipment is: , p(v) = (9) &,2 NDB Over the years many measurementshave been taken which, for this frequency range, show that the attenuationcaused by factory walls can vary between 6 dB and 25 dB. A value of LO dB is assumed in the calculations. The propagation factor for free space is defined as the difference between the field strength El measured at a distance Rl from a source and the field strength E2 measured at a greater distance R2. and p(v) < 1.8 x 10 -4 4.5 Probabilityof Interferenceto NDBs. Assuming a high reliability for NDBs, in equation 3 P(B/A) = 1. It is assumed that Eor R < &+ NDBs have service radii varying between 10 and 100 nautical miles. For the 163 NDBs in use in the United Kingdom the mean radius, the median radius and the modal radius are all equal to 25 nautical miles. Using this value and d = 1.8 km, then,since the assumptions made have erred on the side of safety A E 4 $ for R > Tr Hence P = P(B/A) - P(D/A and C) where E is the field strength, R is the distance Erom the source, and h is the wavelength. 'Thereforethe propagation factor (PF) is qi.venby: =2010glo P = p(f). p(v) < 4.9 x lo-7 x R, CT+ PF= 60 log R + R\ m 60 lo~$~(&) dB& < dB, R2 < RI & 'Thereforethe probabilityof an induction heating equipment interferingwith the operation of an NDB is - = 90 - 10 - PF dB(uV/m) (7) Using equations 4, 6 and 7,the distance (d) at which the level of the spurious radiation equals the threshold level is found to be: d = 9.4 F2 c.03 < /2
Thisprotection is required against an interferenzesignal whose freguency is stable to within - 1 Hz of the tone spectral line in the ILS tone filters. The second level of protection ratio is 23 dB, which is required when the interferencefreguency is stable to within - 50 Hz. In this case the interferencemechanism results from the action of the interferencesignal in the ,tonefilter rectifyingcircuitry. The third level of protection ratio is 6 d0, which is required when the int rference freguency is stable to within 7 - 12 kHz. In this case the interferenceeffect results from the action of the interferingsignal on the IF detectors. ,rherange of 40 dB between the protection ratios quoted indicates that the protection required by the ILS system is very dependent upon the characteristicsof the interfering signal. !Xi/59 60/69Xl/79W/8990/99 Lf3.d at 30 m dB(uv/m) Fig. 1 : Histogram of measured spurii levels from induction heating equipment. 5.2 Characteristicsof ISM Interference. The fourth harmonic of 27 MHz dielectric heating equipment falls within the ILS band. The characteristicsof the spurious signals radiated by equipment were determined by visitingfourindustrial sites and making measurementson 42 separate dielectric heating equipments. 'Theequipments split into two main categories; those which operate continuously (used in processes such as baking and glue drying), and those which operate intermittently(used in plastic welding). The detailed results have been reported C131. From Figure 1 a level of 90 dB(uV/m) at a measurementdistance of 30 m from the source is exceeded in 20% of the readings. 5.2.1 Equipment which operates continuously. Almost 50% of the equipments upon which measurementswere made were of the type which operate continuously. The actual fundamental frequency used is adjusted to maximise the transfer oE power into the load. The fundamental frequenciesfell within the range 26.06 - 29.88 MHz. 'Thestability+ofthe fundamentalfrequency yas within - 10 kHz, short-term and within - 20 kHz over a five day period. It is concluded that a satisfactorylimit for the levels of spurious radiation generated An assessment 'was made of the fourth harmonic of the radiation polar diagrams of the dielec- Many Regulatory Authoritieshave already accepted CISPR Recommendationsconcerning interferencefor equipment other than ISM in which the limits are implementedon the basis that up to 20% of the equipments in use may exceed the stated limit. 44~2 237 - tric heating equipments, four measurementsat a given distance being made in orthogonal directions. At a 30 m measurementdistance the Iminimumvariation in a polar diagram was 10 df3and the maximum variation 35 dB. The typical variation was 25 dB. At loo metre Imeasurement distance the minimum variation in the polar diagram was 6 dB and the maximum variation 35 dB. The typical variation was 15/20 dB. 'Theconsiderablevariation between the measured polar diagrams of similar elguipments indicates the strong influence of surroundingstructures on the radiation influenceis not present in pattern. ,I'his test-sitemeasurements. 'Thelevels oE the measured interference Eield strength at 30 m varied between 53 and 100 dB(uV/m). At 100 m the measured levels igere between 27 and 69 dB(uV/m). 5.2.2 i___.____~-. Bquioment &i-h operates --i.ntcr,ni.tte!ntl.y The plazic welding equipment radiated spurious signals intermittentlywith typically a 5 second "on" time and a 20/30 second "oEE'*time (while new material was being inserted). The nower of the eguipment was in the range 3 - 10 kW. The fundamental frequenciesspanned the range 25.9 - 30.0 MHZ, plus one machine operating on 20 MHz. The EundamentaL frequencywas not stable, the drift rate varying 'between11 kHz/sec to 200 kHz/sec. rypically the drift rates fell within the range 50 - 75 kUz/sec., The radiation polar diagrams at the 4th harmonic were estimated by making four measurementsat a distance oE 30 metres in orthogonal directions. The minimum variation in a polar diagram was 10 dB and the maximum variation 48 dB, typically being within the range 20 - 35 dR. At the 30 metre distance the measured field strength varied between 23 dB(uV/m),and 76 dB(uV/m). 5.3 Interactionof ISM Interferencewith ILS receiver. Some initial experiments in which typical ISM interferencewas simulated suggested that the ILS receiver could operate satisfactorily in interference Eields up to 20 d0 higher than the wanted field. Therefore the interferenceeffects of the spurious signals generated by actual dielectric heating equipment on the operation o.Ean ILS/VOR system was evaluated at two typical industrial locations. One location had 12 continuouslyoperating dielectric heating equipments while at the other location there were 13 bridge-typeplastic welding equipments. 'Theexperimental arrangement is shown in Figure 2. The monitoring equipment was used to identify the harmonic signals generated by the dielectric heating equipment within the ILS/VOR frequency band. The amplitude and frequency of the signals was measured with the calibrated antenna connected to the spectrum analyser. The calibrated antenna was then reconnected to the ILS/VOR equipment and the appropriatechannel selected. Phe results obtained cl41 showed that the protection ratio required for interference oroduced by continuouslyoperating dielectric heating equinment is of the order OF 2 - 4 dB. 'EhereEore,Eor such eqipment the protection ratio of 6 dB mentioned in Section 5.1.should afford complete protection. The results for intermittentlyoperating equipment indicated that t'neinterferenceeffects are present Eor a very short time, sign.iEicantly less than 0.5 sets. %'hereEoreit is diEEicult to assign dn adeguate protection ratio but certainly the nrotection ratio is significantlysmaller than that Eor continuouslyoperating equipment. 5.4 InterferenceLimit During the crl'ticaltime when an aircraft is just about to land the specified minimum field strength for the ILS is 46 dBiB(uV/m). Using the protection ratio oE 6 dR, and making an allowance of attenuation through the Eactory wall or roof of 10 dB, the limit for TSM interference,measured at 30 m, is 50 dB(uV/m). In the derivation OE this limit account has been taken of:the characteristics of the ISM interferenceand of the response of the ILS system, but no account has been taken of other factors such as the probabilityof ISM eguinment being close to the airport, the frequencygenerated being within the passband of the ILS receiver or of the equipment being operated during the critical time. 'CheILS is a critical service and therefore it is wiser to rquire that I?(I)= 0. 'PIUSthe limit OE 50 dB(uV/m) at 30 m will.give complete protection to any aircraft flying to within 30 m of a dielectric heating equipment. A2 cl S/A Al calibrated PU dipole S/A A2 monitor whip SG A O-60 dB atten- ILS uator S 6 dB splitter ILS RX DC power unit spectrum analyser signal generator ILS/VOR tone generator. ILS/VOR receiver Fig 2: Test setup to evaluate interference to ILS/VOR receivers. - 238 - 6. CONCLUSIONS The work of this paper illustrates two of the important areas in determining radio interference limits. It is necessary to find a value of the protection ratio required by the radio service for the type of interference considered. The other is estimating the probability of interferenceoccurring iE a particular limit is implemented. In addition the importance of knowing the history of interferencecomplaints in determining suitable limits is shown. Similar work is continuingwithin CSSPR, most notably with respect to Data Processing Eguiplnent. It is hoped that by combining this type of theoretical approach with practical experience, the reassessmentof the ISM limits being undertaken by CISPR, in conjunctionwith CCIR, will gain recognition!and result in the limits being applied in practice. I. 8. REFERENCES Cl1 CISPR Publication 11, 1975 and amendment No 1, 1976. c23 CCIR IWP l/4 - Doe. l/186 (Rev.11 - E of 23 October 1981, Progress Report by Chairman with technical annexes 5 - 18. c31 CISPR/B r-41 CCIR RQt. 829, 1982. c51 Radio Regulation No. 1.60. [61 CISPR/B/WGl (Whitehouse/UK)8, February 1983. c71 Radio Regulation 2857. C81 Radio Regulation 2854. r.91 CAP 208 Vol.1 pt.5. Cl01 CCIR Rpt. 928. II111 Dept. Trade & Industry DRr ,Yechnical Memoranda Nos. 122 to I28. L121 Royal Aircraft Establishment. cl.31 ERA Technology Ltd Rpt.3572/5, June 1982. [14] CISPR/B/WGl (Whitehouse/UK)9,May 1983. (Secretariat)35, May 1984. ACKNOWLEDGEMENTS I wish to thank the Director of Radio 'Technology for permission to present this paper. - 239 45 - H3 STATISTICAL EVALUATION OF THE EMC SAFETY MARGIN AT SYSTEM LEVEL BY B. AUDONE, R. CAZZOLA and G. BARALE AERITALIA - Avionic Systems and Equipments Group Caselle T.se - TORINO - ITALY ABSTRACT: EMC equipment tests are well es@ blished and defined, while EMC system tests are still rather vague: in MIL-E-6051D dea ling with EMC system testing only general guidelines are given with the definition of the safety margin. In a complex system such as an aircraft, the problem of the electromagnetic compatibility and safety margin have become more important especially with the introduction of electronic equip ments into areas of the aircraft which directly relate to flight safety. Interfe rence effects are not always repetitive and in many cases the malfunctions change ran domly around an average level. Therefore a statistical evaluation of a safety margin was developed for the EMC investigation tests so that a number of parameters, from various avionic equipments could be monitored, when the aircraft onboard emissive equipments are activated. By use of this technique each equipment parameter may be monitored over a set period of time. Comparison can the refore be made, from this point of view, between each parameter with and without the emissive equipments activated in order to evaluate any drifting of any parameter due to EM1 and obtain an EMC safety margin. 1. INTRODUCTION In many works the need of a statistical model for EMC testing has been emphasized 111, 121, 131, because the interference effects are not always repetitive, but in many cases the malfunctions change randomly, around an average level. In a complex system as an aircraft, the problem of the definition of an EMC safety margin related to a stati stical model is very important. At the present EMC system tests are still rather vague and only general guidelines are given about the definition of a safety margin. The purpose of this work is to develope a mathematical model to perform an analysis of random interference effects from the statistical point of view and to define a safety margin related to the chara: teristics of the equipment under test. Some considerations about the degree of uncertainty related to the safety margin evaluation are also given in order to reduce the error probability during the execution of the test at system level. 2. SAW'LE VALUE AND PARAMETER ESTIMATION The two basic parameters of a random variable x which specify its central tendency and dispersion are the mean value and the variance u and dL: X /X (1) -co where p(x) is the probability density function of the variable x. An exact knowledge of the p (x) function will not generally be available. Hence one must be content with estimates of the mean value and variance based upon a finite number N of observed values: (3) The hats (-) indicate that 3 and axL are /. used as estimators for the mean value and variance of the random variable x. In order to have a good estimator, it must be: a) A unbiased EC63 =9 (5) - 240 - n where # is an estimator of @. b) efficient Qlx 2 1 is the Chi-square distribution function n with n=N-1 degrees of freedom.Moreover the sampling distribution of the sample mean value x is given by 141 : n where fl is the estimator of interest and is &y other possible estimator of @. @I where : - r c) consistent where N is the number of observed values and 3 is the estimator of $4. It is desirable that the expected value of the estimator be equal to the parameter being established (estimator unbiased) and also that the mean square error of the estimator be smaller than for other possible estimator (estimator efficient). Moreover it is desirable that the estimator approach the parameter being estimated with a proba bility approaching unity as the-sample size becomes large. The estimators u and 2 2 /x X are estimators unbiased, efficient and consi stent for the mean value and variance of a random variable x 14). 3. SAMPLING DISTRIBUTIONS Consider a random variable x with a probabi lity distribution function p (x). Let x 1, x2, ......... ...... ..... .... XN be a sample of N observed values of x. Any quan tity computed from these sample values will also be a random variable.For example,c3nsi der the mean value x and the variance S of the sample.If a series of different samples of size N were selected from th? same random variable x,the value of x and S computed from each sample would generally be different. Hence x and S are also the random variables with a proba P(x) and Q(S ).These functions are alled 2 "Sampling distributions" of x and S . If the variable x is normally distributsd with a mean of u and a variance of us' , the sampli?g diitsibution of the sampleX variance S is given by 141 : where : -T is the Gamma function is the Gamma function t = F(X-,uxJ/S nn= N-l Pit ( is the Student distribution function wi& n=N-1 degrees of freedom. 4. CONFIDENCE INTERVALS The use of sample values as estimators for parameters of a random variable has been discussed previously.However those procedures result only in point estimates for a parame ter of interest.8 more meaninful procedure for estimating parameters of random variables involves the estimation of an interval,as 02 posed to a single point value,which will i" elude the parameter being estimated with a known degree of uncertainty.Such an interval can be established if the sampling distribu tion of the estimator in questi n is known 9 Y based upon For the case of the2variance d a sample variance S computed fcom a sample of size N,a confidence interval can be esta blished as follows : where: r I %;d\t The degree of trust associated with the confidence statement is 4-d and it is called "confidence coefficient". Furthermore, if d 2 is unknown, a confidence interval can s&l be established for the mean value ux based upon the sample values x and S as / follows: - 241 45 - H3 interference generators (emissive equipment switched ON).When no interference occurs at < 0) one may consider system level (IM the effect of thsYgmissive equipment (step c). This measurement will still be referred to the confidence interval ILo,Uol and ILo',Uo') defined in step a) (see Fig. 1). Therefore an effective interference margin can be defined as follows : where: n=N-1 “Cm:dl2 The degrke of trust associated with the and it is confidence statement is l- d called "confidence coefficient". = 2010glo(Lo/L2) , for%iN,f;ji6(13a) IM EFF 5. TEST PROCEDURE AND SAFETY MARGIN EVALUATION The procedure for estimating the interference margin related to a parameter of interest has been subdivided into three steps: a) evaluation of sample mean value, sample variance and confidence intervals for the parameter of interest by means of the specification characteristics of the equip ment under test b) evaluation of sample mean value, sample variance and confidence intervals for the parameter of interest with minimum number of equipments switched on and interference generators switched off (minimum noise co" dition) cl repetition of step b) with the same number of equipments switched on and interference generators switched on, one at a time on the aircraft. The step a) allow to determine the cop fidence intervals /Lo, Uol from the specification characteristics of the equip ment under test. By means of the test procedure recorded in the step b) one may verify that the sample mean value and the variance, with minimum noise condition, falls into the confidence interval est.2 blished in step a). Fig. lA, B shows a comparison between the confidence interval ILo, Uol and ILo', Uo'I related to the specification limits of the equipment under test and the confidence interval ILl, U11 and ILl’,Ul’I obtained, for the same parameter, with minimum noise con dition. An interference margin at system level may be defined as follows: IM = 2010glo(LO/L1), for XNC ZO SYS (12a) = 2olog10(LO'/L11), for6:6; IM SYS (12b) IM = 2010glo(u1/uo), SYS for Qx, (12c) IM = 2010glo(U1'/UO'), forCi,$ SYS (12d) = 2ologlo(Lo'/L2'),forc+,<6+13b) IM EFF h11.3O = 2010glo(U2/Uo) , for%,;<(l3c) IM EFF IM = 2010g10(U2'/Uo'),for~$&.3d) EFF 0 An interferent situstion occurs when IME& 0. In this case [~2,U21 and (~2',U2'1 are the confidence intervals computed with emis sive equipment switched on.Moreover L. and U. are the lower and upper limits of the c&fidence interval for the sample mean distribution function,while L! and U! are the lower and upper limits of'the coifidence interval for the sample variance. I I I I I Fig. 1 A I I I Example of curves sampling distribution function : (1) not interferent case(L *L 1,2 0 and U ._GU ) , (2) interferent case 1,z CL1 *‘< t 0 Lo) Fig. 1 B %'P4*,u;ri+, I’ I' I I Example of curves sampling distribution function : (1) not interferent case(L' B L' 1,2 0 and U' 4 U' ) , (2) interferent case 1,2 0 (L; 2L L' ) 0 An interference situation (IMSys> 0) in the minimum noise condition is usually a clear symptom of an integration malfunction to be solved at system level. It is necessary to remove the cause of interference at system level before considering the effect of the 6. PRACTICAL IMPLEMJXNTATIONOF THE THEORY The evaluation of this confidence interval in easily performed for a fixed level signal whose level and accuracy are shown. This si gnal may be represented as follows: - where : K = nominal signal level AK = maximum deviation IM All the signal values are included in the ra; ge IK - AK ; K + AKI. This range will be assumed as the confidence interval of the sample mean distribution , with a confidence coefficient 1-d equal to unity and a sample mean value equal to K. ILo,Uol IK-AK;K+AKI = - = 201og "; IM loSTAT L' 2 (14) a(t) = K + AK 242 (15) The maximum variance related to the signal of interest may be evaluated considering the following choice of signal samples: S1 = K S2 = K+AK S3 = K-hK Therefore: STAT = 201og 10 = 201og 1o IM STAT ,6;++$ u2 Ul , x U2' Ul' t 62,+,2CN2(18d) ILo’, UO’( = 10; 4~~1 (17) and the confidence coefficient 1-d is equal to unity. When the equipment specification does not allow to obtain the mean value, variance and confidence interval performed in the previous paragraph, the measurement procedure can be carried out referring to the confi dence interval calculated with minimum noise condition. This is a worst case situation because the confidence intervals [Ll, U11 and ILlI, Ul' I should be usually narrower (IMSYsI fl) than the confidence ontervals [Lo', Uo' 1 and ILO’, Uo'I determined by means of the equipment specification. In this case one may define a statistical interference margin,as follows : L = 201og 1 IM STAT 10-,?N+*" .L 2 2Irl (18a) (18~) (19) High probability of an interference exists when IM STAT' *' The test procedure summarized previously requi res the execution of a large number of measu rements. In case it is necessary to reduce the measurements number,only one mean value (or variance) measurement can be made during the test. The effective interference margin can be defined as follows : ,; ,z 0 The minimum variance value is related to the case in which all the samples are equal to K. Therefore the minimum variance value will be equal to zers and the mean variance will be equal to AK /2. Therefore the confidence interval of the sample variance related to the specification limits is as follows: N = IM - IM IM EFF STAT SYS E IM = 201og N+I EFF 10U N=3 ax N+I The statistical interference margin is rela ted to the effective interference margin by means of the following relationship: L IM = 201og EFF lo+-x N+I where: (18b) IM = 201og 10 &?EFF N+I 4;; o ax N+I o (20a) (2Ob) 2 (2Oc) ‘&+&, O?N+l IM = 201og lo ___, Cr2N+I EFF Uo' ~~N+I';~(20d) In this case the statistical distribution of the values x 02N+I will be neglec 2r'I and ted (see Fig. Fig. 2 Example of execution time reduction (1) not interferent case:* ore I?+I 3 1 L and x arc I U 0rU' N+I 0 0 2 (2) interferent case: x or flN+I ,( L L 0rL 0 :, 0 or L' 0 - 243 The mean value of a sample of N independent /ux observations of a random variable is assumed as an estimator of the true mean value u . Now the sample value u will not come bug exactly equal to u 6" ecause of the sampling /O variability associated with u .In order to / establish the probability of &is error,it is necessary to specify some deviation of the true from the assumed parameter u . parameter ?O /x If the true mean value were in fact = +d (21) PO Px - an error would occur with probability (3 ,if the sample value u falls below the upper li mit or above the ix ower limit of the confidence interval (see Fig. 3). By means of analytical analysis 141,the probability (3 is related to the samples number N,as follows : S(tm;dlL d, where : Ctfi;e) ' (22) I S = standard deviation n = N-l 1-d = confidence coefficient d = maximum deviation The previous considera ions are also valid when the sample variance Q' an esti .x mator of the true varifn:: assumed :f120=df a + d; .It follows that : n o the samples number N. Another error can occur because the area related to the confidence in terval is not equal to unity. The parameter under test can assume values out of confidence interval with probability d . Therefore dis the probability of this second type of error. The interference margin is computed with a un certainty related to an error probability equal to : Prob IERRORI=Probl(TYPE 1 ERROR) U (TYPE 2 ERROR)1 = =0(+(-J-d? 45 - in this section to emphasize the advantages of the statistical analysis technique summarized in the previous pages.In some aircraft there is the possibility of using the onboard compu ter to carry out the analysis previously de? cribed. Each parameter is sampled with and wi thout emissive equipment activated with the following input data : - samples number : 500 - confidence coefficient : 95% - time repetition or sampling :30 s - data requested : x and S. A performance safety margin typical of the pa rameter under test is calculated according to the expressions (20).Some examples of the test results are reported in Fig.6 4,5,6.In this case the system under test is the air data co2 puter (ADC);the emissive equipment is the UHF transmitter. The ADC parameters controlled by Test Integration Program (TIP) are : - True air speed (TAS) - Calibrated air speed (CALAS) An interferent situation occurs (P.S.M.) 0) when the UHF transmitter is activated. Thetest has been performed as described previously in order to reduce the test execution time. In absence of specification data the P.S.M. has been computed with reference to the confidence interval evaluated with minimum noise condition G,. Fig . 4 (24) Fig. 3 Diagram of the error probability area. EXAMPLE OF THE STATISTICAL APPROACH 7. Some experimental results shall be described H3 Fig . 5 - 244 In some cases the analysis may be performed by using a dedicated instrumentation. The di gital signals on the data bus are analyzed with a serial bus analyzer in the monitor con figuration. All the data words transmitted on the bus are checked and selectively cap$ red for the statistical analysis. The serial bus analyzer can be programmed by the user in order to perform a capture actor ding to the flow-chart reported in Fig. 7. Two factors have been used in order to select a proper data word in the message : - terminal address - data word position in the message Fig. 7 Data acquisition program flow-chart For instance Fig. 8 shows a 32 data words message transmitted by a remote terminal on the basis of a predefined command word. When the program runs on the bus analyzer during the data bus activity,only the messa ge corresponding to a set command word and only a data word into this message,defined by its message location,are captured. All the captured data words are placed in the inter nal bus analyzer RAM memory. At the end of the capture phase all the samples in the RAM memory are transferred to a host computer via RS-232 or IEEE-488 for a statistical analysis, according to the previous theory. - Fig. 9 shows a memory page in which it is pas sible to see the captured data corresponding to the sixth word in the messages of Fig. 8. These data words are not equals for each sam pling,but the values change sometimes. Fig. 9 Example of memory page:the data words are stored in the memory by inverting the most and the less significant bytes. This test procedure allows to discover all the interferent situations produced by spurious pick-up in the area before the encoding section (external sensors,black boxes or remote termi nals,interconnecting cables and so on) where the signals are coded according to the 1553B protocol and transmitted on the line. When the interfere:iceeffect is directly related to the on line signal(for example in case of impulsive noise due to relays enclosure),this causes a change on the 1553B signal waveform which results in a transmission error. Also in this case a sta tistical analysis of error generation can be per formed by the bus analyzer by activating several times the interference source. Fig. 10 emphasizes a typical data transmission affected by errors generation. Fig. 10 Updating of terminals activity with tran mission errors presence on the data bus. Fig. 8 Example o~~~~~~~~~~F~s~~P~~:D~ a. CONCLUSIONS The test procedure described in the previous sections may be performed monitoring some pa rameters of interest one at a time. In many cases the value v under examination is rela ted to the other parameters means of a functional relationship.For example: is the (25) - 245 when 8 is interferred (IM > 0), the following situations are pE%ble: - the interference is due to the transfer function dfl?f2 -. . ) of the equipment. - the interference is due to one or more parameters In this f$N. case the equipment deos not. produce further interference effects. - the interference is due to the parameters and to the equipment transfer function at the same time. Therefore in order to detect the cause of an interferent situation it is necessary to show the transfer function of the equip ment under test and to test, with the emis sive equipment activated, some or all the parameters $ . . In many cases these are not always accesstble for the test or the transfe _ function g ( di) is unknown. Therefore it is necessary in this case to determine an analysis technique which allows to discover the actual interference source . This problem has not been solved at the present and the purpose on this section is to emphasize the need to go back to the effective cause of a malfunction, so that this will be avoided easily. 45 - ------------- H3 -- REFERENCES Ill C.W. Stuckey and J.C. Toler, "Stati stical Determination of Electromagne tic Compatibility", IEEE Trans. on EMC, Vol. 9, pp. 27-34, September 1967. 121 D. Middleton, "Statistical-Physical Models of Electromagnetic Interference", IEEE Trans. on EMC, vol. 19, pp. 106-127, August 1977. 131 H.P. Hsu, R.M. Storwick, D.C. Schlick and G.L. Maxam, "Measured Amplitude Distribution of Automotive Ignition Noise", IEEE Trans. on EMC, vol. 16, PP. 57-63, May 1974. I41 J.S. Bendat, A.G. Piersol, "Random Data: Analysis and Measurement Procedures", Wiley-Interscience, 1971. I51 MIL-STD-1553B, "Aircraft Internal Time Division Command/Response Multiplex Data BUS". _. - - - - -- - - - - - --- - 247 46 - H4 THE STATE OF ART OF TV RECEIVER IMMUNITY AND RECOMMENDATIONS FOR APPROPRIATE CONSTRUCTION, DEDUCED FROM TEST STATISTICS R. Bersier Swiss PTf, General Directorate R 8 D Division CH-3030 Berne, Switzerland SUMMARY The inanunities of 16 TV receivers of recent design were tested by the "current injection" method against disturbing AM and FM sources in the HF and VHF range. From the results good state of the art immunity levels are established. The constructional features of the tested receivers that could influence the immunity are tabulated for the antenna input, RF-tuner, IF and mains circuits. From these findings reconunendations are made to improve the construction. 1. INTRODUCTION Recalling the problems concerning the RF immunity of TV receiver installations. During the last years there has been a monotonously increasing number of complaints concerning the interfered TV reception, due to the lack of immunity of the TV receivers or of the installations. These interferences may be divided into the following two principal classes a) and b): a) Interference due to the insufficient TV receiver immunity in the short waves (SW) range. The deterioration of the situation is surely due to the increasing number of transmitters (amateurs and citizen band), that are operating in residential areas. To this interference class belongs the problem of the increased licenced transmitter power and the substitution of the amplitude modulation (AM) by the frequency modulation (FM). In order to clarify the situation, comparative tests were modulation. made with both types of b) Interference due to the insufficient immunity of the TV receivers or of the cable TV distribution systems (CATV) in the reception channels. It is well known that it is not possible to reuse in medium or large CATV systems the same channels that are occupied by local or powerful regional TV transmitters: Due to the bad immunity of the TV receivers and the CATV itself, qhost pictures are created by direct irradiation of the TV or the house distribution part of the CATV. A similar problem arises for the special TV channels (out of the TV bands), if the transmitters of various fixed or mobile services, operating in these frequency bands, are placed in residential areas. up to now the problem was solved by the appropriate choice of frequencies in the CATV, i.e. free channels in the TV bands I and III and non interfered special TV channels. But the problem is becoming acute and difficult to circumvent because of the constantly increasing number of TV programs to be transmitted and of radiocommunication services, located in the interband frequencies of the CATV (104-174 MHz and 230-293 MHz). Therefore it is urgent to improve the innnunity of TV receivers and CATV installations. 2. TEST METHOD The immunity tests of the TV receivers were made by the "synthetic" or "current injection" method, that is already described in [I], [21, 131 and \41* Recalling the principle of the current injection method. This method simulates the dominant effect of the disturbing electromagnetic field on a realistic installation, by injecting an asymmetrical current from a real source (Ri = 150 Q) through the connected cables to the TV'S chassis. The immunity is specified by the electromotive force of this source (E.m.f. in dBpV) that creates a just perceptible interference in the picture or in the sound. Fig. 1 depicts the test set-up used. On each cable (antenna and mains) of the TV receiver under test a coupling unit is inserted. The disturbing common mode current is successively injected on each cable through a source having 150 Q resistance, the other cable being connected to the ground plane through 150 8. The relation between the E.m.f. value that creates an interference by the current injection method and the electromagnetic field that creates the same interference in a real installation was established experimentally, ~;~;o'l~;e'h~sso&l.inq relations may be : 1 to 40 MHz: disturbing field(dBnV/m) 3 E.m.f.(dBpV) - 7dB 50 to 230 MHz: disturbing field(dBpV/m) I E.m.f.(dBnV) 3.SCOPE OF THE TESTS The immunity tests were made on 16 TV receivers with coaxial antenna input, system PAL B/G. The receivers were of recent design 24e receiver is 22 dB outside the TV bands, 34 dB at the IF and 34 to 52 dB in the reception channels. b) The interference effects are more pronounced with AM than with FM (difference of 4 to 7 dB in the median values of the disturbance source E.m.f. that create a just perceptible interference; the values refer to the carriers according to sec. 3). c) 50 % of the tested TV receivers may be interfered by approximately the following field strengths, when using AM: 115 dBnV/m (0.56 V/m) in the range 15-30 MHZ 109 dBnV/m (0.28 V/m) in the range 68-174 MHz at the IF 81 dBnV/m (11 mV/m) N 52 dBnV/m (0.4 mV/m) in the reception channels (1981-82) and originated from different manufacturers, they are denoted in this paper by the letters A to P. The immunity was measured in the following frequency ranges: a) 15- 30 MHz:range of interference from SW, b) 68-174 MHz:range betweenTV-bands I and III, c) 32- 40 MHZ: TV'S intermediate frequency(IF) d) reception channels: 3 (54-61 ms), S7 (146-153 MHz), 7 (188-195 MHz) and 12 (222-230 MHz). In the first three frequency ranges, the interferences produced by AM were compared to those produced by FM. For this, the disturbing signal was modulated with approximately 1000 HZ in the following ways: -by AM, at a depth of 80 % and then - by FM, at a deviation of 5 kHz. 5. INSPECTION OF THE TV RECEIVERS' CONSTRUC- Note: When applying FM, the interference pattern appears clearly more annoying with a small frequency deviation than with a large one. The E.m.f. of the disturbing source was measured using an average detector; therefore its indication corresponds to the rms value of the carrier and is not affected by the modulation. For the tests in the frequency ranges a), b) and c) the unwanted interferences that could be created at the IF or in the reception channels by the harmonics of the disturbing source were prevented by the use of appropriate low pass filters or by the suitable choice of the reception channels (see notes in the diagrams). 4. INSPECTION OF THE IMMUNITY TEST TION OF T3 Fig. RESULTS F w : wanted signal signal T2 Gl Am F T2 Measurement sotup for -1 Interference RF signal generator, 1.5-230 MHz. Broadband power amplifier (Sh: Shielded box.) Low-pass filter. Power attenuator, 6-10 dB, 50 fl. thp immunity THE IMPROVEMENT r--z-; 3 Tl ----f-3cq~Gl I Metallic ground plane Mains coupling unit Rntenna coupling unit TV test pattern generator FOR IMMUNITY 5.1 Influence of the screening of the antenna input circuit and of the tuner. Pursueing the disturbing current,injected on the screen of the coaxial antenna cable, we encounter successively the antenna connector, the optional insulating capacitors, the cable to the tuner and finally the input of the tuner. Then the disturbing current is distributed on the screen of the tuner before arriving on the chassis ground circuit of the TV receiver. The value of the interference voltage produced at the input of the tuner will be directly proportional to the sum of the transfer impedances of the components of the input circuit; that is why these impedances should be reduced to a minimum. The care taken to the screening of the tuner (elimination of slots, well contacting covers) and the performance of the various feed throughs (filtering) will be decisive in preventing the penetration of disturbing signals. The above considerations are reflected in the diagram 4, representing the immunity in the reception channels: i = interference pJ-@-P 1: RECOMMENDATIONS Fc I Pl M A P AND THEIR The constructional details that may influence the immunity are indicated in table I. Comparing these to the diagrams 1 to 4, we make following observations: The E.m.f. values of the disturbing source that created a just perceptible interference are indicated in the diagrams 1 to 4 in a statistical representation. The corresponding values of the field strengths that would create the same interference in a real installation are given on a parallel scale. These values were obtained from the relation given in sec. 2. The inspection of the diagrams 1 to 4 shows: a) There is a large spread in the results: The difference between the best and worst TV I - test of L- - I T -_I Sh TV receivers by the current inj and (Utij ). Thus, in order to determine a conventional function of probability distributions Fd+ it is first necessary to find a composition of lognorma1 laws. As shown in [I] , a determination of a composition of these laws is unfeasible in an analytical form because a characteristic function of these laws is expressed by an infinite series. It is proposed in [I] to use an approxinlatiorlof lognormal laws by means of gamma functions with the aim of finding of an approximate solution of this prob?.im. In so doinn. wheref([& P and (r+l) are tl;anmla functions; are parar!letersof+a probability distribution. is aasily found in a similar way. In order to estimate the error of an accepted approximation the values of IJ,, were calculated, which cosrespond to comparatively high levels of probabilities, F ~u,&o.o - 0.95, that are used fox es! imation of quality of radio communication. To this effect, by a method of numerical integration on a computer, values were found of g ability distribution functio%R6° #Y mormal laws with equal values of-standard deviations and with K=2;4;8;16 and of corresponding probability distribution functions which were obtained as a result of approximation, The results of calculations showed that within a wide rank;e of values of standard deviations of effective interference voltagfes, produced by individual source~,W{U~i~~ 6 - 15 dB and that the maxirrml differences between the values of UQ+ccdo not exceed 1.0-1.5 dB. In order to determine a probability distribution function F,,it is , 47H5 - 257 - necessary to find a compositior$ of distributions F_C U&and account the accepted approximation the characteristic function or a probability distribution Pti,K is whcrt: [p,,(1(;+4),aJ and Ifl&z*i),a21 are, respectively, the generalized parameters of probability distributions of values of interference from individual sources of group one and group two at the input of a radio receiving device. The values of the parameters aredetermined depending on values miandWL from graphs of Pig.A. Taking into account, firstly,Formula (r), secondly, the probability distributions for the number of sources whose interferences act at the input of a radio receiving device, and thirdly, those probability distributions of the nurtlberof interfcrences belonging to group one and group two, which were accepted in Section 3, we shall obtain an expression for a probability distribution function of effective voltages of total processes of interference which penetrate out of electricity supply networks and through the field: cpeq3’ when K=d) ; p and g are, respectively, the probability of appearance of interferences of group one and group two among the total number of B interferences which act at the input of a radio receiving device. In a particular case, when at the input of a radio reoeiving device the interferences of only one group act, for instance, the interferences radiated by electrical equipment, the formula is simplified: 5. Permissible values of total interference processes at the input of a radio receiving device can be determined from Formula (6) and condition where F-'(d) is suci a value of the argument of a function which is determined by parameters which are included in expression (6); In order to estimate a possibility of determination of permissible values of man-made radio interference from individual sources let us use a simplified Formula (7) taking into account condition (1) and a formula for b which was given above. when 0 < U2 ej& ka,pd+(N-K)aepg In Formula (6): cf i is the number of combinations of N things K at a time; is anjaverage number of interferences which act at the input of a radio receiving device; z = 7-+& - kca4&- Wic)a&g , In the CISPR, the limit$ o?interference are expressed in quasi-peak values. 'Therelationships between effective and quasi-peak values (acco. rding to the CISPR Publication 16) are in sufficient detail treated-in work [4] It is ihown there that the following conversion factor can be introduced: Kr * 20 1% .A& - 20 1% Qyp "i eff - Kr * 3 i 2 dB for a frequency band 1 O-150 kHz; Kr m 15 f 2 d% for a frequency band 0.15-30 MHz; Kx - 1'7 2 2 dB for a frequency band 30-1000 MHZ. When determining the permissible values it is also necessary to take account - of the difference between the bandwidths of the radio receiver and of the xadio interference measuring set (A fz, and A f+,e,res- 258 - 6. Discussion of the results obtained The results obtained enable to c alculatc the permissible values of total man-made radio interference processes and to establish requirements for interference from individual appliances. References pectively), - of the probability level at which the limits of man-made radio interference are specified. Taking this into account we shall obtain fox quasi-peak values of interference: CISPR/AWG2(Pevnitsky-USSR)2, April 1978. On the draft CISPR Report on Study Questions Nos. 54/l, 55 and 77. International Electxotechnical commission, International Special Committee on Radio Interference (CISPR), 20 p. Pig.2 sho& the results of calculations of Fq(d) which are included in Formula (8). AJl .KamaKoB F;:) ?5 '~acnpexeneHm Bf3pofiTHocTei2 Hanpmemi; pamonoMex Ha axoAe npuemmrca nprn owospeMeHHOM J303fie~cT~MM CJfyYaBHorC YMwa ~~~~MTHuX CMrHeaOB". Tpy~qa HlilVP 181 4) 19B3r. :i.E.iiiop C~wmcwdecwe MeTow am JIMSa M KOHTPOJI" ICaWCTBa M HQeEHOCTM. ~OH,PiZ~~O, ~:OCtiB+%?. B.lI.rlemI4r~Kr4il"I{ sonpocy 0 COOTHOWf?HMJ3[X MelEJQ~ KBa3WlUlCOBHMM M e$$eKT'GIBHHMM BHaYeHMRMM KB%3MUMl-@bCHhrX I'IffOIJeCCOB pazqvlonmex 0 Tpym - 259 48 - THE APPLICATION AND DEVELOPMf8NTOF Em Q. ~6 IN CHINA Chen Y.C. Zhu China AtiationResearchIn&.tute for Standardiaation Beijing,China This paper introduces comprehensively the application and development 70's to the early 80's, China possesed of EMC, including the new efforts in engineering development, measurement teohnique,standards and SpeCifiCatiOnS, EMC training in industrial, Scientific, which had medical, aeronautical and ship building area in China, and EMC international cooperation. ried out for 4 years to study and analyse the condition of electromagnetic pollution and interference in China. ~&XXXHF plastic hot-jointing caused a great interference for lack of perfect shielding, ground.. ing and strict spectrum OOntrOl.InVeStigation and measurement had been In 1970'9, electromagnetic ference caused General machines by various car- inter- radio-fre- quency equipment used in industry,science and medicine interfered in quirements for anti-conducted interfew 2596 of coverage area of TV broadcasting of main nce was placed on aeronautical industry, large and middle cities, so that the TV electronic industry, and ship building etc. in China. But for radiated inter- sets could not get clear picture.During ference, the requirements for electromagnetic susceptibility didn't come to of aeronautics, ship building,broadcasting and TV,labour and environment pro- In 50's_60's of this centry, the re- a decision to prepare standard then. Since 1970's, the application and df+ velopment of EMC in aeronautical, ship this period the departments concerned tection, and railway transportation had made a number of live measurements and statistical analysis of electromagnetic building, television, communication,and interferenoe, prepared or revised stan- electric railway dards, in which allowable transportation areas have progressed rapidly. In this period, limits and ference, the performance of some equip- test methods were established. In the same time, EMC engineering design and measurement technique had further deve- ment loped. The quantity of researchers because of intra- or inter-system inte* degraded to such they could not an extent that work normally. Electro- magnetic transmission with large power subjected the inflammable and explosive materials to a very hazardous situation. In order to eliminate oocured electro- engineers who engage in EMC had increased greatly. A lot of progress had been made in respect of test and measurement, standard and specifioation,analysis and prediction, research and development. magnetic interference,great expenditure and time must Sometimes be spended. For example, according to statistics, from and Research and Development - 260 In 1980's China had already applied the EMC analysis and prediction techniques to aeronautics, telecommunication and 90 on. When the multiplex system was used in aircraft, it was found that EMI caused by electronic and electric equipment is the main cause for operation failure of this system. The way in with multi-dimensional random process can be resolved with computer was - mated only by describing with quasi Peak, when estimating interference characteriatics of analog and digital system. The MD COtmmiCatiOn measurement has been adopted in China. The develooed measurement instrument in APD conjunction with microprocessor have many functions such as control. computation, sampling, data processing, display, typing and adjusting measurement time. The results found, thus the code error rate caused by conducted, radiated interference and of measurement may be displayed by spike can be predicted, providing thereby a scientific basis for reliability design of the system. in decimal number. It also can be plot- Amplitude or be typed out by microlattice hrpewiter ted out by x-y plotter or recorded tape or magnetic disc. probability destribution (At?D)measurement interference is of electromagnetic the most advanced mea- surement method at present. It is also an effective way for The structure and establish has been proved that effects of inter- ference could not It be effectively esti- Compar&tar I 1 bAefOJ”nce Amp I Vo/tayO AI IF Fip i. blocK d'lagram APD instru- input signal: intermediate frequency output (-" 2.5MHz) or peak detected output measured item:It can measure Am and NAD simultaneously to 10-4s for APD and to IHz-40KHz for NAD measuring level: inl2steps, sensitivity of the corn... parator: 1.5mv interference mathematical statistics. digram of AFED The main performance of ment are as the following: the effects of interference on communication system. This method can be used to find out the nature of interference source block on instrument is shown in Fig 1. people to study model by LDE of APD measuring instrument - 261 measurement times At IF of I.67 ~~IZP the time for APD measurement can reach 1285 seconds; at IF bandwidth of 12oKRz, up to140 seconds for NAD 48 - f erence equipment, susceptibility test signal source, and RMI/RFI dattiacquisition system, manufactured by companies in West Germany, U.S.A. UK are S,Scifications and Standards _-_)___~,_,I-,_____-_-~__"_- BYD circuit, including 137 digital chips used as comparator. With interference produced in Cities and on electrified railway,ApD measurement have been successfully conducted. The hazards of electromagnetic energy such as electrostatic hazard, hazard to inflammable and explosive materials and effects on animals are subjects under studies in labour and environment protection and medicine areas. The EMC techniques are adoped in hardware design In the early 60's the standards electric equipment, and standards on their test method were released. SPecifications and standards have developed rapidly, composing a complete series of ErilCstandards and specifications. Among them, the primary specifications are China has completly adopted the measuring instrument specifications prepared by CISPii,All publications No.l-16 of CISPR were translated into Chinese RF anechonic chamber of various typies and sizes with conical or rectangular shapes were built. Their working frequency ranges are laid on P through X bands.These cell.3can be used to perform simulating measurement of air space, to study RMC of system,to test the radome and to measure antenna In the meantime many related national general advanced standards ment, the measurement is conducted in standards inter- and world have been studied in detail. For radio-frequency equipment system and radar cross section etc. With respect to El% of marine environ- following: tl3 Requirements for electromagnetic interference and electromagnetic susceptibitity, Measurement methods, Definitions and System-of-unit, Lightning protection, Static charge protection, Electromagnetic compatibility of system ( EMCS ), Specification3 for electromagnetic interference measuring instruments, EhlC requirements for electronic instruments, Safety standards of microwave radiation. Measurement Technique A number of on the conducted radio interference characteristics of marine r?ndairborne instrument3 and radio apparatus, household by machine building industry. and issued, also im- ported. The midium size chips are used thmughout this ~6 dustry, science and medicine, electrical equipment, radio in in- household and T.V. the test field simulating the seas. The receiver, mobile vehicle Coupling measurement and the characteristics of radiation patterns are pre- and high-voltage dicted through simulating by scale down model. publication No.16 of CISPR are applied. The TEI'II cell is already used to measure E!{Cto study radiated interference system, power the documents and igniter, transmission of II%:,CISPH end For zhips,"The Proposal of Ship EMC" is used. This proposal delineates cri- teria to be applied to verification of EM1 and anti-interference. Test freq- and charactristics of radiating source. Nowever,this cell is mainly used to calibrate measuring equipments. As f Or measuring equipments,in addi- uency ranges from 0.01 to 1000 MHz. For aeronautical equipment, the documents "Terminology of EMI and EMC" tion to domestic products, and some equip- ment, including electromagnetic inter- "EMC Requirements for Aircraft and Test Equipment" are Method applied. - 262 - These standards specify in detail the emission limit and susceptibility of conducted and radiated electromagnetic interference and their test method. The item to be tested are 13, including conducted interference of power, signal and control wires,and anttena end: conducted sensitivity of power wire and control wire: mutual modulation, cross modulation, and sharp peak signal: and radiated interference and its sensitivity. The equipment are classified intO categories as specified in these specifications, so that different items can be selected to test according to category of the equipment. The requirements for compatibility of system delimitate a 6 dB safety factor of electromagnetic interference, but for inflammable and explosive materials this factor must be 20 dB. China takes great interest in adopting international general standards and world advanced standards. Some EMC standards of IEC, CISYK, MIZ, ISO, HTCA and BS etc. have been widely issued and selected for use in purchase contracts. the standard correctly. The short terms lectures are jUSt like lectures given in college. The main contents .forthe lecture contain introduction of EMC, hardware engineering design,standards and specification% methods of measurement, analysis and prediction ( software ). The Person8 attending each lecture counted 100 to ZOO. The coverage of academic discussion ard wide, the form are different. Since discussion has 1980, national academic been held twice. Usually the meetings of different industries are held more frequently. People who attend this kind of meetings are in hundreds. Local EMC symposium are also held more freqently. In these symposium, protection of EM1 caused by HF plastic hot-jointing machine, the interference of electric locomotive on receiver, mutual interference of inboard equipment, susceptibility EM1 measurement, and measurement of safety factor were disscussed. International Cooqaration Since the late 70's, China has EMC Training already begun to exchage EMC technique Over the years, a EMC tranining has with many countries and international been given to the people who are engaorganizations taking part many times in ging in design, test,production control activities held by CISPE, ISO and IEC. in various ways. The scope of training The chairman and experts of CISPH and includes basic theory, guideline8 for SAE visited China and delivered lecture engineering design, test method, stanand heldacad ozic discussionwith &he Chinese dards, specifications and project plan and control.EI% training has made those experts. The Chinese experts and engiengineers understand and use EMC technineers are well informed of transactions ques correctly at each phase of design, of EMC symposium and exhibition and EMC production, test and maintenance. proceedings of IEEE, They are very Methods of training are interested in these information. Et$g;Elish booklists and magazines As described above, China has imported several complete sets of EMI/EMC to givi short terms lectures, to test, demonstrate and exchange equipment made by some campanies of the views on the-spot, U.S. UK and Westen Gemany etc. OF all to give instructions in interpretaothers they are H/S carp, HP carp, tion and application of certain stanE.T.N. COHP, SINGER Corp. The equipment dard or specification, to Solve problems of EM1 with EMC imported include interference field technique, strength meter, spectrum analyzer, to introduce application and popilari.verioua signal sources, detector, data eation of new EMC technology abroad, acquisitionand analysis to hold EMC symposium. system and Before and after a new standard is software pertinent to them. Their operissued,we usually hold meetings to diaation frequency is up to IH GHz. They cuss some technical problems. At the have been used in many departments in meeting, the people who prepared the China such as aeronautical and shipdra.fts of the standard shall explain building industries, broadcasting and and verify some important problems by T.V., traffic and tele-communication necessary demonstration and calculation etc. to make persons concerned understand - 263 49 - I1 NEW WAYS FOR INTERFERENCE COMPUTATION AND MONTE-CARLO-OPTIMIZATION TO GUARANTEE THE COMPATIBILITY OF INDUCTIVELY COUPLED LINE SYSTEMS H.-J. Haubrich Vereinigte Elektrizitdtswerke Westfalen AG Dortmund, Fed. Rep. of Germany Summary The electromagnetic compatibility of the components of power supply systems is a criterion participating more and more in the decisions of network planning and operation. In the following problem cases of electromagnetic and electrostatic induction by alternating fields at power frequency are treated, caused by power lines of the electricity supply system to lines of the own and of external energy or information transmission systems. The basic idea is to transform the,inducing voltage Up and current Ip of the field producing conductor P (e.g. the high voltage line) into an injected current IpV with help'of the mutual inductances and capacitances LPV and CpV between the conductors P and V. Line V can be an element of an extensive meshed network V consisting of pipe-lines, telecommunication lines or high voltage lines. Tnis network is modeled by its admittance matricYV, thus enabling an easy representation of any optional topology. With respect to a possible impairment of technical installations not belonging to the electrical power system, special attention has to be payed to three-phase overhead lines as a source of electric and magnetic stray fields. They leave the closed electric substations and meet with practically all other line systems when crossing wide regions. The principle of trace bundling followed by the licensing authorities in Germany forces the power lines to have long parallel runnings with a corresponding strong coupling to other lines. High voltage overhead lines are technically and economically predestinated for high transport capacity. Due to their high operating voltages and currents and extraordinary high short circuit currents they can produce interferences in quite a wide range. @mpensator Fig.1: Electrically coupled lines with given inducing directions The calculation scheme is independent of the feeding conditions and of the number of disturbing lines P: the resulting source current is obtained by geometric addition of the single components. The view of the acceptance or inadmissibility of these interferences requires the quantification of the physical effects. The analytic simulation leads to reliable forecasts with respect to system stresses and suitable countermeasures. A universally valid algorithm for the calculation of interference voltages and currents in networks of optional topology or with several exposures to power lines is presented and applied to practical problem cases. In combination with the Monte-Carlo method, one can optimize the protective earthing of pipe-lines against dangerous induction voltages. I A universal algorithm for interference calculation The capacitive and inductive interaction of coupled lines according to fig. 1 can be calculated with the uniform algorithm shown in fig. 2. i_______+ Coupling I Matrix v-c Fig.2: Computing scheme with injected currents 264 When several lines of the network V are simultaneously involved in an exposure, current sources have to be added at the boundary nodes of all interference sections. The resulting network equation - Both above mentioned constraints can't be simultaneously be fulfilled within the permitted mistuning range. There is a demand for additional provisions to guarantee the compatibility of both voltage levels on the same tower, e.g. by transposition of the 400-kV-circuits [3]. is well suited for the computer calculation of the required nodal voltages WV to earth and the induced branche currents IV. A necessary presupposition however ist the existence of homogeneous line sections with uniform exposure to the inducing power lines which have to be approximated by sudivision in case of need. Additional extraneous earthed conductors C, e.g. earth wires, metal cable sheaths or special compensation conductors, are often involved. Their induced currents IC, determined in the same way as IV, additionally act upon V by the coupling C-V, diminishing the interference in case of zero-sequence components IR, but partially also increasing in case of symmetrical operation of the disturbing three phase system P. The reaction from V to C cannot be generally neglected; in case of a close proximity V-C, the dashed feedback in fig. 2 has to be regarded. Point Fig.3: Neutral point displacement voltage Uo(Ir) of a llO-kV-network with capacitive coupling to 400 kV lines Electrostatic induction in high voltage networks with resonance earthing Resonant earthed power systems must fulfill two main restrictions with contrary demands for the tuning of the arc suppression coils: 1. in case of a single phase-to-earth fault, the residual earth current Ir should still ensure the self extinction of the arc (e.g. I, < 130 A for the llO-kV-voltage-level cl]), Electromagnetic induction in pipe-line networks Modeling The mathematical model of a line with uniform exposure to the magnetic field of a power line is derived from the line element ds in fig. 4a. 2. under normal operating conditions, the neutral point displacement voltage U, should be kept as small as possible (e.g. U, < 10 kV in llO-kV-networks). _,Fig.3 shows the calculated geometric locus [2] of the voltage vector U, in an IlO-kV-network when mistuning the arc suppression coils during normal system operation. U, results from the capacitive induction by two 400-kV-circuits installed on the same towers with two circuits of the IlO-kV-network for a length of lpv=53 km Resulting from the geometrical unsymmetry of the phase configuration the 400-kV-circuits transfer a high zero sequence voltage into the llO-kV-network even if the three-phase inducing system is balanced. The equivalent injected currents are easily found, when the 6x6-matrix Cpv of the mutual capacities CPV per unit length is known: i,, = jw lpv* C,; u, (2) Fig.4: Equivalent circuits of homogeneous lines with inductive interference a) line element ds with the induced e.m.f. EpV b) uniformly exposed long line with the induced current source IpV YV:propagation coefficient of line V WV:surge impedance of line V 265 The induced electromotive force EPV per unit length E PV iEn = c IPi'ZPVi (3) P - 49 The exact line equations of a posed section lv coshyv 1v LI U Vl E includes the contribution of the np inducing lines P. When calculating the mutual impedances ZPV between the lines P and V by the Carsson [S] formulae even oblique exposures may mostly be handled as equivalent parallel lines in the distance Jal'a2 (fig. 5). PV uniformly ex- WVasinhYVIV uv2 . = Wv'IvPyv I1 sinhyv 1v w WV'coshYVIV I L v' I+% v2 y, I yield immediately the required injected currents IPV (5) = EpV/(~V.WV) representing the inductive interference in fig. 4b. After dividing in line sections of quasi uniform exposure any network can be composed by such basic quadripoles which define the nodes and branches of the passive network model. y,. i iwm lil middle gmn.tric dtrtmc. lioo w - Fig.5: Failure F of the e.m.f. induced in a conductor V with earth return if modelling its oblique exposure to the disturbing conductor P as paralleliSm(frequency: 50 Hz;conductivity of the earth: 50 R m) Real example The efficiency of the described algorithm shall be demonstrated by the example of an underground pipe-line network whose three branches are exposed to the magnetic induction of a 400-kV-double-circuit line (fig.6). L*>,‘\ \ /’ /’ 1;=1,2kA=I,b vv=(O,O22+j 0,047)/km W,=h3,0 +j 2,4 ) R Fig.6: Real example of a buried pipe-line exposed to the steady-state magnetic field of a 400 kV overhead line. Construction of the sections with quasi uniform induction. 0 : node number; - 20 --:distance in meter a C - The length of exposure reaches more than 10 km. Following the discontinuous line configuration and coupling impedances along this length, the pipe-line has to be divided into numerous uniform sections corresponding to the vertical propagation of the power frequency magnetic field. The dotted reference axes show the way how to determine the nodes of the pipe-line model. The induction by a single-phase-to-earth fault was proved to be less critical than by the balanced loading of the 400-kV-circuits. That induces pipe-line voltages to earth exceeding by far the permitted value of 65 V (fig.7, curve a). Without suitable protection devices maintenance staff working on the pipe-line would be exposed to danger. 266 - of this so called Monte-Carlo method includes the costs CF for the protective earthing at m feasible ground connection points (fig. 8a) and the constraints uVk< 65 V, the violation of which is punished at all regarded nodes nV by a penalty function PF according to fig. 8b. CF CF \ L ---_-_---__-_ w 0 a) conrtrrlnt i. 0 R 65V t11 “V Fig.8: Examples of the cost function CF and of the penalty function PF for the MonteCarlo optimization UV 0 65 V Start Define Fig.7: Calculated induction voltages line-toearth of the pipe-line act. to fig.6 I Calculate OF Ir,) - two) I a) without any countermeasures + PF(u,) -t b) with an optimal earthing resistor of 35 51at node number 10 Chance variation of rg - I I Calculate Optimization of the protection devices 1 vector r. starting r OF(r) I 1 I no Protective earthing is usual and well suited to damp the dangerous potential rise induced by the steady state magnetic field. The search for the most effective earthing points and earthing resistances may be very complicated since the low-resistance earthing of the normally well isolated pipes jeopardizes the efficacy of the cathodic corrosion protection. no Nowadays computer-aided methods allow an automatic optimization of the protective earthing by lumped resistors. For the case in question, the method of statistical trials has proved to be useful [4]. The objective function OF = C CFi + C PFk is m; k.s nV (6) Solution CL3 r Fig.9: The Monte-Carlo search process for the optimal resistive earthing of induced pipe-lines - 267 CF and PF are functions of the earthing resistors Ri, being the control variables of the optimization process. Fig. 9 marks the main steps of the iterative Monte-Carlo solution. An arbitrary chosen starting vector I,=(RI,...Ri,.. .Rm) delivers the reference value OF(r,), that shall be improved by a following chance variation of all random variables Ri within a given margin. A sufficiently optimal solution is reached when subsequent iteration steps don't bring any more remarkable improvement. In numerous applications the convergence was reached after about 20-n" experiments, bad trials inclosed. The application on the above shown problem case yields a 35-R-resistor connected to node 10 as the optimal protecting measure. 49 - I1 References [l] VDE 0228, Tail 2/7.75: VDE-Bestimmung ffir MaRnahmen bei Beeinflussung von Fernmeldeanlagen durch Starkstromanlagen, Beeinflussung durch Drehstromanlagen [2] Poll, J.: Sternpunktverlagerung in gelbschten llO-kV-Netzen. Elektrizitatswirtschaft 80 (1981) H. 22, S. 810-813 [3] Brandes, W.; Baubrich, H.-J.: Sternpunktverlagerung durch Mehrfachleitungen in erdschlui3kompensierten llO-kV-Netzen. Betrieblithe Erfahrungen und AbhilfemaRnahmen. Elektrizitatswirtschaft 82 (1983), H.ll, s. 400-405 This mathematical solution is at least a helpful approximation, even if the implementation may be modified or adjusted by further practical considerations. [4] Schwefel, H.P.: Numerische Optimierung von Computer-Modellen mittels der Evolutionsstrategie (Basel,Stuttgart:Birkhauser 1977) The high efficiency of the described optimization method becomes just obvious in such cases where the induced voltages UV can only be limited sufficiently by earthing the pipe-lines at more than one point. [5] CCITT: Directives concerning the protection of telecommunication lines against harmful effects from electricity lines. The International Telecommunication Union 1963 - POTENTIALS BURIED 269 AND CURRENTS CABLE EXPOSED 50 - 12 ALONG AN EARTHED TO ELECTROMAGNETIC EFFECTS OF A POWER LINE UNDER FAULT CONDITION W.Machczydski Technical University of Poznafi Poznad,Poland Disregarding the influence of the current flowing to soil through the earth electrodes on the cable subjected to electromagnetic effects of a nearby a - c transmission line may lead, in certain conditions, to improper results in calculating the potential distribution. The paper presents the method of calculating the potential distribution along the earthed underground cable, taking into account the additional conductive influence of currents flowina from the earth electrodes of the cable a on the potential and current distribution. The method is illustrated with examples of calculations. INTRODUCTION The classical EMC problem in wire telecommunication is that dealing with disturbances caused by electricity lines. The resulting effects of electromagnetic interference of overhead high voltage a - c power lines can range from noise on communication lines to equipment damage or even personnel hazards. The growing use of a - c transmission power lines located in joint use right-of-way or located in close proximity with buried cables and the increasing levels of voltage and current capacity make the determination of potentials excited along the cables an important task. Where calculations indicate that the possible hazard could exist, the precautionary measures should be taken to minimalize the potentials on underground cables subjected to electromagnetic effects of a nearby a - c transmission lines. As is known, the use of properlydesigned earthing systems permits the maximum mitigation of cable potentials. However, in calculations of currents and potentials along the earthed underground conductors usually the additional conductive influence of currents flowing through the earth electrodes on the protected earth return circuit is not taken into ac_ count. Disregarding this fact can lead, under certain conditions, to improper results of evaluation of the zone of dangerous potential on the earthed conductor. The purpose of this paper, which iS continuation of considerations given in [31, is to present calculations and formulas applicable to the analysis of shield potentials and currents excited along an underground earthed cable, by 50-Hz a - c power transmission line sharing a joint right-of-way. The calculations take into account the additional conductive efect of currents flowing through the earth electrodes on the potential distribution along the cable. It is assumed in the paper that the earth is a homogeneous, isotropic medium of finite conductivity, that the underground cable is infinitely long and that the system considered is linear. It is also supposed that the currents and potentials vary with the time as exp (jut). Therefore the alternating component of the short-circuit current is taken into account. GENERAL EQUATIONS In the analysed system, shown in Fig.1, part of the current flowing along the buried cable enters the earth through the connected earth electrodes. Since the cable is in the current field of the earth electrode, a part of the current flowing from the earth electrode to the earth flows back to the cable. Currents and potentials along the cable are to determined as the superposition of two additional states [2,3,4], that is: - current energisation of the cable by the current leaving the cable in point x = xk (k = 1,2,...,n), - conductive energisation of the cable by the current flowing from the earth electrode located in point x = xk (k = 1,2,...,n). - 270 - The current 1,k flowing to the earth through the earth electrode is determined on the basis of Thevenin's theorem, hence VT(x,_ ) lek(xk) Fig. 1: Infinitely long underground earthed cable Hence I(x) = IO(x)+k;l v(x)= vO(x) Ii(x) + k$ + '; v;(x) k=l 1; (x) (1) n z vi k=l + (x) (2) where index "0" means the primary current and potential, that is the current and potential excited along the cable when the earth electrodes have been disconnected, indices "1" and "2" - current and potential along the cable for current and conductive energisation, respectively. The shield currents and potentials excited on the buried cable by an incident electromagnetic field can be calculated using the distributed source transmission-line anal sis technique discussed in [1,2,5 r . For the current energisation we have I I;(x)= - sign(x-xk)+ V:(x)= ZOIek 2 e e , -YlX-Xkl and for the conductive -ylx-Xkl = Z + Z K ek ck + 'ink (7) ’ where Zck is the impedance of the cable joining the earth electrode with the cable under consideration, Zink - the input impedance of the earthed cable when the earth electrode k has been disconnected - Thevenin source impedance and VT(xk) - Thevenin source voltage. The voltage between cable sheath and the cable conductor is obtained from the relation [21 u(x)= dI(v) “P zS 2v _m 7 _y e Ix-v I C dv, C (8) where Z, - internal-surface impedance of the cable sheath, yc - propagation coefficient of circuit involving cable insulation and cable conductor, and the current I is given by eqn.(l). EFFECT OF SHORT-CIRCUIT CURRENT POWER LINE ON THE EARTHED CABLE IN A One of the methods used to protect earth return circuits against the electromagnetic effects of a power line is the earthing of the circuit under protection in these places where the highest values of potentials take place, e.g., in points opposite to the ends of approach section of a power line, as seen in Fig.2. (3) (4) \ energisation (5) Z where v Z el e2 : - characteristic zO cable, impedance of the Y - shunt admittance of the cable, Y - propagation coefficient of circuit involving the earth and cable sheath, 0 - earth conductivity, S - distance from the earth to the cable, Y!rQ - Sunde's functions [5]. electrode Fig.2: Underground cable earthed at points of maximum inductive influence of a power line A more general case of the influence of a short-circuit current of a power line on a nearby underground cable is shown in Fig.3. In the system presented in Fig.3 the fault current IO of a power line influences both inductively and conductively on a nearby cable. - 271 5012 - It should be pointed out that in case of the system presented in Fig.2, that is, the case of inductive influence of a power line, the primary currents and potentials along the cable under the influence are drawn from rewith the ommission lations (9) - (ll), of the terms containing functionsyand R (representing the conductive influence of a power line). EXAMPLES Fig.3: Underground earthed cable in the vicinity of a - c power line with earth fault To simplify considerations, it has been assumed that the power line consists of only one overhead conductor with the resultant earth fault current, while the fault current flows into and out of the soil through point earth electrodes placed on the surface of the earth. If the line is energised at the point x = L, and the earth fault occurs at the point x = 0, the primary current, potential and voltage between the cable conductor and the sheath of the cable take the forms [1,2,51 z I IO(x)= + -YlXi [sign(x)(l ? . 11 (1 - e - sign(x-L) + 2 wyx,w -ylx-LI )I -Y[v(x-L), z12z010 vO(x) = )+ - e + usI> (e -ylx-LI (9) -Ylxl I+ -e 2z11 + g- {n(vx,us) z12Zs10 uO(x)= 2Zll(Y2 _ e-Yclx-Ll)._e + YZslo 2nr -R - Y [ g -Yi) (e -yc Y rYSl- {Q(YX,YS) -+(v,X,Y,S) n[u,(x-I;), Y,SlH, (10) Ix'+ c -ulxl+ (Y2-Yc2) [y(x-L YSI) R[Y(x-L), - e -ylx-LI]+ -t -t (11) - self impedance of cable 212 - mutual impedance betwee; overhead conductor and sheath of the cable. Finally, the resultant current, potential and voltage distributions along the buried cable being earthed through earth electrodes, due to electromagnetic effects of a nearby power line, may be obtained according to eqns (l), (2) and (8). ;;;=FhZ1l A 6 cm diameter cable, 0.5 mm thick tubular copper shield is subjected to a unit current in a power line. The cable is earthed at points x = 0 and x = L through the earth electrodes located at distances 2 m from the cable'. The horizontal distance between the cable and the power line is 10 m. Height of the overhead conductor of the power line above the earth's surface is 10 m. The remaining data concerning the parameters of the cable are as follows: zi = 0.184 + jO.003 R/km, Zs = 0.184 R/km, unit-length capacitance between sheath and conductor of the cable - C = 0.1 pF/km, unit-length leakage conductance of cable insulation between sheath and conductor of the cable G = 0 S/km. Results of calculations of potentials and voltages due, in the buried earthed cable, to electromagnetic effects of a power line are shown in Figures and Table (moduli per unit current). Calculations have been carried out for the case of earth fault occuring in the close proximity of the cable (conductive and inductive influence of power line) - Fig.4, and for the case of inductive interference only (the earth fault is remote from the approach section of the power line) - Fig.5. In Fig.4 the potentials are plotted as function of x, whereas in Fig.5, the influe,nce of cable shunt conductance on the maximum value of potential of the cable is shown. In Figs 4 and 5 the curves IV'1 represent potential for the case where the additional conductive effect of the earthing currents was not taken into account. The influence of earth conductivity and shunt conductance of the cable insulation on the values of maximum voltages (at the point x = 0) between sheath and cable conductor is shown in Table. In the Table entry with prime corresponds to the case where the additional conductive effect was disregarded. CONCLUSIONS The results of calculations show that the earthing of buried cable reduces significantly the value of potential excited by electromagnetic effects of an a - c power line. The values of Potentials obtained on the assumption - 272 - IVI hV1 13(-j -0,s 0 Fis.4: 0,5 Potential I,0 I,5 distribution 2,O along earthed 2,s underground 3,0 h-d cable that no additional conductive influence exists may be smaller by 20% (for maximum values). The influence of the additional conductive effect on the potential distribution along the protected cable becomes obvious in case when the cable having an insulation of great conductance is buried in the earth of low conductivity. The earthing of the cable may result in increasing of voltage between sheath and cable conductor, which is noticeable for well insulated cables. The influence of the additional conductive effect on the voltage between the cable conductor and the sheath practicaly does not exist. IVI hV1 300 0.1 1.0 REFERENCES Gi [ S/km1 10 [ll. Fig.5: Maximum potential vs cable shunt conductance for various values of earth conductivity Table. Maximum voltage between and cable conductor cr=5.10-3S/m sheath cr= 10B2S/m ( IUl / mV 1 90 86 83 1 90 85 79 1 1 IU'II mV / 95 91 841 93 87 79 1 Krakowski,M.: Currents and potentials along extensive underground conductor.Proc.IEE,Vol.llS,No 9, 1299 - 1304 (1968). lI21.Krakowski,M.: Obwody ziemnopowrotne. WNT,Warszawa, 1979. Electromagnetic [31. Machczydski,W.: effects of a - c transmission lines on extensive conductor earthed through impedances. Seventh International Wrockaw Symposium on EMC,WrocZaw, June 18-20, 1984 485-494. J.W.: Tieorija i ras[41. Striiewskij, czet wlijanija elektrificirowannoj ieleznoj dorogi na podziemnyje metalliczeskije sooruienija. Izdat.Lit.po Stroitielstwu, Moskwa, 1968. [51. Sunde,D.E.: Earth conduction effect in transmission systems. Dover Publication, N.York,1968. - 273 5113 - COUPLING AND PROPAGATION OF TRANSIHNT CURRHNTS ON MULTICONDUmR TRANSMISSIONLINE?3 J.L. ter Haseborg*, H. 'Winks*, and R. Sturm** *TechnischeUniversitatHamburg-Harhurg Hamburg Germany **NRC Defense Research and Development Institute Munster Germany In order to estimate the protection efficiency or to realize an opt.imumprotection for sensitive electronic devi~cesrespectively by special protection circuits aqainst guided transient currents - e.g. caused by liqhtning (~23) or NEMP - the variation of the time dependent or frequency-dependentshape of the interfering currents by the transmission line has to be known. Particularlythe edge steepness of the surqeinfluencesthe response of ft;?yl protection devices, e-q. gas arresters tronsmissio n - protec4 on circuit .---- ! , L---- I -----J--f*--’ electronic deviceNine tefmination) to be protected A i-._._._ A shielded multiconductortransmission line is considered.The coupling process between cable sheath currents and conductor currents as well as the propagation of sheath and conductor currents are described analytically. Starting from this description a computer code is developed. Introduction There are two applicationsfor the cornputations concerning coupling and propagation of transient currents on multiconductortransmission lines. The first application is explained by Fig. 1. The coupling and propagationof incoming pulses,runningto the input terminals of protection circuits as well as the propagation of residual pulses at the output terminals of responding circuits, running to the input terminals of the electronic device to be protected, are of special interest. Referring to Fig. 1 the shape, particularlythe edge steepness, of the pulse@determines the response (dynamic threshold voltage) of the protection circuit. Often the electronic device to be protected and the protection circuit are not directly interconnectedbut separated by a line of the length 1 as shown in Fig. 1. Starting from a definite residual pulseOat the output terminals of the responding protection circuit the transmissionline constants in connectionwith the line termination dertermines the total pulse@at the input terminals of the device to be protected.Assuming the worst case the amplitude of the total pulse @nay be two times larger than the amplitude of the pulse@ . The second application,which is also typical for many cases in pracitce, is shown in Fig. 2. F’ig. 1: I * ._._or LRMP-induced and residual pulses@on transmission lines Propagation pulses@ of NEW- EMP termination 1 input terminals 1 S termination .-.-. 2 input terminals 2 Fiq. 2: Transmission line terminated at both ends with arbitrary impedances A multiconductortransmissionline (length 1) is terminated at both ends with arbitrary impedances. When the line is excited by an electromagneticfield (NEMP or LFMP) the total currents I Cl1 *.* ICln Or *c21 *.* ICZn respectivelyat the input terminals of the termination 1 or 2 respectivelyare of interest. However,quiteoften these terminals are not accessible and therefore it is impossible to measure the voltages or currents respectively at the pins of the terminations.Starting from an induced current as lumped source on the cable sheath the computer code developed allows the calculationof the total currents at the input terminals 1 and 2 assuming arbitrary terminations. In this paper results concerning the second application of this computationprocedure will be presented and discussed. The first application,thatis coupling and propagation of induced pulses on transmission lines as well as propagationof residual pulses, caused by responding protection circuits (s. - 274 - d1 -_=-y '* u dz - Fig. l), will be the subject of another publication. Theory basis for the calculationsis the transmission line theory, particularlythe transmission line equations in the frequencydomain. This theory is applied on the complete transmission line, that means on the inner conductors as well as on the cable sheath (shielding).Generally it is not possible to describe the sheath currents e.g. the currents on a braided shield by means of the transmission line equations. Starting from the transfer impedance which describes the coupling between sheath currents and inner conductor currents an equivalent cable sheath has to be found showing the same frequency-depndent transfer impedance as valid for the real sheath and which allows the application of the transmission line equations. A (4) the transition to ntl conductors provides the equations for multiconductortransmission lines in matrix notation: (5) d [II - dz = -[y’[ Q’] . (6) The insertion of (6) in the derivative of (5) yields the wave equation for multiconductor transmission lines: d2[El = [z’ 1 [y’ 1 [El l Fig. 3 contains e.g. for a braided shield typical curves concernin the tran f r imp?3_)have dance [2]. Computations7 s. Kaden BeI shown that it is possible to realize a cage as in Fig. 4, showing a freguencyrepresented depndent transfer impedance,which is largely identicalwith the transfer impedance of the real sheath. l (7) dz2 Concerning a definite braided shield, Fig. 5 shows a good agreement between the transfer impedance of the real sheath (solid curve number 1) and the transfer impedance of the equivalent cage (dashed curve). The values are dependent on - diameter of the cage - number of cage conductors - diameter of the cage conductors. This transformation: real cable sheath + equivalent cage enables the applicationof the transmission line equations not only on the inner conductors but also on the cable sheath. Referring to the equivalent circuit, shown in Fig. 6,it is no problem by means of the transmission line theory for multiconductortransmission lines, besides the considerationof inductive and capacitive couplings, additionallyto take into account the line losses, that means ohmic losses in the conductors and in the sheath as well as frequency-dependentlosses in the insulation. Generally the transmission line constants R', L', M', G'and C' - referring to Fig. 6 which are the elements of the matrix [z'] or [Y'] respectively,are measured. In thys case &ese constants are computed according to [4], [5]. Starting frcnnthe transmission line equations for a two-wire line in the time-domain: au xi= - a i $ u (R'+ L' -& ai. ,,a YiZ= - (GtC (2) or in the frequency-dcanain respectively: dV z=-z '. I (3) 2 2 IO0 10“ 2 lo0 2 5 10' 2 MHz 10' fFig. 3: Typical curves showing the transfer wance of braided shields [2]. co: optical coveraqe 5 By means of this multiconductortransmission line theory an arbitrary nuker n of lines can be considered. In (7) the voltage of one line is coupled with the voltages of all other lines. In order to solve this coupled differential equation system a linear transformation provides decoupled wave equations for multiconductor transmission lines: [El = [VI- [WI , (8) 5113 275 - The matrix [v] has to be chosen in such a manner that the insertion of (8) in (7) results in decoupled differentialequations. Insertion of (8) in (7) and multiplication with [VI-' provides: - [g’ ] [y’ I [VI = [VI Cr21 This equation represents an eigenvalue problem [6]. The decoupled differentialequation system is shown in (12): d2[W] __ = [r'l [WI . dz2 In order to obtain for [W] decoupled differential equations it is required: [g-l [g’J[y’I[vl = [r21 (lo) whereby [r'] represents a diagonal matrix. '(11) (12) The components of [w] are designated as natural waves of the multiconductortransmission line. 'The square roots of the elements of matrix [r'] are the propagation constants of the natural waves: (13) Yu = au + jfiu whereby a or (3 respectivelyrepresents the attenuatign con&ant or phase constant res_u. The general pectively of the natural wave W solution of the wave equation contains an incident and a reflected guided wave: able ‘al 1 inner shield) consist conductors Fiq. 4: Cage conductors as equivalent cable sheath concerning the transfer impedance The uantity z marks the location on the line, and i! W.(O)] or [W (O)]respectivelyare the incidenEior reflec& waves respectivelyat z=O. By means of equation (14) and various matrix operations the equations for multiconductor transmission lines can be obtained: for the line currents the following expression is valid: This formula shows clearly the two parts belonging to the incidentorreflected wave respectively. Generally it is not easy to cqute the matched termination for a multiconductor transmission line, because terminating impedances between all conductors are necessary. In case of the computationprocedure presented by means of various matrix operations the matched termination can be found. Matched terminationmeans, no reflected waves are existent, therefore the following equation is valid: [v]-‘*[u(O)]-[r]-‘*[rs]-‘*[~’ ]*[&(O) ]= 0 . (16) After various mathematicaloperations the Ymatrix for the matched terminationof a multiconductor transmissionline is obtained: I&J= fCnzl Fig. 5: Transfer impedance (normalized), - solid curve: braided shield - dashed curve: equivalent cage Equation (lo), multiplied with [v], provides (11): C3’ l-l~[~l~[rl~[~l-l , (17) Computation results In this paragraph results - referring to Fig. 2 - are presented.As already mentioned above problems and results - referring to Fig. 1 - will be the subject of another publication. Concerning Fig. 2 a multiconductortransmission line is excited by a spatially short, pulsed electramagneticfield. In the present case the computationsdo not start from this field but from a sheath current Is, which can arbitrarilybe assumed, as lumped source lo- - 276 - a) termination 1 termination 2 600R p@]g 600R b) piq. 6: FQuivalent circuit of multiconductor transmissionline of infinitesimal length dz cated at the line center. The parameters of the dissipative transmission line investigated show the following values: - lengthl=lcom - 4 inner conductors - transfer impedance of the sheath,servingas shielding, according to Fig. 5 curve no. 1 - the puls-shaped sheath current I assumed has a rise time t =50 ns and a h&f-amplitude pulse duratign tf=5,5 us. These values are valid for figures 7 and 8. Referring to Fig. 2 the curves of Fig. 7 show the currents I_ (termination1) and termination ICI 1 ll!flI l/L_ 1.1 85 0 -.OS 0 2 i 6 tcps1 0 2 4 6 termination t bsl 2 G .05 0 L------ 2.1 -.05 0 2 L 6 tIpsI 0 2 L 6 tIpI 0 2 1 6 t&4 a) terminations 1 and 2 600R 600R b) termination 4 termination 0 Fig. 8 1 4 Fig. 7/%: 10 c 20 30 t lps1 -.L D 10 20 30 tIpsI Coupling and propaqationof a pulse (t,=50 ns, tp=5,5 ps) on a line ter- minated at both ends according to Fig. 2, a) line terminations b) 1.1 incident guided wave, 1.2 and 1.3 reflected by termination 2 at termination 1 2.1 incident guided wave, 2.2 and 2.3 reflected by termination 1 at termination 2 c) total currents at terminations 1 and 2 (termination2) on one of the four inner 2 IC2 conductors.Both ends are terminated with 6coR symmetricalas shown in Fig. 7a. In Fig. 7b the incident and in each case two reflected guided.waves are representedthat means at both ends two reflectionsare considered.Fig. 7 c shows the total currents at both terminations. Fig. 8 shows the corresponding currents for the case that the end"l"is terminated with 6cxQ symmetricaland the end/'2" is short-circuitedaccording to Fig. 8a. In figures 7 and 8 the currents are normalized on the amplitude of the primary sheath current Fiq. 7: IS - Apart from the transmission line COnstants the pulse shapes are dependent essentially on: - line length - line terminations. The pulses are largely characterizedby line resonances which are dependent on the line length. In [7] a similar example has been ccmputed, that is a lossless transmission line consisting of two conductors excited by a half sine wave of width 20 ns. Conclusions Here only a small number of results can be presented. We have carried out a lot of canputations and besides line length and line terminationswe also have varied the primary sheath current pulse I particularly its rise time. The smaller the &se time the larger the pulse coupled into the inner conductors.Responsible for this effect is the frequencydependent transfer impedance.Referring to the transfer impedance shown in Fig. 5, taken as a basis for the computations,particularlyrise times less than approx.severalhundred nanoseconds cause comparativelyhigh pulses on the inner conductors. The research work concerning coupling and propagation of transients on multiconductor transmission lines is going on. References [l] ter Haseborg,J.L.;Trinks, H.: Protection circuitsforsuppressing surge voltages with edge steepness up to lo KV/ns. 277 5113 - 5th Symposium on electromagn.Cornpat. e, Zurich, March 8-10, 1983 [21 Homan, E.: Geschirmte Kabel mit optimalen Geflechtschinnen. NTZ, Heft 3, 1968 und Schirmung in [31 Kaden, H.: WirbelstriSmE! der Nachrichtentechnik.Springer-Verlag, Berlin, Gijttingen, Ileidelberg, 1959 [41 Clements,J.C.;Pau1,C.R.; Adams,A.T.: Computationsof tlX?capacitancematrix for systems of dielectric-coatedcylindrical conductors. IEEE Trans. Electromagn. ccXnpat.,vol. FMC-17, no. 4, Nov., 1975 of [51 Pau1,C.R.; Feather,A.E.: Ccanputations the transmission line inductance and capacitance matrices from the generalized capacitancematrix. IEEE Trans. Electromagn. Ccanpat.,vol. W-18, no. 4, Nov. 1976 [61ter Haseborg,J.L.;Trinks,H.: Transient response and protection of multiconductor transmission lines. InternationalAerospace and Ground Conference on Lightning and Static Electricity,Orlando, USA, June 26-28, 1984 [71 Agrawa1,A.K.;Price,H.J.;Gurvaxani,S.H.: Transient response of multiconductor transmission lines excited by a nonuniform electromagneticfield. IEEE Trans. Electromagn.Ccmpat., vol. EMC-22, no. 2, May 1980 - 279 52 - 14 Response of a S it-w le-ConductorOverhead Wire Illuminated by an InhomogeneousPlane Wave - - F. PALADIAN, J.P. PLU?4EY, D. ROUBERTOU, .I.FONTAINE University of Glermont-Ferrand France 1. Introduction In this paper, we obtain the time domain response of single conductor overhead wire illuminated by an inhomogeneous plane-wave. For this purpose, we develop an E- integral equation formulation for the current in the frequency domain and discuss a numerical procedure used to solve this integral equation, based on the application of the method of moments and the finite difference technique. This method presents several advantages over he transmission line theory : first, coupling between horizontal and vertical wires is not taken into acount with this theory, and then, results from transmission line theory fail to exhibit the resonances for the structure. At last, a Fast Fourier transform (FFT) algorithm is used to convert the frequency domain results into the time domain response Numerical results are presented for the current induced on the structure placed over perfect g.round. 2. General formulation Figure 1 shows an overhead single conductor, horizontal with two vertical terminations penetrating the ground. The angle between the vector 2 for the plane-wave and the horizontal line is $. Regions 1 and 2, respectively the soil and the air, are characterized by (cl = eIcO, 1-11 = 1-I,, oI) and (e2 = Ed, u:! = u,, u2 = 0) where E and po are free-space parameters. 0 we consider "thin wire hypothesis" that is to say antenna radius is much smaller than vertical and horizontal parts lengths. We consider an observation point P, with c2ordinates (X, Y, Z), which is labelled by OP = r. If we design the Hertz vector potential by t, the scattered field at P is given by : 1. - if P region 1 : it,(;) = (k:+VV.) I $1) 6, +.( r )d;' (1) structure - if P g,(G) region 2 : = (k$+VV.) I $2) + +.I (r, r )dg' (2) structure where k represents the wave number : k? = w2p - jwpiui 1 i Ei i = I,2 Designating &he unit normal as fi and the incident field as El, we obtain the integral equation : fi. (@c:, + it(Z)) = 0 (3) From each equation (1) and (2), we obtain as many equations as there are vertical and horizontal dipoles respectively in regions 1 and 2. (for the studies case, there are two equations if P belongs to region 1 and three equations if P belongs to region 2, so we obtain five equations). In equation (l), we consider z component of $1 in region 1 and the kernel depends on we consider the source as : - either a VED (vertical electric dipold or a HED (horizontal electric dipole) in region 2. - or a VED in region 1. In equation (2), the kernel depends on we Sonsider either x-component or z-component in region 2 and it is different when of E2 the source is Figure 1 ,: Geometry of the structure The time variation for all the expressions given in this paper is represented by the time factor exp (jut). We take into account following hypothesis : first, vertical and horizontal lines conductivity is supposed infinite, and second - either a VED or a HED in region 2. - or a VED in region 1. The kernels of equations (1) and (2) are composed of Green's functions and Sommerfeld integrals (cf. annexe). The numerical solution of equations (1) and (2) is obtained via the application of - 280 - the method of moments (2); we used Point Matching method that means currents are considered as constants along every patch of length A on horizontal or vertical wires. Application of the method of moments allows one to write integral equation in terns of the following matrix equation : k: f Zj+A/2 zj-A/2 +(z hg>i”j + (zhg).I. hg or:Z..=(Z 1J 4 source term or (Zmn)(In) = - (Eim) (I”) represents unknown current matrices. & In each sub-matrix, every component is broken up in three terms : a source term (when we consider region 2 as infinite), an image term (when region 1 is considered as a perfect ground) and a corrective term composed of Sommerfeld integrals (when region 1 is considered as a lossy ground). For example, the expression of the incident field is given when the observation point is along the horizontal part of the structure in region 2. Ba;\;osEormalism(2) is used with the time factor exn _ (iwt). _ E;(X,h)+$?o + $Q !; I,W’k& ~hI,(0,z9 + $$$ x=x V12dZ' X1=0 Z=h (G22-G21+k&)dZ' & II'Ix(X',h) ($ x=x X'=O Z=h + k;)(G,,-G,,) 0 x=x + a2 k2 2 (w + +f$$ + *22 I V22 I,’ L&Z’)~ dX’ a2 Z'=h z=h /-hrz(L,z,) & x=x “,‘r; (G2,-G21+k;v,,)dz’ V12 dZ' where expressions of Green's functions (G22 and G21) and Sonunerfeld integrals (Vl2, V22 and V22) are given in Annexe. In this expression, every integral gives us expression of a sub-matrix of the third line of (4) ; we obtain respectively Zhg , Zhg, zhh, Zhd and zhd . For example, we obtain : ij ho W Zj+A/2 f zj-A/2 a2 hg)C .. ! lJ corrective term & f(X,Z) = f(X+A,Z) + f(X-A,Z) - Zf(X,z) A2 f(X,Z): = f(X+A/2,z+A/2)-f(X+A/2,2-n/2) A2 _ f(X-A/2,Z+A/2>+f(X-A/2,2-n/2) A2 The first purpose of this paper is to obtain the current value at the point M in the frequency domain. Then, these results are used to investigate the transient behaviour of the antenna current mainted over perfect ground. As a first step, the transfer function (impulse response) of the structure current is computed using the previous results. This transfer function is then multiplied by the spectrum of the EMP (electromagnetic pulse), and finally the Fourier inversion is performed numerically via a Fast Fourier Transform (FFT) routine to obtain the transient response E(t) = El(e-et- eeBt) and by a Fourier transform, we obtain : E(f) = E,(& - &) - j E1(;&-&'& ;I;" 0 zhg = Xi EMP form is given by : x=x + * ) lJ image term = ;': ; By means, it is possible to reduce calculation times by taking into account all symmetries of each term. To avoid difficulties that come from differential operators inside the integrals, we have chosen the finite-difference scheme for performing the differentiation outside the integrals. The two following formulas are used : (zmn) represents the impedances matrices. where X a2 V22 dZ' axaz x=x; axaz Gz2dZ' X'= 0 Z=h At a point M on the structure (cf. figure 1) the current transfer function may be defined by H(f), which is the deltaresponse of the structure at point M. The transfer function can be constructed discretely from previous results. Using the superposition theorem, the transform of the current at point M due to the EMP E(t) can be written as : I(f) = E(f).H(f) and the transient response of the structure current at point M can be expressed as : 4-m i(t) = 1 I(f) ej2xft df - 281 3. Results a) In the frequency domain : Results are given for a case closely related to Taylor and Castillo's (4) with a different height of horizontal part to take the same patches length on horizontal and vertical part of the structure. We used 40 patches on horizontal part and 1 patch on each vertical part : this choice ensure that moment method solution converges. Response in the frequency domain of the structure is given for a perfectly conducting ground, when vertical parts are each loaded first by 100 R, second by the caracteristic impedance Zc. 52 - 14 lines theory is equal to zero that is verified between resonance frequencies (5) * III (Al , 4.p 2f.QMh.z 3.19 2.18 ld? zc = 120 Log ($) = 635 n First, currents are computed along the structure for a frequency corresponding to a wave length equal to the total structure length (figure 2). The plane-wave incidence is 90'. So the excitation of the structure is symmetric about X = $ ; theLcurrents are of course symmetric about X = 2. Figure 3 : Response in the frequency domain at. point M Figure 4 : next page b) Time domain results. Previous results are used to obtain the time domain response of the structure at point M. The structure transfer function H(f) is computed at the total of 1024 frequencies in the range of : 0 Figure 2 : Response in the frequency domain along the structure. Limits of moment method are given by L > X/6 These conditions are directly linked to the basis functions choice and they ensure results convergence ; they have been specified by many testings. For this case, we may obtain results for : 3 MhZ < f i 40 MhZ Currents are computed at point M for this frequency range. Resonance frequencies are given by : where < f < 40 MhZ With this choice, all the sharps peaks are taken into account. f = 3 MhZ represents the lower limit of moment method so transmission lines theory is used for 0~ f < 3 MhZ and moment method for 3 MhZ < f < 40 MhZ. The transform of the EMP is then multiplied by the structure transfer function to obtain I(f). Note that at f = 40 MhZ, the resultant frequency domain currents have decayed sufficiently so that zeros can be added for f > 40 MhZ. Now that the entire frequency domain is constructed numerically, a Fast Fourier Transform routine is employed to obtain the time domain response i(t). The ringing effect linked to the sharps peaks of the frequency spectum makes plotting i versus time response difficult. So, figures 5 and 6 respectively shows the time domain response only up to 1,25 us and the envelope along the all range of times. Ii1 (A) 800f 700.. 800. 500' fr = mcl2L 400. c is the velocity of light m=l ) 2, 3, . .. 300. 200,. In this case, for I$ = 90", the resonant current modes corresponding to odd values of m are not excited : the excitation of the structure is symmetric about X = L/2 and these resonant current modes are not excited (figures 3 and 4). Current value given by the transmission 100 0.75 ’ 1 Figure 5 : Time domain response at point M for 25 ns ,< t,< 1,25 us - 282 - W -y2(z+z')(Y2-y~) =2fme 22 W 21 k:Y, + k;Y, J (A,-> Ad), ' (Y2-yl)Jo(Xr) AdA = -Y2(z+z') U 200 '-I Jo(Xr) Xdh 22 = Yl+Y2 0 100 BW22 Figure 6 : Envelope of time-domain response at point M 4. Conclusion Our purpose is to obtain time domain response of the structure by using Fast Fourier Transform.Frequency domain results show that it is necessary to resort to an integral formulation of the problem. Transmission lines theory actually fails to exhibit the resonances for the structure that would affect time-domain results computed by FFT. Results obtained for a perfectly conducting ground : this case is unrealistic but represent a step in the solution to the problem when the ground is imperfectly conducting. a) Green's functions .-jkzR2 .-jk2Rl G22 = ~ G21 =R2 e-j w21 = we have + G12 =- e-j x-2 = x2 f y2 order References 11) A.J. tion Poggio, E.K. Miller, 'Integral equasolutions of three-dimensional scattering problems', in Mittra, R.(Ed.) : 'Computer techniques for electromagnetics' (Pergamon, 1973), pp. 159-263. (21 R.F. and h = 1~') (4) m Jo(Ar) XdX 0 k;Yl + k:Y, co "21 eY1z-Y2z' 321 Jo(hr) XdX 0 kZY, + k;Y2 e-Y2z+Ylz' "12 = 2 jm 0 V 11 L 2 Jrn 0 i = I,2 Jo represents Bessel function of zero (3) b) Sommerfeld integrals "22 =2j ki)l’:! Harrington, 'Fiel Computation by moment methods', Ed. Mac Millan, New York, 1968. RI = (r2+ (z + h)2)1'2 where : (X2- * with : R2 = (r2+(z - h)2)1'2 and : r represents horizontal distance from the observation point to the origin 0. kl RI RI = "21 * Gll =R2 Yi (& - &'i with real (yi) > 0 Rl kl R2 = k; V22 - U,, k2,W2, = -$ (2G21-(k;+k;) "22) Annex Expressions of Green's functions and Sommerfeld integrals used in Ba?los's formulation. az (5) A. Bazos, 'Dipole radiation in the presence of conducting half-space', NewYork : Pergamon Press, 1966. C.D. Taylor, J.P. Castillo, 'On e'lectromagnetic-field excitation of unshielder multiconductor cables', PEEE Transactfons on electromagnetic compatibility, vol. EMC-20, NO. 4, November 1978. A. Albert, Jr. Smith, 'Coupling of external electro-magnetic fields to transmission lines', John Wiley and Sons, New-York (Eondon) Sydney (Toronto,1977). Jo(Xr) Adh k:Y, + k;Y1 eYl(z+z') Jo(hr) XdX k; Y2+ k;Y, Figure 4 : Response in the frequent domain af point M. - TIME DOMAIN SCATTERING 283 5315 - BY THIN WIRE STRUCTURES A HOMOGENEOUS F. MAUMY, B. GECKO et 0. DAFIF Universite de Limoges - U.E.R. des Sciences de Communications Optiques et Microondes (LA 356 du C.N.R.S.) 123, rue Albert Thomas - 87060 LIMOGES CEDEX Laboratoire ABSTRACT This paper electromagnetic shows pulses of any form, terized by a conductivity the above capital current stage induced equations, the by thin o.The is the on the wire, taking scattering wire a homogeneous into of problem from is resolved by integral in -Figure which of the k’ charac- determination account The space-time presence space of Some I- applications domain of and electric are chosen in telecommunications or railway systems. theoretical to responses through on these installations. perfectly was treated wire illumination of induced currents scattering problem structures [ll, of a perfectly [2], with takes a finite into account conductivity. Fresnel’s approximation, in integral equations verified induced on wire [41 and structure. the the simulator as antennas earth, with The of method of interpolation is also applied radiation or essay transmission of waves. moments used in pulse emission 11.1. Principle the P the any in the (I?“, Gal, the superposition at a point (gd, of a wire perfec- half-space, t which is respon- Ad). The field is the M of the it (?$I) expressed on : by (2) (3) determined electric half-space conditions called as the half-space Knowing @, field a current operator wire arrives wave half-space, in = L[I(M,t)l ting I% (1) Ed(P,t) boundary in it’s written integral the obstacle, + Ed(P,t) equations. point (I?, half- this to field Maxwell’s larly total structure = Ea(P,t) an wire @“, fia) illuminates of with the a reflected point due on its surface field by 8 radiated and therefore, wire surface, on the the at particuone applies perfectly conduc- : in presence pulses s’;‘?(M,t) = 0 where s PROBLEM of the method any in + Er(P,t) wave (4) is the at the point II - THEORETICAL wave At of diffracted is incident is = Z’(PJ) point L Without creates , treated earth. it conducting sible is Therefore, current are to some electromagnetic line. a establishing solution. This work problems by the Lagrange to the numerical field”, a It consists, the the obstacle, QP,t) paper 1). “applied induces conducting plane without tly [3J, the (figure half-space problem ground, When the ground. ground fir) Ea(P,t) structures pulses Transient in presence This in of the knowledge conducting or dielectric real work electromagnetic proceeds by the the both INTRODUCTION The of the equip- above when on protection ( &?I Iair P,,rr ,r,,,r,,,,,,,,,,,,,,,, earth the the earth. ments :;‘:y_e&+ structures ground established directly domain, ABOVE GROUND Using lows to unit tangent vector to the path M. obtain (2) and (3), this an integral condition verified (4) alby I(M,t): - z.{L[I! M,t)] + ca(M,t)) 11.2. Applied field, First, in we the the order time the reflected our course of action is the following. consider the reflection domain. domain by aninverse the air/soil we use Laplace the R(t) takes and the function the earth we present the = $$ I (e-d) is at e-T, (ii) - 02 _(6-d) the horizontal -t d , + (8-d) IO $1 with e where tion -- Bt e * 0‘; B= CI E E r sin’ obtained Thus, consists of two the dielectric, of order RTM(t) reflection parts, the other 0 2 shows dence angle of the half one of which R(t) is due to (8) the dependence is = E’(o,t) 11.3. Integral The field poles on the the computed at any from point the P plane waves gd = L(I) being boundary conducting the wire integral condistructure equation verified 1: to *(t) et T*(t) are vectors on the earth, [T*(t)] = [R(t)l.[6*1 where (To, d o*) are respectively the source dipole R(t) is the by the which express connected with [R(t)] [5], [6] by : matrix = [R(t)l.[:o*l the unit and to the image reflection dipole, matrix of the connected vectors to dipole. field, with radiated the functions sin 4’ o II61by : et RTM(t) [R(t)]. - RTM(t) cos - RTM(t) sin 4’ i the (M~,A,P) by adding of all the electric . Frequency have by domain using the \iq*( image) - RTM(t angle of incident of the Fresnel’s been validity obtained comparison ground, condition : zones Fresnel’s plane coefficients [7] with of reflection in the approached coefficients frequency exact domain method using integrals. For an 1 o domain Sommerfeld’s ‘\ cos $’ 3). The earth RTE(t) 0 (figure wire. 3 - RTE(t) method up di- - polarization 11.4. Validity is obtained @’ 0 is method - Figure to obtain reflection field 6 (t - +G ed(P,t) A is suppo- of R&t) on the inci- field dEd(P,Mo,t) the 3). of the [@I equation the contributions perflectly reflections 4’ (3 one RTE(t) at the interface j$‘(p,t) on from of the field application by the current conductivity. reflected space be to the conductor. and the earth The can function R(t) = Rc 6(t) + R&t) Figure allows the and where polarization. the [_5], the on the a tb Bessel functions expression for a vertical reflected spring at the point coefficient a 0) b’f3 e =- similar one at M; (figure the at formed dT and 1. A by seems The diffracted reflec- (11 (9) 0( IO and Ii are the and emited MO, is . s(t) a = cos 8;bnJcr-sin* i with is domain, e-av2 e-d(t-‘d _- Bt 2 ] + sb b*B d=jyr-g done known polarization. $ situated time at half- frequency (RTE ou RTM). (@.) 02 1 dipole which which the used in situated wave wave expressed only emdt - a?_$- wave in very consists, dipole straight oneself the The reflection It is written RTE(t) image characteristics R(t) a ground product incidence. for one the (6) [61. Here, analytically tion of by the P by seats earth, [2] supposing sed into’ account angle in reflected = R(t) (;) @(o,t) g’(o,t) it transform. interface, by the convolution is given in one the method point coefficient Then, above domain determine Near field When space to frequency - (5) field computation In electric = 0 284 the electric height dipole, h must situated verify the above following I - 0 01. ,‘.‘*‘..‘*“.“’ 150 100 “, 1, 50 ,‘, 208 285 5315 - Er.“,l”..l.““““““’ 0 50 250 100 150 200 240 t(ns) t,(ns) _d--.5 -5 - -4 Conduct I i vi ty: lG3 mho/m t t -I t t t (, . . . . ,. ‘. .’ -1.5” _-3 Conductivity: -2.5 t 5 (*la mho/m 5 -25 i ) 18 (Xl0 (- ) 50 100 150 ( 250 2h0 t(ns) -rib -108 -100E Conductivity: Conductivity: 18 -2 mho/m -150 -1500 -200 -250 IL ( - Figure ANGLE Permittivity OF 2 - INCIDENCE E, = 10. . . . . 60" --_ --- 450 300 00 REFLEXION ANGLE AND OF FUNCTION CONDUCTIVITY INCIDENCE RTE (t) oINFLUENCE (degrees) : 18 -1 mho/m - - h>--X 07 Ee = Er +-z-JWEo JF7;1 from ‘thlis height condition, [S] a frequency 0,7)‘cZ h2 f > (12) it is possible : condition using 1 Fresnel’s available so magnetic pulse are reflection that all the wire behaves like a is The seems However first penalizing high For I Moreover, (sod our attention border effects, equation TM by the sampled numerically method resolved sequence [ 21. proceeds in the sample t-R*/c The original Ru = by of stage the (so,+- of ximed I(so NS where of this time $$so,T) of time 1, uq = I seen to =-2 can term: dT (15) be intervals s;+ 1 segments and where AR(*) to (221 < 0,5 avoid interpolation (j+m)-th term appro- (16) all . I time at into interval. time step sampled tj current i ,i, Ri$s+l by, :At/2 may be values up 2 (sl;,t?] dt” = J!l For reasoning (23) Pis easily the reflection to be a constant interval. t”) . TM within function is con- space segment each This approximation is satisfacto- ry when u < Id” mho/m (fig. 2). On the contrary, R;j-$($!t”) will be interpolated in time by a Lagrangian polynomial ,At/2 v+2 mZv R$j-s+l TM and NT The spacetime then be written otherwise t,$?m)dt!l Ii+1 s+m 1 ’ At/2 dependent convolution (24) term can as j+m +1 0, ) > 0,5 At/, ti’ = to - tj) of space AR(*) = c(tR$’ for include Pijbi,tj)=Pijz sidered variation (21) [2]. problem convolution NT jgl Iij (sY,tj’) u (s?) v(tjl) NS is the number the number future and time s” zz s - s. i 0 I’ with for want [4] repeated bv : = igl we to ti of moments spacetime (20) p. 1+1, jtm I. are the current and convolui+l,j+m and Pi+l,j+m rion term values at the center of the (i+&tk space TM current (191 1 I-1 the (14) moments a of a in two dimensions and is .: ds 0 0 function Q,m! B.. ‘J 0 * -$-y(so,to*)} (18) (I,m) I, i+l, j+m “(9 =-I because Its integral vc;j, Bij v’+2 JzV, “(‘1 the i and where +I v(l)+2 (so-si+ )(t(*)- t.+ ) (1 m) p 0 B!$j = II1 p=-1 pT$)(Si+l-si+p)(tj+m-tj+qJ very the low frequencies we restrict segment response isn’t TM It’s . Method V+2 rn% and the Gigg-0 (so,to)- K&o)$$g+o,to*)segment the convolution and the Us;) interpolation pulse. wire. pij (sY,tT) +1 =,5-l pij(“,+ domain. the 21 +1 = c 1=-l MHz dimensions it of space interpolation Lagrangian Iij(s;,t;l Fresnel’s [9]. by in neglecting G P(So,to*) filter lengths II z) in the form we can put Iij(sy,ti’‘f and pij(si,tj in taking finite otherwire t; = t* - t. I 0 1 . Lagrange be SOLUTION so like an infinite cO with < _nz’ At the = i;; to in time Therefore, simplicity, wire than uncorrectly III - NUMERICAL J intervals. with electroma- using generally, to treat :.Rs,t,=g Ai and time mho/m,the higher pass moments. a long with method electro- But method obstacle of the electromagnetic to of (13). inapplicable interested, the verify correctly. approximation in frequencies frequencies wave treated one the coefficients the spectre for h = 10 m, E = 10, d = lo-’ r condition shows that only frequency gnetic necessary 0, i p(so,to*) be It;1 Similarlv. (13) E r 1, I = to deduce - 3,24 102’ (5’ would for example - vet!;, . Time domain It 286 V+2 sZ1 r=C-1 t&l .(r,t) i+, R$Mj+m-s+l I~+,+~ s+t , (25) 287 where v’ has been taining and terms used later in place than v of to avoid 5315 - radiation ob- time t. v-t2 in pulse [l] JECKO (26) B. “Diffraction sionnelles developments and (18), (23) into (14) allows a currents current matrix at convolution the integral expression time-step v to for be equation the unknow obtained. It is (27) Ziu is the between v = 1, NT u = I, NS SELDEN in the matrix the of the structure mutual interactions segments. It is space” 0. It Ii v are specified c&entvalues a repeated to sequence bJ’ obtain of $1, Ii2, matrix etc... a with in method long frequency domain frequency [lo]. The by lines approximation the are using solution the 4) problem inverse validated finite which differs work an results method rencies The the time wire is is obtained evolution h, the conductivity tivity greater the current ted its is (Figure wave which intensity, 6 and the the the current different height later is, the stronger the reflec- the diminution and later. conductivity and of Figures permittivity V - CONCLUSION proach, of the by wire of a treated work directly electromagnetic structures particular study [8]. in waves of any homogeneous developed constitutes about This an in the It first time scattering form, ground. method the space ted (1972). de “Field the JECKO 3eme (France) computation Mac Millan B. “Diffraction filaires Colloque Company en june 1983. F. “These de d’1.E.M. presence National Electromagnetique. University T.K. sur la Tregastel Doctorat of domain, problem presence continues infinite wire, is applied ap- the already to the “Analysis thin-wire ground antenna planes” M. New Brunswick 3eme june values influence. This half- vol.50 Limoges of structures [71 SARKER 3eme o and the permit- instigates antennas conductivity Doctorat R.F. Z&me Limoges de 3eme (France) cycle” arrays SC. G. over E., “These de of “These University J.P., imperfect Universite of Doctorat de Limoges (France) 1984. de of Internal Doctorat Limoges 1983. [IO] BERENGER orien- (1971). University january, cycle” of arbitrarily Canada M.P. [91 LEBORGNE 5). Effectively, arrives 7 show of for of the height E. The des sol” no 2-84, a little. observed G.3. (1983). et [81 GOURDY diffe- IV.2. Influence of principal parameters on du BURKE (1968). 0. II61MAUMY (Figure treat execute The Transform. forthe compared which and Ann. to be published. applied . it’s wire methods the is mdtalliques” of Physics de Method” New-York cycle” several Fourier by february (France) operations. IV-I. Validation of the method The impul- of wire a University par I. IV - APPLICATIONS about J. Compatibilite is possible E.M. “Analysis “These by Moment time ) . ..) v - 1. j=l by currents of Ezv and the known in terms earth. A.J., of Canadian 151 DAFIF the E.S. 141 HARRINGTON independent. Finally, the (1983). POGGIO presence cycle” x=I,...,v-I n”5-6 E.K., and n”8-83 i = 1, NS t.38, [2] MILLER term [3] DAFIF with of d’ondes par des obstacles Tdlecomm. of presence BIBLIOGRAPHY C\l,m) Substitution in the report. de (France) - 288 - 3500 F 3000 2500 1500 t t 1500 -3 506 mhoim I t,(ns) 01 0 - I’.‘. “‘. 50 Figure XXX coo 4- Current wire 158 evolution on the - (ct.) I e I’.“” 180 10 20 30 40 50 . 60 e . 1 70 00 (In) Figure 5 - Height influence can be used to verify the condition 02) in time domain. Here, our method using the Fresnel's approximation (-) is compared with a rigorous solution (---) f‘ourlded on the Huyghens's principle (81 Fresnel's approximation Lines approximation Finite differencies I(t) I(t) 500E 2508 - 400E ///,,/// N/Y AH 2000 . 3000 1500 - 2006 100a - 500 * 100e t(m) a~.."'..."*.'." 0 - Figure 56 6 - 100 Conductivity E 150 influence - Figure 1 10 20 7 - . I I 30 40 Permittivity 1 ., 50 I. I 60 70 influence ., 00 et(m) - NOISE SOURCES 289 5416 - AND INTERFERENCE VALUES IN HIGH VOLTAGE SUBSTATIONS H. RGhsIer A. Strnad Energie-Versorgung Stuttgart, Abstract - Noise sources in high voltage substations may cause severe interference problems in the secondary circuits. The knowledge of the different noise effects sources and an analysis of the interference allows to estimate the overvoltages to be expected and their frequency of occurence. By a proper design of both HV equipment and secondary wiring and by the application of voltage limiting devices the noise voltages can be limited to an acceptable level. 1. Introduction The phenomena of electromagnetic transients affecting electric and electronic devices has been known since the beginning of electrical engineering. The installation of sensitive electronic equipment in high voltage (HV) plants and the problems associated have led to an intensive work in the field of interference on an international basis. Concerning HV substations a lot been made mostly examinating occured. However, up to now description of the noise sources values. of investigations have interference problems there ist no detailed and the interference It is the aim of this study to describe the noise sources and to analyse the interference values with special regard to their shape, the maximum of amplitude and steepness and the impedance of the noise source. 2. Electromagnetic The noise substations noise sources in HV substations sources which have to be regarded are described below: in HV - switching in primary circuits (i.e. on the HV level) Switching of disconnectors or circuit breakers is a frequent source of noise in HV substations. The guided waves are transmitted by the current transformer (CT) and the voltage transformer (VT) to the measuring und protection circuits. Current flow on cable-screens, produced by guided waves and magnetic fields, and currents fed into the earthing system through CT and VT generate common mode voltages which also influence the secondary circuits. - atmospheric events A lightning stroke generates travelling waves on the HV line. These waves can be produced by a stroke to the conductor, to the earth shield wire or the tower. The shape of the travelling waves depends on the amplitude and the shape of the lightning current. A flash-over of the insulation can be caused by a lightning stroke to the line or to conductors in the substation, or by insulator contamination. Whatever is the Schwaben AG West Germany cause of the flash-over, it will produce electromagnetic waves which affect the secondary circuits. The. lightning current fed directly or via an arc into the earthing system may result in high potential differences within the earthing system. - earth faults Earth faults caused by the described events, by switching overvoltages, conductor galloping or faulty switching are to be regarded mainly with respect to the electromagnetic waves radiated. - switching in secondary circuits De-energizing of inductive loads generate transient high frequency overvoltages in secondary circuits. - electrostatic discharge Electrostatic charged persons cause a very steep current with a rise time of some nanoseconds when touching earthed equipment. - radiotransmitters (walkie talkies) The high frequency field generated by radiotransmitters, including those which are used by maintenance staff, can influence sensitive electronic equipment. The described noise sources affect the secondary circuits. It has to be distinguished between the interference by conductive (direct), inductive and capacitive coupling on the one hand (guided waves) and interference by radiated waves (interference fields) on the other hand. The influence of radiated waves on the secondary circuits working as antenna gets important for high frequency events in the MHz-range. 3. Interference values in open air substations and GIS Table I shows the maximum interference values measured at the secondary equipment of open air substations and GIS (gas-insulated substation). The figures are valid for careful installation regarding the described measures, however without application of voltage limiting devices. The interference values and their origin are dealt with following. 3.1 Guided waves 3.1.1 Switching in primary circuits The most frequent event producing interference is, the operation of disconnectors in HV substations. During one operation interval there are up to 100 discharges of which the most critical follow each other in not more than 10 ms. - affected circuit 290 protectioncircuit - controlcircuit auxiliary supply circuit type of interference guided wave radiated wave guided wave radiated wave guided wave radiated wave dimensions kV mT kV mT kV mT 0.3 1 0.3 1 0.3 1 atmospheric event 2.5 1.5 2.5 1.5 2.5 1.5 switching atmospheric earth fault in primary event circuits 0.3 2 1 0.1 1.5 1 .o Noise sources E earth fault z 21 2 $ ?i frequency range open air substation 0.05-10 kHz 0.024 MHz MHz __I :: (11 0.3 1.5 1 0.1 1.5 1 0.3 1.5 1 0.1 1.5 1 CJ GIS 0.24 MHz 0.2-100 0.05-10 0.024 MHz kHz Note: The electric field is not considered; due to the screening always existing there is no serious interference. The figures are valid for the common mode voltages. In measuring transformer circuits common mode voltages and transverse voltages are identical Maximum interference values to be expected in I-IV open air substations and GIS with a rated voltage of 123.....420 kV Table I: The situation is shown in Fig. 1. If, e.g., an opened circuit breaker is isolated by a disconnector from the plant alive reignition occures as soon as the restriking voltage across the disconnector contacts is exceeded. The shape of the wave transmitted into the secondary circuits is characterized by a steep front with a rise time of 100 . . . 200 ns in open air substations and 5 . . . 20 ns in GIS. The high frequencies in the MHzrange are able to influence secondary equipment seriously or even to destroy it. The maximum interference depends on the first ignitions when closing and the last ignitions when opening the disconnector, i.e. the interference rises with the restrike voltage and hence also with the rated voltage of the substation. The measured frequency range of the described events is 200 kHz . . . 100 MHz for open air substations and GIS. primary circuit secondary circuit voltage of the arrester and thereby on the steepness of the incoming wave. In open air substations often only power transformers are equipped with lightning arresters; that is, why the measuring transformers (VT and CT) in the line section of a substation can be stressed by high overvoltages compared to the rated voltage. Moreover these high voltages may cause a flash-over of the insulation. When steep overvoltages or the breakdown of the insulation is transmitted by the measuring transformers high transients may occur in the secondary circuits. The frequency and the amplitude of these events depend on - the transient reponse of the measuring transformers - the statistical distribution of the lightning parameters - the lightning protection facilities of the HV line and the substation - the exposure area of the HV line - the isoceraunic level The amplitude of transients in secondary circuits caused by lightning phenomena and their frequency can be estimated (see Appendix). 3.1.2.2 U, striking voltage u secondary Fig. 1: 3.1.2 voltage Voltages in primary and secondary circuits when opening a disconnector (schematic) Atmospheric events Amplitude and steepness of lightning currents show a statistical distribution [ 1,2] ; consequently the overvoltages expected can only be determined by statistical investigations. Establishing an acceptable risk will fix the measures to be applied to the secondary circuits. 3.1.2.1 Lightning stroke to the HV line ,Lightning overvoltages are limited on the HV level by lightning arresters. The maximum overvoltage influencing the secondary circuits depends on the threshold Lightning strokes HV substations to earthed components of Lightning strokes to the structures or earth conductors of the substation or to the earth conductor of the line close the substation produce high transient potential difference in the earthing system. These potential differences cause currents over the screens of secondary cables and the cabinets and earth connections of secondary equipment. Concerning the secondary cables these currents produce voltages whose amplitude depends on the construction of the screen, the length of the cable and the amplitude of the current in the screen. It is possible to estimate the transient voltages caused by lightning phenomena for a certain substation by a combination of measuring technique and mathematical methods [3]: When discharging a charged overhead line into the earthing system of a substation the discharge current and the voltages at selected points of the secondary circuits can be measured. The knowledge of these signals allows to compute the transient response of the systetns. The frequency of lightning strokes to earthed components can be determined by - 291 5416 ‘- z considering the actual exposure area of the substation and the isokeraunic level. The maximum transient voltages to be expected at the selected measuring points can then be calculated by bringing in the statistical distribution of the lightning parameters. Fig. 2 shows the result of such an investigation; for the measuring point M 3 the voltage of 2700 V is exceeded once in ten years, the corresponding value for M 4 is 560 V. K TK Fig. 4: 1 Interference by guided waves. H t The impedance 3.1.4 of the noise source 10-l The energy content of a noise voltage is of great importance and has to be considered especially when electronic components are stressed. Consequently for a certain shape of the noise voltage the knowledge of the source impedance is necessary. For a generator constructed to test the interference withstand capability of secondary equipment the proper simulation of the source impedance is also an inevitable demand. _j_ a 10-a IO2 Fig. 2: 103 V T----) “2 IO4 A model of an interference in Fig. 4. Here are Noise voltages ii2 in a 420/123 kV open air substation caused by lighting strokes to earthed components H frequency of strokes producing a noise voltage which exceeds b, M3, M4 measuring points the impedance %Q the voltage u1 z, Switching in secondary circuits Overvoltages produced by de-energizing inductive loads occur because of reignitions during opening the relay contact. These overvoltages are typically sawtooth shaped followed by a damped low frequency oscillation. The amplitudes of the overvoltages can exceed 5 kV, the risetime of the spikes being in the order of some nanoseconds. The frequency of the damped oscillations is normally below I MHz, but sometimes comes up to 20 MHz. Figure 3 shows an oscillogram of the transient voltage generated by de-energizing a relay coil. The application of voltage limiting devices allows to reduce the transient voltages to less than 1000 V. impedance time the frequency the secondary dependent equipment of the secondary of the secondary input cable impedance at the end of the secondary the resulting -za is shown of the noise source the travelling the voltage u2 waves of the noise source the characteristic cable ZK ‘K 3.1.3 by guided impedance of the noise of cable source. The characteristic impedances of secondary cables (conductor-conductor; conductor-screen) come up to 30 ..* 100 ohms nearly independent of frequency; however, these figures are not valid for the circuit screenearth return. In most applications we have zK L Z+. During Z+ 2 TK = zSQ; we then After 3 T, approximation for ist charged 2 T~L t L 4 TK to + ZSQI. by the an alternating Z+ = z,, the cable u2 = 2 UIZK/GK Caused find reflections behaviour, with at the cable the final value T = z,, to zsQ ’ CK and ends Z_ shows being CK = 2 -.SQ’ T~/Z~, the is sufficient. U2 1 J --Y- 200 V 3.2 +I I- 3.2.1 2 ms Fig. 3: Transient voltage zing a relay coil. During t L 2 T the characteristic impedance of the secondary cab lK e represents the impedance of the noise source. In HV substations the length of secondary cables is about 30 m to 100 m; assuming a travelling wave velocity of 0.16 m/ns we find for t = 0.375 US . . . 7.5 ,.us Z. = ZK. In most cases the I maximum of the noise voP tage- will appear within this period. generated by de-energi- Radiated waves Switching in primary circuits The magnetic flux density produced by switching in 420 kV substations comes up to 0.1 mT nearby secondary cables a range of in frequency - 200 kHz . . . 100 MHz. For the 10 kV/m have been measured. electric field up 292 to In open air substations e.g. the disconnector switches the current transformer and the line section between disconnector and circuit breaker. When neglecting the resistances one finds fo: thf current from CT to earth with C as the caused by discharges I = uLE /m conductor-earth capacitance of the CT and L as the inductance of the circuit. For 420 kV substations it can be estimated C = 1 nF, L = 50 uH . . . 100 ,uH resulting in “; = 1085 A . . . 1534 A. Thi.$ current produces in a distance of 5 m a maximum magnetic flux density B = ‘i.p,/2n r 1 61 ,uT. When operating a discon;ecFr in GIS the correspon, = uLE/ZLd = 1372 A with is current ding 2 = 250 ohms as the source lmpe ante of a 420 kV okerhead line. This current will flow at the GIS bushing along the enclosure and via the earth wires to earth. 3.2.2 Lightning strokes substations to earthed components - By a proper choice of the material (wool, antistatical treatment) and/or regulation of the humidity the above mentioned value can be met. 3.4 The shape of the noise voltage Transient voltages in HV substations are normally strongly damped. To simulate a noise voltage by a damped oszillation according to[ 71 will cover a lot of the interference occuring. However, the rise time of of 75 ns fixed in [7] d oes not meet the requirements GIS; here rise times of 5.....10 ns have to be considered. The amplitude of switching overvoltages varies during the operation interval. Secondary equipment may show maloperation when voltage limiting devices are applied and the transient voltage amplitude is right below the limiting voltage. 4. Measures to reduce interference HV substations values in in HV 4.1 Primary circuits Lightning strokes to earth shield wires and structures will normally lead to a distribution of the lightning current which is fed via a number of injection points into the earthing system. However, if lightning rods are installed in form of steel masts erected on the ground the current is injected at one point into the earthing system. The maximum magnetic flux densities can be estimated as follows: There are only a few steps possible to reduce the noise sources in primary circuits. This stems from the fact that there are few technical solutions which can be implemented at reasonable cost. Therefore most of the measures listed below are necessary to protect HV equipment, but they have also a beneficial effect on interference in secondary circuits. From the statistical distribution a lightning current amplitude of 200 kA can be withdrawn as a 95 % value, i.e. in 5 % of all events this value is exceeded 3 . Concerning 4 injection points into the earthing system the magnetic flux density in a distance of 10 m is B = 1 mT. For only one injection point the distance between the lightning rod and the relay kiosk may be about 15 m; in this case the maximum flux density is B = 2.65 mT. The lightning current resp. the transmitted field has a frequency range up to some MHz. Nearby HV substations, especially nearby GIS, overhead lines are equipped with two or even three earth shield wires. This measure reduces the frequency and the amplitude of strokes to the conductor. By this the influence of lightning strokes on secondary circuits is reduced, too. 3.2.3 Earth faults The highest 50 Hz flux densities are to be expected for earth fault currents. In a 420 kV substation for a fault current of 35 kA the magnetic flux density is B = 0.1 mT considering a distance of 7 m. 3.2.4 Radiotransmitters (walkie talkies) The intensity of interference fields produced by radiotransmitters depends on the distance between antenna and secondary equipment and on the transmitted power. An acceptable maximum value seems to be 10 V/m [4,5]. 4.1.1 4.1.2 3.3 Electrostatic discharges The electrostatic charge can be limited to a value of 5 kV also for substations. The voltage people can be charged up to depends on the material of the carpets in control rooms and on the relative humidity; for lower humidity the maximum voltages rise strongly 161. against Configuration lightning of earthing strokes. system An earthing system properly designed for 50 Hz stresses needs only slight modifications to reduce transient voltages. Firstly single tee off connections have to be avoided in general, and secondly bare conductors should be layed parallel to long cables and be connected to the earthing grid. 4.1.3 Measures at CT and VT The transient voltages transmitted via the measuring transformers to the secondary circuits can be reduced by careful earth connections within the transformer and additional screening of the secondary windings 191. 4.2 Secondary 4.2.1 The permissible distance d depends on the transmitted power P and is d = 1.6*/-P/E [6]; for E = 10 V/m and P = 2 W the permissible distance between antenna and cabinet is d = 23 cm. Protection circuits Screening of secondary cables The best system to reduce interference in the circuits is the adoption of screened cables. In the ideal case of continuous and perfectly homogeneous screens with no resistance, the protection against the external high frequency electric and magnetic field would be perfect. Because of the practical performance of the cable screen however, one has to consider some points: - The screen should be almost continuous and with low resistance (a few ohm/km). The screen should have a low coupling impedance within the interference frequency range. ‘-’ - _ Earthing of the screen should have a very low impedance, that is, the earthing conductors should have adequate section, minimum length an optimum contact arrangements. In some cases it may be necessary to earth the screens at the inlets to the relay rooms or in the equipment cabinets so that the currents circulating in the screens do not affect the unscreened circuits; whenever interferences are due to induction, earthing at both ends is suitable. 4.2.2 Configuration of secondary circuits The cable route should run as far as possible not parallel to bus-bars or power cables. - 293 from and The forward and return conductors of the same circuit should be run in the same cable; twisted pairs or quadruple cables should be adopted whenever possible (for instance for very low current circuits and data lines). 5416 - 4.3 Introduction of new technologies An essential reduction of interference can be achieved in the future by - by guided waves application of optic fibre cables development of electronical measuring tratX.fOrmerS transmitting the digitized signal by fibre optic cable from the HV level to the protection equipment It is the advantage neither conductive, nor radiated waves sion. Consequently mary events will be of the fibre optic cables that inductive and capactive coupling are able to disturb the transmisthe noise voltages caused by prineglectable. Today, CTs or VTs with digital output are still in the experimental stage and their introduction on a wide-spread basis is not expected in the near future. 5. Conclusions - All the screened cables should run ,as close together as possible in order to benefit from their mutual screening effect. - Laying of bare conductors parallel ,to the cables; the conductors are to be connected to the earthing network at the two ends and, if possible, at a few points along their route. - for D.C. auxiliary supply cables guration is better than a ring. 4.2.3 Use of voltage limiting a radial confi- devices Voltage limiting devices should be installed inside the protection and control equipment; their installation outside the equipment in new plants should be avoided. However, their adoption might be useful in plants already in operation, to allow installation of standard devices with low EMC limits. The most common - - - - voltage limiting A proper and careful installation leads to interference values in secondary circuits of HV substations which are below 1000 V; an exception are the noise voltages caused by lightning strokes and by switching in secondary circuits. As the lightning current amplitude and steepness is statistically distributed the acceptable risk determines the measures to be taken. Ry installation of voltage limiting devices the amplitude of the noise voltage can be held within the acceptable limits. The same concerns the noise voltages produced by secondary switching. The future use of optic fibre cables will result in an essential reduction of interference. Future measuring transformers, equipped with digital electronics and fibre optics will transmit only small noise voltage into secondary circuits. devices are: Condensers or RC circuits that reduce HF overvoltages and are adequate both for circuits coming from the switchyard and for supply circuits. An interesting application consists of filtering the auxiliary D.C. power of circuits and equipments at the lowest EMC level, supplied by a battery which is feeding also circuits and equipment (e.g., HV breakers and disconnectors) at the highest EMC level. LV arresters discharge overvoltages at high energy content in A.C. and D.C. circuits having voltages -5 48 V and have such a time delay that they are not adequate for steep and HF overvoltages. They require a low-impedance earthing to the equipment they protect; generally they are installed on telecommunication lines. Zener diodes are advisable for overvoltages at low energy content only and therefore should be used with great care. Transzorb diodes are electronic components similiar to zener diodes with very good characteristics such as very small delay time and leakage current and constant voltage limitation up to some hundreds of volts. With a series-parallel combination they can be used for high voltage limitations and for high energy transients. Varistors show a restistance that is inversely proportional to the applied voltage; utilization of the zinc-oxide type is spreading (there is a tendency to replace zenerdiodes and RC circuits with these varistors) because of their short delay time (5 25 ns) and high impulse current (up to 25 kA). Appendix Estimation of transient secondary lightning strokes to the conductor. voltages caused by a) 420 kV open air substation For a shielding angle of 45” (I earth shield wire) the maximum lightning current amplitude reaching the conductor is I - 36 kA [I]; the measured transient secondary vol aie caused by switching in primary circuits (U = i ‘-I-l /p) comes up to U = 500 V.To estimate h&w oftenmU = 2500 V is excee ?led at first the corresponding prim&y voltage Ul is determined: resulting 2.1710 ‘Bl= kV = 13.7 in kA ZW of the with 2 = 250 ohms as the surge impedance overh&x line. Lightning currents between 13.7 kA and 36 kA can produce secondary transients of 2500 V and more. It is known from measures that the steepness S of the primary voltage ist proportional to the amplitude of the secondary transient for S 2 10 kA/,us. Since 97 % of all strokes have a steepness 2 10 kA/ us anyhow this value is fixed as a bottom limit fo t/ the steepness considered. 234 125 T;: i :: 1 100 ! ! / Fig.Bl 50 25 25 0 50 0 100 150 200 250 3;0 I L= 0.8 1 1 kA/ks-z w/km 1 = 3.81 km For this value of L the maximum steepness stressing the substation is 10 kA/ us. Along the collection length L the steepness of t h e lightning . strokes causing 10 kA/ us at the substation will differ; the relations are sh d wn in table A for different length 1. P describes the probability that the corresponding steepness ist exceeded for stroke currents from 13.7 .. . . . 36 kA. These figures can be derived from the presentation in fig. 6 showing a typical distribution of amplitude and steepness of stroke currents when considering their correlation 131. The integral of the probability P along the collection length L comes up to P, = 24 %, i.e. along L 24 % of all strokes fulfil 13.7 kAL IB” 36 kA and cause S b 10 kA/,us at the substation. The frequency H of secondary transients 1 2 500 V is then H = B-L. P;k.NB with collect.ion width LB isokeraunic level N”, z “1!9;a km21 [l] k regards which part of the strokes within the determined collection area really hit the conductor. Following GOLDE [lo3 it can be estimated k = 0.25. H = 0.04 ’ 3.81*0.24.0.25 = 0.017/(a.circuit) - 1.9/(a=circuit) This results in a transient secondary voltage h 2 500 V every 58 years for each circuit; a substation with five circuits will be stressed every 12 years with such a transient. Table A S (kA/,us) 65 60 50 40 35 30 25 20 15 10 l(km) 3.81 3.75 3.60 3.38 3.21 3.00 2.70 2.25 1.50 0.00 PI%) 0.0 0.2 0.8 2.4 4.6 8.0 12.6 23.6 32.6 39.2 40 35 30 25 20 15 Table B S(kA/,us) 43 I(kd 1.53 1.50 1.43 1.33 1.20 1.00 0.67 0.0 PC%) 0 0 0.2 1.0 1.6 4.2 10 10.2 14.4 0 50 100 150 200 250 I [LA] From Fig. A the 95 % value of the steepness corresponding to IB2 = 36 kA can be derived 5 = 65 kA/ US [3]. Regarding an attenuation of 0.8 ps/km along t!he line the collection length L becomes IB2 * 10 0 300 bA1 b) 420 kV GIS Considering two earth shield wires the maximum - 16 kA, the stroke current to the conductor is I 95 % value of the steepness is 43 kAv2ui. The measured transient secondary voltage cause 6 by switching in primary circuits is U2 = 1000 V. To estimate how often a secondary voltage of 2000 V is exceeded the corresponding primary voltage U - 684 kV and the stroke current j-,1 = 5.5 kA have take considered. The collectlon ength 1s L = 1.53 km, from table B one gets pZ= 7 % and the frequency of transient voltages exceeding 2000 V is H = B*L* p,- k*NB = 0.025 * 1.53. 0.07 * 0.25. 1.9/(a.circuit) ZZ0.0013/(a.circuit) i.e. a substation with five circuits will be stressed every 157 years with a secondary transient A 2000 V. References PI Anderson, meters for R.B.; Eriksson, A.J.: Lightning paraengineering application. Electra Nr. 96, 1980, 65-102 von PI RGhsler, H.; Strnad, A.: uberspannungsschutz metallgekapselten gasisolierten Schaltanlagen im 420-kV-Netz. Etz Archiv 6 (1984), 233-238 PI Fischer, M.; Strnad, A.: Bestimmungen der bei Blitzeinschlsgen zu erwartenden transienten Uberspannung in Sekundgrkreisen von Hochspannungsschaltanlagen. ElektrititZtswirtschaft 82 (1983) 87-91 P+JAnders, R.; Campling, A.C.: Interference problems on electronic control equipment in power plants and substations - installation and interference tests. CIGRE-Report 36-05 (1980) G.: Disturbances produced by transr51 Champiot, ceivers and walkie-talkies. Electra No. 83 (1982), pp. 103-110 conditions - Part 5: Electro[61 IEC 654-5: Operating magnetic compatibility. Draft-Publication 1983 Impulse voltage withstand tests and PI IEC 255-4: high-frequency disturbance tests. Appendix E. PI Riihsler, H.; Strnad, A.: Die HBufigkeit riickw;irtiger Uberschkige und ihre Reduzierung. Elektrizit8tswirtschaft 82 (1983), 386-390 von Sekundsrkreisen in PI Strnad, A.: Beeinflussung Hochspannungsschaltanlagen bei rasch ver;inderlichen VorgZngen im Hochspannungskreis; Thesis (1982) Technical University of Darmstadt, West . Germany [IOj Golde, R.H.: Lightning, Vol I. Academic Press, London 1977 - BALLOON AND SATELLITE 295 - 55 OBSERVATION OVER NORTHERN Takeo Yoshino University l-5-1 Chofugaoka, 17 OF POWER LINE RADIATION EUROPE and Ichiro Tomizawa of Electra-Communications Chofu-shi, Observation of the electromagnetic field variation phenomena of Power Line Radiation (PLR) related with polar substorm activity had been done by two observation balloons and one satellite in the arctic circle of northern europe. The balloons named as B15-1N and B15-2N were launched on March 20, 1982 from Stamsund, Norway and on November 23, 1982 from ESRANGE, Sweden respectively. Both balloons could be obtained several electromagnetic field variation data with various frequencies caused by substorm effect during their flight. By the balloon observation, the field intensity of higher harmonic frequency of Power Line Radiation(PLHR) in the frequency range between 200Hz to 1 kHz which were enhanced by the EM field disturbance in the poler substorm, were obtained the clearly spike-like level increase as harmonics of 50 Hz step in the frequency spectrum data, at 300 Hz, 450 Hz and 600 Hz appeared in the data of B15-1N and at 300 Hz and 450 Hz in B15-2N. However, the intensity increase of fundamental, and 2nd and 3rd harmonics of 50 Hz were not exceed over than 3 dB during the polar substorm. The Japanese scientific satellite EXOS-C "OHZORA" was launched on 14th February 1984. The data in the northern europe is received from satellite at ESRANGE station based on the JapanSweden co-operation program for EXOS-C, and the data with analized form will be able to appear by end of this year. INTRODUCTION As shown in a recent paper [l], the observation results of fundamental ELF wave propagation characteristics of Power Line Radiation to horizontal and to vertical direction were observed by using of balloons and rockets over -Japanese island. One of the purpose of this experiments was the determination of the standard propagation characteristics of ELF waves during the condi- Tokyo 182, Japan tions in quiet solar activity at middle latitude area used as the caribration standard. By this observation resultsI the attenuation constant to horizontal direction propagation which are consisted of a guided mode between the bottom side of ionosphere and the ground (sea) surface observed by balloon experiments, is approximately 1.2 dB/lOO km at 50 Hz and 1.3 dB/lOOkm To vertical at 60 Hz respectively. direction propaqation which penetrate into the lower ionosphere is-approximatelv 1.1 dB to 1.2 dB/lOO km on eithe; frequency, which'observed by four rocket experiments as described in the recent paper [l]. This work continue after publish of paper [l] to obtain more precisechracteristics on the several items of basical propagation characteristics at more wider areas. For example, a transPacific balloon experiment have been launched on September 23, 1984 from Sanriku test range, Japan, to observe the long distance propagation over the Pacific ocean of Power Line Radiation from Japan. And this balloon flew eastwards and operated 50 hours after launch, and the flight distance was approximately 2000 km on the pacific ocean from Sanriku, Japan. The 9th Japanese scientficsatellite named EXOS-C "OHZORA" was launched on 3.4th February 1984, from Kagoshima Space Center as anouncement in the paper 111 for global monitoring of PLR. The observation results of this satellite will be mentioned later. In the last decade, lot of high power hydro-electric power generator stations have been built in sub-polar regions or in polar regions as Manitoba and Quebec in Canada and in northern area of Norway and Sweden. And these huge electric power have been carried by extreme long distance transmission line with 3 wire system over 2000 km from sub-polar area to main industrial area in the middle latitude,across perpendicular the aurora1 oval. The line voltage of these power lines are - 296 This paper presents the results of balloon observations on the electromagnetic field induced by Power Line Radiation observed around aurora1 oval under geomagnetic 'active conditions. Observations have been done by using balloons, that is why, strong electromagnetic radiation of Power Line field will be suppress the true values of integrated electromagnetic field induced by power line systems. 500 kV or more. At major geomagnetic disturbavce on September 21, 1977, the huge unbalance anomaly current of over 25A to 50A was induced, and observed this induced current at the grounding point of the transformers on the Manitoba line of Canada. And there are many troubles appear on this line such as the breakdown of main circuit breakers and the burnout of transformers wires. The many similar troubles in this time appeared not only on the power lines of the Manitoba lines but also on the other long power line system at subpolar and polar area as Alaska, north Canada and northern europian areas. At same time, the troubles were expanded other field as long distance communications wire and the pipe lines in the sub-polar and polar regions. [2], [3] and [4]. Above mentioned induced anomaly current on power lines induce magnetically saturation of transformers core, and the waveform of current and voltage will be distorted by the non-linearity effect of the saturated transformers. This distorted current will be produced lot of higher harmonic frequencies radiation from power line system to magnetosphere. [5,6]. BL5-2N - OBSERVATION BY BALLOON The balloon B15-1N was launched at 1909 UT on March 20, 1982 from Stamsund in Norway, and the balloon B15-2N was launched at 2059 UT on November 23,1982 from ESRANGE in Sweden. The flight trajectries and altitudes of balloons are illustrated in Fig. 1. The trace of B15-1N drifted eastwards from Stamsund and flight to across over northern Norway and Sweden, and dropped the payload in the territories of Finland as shown in Fig. 1. And B15-2N drifted toward north-east from ESRANGE, and flying across over northern Finland,the west of Kolskiy area of U.S.S.R.and the payload dropped into Arctic ocean at off the coast of Murmansk as shown in Fig. 1. (1982/11/23-24) 70 (1982/ 3/20 U.S.S.R 1lJ 1982/ 3/20 .I 0 15 20 25 30 35 40 45 50 55 6 DE LONGITUDE Fig. 1. The flight trajectries and altitudes of balloons Bl5-1N and B15-2N. The balloon B15-1N was launched at 1907 UT on March 20 and the balloon B15-2N was launched at 2059 UT on November 23, 1982 respectively. - 297 5517 - ELECTRIC FIELD , PRE AMP AMP E PRE AMP AMP HPF U-MAGNETIC FIELD I I 1 L Fig. 2. A schematic block diagram line radiation measurement board the balloon. of the power system on The geomagnetic activity on flight time of both balloons, Kp index was 2+ from 19 UT to 21 UT on March 20, 1982 for B15-lN, and the value of Kp index fluctuate from 6+ to 3-l-from 21 UT on November 23 till 03 UT on 24, 1982 for B15-2N. The on boarded equipments of balloons B15-lN.and 2N have mostly same properties except the receiving gain. Fig. 2 shows a block diagram of PLR detector, and the frequency range of this detector covers from 40 Hz to 1 kHz. The detector consists of a horizontal magnetic field sensor, a vertical electric field sensor, and a multiplexer for switching the magnetic and the electric field signals as shown in Fig. 2. The detected signals are connected to the telemetry transmitter through IRIG-#18 FM channel, which has the maximum frequency responce of 1kHz. The horizontal magnetic field sensor is equipped with a loop antenna which uses a permalloy core (50cm x 6 mm x 6mm) wound 6000 turns with the copper wire of 0.2 mm pl, a flat frequency responce pre-amplifier, and 40 Hz HPF which is required to reject the strong Schumann resonance frequency portion on the spectrum. The vertical electric field sensor is equipped with a short electric dipole antenna which has a copper wire of 10 meter tip-totip, connected with high input impedance terminal of pre-amplifier, and 1 kHz LPF. The multiplexed signal is separated into two amplifiers in which the gain difference is 40 dB to get wider dynamic range. OF OBSERVATION RESULTS OF Bl5-1N Fig. 3 (a) shows the H component variation of geomagnetic disturbance at Andoya during flight the B15-1N balloon. The Figures 2 (b) and (c) g I l-i ;; 2 ::: i 6)v-i g 5 E 5 i _ ‘p 4 ml9 s ; ; '; 6 i z L !: g d : i !2UT 21 20 1982/ 3/20 5OHz I ,60HZ \lBOHz ml9 20 21 .-I Fiq. RESULTS are display of the observation record of on board detecter on the variation of magnetic and electric field intensities respThe contours of 60 Hz ectively. and 180 Hz as indicated by dotted line are shown as refer to the But back ground noise level. the real electric field intensity does not exceed over the system noise level in this figure for the gain set too low against the intensity of back ground noise level in this district. In the Fiq. 3 (b), the magne-, z2UT 7JME 1982/ 3/20 3. (a) The horizontal component of the masnetograph at Andoya, Norwayl (1;) and (c)the maanetic and electric field intensity which observed by balloon B15-1N on March 20 1982. Solid line indicated 50 and its harmonics, and dotted 1 i ne indicated 60 and 180 HZ. I - 298 tic field intensity at 50 Hz has a pea % at 1936 UT with the value of 1.6 x lo-, greater than the (A/m) which is much , _ value of 60 Hz (rerer to back ground The intenoise level close to 50 Hz). nsity at 50 Hz gradually decreases with time until the end of data without any enhancement, even at the time of substorm onset at 2040 UT. The intensities of harmonic radiation at 100 and 150Hz are less than the intensity of background noise through out this observation. The magnetic data between 1909UT to 1920UT should be omitted, because the large swing of contours are affected by the joggle of gondola during the ascending of balloon. Dynamic spectra of the magnetic field in the frequency ranqe between 40Hz to 1kHz is shown in Fig; 4. Spectral peak of 50Hz is clearly identified until 20 13UT, however, after 2017UT the 50Hz sunk into background noise level. Spectral peak at harmonics of 50Hz cannot be identified in the dynamic spectra until 2036UT. At 2040UT, a spectral peak appears at 350Hz, but it does appear after 2044 UT. The remarkable spectral peak appears in the successive spectra between 2044UT until 2121 UT. Also the spectral peak at 600 Hz, peaks at 300 'Hz and 450 Hz appear in the spectra from 2051 UT to 2121 UT, but the peaks of 300 and 450 Hz are not clearly identified. It is noticeable that the peak frequencies are appeared not exactly at the harmonic frequency of 50Hz, occasionally shifted lowerside. These lowerside shift of frequencies of the three spectral peaks is correlated each other, and the amount of shift frequency is proportioal to the frequencies of 300, 450 and 600 Hz. This lowerside frequency shift bill be attributed to the shift of fundamental frequency of 50 Hz. The appearance of these spectral peaks is coincidence with the geomagnetic substorm which started at 2040 UT. Therefore, it is concluded that the geomagnetic substorm not only enhances the power line radiation especially at 300,450 and 600 Hz, but also disturbs the frequency of the generator revolution as 0 heavy unbalance load. Fig.4 Dynamic spectrum of the magnetic field in the frequency B15-1N balloon. REXULTS OF B15-2N Fig. 5(a) is shown the magnetgram record of H-component magnetic field intensity and ULF magnetic pulsation on November 23, 1982 which observed at ESBANGE, Sweden, during the flight of Bl5-2N balloon. A polar substorm are occured at 2251 UT and decayed after 30 minutes. A negative bay of H-component of magnetic field and a sudden comencement of H-component of magnetic pulsation at this substorm were recorded in Fig. 5(b) and 5(c) illustthis figure. rate the magnetic and electric field intensity variations during the balloon and flight at 50Hz, 60Hz,100Hz,150Hz 180Hz respectively. The contours of 60 Hz and 180 Hz indicate of background noise level close to the fundamental 0.5 Freqency range between LkHzl 40Hz to 1 kHz observed by - 299 and 3rd harmonics of power line radiation. The magnetic field data between 2115 to 2220 UT should be omitted, because the large swing of ccpztaurs are affected by the joggle of gondola during the ascending of balloon. It gives a clear evidence that the intensities at 50Hz,lOOHz and 150Hz were not enhanced during substorm. The electric field intensities of 50Hz, 1OOHz and 150Hz had increased from 0115 UT to 0230 UT in spite of low geomagnetic activity, when the balloon flew over the large industrial and mining area near Mulmansk in USSR. The level of enhancement of the magnetic component are much less than the electric component at this time. This record is suggested that the enhancement of power line radiations could be induced by a localized power line system. The Lapland area have not much high voltage power lines. Fig. 6 shows the dynamic frequencyspectra of electric field intensities at this time. r I FREQUENCY (kHz) Fig. 6 I I t 21 22 23 1 I I 00 I I ‘ 01 02 03 -4 7 6X21 I 22 23 00 01 I . 02 I03 Dynamic frequency spectrum of electric field intensities at 0115UT to 0230UT by balloon Bl5-2N. Fig. 7(a) and 7(b) shows the amplitude spectra of magnetic and electric fields around 2250 UT of the frequency range between 4OHz to 1kHz of PLR harmonics radiation during the geomagnetic substorm which started at 2251UT as shown in Fig. 5(a). From the observation results in this figure, it is clear evidence that the spectral peaks at the harmonic frequencies of the power line radiation in the frequency range higher than 200 Hz are enhanced at the time of the substorm onset at 2251 UT in the electric and magnetic fields. The intensity in enhancement at 300 Hz and 450 Hz corresponds to the duration of the enhancement of ULF pulsation included high frequency component as shown in the Fig. 5(a). OBSERVATION OF SATELLITE EXOS-C"OHZORA" The Japanese scientific satellite EXOS-C "OHZORA" was launched at 0800UT on 14 February 1984 from Kagoshima Space Center in Japan. Satellite on boarded the observation equipments for correpondence to MAP program and a 3 channel narrow band receiver for magnetic field detection and a wide band 2: receiver for electric field detector to cs21 22 23 00 01 02 03UT PLR observations. The initial orbital ’ data was; apogee-865 km, perigee-354km, 1982/11/23-24 incrination-74.5'and orbital period-96.9 Fig. 5 Intensity of power line radiation minutes. The data in the northern europe observed on the balloon Bl5-.2N, is received at ESRANGE station, Sweden. (a) H-component of magnetgram and The data of PLR in northern europe is ULF magnetic pulsation.(b) and (c) handling for analysis, and the presenmagnetic and electric field inten- tation of this results will be able to sities record at 50 to 180 Hz. start by end of 1984. As an example, a data observed over east China shows as Figure. 8. .-I - 300 - CONCLUSION ill References Yoshino T,, and I.Tomizawa: Rocket & balloon observations of power lines over Japanese islands, EMC 81,Zurich, pp525-530,(1981). Boerner, W-M., Res. Grant Proposal (1979 and 1980) Akasofu, S-I., and Merritt, Nature, 279, 308-310,(1979) Lanzerotti, L.J., Space Sci. Rev., 34, 347-356, (1983). Helliwell, R.A., et al, J. Geophys. Res., 80, 4249-4258, (1975). Luette, J.P.,,C:G. Park and R.A. Helliwell, Geophys. Res. Let.,4,275278, (1977). Hayashi, K., et al, Nature, 275, By the balloon and satellite observations of power line radiation characteristics, the relationship between the power line radiation and the magnetic [21 substorm must be separate to estimate at the effect of the frequency range [31 below third harmonics, and at higher than 4th harmonics of fundamental fre[41 quency of power line for three phased AC transmmision system. The intensity [51 of fundamental frequency and lower harmonics of PLR observed on the ball[61 oon is not controlled by the geomagnetic substorm. And it is suggested that the enhancement of power line radiation [71 at lower frequencies will be influenced 627-629, (1978). mainly by the localized condition of [81 Park, C.G.and R.A. Helliwell, Science power line networks and power consumpScience, 200, 727-730, (1977). tion in that area. t91 Tomozawa, I. and T. Yoshino, Memoir However, the intensities of the of NIPR, Special Issue 31, 115-123, higher harmonics more than 4th orders (1984). are enhanced by the magnetic field variation of substorm as shown in the Figures 4 and 8. This phenomena was also observed clearly on the ground at the time of geomagnetic b) subsrorm, which increase of 3rd, 6th, 9th, 12th and more higher harmonics of power line. This data appeared in Central Canada in 1979 [7]. Ry the radiation mechanisum described the recent paper 111, the intensities of radiated magnetic field are emitted from a current loop r! which consists of neutQ, ral wire of three phase iT system and the ground, .umust be intensified by these harmonics current. E The intensity of PLHR at g 2 more than higher frequencies can be even induced under the condition of the more weak 1. a.*0 1 0.5 0.5 0.0 1. FRE:QUENCY substorm. [8,91. tra of thdkE &tric _I and magnetic Fig. 7 Dynamic spec The data of this field betwee n 40Hz-1kHz observed by B15-2N. balloon and satellite I ’ ‘ experiments will be able Rev to offer the lot of qls: valuable saggessions on the research field to related effect of power line radiation in the magnetosphere. T--l l-u 1 Acknowledgement The authors 'are deeply indepted to Drs. H. Yarn;-gishi, T.Ono, H.Fukunishi, H.Miyaoka and M.Ejiri of NIPR, and the staff of ESRANGE, and the staff of ISAS to their cooperation and to helpful discussion, and to the assistance of S. Yamakawa in the spectral analysis. I I I L 121 111 111 1s 181 -'&s' ’ ’ * ’ * * . . LONGITUDE 1984/6/2 UT Fig. 8. The RLR observation over eastern China by satellite EXOS-C "OHZORA", 171 - 301 56~1 - THE NUMERICAL ELECTROMAGNETICCODE (NEC)" J.K.Breakall, G.J.Burke, E.K.Miller Lawrence Livermore National Laboratory Livermore, CA 95550 I. INTRODUCTION In the 20 years or so since the first computer-based Moment-Method solutions for EM radiation and scattering problems were presented, numerous computer codes have been developed. Relatively few of them have found applications beyond those for which they were originally intended. Of the latter, furthermore, only a small fraction have received widespread acceptance and implementation. The Numerical Electromagnetics Code--Method of Moments (NEC) Cl] is one of those codes which has entered this select category. There are several reasons for this situation, including: 1) continuity of support and personnel involved in its development; 2) systematic updating and extension of its capabilities; 3) extensive, user-oriented documentation; 4) its ready availability; and 5) accessibility of its developers for user assistance in applications. The result is that there are estimated to be several hundred users of various versions of NEC world wide. NEC has been under development for more than 10 years (in earlier forms it was known as BRACT and AMP). It is a hybrid code which uses an Electric Field Integral Equation (EFIE) to model wire-like objects and a Magnetic Field Integral Equation (MFIE) to model surface-like objects with time harmonic excitation. A threeterm sinusoidal spline basis is employed for the wire current while a pulse basis is used for the surface current, with delta-function weights employed everywhere. Provision is made for connection of wires to surfaces via an attachment, or interpolation, basis. NEC includes a number of features for efficient modeling of antennas and scatterers in their environments. The most recent addition, in the version NEC-3 [2,31, has been the capability of modeling wires that are buried or penetrate the ground-air interface. This paper addresses the current status of NEC-3 and plans for "Future NEC." Some typical results are included. THE SOLUTION METHOD II. Integral Equations NEC-3 uses both an electric-field integral equation (EFIE) and a magnetic-field integral equation (MFIE) to model the electromagnetic response of general structures. Each equation has advantages for particular structure types. The EFIE is well suited for thin-wire structures of small or vanishing conductor volume while the MFIE, which fails for the thin-wire case, is more attractive for voluminous structures, especially those having large smooth surfaces. The EFIE can also be used to model surfaces and is preferred for thin structures where there is little separation between a front and back surface. Although the EFIE is specialized to thin wires in this program, it has been used to represent surfaces by wire grids with reasonable success for far-field quantities but with variable accuracy for surface fields. For a structure containing both wires and surfaces the EFIE and MFIE are coupled. This combination of the EFIE and MFIE was proposed and used by Albertsen, Hansen, and Jensen at the Technical University of Denmark [4] although the details of their numerical solution differ from those in NEC. A rigorous derivation of the EFIE and MFIE used in NEC is given by Poggio and Miller [5]. The thin wire approximation is applied to the EFIE to reduce it to a scalar integral equation. The assumptions involved are: a. b. c. d. Transverse currents can be neglected relative to axial currents on the wire. The circumferential variation in the axial current can be neglected. The current can be represented by a filament on the wire axis. The boundary condition on the electric field need be enforced in the axial direction only. These widely used approximations are valid as long as the wire radius is much less than the wavelength and much less than the wire length. *Work performed under the auspices of the U.S. Department of Energy by the Lawrence Livermore National Laboratory under contract No W-7405-ENC-48. - 302 An alternate kernel for the EFIE, based on an extended thin-wire approximation in which condition c is relaxed, is also included in NEC for wires having too large a radius for the thin-wire approximation [6]. With the thin-wire approximation the EFIE becomes - for F on wires and i,(F) l i?(F) - - & t,(F) I l I(s’)(E x 0’ L -3 2(F) = 2 l I(s’)(k2P . 9’ g(;,;‘))ds’ - - 3 t*,(F) l Is(G) - C ~1 a2 PC:, ; i’)ds’ E Z(s) 1 ZG f :2(G) l [zs(:‘) X V’ g(;,;‘)]dA’ (2) s1 where and g(;, ;') = exp(-jk(r, ;‘I)/\; - ;‘I -i,(i) k =WV’!J E 00 l i+(r) = & f,(T) l I I&')(2 x 0' L g(t,;‘)ds’ I is the induced current, E' is the exciting field, s^ and s^'are unit vectors tangent to the wire at s and s', and i;and P' are vectors to the points s and s' on the wire contour C. x g(F) =:- 3 3p> 1 G n(F) + x V’ ES where t?(P) is $e unit vector normal to the surface at P, H is the exciting magnetic field, and j is the surface current. This equation is separated into the coupled scalar integral equations for components of the surface current. For a structure with both wires and surfaces, the hybrid EFIE and MFIE equations are l 8’ V' g(r,?)ldA' (3) The integral equations (l), (Z), and (3) are solved numerically in NEC by a form of the method of moments. The test functions for the method of moments solution on both wires and surfaces are chosen to be delta functions resulting in a point sampling of the fields known as the collocation method of solution. Wires are divided into short straight segments with a sample point at the center of each segment, while surfaces are approximated by a set of flat patches or facets with a sample point at the center of each patch. Straight segments and flat patches are not mandatory for the solution method but are used to simplify the specification of the geometry and evaluation of the fields. Delta functions are also used in NEC as the current expansion functions on surfaces. For the MFIE this elementary Galerkin's method has been found to provide good accuracy on large smooth surfaces. Due to the nature of the integral-equation kernels, however, the choice of current expansion functions is more critical in the EFIE than in the MFIE. a2 asasl) - - L g(;,;‘)dA’ f i,(i). [?$l)x + Numerical Solution g(;,;‘)]dA’ S I(s’)(k2$ 9 js(G) for F on surfaces. The vectors i,(F) and t"*(P) are orthogonal unit vectors tangent to the surface at F, and the symbol / represents integration over wires while, represents integration over surfaces exclu + Ing wires. X [j,(?‘) ; S,(G) s1 The MFIE for a closed surface s is a(;) 1 z? - 5 (1 The expansion functions for the current on wires are chosen so that the total current on segment number j has the form Ij(s) = A. + Bjsin 3 k(s-sj) Is-ss( + Cjcos kb-sj) < Ai/2 J (4) J 56 303 - where s. is the value of s at the center of SegmentJj and A. is the length of segment j. This expansion das first used by Yeh and Mei L-71 and has been shown to provide rapid solution convergence [S], [9]. It has the added advantage that the fields of the sinusoidal _ currents are easily evaluated in closed torm. The amplitudes of the constant, sine, and cosine terms are related such that their sum satisfies physical conditions on the local behavior of current and charge at the segment ends. On a single wire with continuous radius, obvious conditions are that the current and charge density (aI/as) are continuous along the wire and that the current goes to zero at free wire ends. These conditions applied at segment ends together with the equations from the method of moments are sufficient to determine the current in the form of (4) on all segments. If the wire radius is discontinuous, Wu and King Cl01 have shown that the charge density should satisfy the condition Jl Kirchhoff's law and (5) for each segment at the junction. If an end of segment j has no connections, then f. goes to zero there with no condition on the de&vative. The general equations for fj(S) are given in Cll, Part I. Where a wire connects to a surface, a more realistic representation of the surface current is needed than the delta function expansion used in NEC is normally used. The treatment quite similar to that used by Albertsen et al. c41. In the region of the wire connection, the surface current contains a singular component due to the current flowing from the wire onto the surface. The total current on the four patches about the connection point, with coordinates shown in Fig. 1, is represented as 4 S,(X,Y) = Iofbw’) + 1 gj(x,y) (sj-Iosj) (6) j=l where I, is the current at the base of the wire, and x2+ yy h P(x,y) = 2?T(x2 + y2) =-.-L-L_ s at junction q$) - y (5) 5 =3(x,y) j s j j where a = wire radius, k = 21~/X y = 0.5772 (Euler's constant). Q is related to the total charge in the vicinity of the junction and is constant for all wires at the junction. ,At a junction of several wires (5) is applied on each wire. The continuity of current is generalized to Kirchhoff's current law and provides an equation to eliminate the unknown Q. qx%Y) = 1 (d+x)(d+y) 4d2 .!J2(X’Y) = -+ (d-x) (d+y) 4d .qX’Y) = 1 (d-x) (d-y) 4d2 g4(x’Y) = L (d+x) (d-y) 4d2 To enforce these conditions on total current, the current on a structure with N segments is expanded as I(s) 1 = j&jfj(5) with the expansion functions fj(S) chosen to satisfy the above conditions. On a uniform wire f.(s) is a spline-like function extending over s4 gments j-l, j, and j+l. In general, if several wires connect to each end of segment j, f.(s) extends over segment j and all connected sigments and has the form of (4) on each segment. The constants A, 8, and C differ on each segment and are chosen so that f. goes to zero with zero derivative at the oute$ end of each segment connected to segment j. At the junctions at each end of segment j, fj satisfies Figure 1. Detail of the connection of a wire to a surface. - and (xj,yj) = (x,Y) 304 at the center of patch j. The current in (6) is integrated numerically when computing the electric field at the center of the connected wire segment due to the surface current on the four surrounding patches. In computing the field on any other segments or on any patches, the delta-function form is used for all patches including those at the connection point. This saves integration time and is sufficiently accurate for the greater source to observation-point separations involved. ground surface. The solution is based on the Sommerfeld-integral formulation for the field near the interface. For the method-of-moments solution, the field values are obtained by table lookup and parameter estimation involving a model for the complex behavior of the field transmitted across the interface. The current expansion is modified to account for the discontinuity in charge on a wire penetrating the interface. This numerical treatment for burial wires is described in [2]. Efficient Solution Methods III, CAPABILITIES OF NEC-3 NEC-3 includes a number of features for convenient modeling of antennas and scatterers. These are summarized below. Source Modeling A voltage source on a wire may be modeled by an applied field on a segment or a discontinuity in charge between segments. Alternatively, a structure may be excited by a plane wave with linear or elliptic polarization or by the near field of an infinitesimal current element. Nonradiating Networks and Transmission Lines Nonradiating two port networks and transmission lines may connect points on wires. These are modeled by deriving a drivingpoint admittance matrix from the admittance matrix of the entire structure to avoid modifying the larger matrix. Loading Lumped or distributed RLC loads may be specified on wires. Also, the conductivity of a round wire may be specified and the impedance computed taking account of skin depth. Ground Effects Three options are available for an antenna in or near the ground. A perfectly conducting ground is modeled by including the image field in the kernel of the integral equation. This doubles the time to compute the interaction matrix. An approximate model for a finitely conducting ground uses the image modified by Fresnel reflection coefficients. This approximation is usable for antennas at least 0.1 to 0.2 wavelengths above the ground and doubles the time to fill the matrix. NW-3 includes an accurate treatment for wire structures above, below, or penetrating the The matrix equation is solved by factoring the matrix into upper and lower triangular matrices which are saved for reuse when the excitation or other parameters that do not alter the matrix are changed. Rotational symmetry or reflection symmetry in one to three planes can be used to reduce both the times to fill and factor the matrix. New wires or surfaces may be added to a structure for which the matrix has already been computed, factored and saved on a file. The new solution is found from the self and mutual interaction matrices through a partitioned-matrix algorithm with no unnecessary repetition of calculations for the basic structure. This feature can be used to take advantage of symmetry in a portion of a structure to which unsymmetric parts connect. Input A user-oriented input scheme permits defining straight wires, arcs, and surfaces. Electrical connections are determined in the program by searching for wire ends and patch centers that coincide. Shifts, rotations, and reflections can be used in building complex structures. A solution can be repeated for modified model parameters (transmission lines, loading, etc.) without respecifying parameters that are not changed. output Output selectable by input parameters may include: current charge density on wires input impedance, admittance, and power radiated power, ohmic loss, efficiency radia,tedfield, power gain, directive gain average gain_ near E and H fields maximum coupling (for matched source and load) receiving patterns scattering cross section For accurate results, the lengths of wire segments should be less than about 0.1 A. Longer segments up to about 0.15 x may be acceptable on long, straight wires or - 305 The wire radius, a, rf!lat.iVe t0 X is by the approximations used in the kernel of the electric field integral equation. Two approximation options are available in NEC: the thin-wire kernel and the extended thin-wire kernel. In the thin-wire kernel, the current on the surface of a segment is reduced to a filament of current on the segment axis. In the extended thin-wire kernel, a CUrtWIt Uniformly distributed around the segment surface is assumed. The field of this current is apprOXimated by the first two terms in a Series expansion of the exact field in powers of a*. With the extended thin-wire kernel, the ratio of segment length to radius can be as small as 0.5 before instabilities appear in the solution, while with the thin-wire approximation, the limit is 2. In either of these approximations, only currents in the axial direction on a segment are considered, and there is no allowance for variation of the current around the wire circumference. The acceptability of these approximations depends on both the value of a/x and the tendency of the excitation to produce circumferential current or current variation. Validation of results is an important step in the modeling process. The ideal validation is comparison with reliable measured data for the antenna of interest. Lacking such data there are several checks that can be made on the internal consistency of the NEC solution. One is to vary the density of segments or patches to test the convergence of the solution. Since the solution for wires uses different functions for current expansion and field testing, reciprocity is not assured. Hence, it is useful to check reciprocity in the receiving and transmitting patterns or bistatic scattering. The accuracy Of the computed input resistance can be checked (except for antennas over lossy ground) by integrating the power in the radiated field. The power obtained by integrating the radiated field is insensitive to small errors in the computed current and can, if necessary, be used to correct the computed radiation resistance, input power and values of gain. An important consideration in using NEC is the solution time versus model size since this may limit the amount of detail that can be modeled and the segment and patch densities. For a model using N wire segments, the solution time in seconds on a CDC 7600 computer is given approximately by the formula T = 3(10-4>kN2/M -I-2(10-6)~3/~2 where and (7) M is the number of degrees of symmetry k = 1 for free space = 2 for perfectly conducting ground Jl or reflection coefficient approximation noncritical parts of a structure while shorter segments, 0.05 A or less may be needed in modeling critical regions of an antenna. A reasonable maximum for the area of a surface patch appears to be 0.04 square wavelengths. limited 56 - 4 to 8 for accurate (Sommerfeld) tr:atment of finitely conducting ground. The first term in (7) is the time to fill the matrix and the second term is the time to factor it into triangular parts for solution. For a model using No patches, the solution time is I_ about T = (10-5)k(2Np)2/M + 2(10-5 (2NJ3/M2 since two rows and columns in the matrix are associated with each patch. The reduced time factor for filling the matrix, due to the deltafunction current expansion, is one attractive feature of this solution method. A more complete relation for running time is contained in [ll, Part III. When the Sommerfeld ground treatment is used, a fixed time of about 15 seconds on a CDC 7600 computer is needed to generate the interpolation tables. The tables depend only on the ground parameters and frequency, however, and can be saved for use in any case in which these parameters are the same. IV. REPRESENTATIVE RESULTS NEC has been used to model a wide variety of structures including LP and Yagi-Uda arrays; Beverage, Discone-Cage and multiarm helical antennas; spheres, cylinders, and cone-sphere scatterers. Whip antennas mounted on a ship or a simple box have been modeled using both the wire-grid and patch methods. Some results obtained with NEC-3 are shown in Figs. 2, 3, and 4. The radiation pattern of a cylinder with two attached wires is shown in Fig. 2. The cylinder height is 22 cm, the diameter 20 cm, and the wire dimensions are shown in the figure. Wire 'a' was driven by a voltage source at its base, while wire 'b' was connected directly to the cylinder. This model used the hybrid EFIE-MFIE capability of NEC. The measured pattern was obtained by Albertsen et al. C4l who presented similar numerical results. NEC results for a higher density of patches than was used for Fig. 2 were in worse agreement with the measurements, possibly due to inability of the MFIE to handle the square edges of the cylinder. The current on a monopole and radial-wire ground screen is shown in Fig. 3 for the screen above and below the ground surface. The sum of the currents along the radials is plotted for negative distance, while positive distance is along the monopole. The transition from upper to lower medium propagation factor for current along the radials is seen-go take place within a vertical distance of 2(10 )ho. - ol0 306 1 - I . 270 I 90 180 J 360 OBSERVATION ANGLE Figure 2. Radiation pattern of a cylinder with attached wires, with wire a excited. In Fig. 4, a buried horizontal pipe is illuminated by the field of a distant vertical tower. The plot shays the magnitude of the scattered field E (4 is normal to the direction to the transmittifig tower) relative to the incident field E, and represents measurements made by moving a probe in a raster-like fashion over the search area. Both active and passive geophysical searches have been modeled with NEC3. V. -16. ’ -.5 ’ ’ -.4 n -.3 -.2 Distance -.l 0. .1 .2 (Wavelengths) Figure 3. Real part of the current on a quarter wave monopole driven against a ground screen consisting of six evenly spaced radial wires with a screen radius of 0.5 h0 and 2_ = 16 j0. The height of the ground screen above the interface is s. The sum of the radial currents is plotted to obtain a continuous current at the monopole-screen junction. It can be seen that transition from the upper to lower medium wavelength occurs6largely within a vertical distance of 2(10 )ho. CONCLUSION NEC-3 is a versatile code for modeling antennas and their environment including transmission lines, networks, loading, and ground effects. It has become widely used, due mainly to its convenient operation, documentation, availability of the code, and continuing support. NEC is written in Fortran IV, consists of about 9,000 lines of code and requires about 95,000 words of memory on a CDC 7600. These are also versions of the code on IBM 3033 and DEC VAX-11 computers. In a project now beginning, NEC is to be revised with the new code to be called Numerical Electromagnetics Engineering Design System (NEEDS). In NEEDS, we hope to develop a truly user-friendly code, taking advantage of both user experience with the present code, and new concepts in software design. One goal is to reengineer NEC into a tool that is more transportable. maintainable. and uodatable bv utilizina a 56 Jl Buried Pipe Figure 4. Typical signature (defined by the magnitude of E relative to E ) of a buried horizontal pip& (length L = 56 m, diameter d = .2 m) buried 5 m below the interface for Z_ = 16 - j80 and f = 200 kHz. The excitation is provided by a vertical electric source 10 km distant. Only the left half of the field, which is left-right symmetric, is plotted. The peaks are associated with the charge accumulation on each end of the pipe. more module structure. The other primary goal is to incorporate features that make NEC easier, and therefore more productive to use. This will be done through introduction of more diagnostics and error checks, more use of standard geometry modules, incorporation of interactive graphics, etc. In addition, a number of new modeling capabilities are planned for NEEDS, including: 0 l l l Electric field integral equation for surfaces with the capability of connecting wires to surfaces. Coupling to GTD codes for reflector antennas and scattering by plates and cylinders. Improved numerical precision on 32-bit computers. Modeling insulated wires in the ground or air. l l l l Propagation calculations for irregular ground. Further exploitation of symmetries and repetition patterns to reduce solution time. Improved model for voltage sources. Built in optimization capability for design applications. Plans for NEEDS were developed, in part, . through a survey of the needs of current NEC users. It is hoped that this interchange between NEC users and developers and among users can be continued through a NEC/NEEDS User's News Letter published periodically. The news letter would provide a forum for exchange of user information gained in practical applications as well as dissemination of information on new developments in the code. Through this process, we hope to develop an expanding set of modeling guidance in a variety of applications. - 308 - REFERENCES Cl1 Burke, G. J. and Poggio, A. J., "Numerical Electromagnetic Code (NEC) - Method of Moments Parts I, II and III," NOSC TD 116, Naval Ocean Systems Center, San Diego, CA, 18 July 1977 (NEC-1) revised 2 January 1980 (NEC-2). C81 Neureuther, A. R. et al., "A Comparison of Numerical Methods for Thin Wire Antennas," presented at the 1968 Fall URSI Meeting, Dept. of Electrical Engineering and Computer Sciences, University of California, Berkeley, 1968. c21 Burke, G. J. and Miller, E. K., "Modeling Antennas Near to and Penetrating a Lossy Interface,' IEEE Transactions on Antennas and Propagation, Vol. AP-32, No. 10, pp. 1040-1049, October 1984. cg1 c31 Burke, G. J., "Numerical Electromaqnetics Code - Method of Moments User's Guide Supplement for NEC-3 for Modeling Buried Wires," Lawrence Livermore National Laboratory, UCID-19918, October 1983. Miller, E. K., R. M. Bevensee, A. J. Poggio, R. Adams, and F. J. Deadrick, "An Evaluation of Computer Programs Using Integral Equations for the Electromagnetic Analysis of Thin Wire Structures," UCRL75566, Lawrence Livermore National Laboratory, CA, March 1974. DOI Wu, T. T. and King, R. W. P., "The Tapered Antenna and Its Application to the Junction Problem for Thin Wires," IEEE Transactions on Antennas and Propason, Vol. AP-24, No. 1, pp. 42-45, January 1976. c41 Albertsen, N. C., Hansen, J. E., and Jensen, N. E., "Computation of Spacecraft Antenna Radiation Patterns," The Technical University of Denmark, June 1972. c51 Poggio, A. J. and Miller, E. K., "Integral Equation Solutions of Three-Dimensional Scatterinq Problems," Chapt. IV in Computer Techniques-for Electromagnetics, edited by R. Mittra, Pergamon Press, NY, 1973. Ccl Poggio, A. J. and Adams, R. W., "Approximations for Terms Related to the Kernel in Thin-Wire Integral Equations," UCRL-51985, Lawrence Livermore National Laboratory, CA, December 19, 1975. ‘This document was prepared as sn account of work \penwrrd the I’nitcd Stales (;evernment. Wversity of (‘alifornia press or implied, curary. Neither hy sn agency of the (‘nited State\ Govcrnmmt or awumcs any legnl liahilily complctenesc. or uwfulncw or respwnihilily of any inform&m, Yeh, Y. S. and Mei, K. K., "Theory of Conical Equiangular Spiral Antennas," Part I - Numerical Techniques, IEEE Transactions on Antennas amropagation, AVol. P- , p. ex- for the w- appsralw. preducl, or process disclosed, or rcprewnts thrl its USCwould not infringe privately owned rights. Referenre herein tu any specific commercial productr, procre. c71 nor the nor any of their empleyecb, makes any warranty. hy trade name, trademark. manufacturer. constitute or imply its endorwmcnl, or otherwise. rrcommcndatien, Stater Government or the University of ~‘alifornia. or wnice doe\ no1 ncwwrily or fweriog hy Ihe I’nitsd .I'he views and opinion% or awlhers expreasad herein de not necessarily Qrte or rellecl thow of the I’niled States G~~~rnent dorwmeot thereof, and shall not IJP uwd for adwrti+ing purposes. or product cn- - CClvB?UTER-AIDED ANALYSIS IN VHF-FM 309 57~2 - OF ELECTROMAGNETIC BROADCASTING D. J. Bern, J. Janiszewski, Technical University Wrockaw, The paper method presents a computer-aided compatibi- for electromagnetic lity analysis casting (EMCA) The method networks. developed in VHF-FM on the basis mathematical model broad- has been of an original of the VHF-FM net- NETWORKS R. Zielidski of Wroclaw Poland The coverage The protection considered, is determined called compatible mitters which compatible tic environment dition requires of compatibility all systems information should compatibility of various signals. tions provide kinds Simulation us with the performance making external interferences. number ters distributed frequency achievable ceivers, with a and, at from co- transmit- predetermined method presents a computer-aided for electromagnetic compatibi- (EMCA) in networks. VHF-FM The computer in FORTRAN broadprogram 1900. munication ponents mitter consists of transmit- should be area. of done in area, i.e., a good reception domestic of the system elements In a computer of and allocation normal value) adjacent-channel Models for possible such a way that the service the region where in the ever a definite channels the field of distur- of radio This distribution and casting and analysing network (protected ters do not exceed investiga- allowances A broadcasting value has been written broadcasting networks which is not less than the minimum lity analysis this characteristics and existing a certain In the an effective for forecasting planned signal area. The range of a transmitter The paper an undisturbed of a desired presence method network so of the trans- this the same time, interference channel the level. of be observed. means to signals or processing case of a broadcasting bing regard using electrical for transmission reception that the conwith by ranges service an area over usable of the electromagne- i.e., the ratio Of area to the total area strength Introduction ratio, the service means work. COMPATIBILITY is re- model system of a radiocom- the following can be distinguished: models, antenna com- trans- models, recei- ver models, propagation models, and a geographical environment model. These models describe system components all properties in compatibility Propagation which are of the important analysis. model. The propagation curves based 370-4 on CCIR Recommendation are used for calculations of the field strength. strength to path covers the greatest poss'ible valent Part of a given geographical area. a parameter They relate length with transmitting antenna for various field the equiheight percentages as of - 310 - in various time from 50% to 1% They represent tic regions. strength exceeded and apply model for vegetation Pr = -119.6 allowance and to the limit value and man-made structures surface. paths and However, terrain the territory cluded is enough the transmitter follows analysis in- patibility the considered for emissions as a determined Power nominal distribution spectrum com- showed are beyond range. a frequency-dependent ratio The receiving characterized antenna. (ij protected by part of the system - rural is para- of the field emission: areas, 6Q dBp - urban 70 dBp - great emission: - urban value (monophonic cities; 54 dBp - rural areas, (ii) noise mismatch transmitting and receiving the omnidirectional mismatch power are not taken adopted model value of the the reception absence is correct of interference ble field strength). value 66 dBp cities); field prefer- Tab. 1. The antenna since in is the it does Table in the Transmitting antenna vertical horizontal circular linear Attenuation coefficient Receiving antenna horizontal vertical linear circular for the effective ing conditions sary height equal to e.g., when to know smaller represented 3 made of ,the recei- 10 m. in cars, In some the it receiv- .is neces- field strength and then for an additional is introduced. by antenna can a probabilistic in the case of an nal antenna 18 testing the heigths correction CdB] curves were \ be model omnidirectio- takes the form z usa- 1 Signal impairment due to polarization mismatch which for (minimum antennae of EMC analysis The transmitting is the minimum of the is in into account cases, level at the receiver The protected between presented method areas, input. strength case gain of the receiving ving antenna stereophonic 74 dBp - great the polarization 48 dBp areas, of an omnidi- In Polarization [2]. by two additional recommended or the model rectional CCIR propagation strength which [4] antenna to meters: ._ by in the transmit- introducing protection a directional and the receiver are taken into account According of power[l] . selectivity models. to the needs one can use the model is equal output is described not play any role. The emission value antenna red. The consequences of polarization of the to the internal of the fundamental ter output The Cl1f emission frequency the transmitter It takes into account of the system which that the harmonic treated model. from the character related of the nuisance models by two deterministic CCIR only the fundamental power has been and the receiver. model dBm E = -1.6 dBu The receiving deover to use a common model transmitter This A Ah, irregularities and receiver [3) Antenna in the model. Transmitter susceptibi- field strength on irregularities. of Poland, cor- it takes map of the parameter scribing by lity threshold land-sea mixed terrain to the receiver pro- introduced consideration figure was aSSU- to 4,5 dB which and does not make into responds noise The discussed attenuation the Earth's digital med to be equal ver- to both horizontal field The receiver field at 50% of locations, tical polarization. pagation clima- the [l] : - 311 57~2 - where e is the mean value of the power gain in dB, or by means of a determini- The other method assumes that there is no time and spatial correlation stic model in the form of a measured radiation pattern. between the fields and that the wanted field strength is dominant everywhere. The point here is to determine - for EMC analysis a given percentage of time - The fundamental task of the electromagnetic analysis is to determine the probability p that the wanted signal level exceeds the level of a resultant interfering signal by a strictly defined value over a given percentage of time P= PC(EU - EI) > ACT,)] (2) where EU - wanted field strength in dB EI - resultant interfering field strength in dB protection ratio for Ta ACT,) percent of time in dB. EU is a two dimensional random variable of Gaussian distribution with median E(50,50) and standard deviations UT, DL. From the analysis of the CCIR propagation curves the standard deviation of spatial distribution UL = 6 + 0.69a - - O.O063(Ah/h) and of time distribution (3) YT = 0.5I0.429 [E(1,50) - E(50,50)] + + 0.781[E(10,50) - E(50,50)]) (4) were found. tial probabilities of undisturbed reception in the presence of successive L. . The probabi7 lity of undisturbed reception in the presence of all interfering signals is equal to the product of probabiliinterfering signals ties for individual interfering signals = II Lj LE j The calculation of the omnidirectional antenna the standard deviation of spatial distribution of the field strength is calculated from the formula (5) is the standard deviation caiculateci from formula (31. There are two methods to determine the probability distribution of the random variable EI. The first, which is more laborious and sometimes gives no results, consists in defining the distribution function as the convolution of successive interfering fields. - Lj probabi- input: usu = HA + ESU(50,50) + PPATRU + (7a) + 'EFHU + pPoLu usI = HA + ESI(50,50) + PPATRI + (7b) + 'EFHI + pPoLI where U - signal median at the receiver S input (index U means wanted signal and I - interfering signal) HA - effective lengthof the receiving antenna ES(50,50)- median of field strength 'PATR at the receiving antenna - correction resulting from the radiation pattern of the receiving antenna model of an where 0L (6) lity starts with the determination of the signal medians at the receiver In the case of the probabilistic u;,=&Y the spa- 'EFH - correction due to the change of receiving antenna height pPoL - correction due to polarization mismatch All quantities in eqs. (7a) and (7b) are expressed in dB. The median of the difference of the wanted and interfering signals AUs = Usu - USI = ESU - E is SI + Ap (8) where AP = (PpATRu - PpATRI) + (ppoLu- 'POLI)+ ('EFHU - 'EFHI) ('I The difference in the levels of the wanted and interfering signals exceeded in T percent of time and in L percent -' - 312 - is - according of places distribution - described AUS(L,T) to the normal which by last but not least, (10) = AUS - H(T) - H(L) siderations. H(T) = F(T) _I 0TU (lib) with F(x) being distribution T > ratio depends ference between fering of the protection (11) to eq. eqs. and inter- (7), (8) (13) ble (14) the spatial lity of nondisturbed The following reception parameters in EMC probability reception [53: of places is fulfilled with Compatible the value exceeds strength included) of the wanted the minimum signal usable segment (MODY), set of calculation by the as new transmitters introduced user may become The - components median presence of many (in point. a the sources) The SIP0 version one; it selects transmitters to EMC performs interfering is a simplified the previous also the of analysis observation this he wishes procedure where at a given which well by the of The user determines POINT net- transmitter of only at the observation point. with LVS (16) being of the spatial reception. user of takes into account transmitter Any data file as full interference L* b LVS limit value who from the permanent useful set a particular for analysis. are possible. (15) is the on disc, one is the user analysis and the relation of undisturbed data file stored Ta = EU(5O,5O)>,EC output The first data source permanent Seven procedures EC algO- of data are used in the perform. field steering particular and data input and rithms, procedure is fulfilled realizing kind of analysis range of a transmitter the area (boundaries a supervisory (Fig. 1) where 99%. in systems. data base with network. of undisturbed - percentage (2) system analysis compatibility MASTER, selects describing was segment work Lj. program 1900 for the ana- sound broadcasting and the second probabi- sound broadcasting are of interest Spatial VHF-FM program. the value H(L) one can now determine condition is accepta- =1-T TX computer FORTRAN segments. TWO sources of time interference VHF-FM in procedures - - A(T) TX is the percentage easily and - ESI(50,50) + AP + H(TX) Knowing NEWEMCA described of the lysis of internal (12) one gets during which method, written It contains H(L) 25O arrival in - 316 - degrees. The ITU Radio Regulations make no and geostationary between distinction nongeostationary satellites with regard to the application of PFD limits. The potential interference from geostationary satellites is receivers. terrestrial to continuous However, a nongeostationarysatellite appears to be in a different location, in the orbital sphere, to an observer on Earth. This is due to the relative motion of the Earth with respect to the satellite. Therefore, the potential interference from nongeostationary satellites is intermittent as a function of time in comparison with the potential interference from a geostationary satellite, which is continuous. The recognition of this phenomenon and the fact that the CCIR interference to noise for criteria terrestrial radio relays is time dependent should allow the evaluation of two different sets of PFD limits. Two sets of PFD limits were established by this analysis, one for for one geostationary satellites and Earlier nongeostationary satellites. analyses ([l] and [2]), as summarized in this paper, indicate that ITU limits may be raised without jeopardizing Fixed Service systems' These also operations in the 2 GHz band. indicate that the PFD limits set to protect terrestrial analog systems will protect digital systems. Determination of the PFD limits, using the models discussed, was based on noise criteria set by the CCIR for Fixed Service line-of-sight mode of systems using a Technical characteristics of propagation. terrestrial systems in satellites and prepare input used to services were The parameters for the analysis models. input data were obtained from Government and non-Government sources. The models used in this analysis are the Geostationary Model Ratio Power Noise (NPR), (GM), Nongeostationary Model (NGM), Bit-Error-Rate (BERASK), Amplitude-Shift Keying Bit-Error-Rate Phase-Shift Keying (BERPSK), and Bit-Error-Rate Frequency-Shift Keying (BERFSK). GROSTATIOMARYHDDEL (GM) The GM is a modified version of the algorithm developed by BTL ]3]. The model makes use of trendline characteristics. A microwave communication circuit is trendline that generally defined by a consists of a number of repeater stations (radio-relays). In this model, the azimuth of each trendline was assumed to be random, varying uniformly between 0 and 2 II . In addition, for each radio-relay station in a was trendline, the antenna-pointing angle assumed random, with uniform distribution between f25degrees of the trendline. The modifications to the algorithm took into account the effects of desired-signalfading, a microwave radio-relay system's frequency plan, and desired and undesired transmitter including characteristics, emission modulation type. Of these, the effect of the latter two modifications on the PFD limits was found to be more significant. Despite the highly directive antennas now available (and even used in some of the a certain fraction of trendlines), a trendline station's transmitter power may radiate in directions other than those intended. This undesired radiation is worse when less directive antennas are used in a trendline. Cost considerationsoften make it necessary to use less directive antennas for systems designed to operate in the 2025-2300 MHz frequency range. Depending upon the coupling involved in the reception of the undesired radiation, the interference is called over-reach, adjacent-section, or same-section interference. At every repeater site, transmitter and receiver frequencies are separated by Af ( Af is often larger than 40 MHz) in order to avoid such interferences In a system design, frequency engineering is generally used in conjunction with the selection of an appropriate antenna in order to mitigate harmful results from these interferences. The selection of a frequency plan in the design of a microwave trendline is a result of a trade-off among various factors such as economy, quality of performance, and desired interference levels. Importantly, the impact of the potential interferencefrom satellites to stations in a trendline is dependent on the frequency plan used in the design of the trendline. The use of a single frequency in the design of a multihop trendline is not practical, Highly directive antennas are needed in a trendline that is designed to operate with a two-frequency plan. However, four- and six-frequency plans are common in the 2025-2300 MHz frequency range. For a worst case analysis, the two-frequency plan was considered. Another modification to the GM program was the change in the relationship for the ratio of baseband interference to thermal noise. The original relationship used in the program was more appropriate for satellite signals using a relatively high index of modulation. In the 2 GHz band, satellites generally have a medium index of modulation, and a suitable relationship for the inputoutput interference-to-noise ratio may be written as: iC n C i4 = k(Af,m) n4 (1) where: k(Af,m) = a function dependent on modulation indices of both the undesired and desired signals I4 = interference power level, in 4 kHz band, at the any receiver input n4 = noise power level, in any 4 kHz band, at the receiver input i, = interference in channel nC = noise in the receiver channel the receiver - 317 The original GM algorithm made the assumption The function k may be written as that k=l. t21: '41nBW1 (2) k(Af,m) = Bwn where: i41n = BWl = BWn = levels normalized interfering signal of the bandwidth of = noise power Note that if interference is noiselike, k in For unity. will approach Equation 2 narrowband interfering signals with a low index of modulation, the function k is less Hence, characteristics of the than unity. the radio-relay satellite's signals and spectrum have significant transmitters' effects on the evaluation of the function k. The input parameters for the GM program are: IX NF - Random number generator starter - Type of frequency plan (1,2,3,...) VK - Value of k function N - Number of stations in a trendline NTR - Number of trendlines XL - Latitude of the first station (deg) DXLL - Latitude of the last station (deg) XL1 - Latitude increment (deg) ss - Satellite spacing (aeg) DBRNCO - Allowed noise level in receiver DBL - Feed loss (dB) TS - Receiver noise temperature FGHz - Frequency (GH2) _ The typical antenna pattern for a radiorelay receiver and the satellite PFD limits are two sets of important input data. The antenna gain pattern for a radio-relay is similar to that given by the CCIR [4]. G(0) = C dBi O0 02 80.0 the for the ratio receiver when the interference is noise (npr) = noise power ratio when the noiseiis the interfering signal (npr), The constants FVl, FV2, and FV3 are the input parameters (in dB) and EVA1 and EVA2 are input data (in degrees). A sample program output is shown in The Figure 2 for illustrative purposes. curve was calculated using the PFD limits Note given in the ITU Radio Regulations. that the noise level corresponding to the 90% Since the CCIR allowable point is 550 pW. noise power level shown in Figure 1 is 1000 pw, the PFD for the example in Figure 2 may be raised by 2.6 dB. 100.0 bandwidth of the l.th segment of the spectrum density of the interfering signal noise receiver 58~3 - 60.0 40.0 20.0I - 0.cI I 80 I t I I I 1220 810 460 INTERFERENCE (PW) Figure 2. Cumulative distribution of interferencefrom geostationary satellites to a 40 hop radiodelay trendline. The modified GM is useful for the determination of PFD limits in any shared frequency band. However, care should be exercised in the preparation of input parameters. The input parameters for the GM should be based on allocations in, and the technical characteristics of the equipment using, the frequency band. Tn an earlier analysis [2], it was shown that PFD limits for geostationary satellites can be raised by '10 dB based on the use and technical characteristics of equipments in the 20252300 MHz band, CQMPUTATION OF NOISE-POWRR RATIO (npr) The algorithm used in the computation of NPR is based on the method developed by Pontano, et al. (51. The equation for the determination of npr is: (3) npr = ml 2 fml HP(f) 2 (5) (1 - E) I(a,b) fch2 where: rms modulation index of desired signal fml = maximum baseband frequency of desired signal HP(f) = frequency response of preemphasis network E = ratio of minimum to maximum baseband frequencies f ch = midfrequency of baseband channel under consideration m, wherd A, B, C, el, and 0 are input parameters to the program. T8l e PFD limits from the satellites to Fixed Service radiorelay receivers that have been accepted by the CCIR are in the form of: FVl 0 < 6 1 = lines. x = wire length n = twists Per length unit A = wavelength of the incident where wave cables. coaxial for exists also DMC are currents then surface Field-coupled produced on the outer surface of the Coax. By the transfer imPedance of the coaxial line, a differential-mode voltage is produced. cable-to-cable couplinq. By field Near inductive and/or means, from a capacitive nearby source cable carrying either raw Power or signal emissions, cable-to-cable coupling produces unintentional emissions on a victim cable. The following figure shows the network involving coupling between culprit and victim circuits: Conducted coupling Paths we limit our study to As a first step, the common-impedance coupling. Another COUPling mode (Power supply) exists. Within the framework of our study we thus suppose the supplies ideal. common-impedance coupling exists when two, networks or equip ments circuits, or more, share a commun section of a ground plane due also exists grounding. It to multi-Point current high frequency goes whenever a through the ground plane. In the study the ground plane is considered as an homogeneous plate. !l’wo theories are developed: l finite thickness plate l infinite thickness plate The calculation of the dissipated Power and of the electromagnetic energy accumulated in a Prism (a,b) leads to the following exPressions of the ground plane impedance: *finite thickness plate. 1X dx The equivalent follows: scheme may be presented as a R= (a/b). L = (a/&) 2~6 . 206 *infinite - -. sin(e/s) + sh(e/6) Cch(e/6) - cos(e/6)1 sh(e/6) - sin(e/6) Cch(e/6) - cos(e/6)1 thickness plate. . . --_ dx nr I *#It=2 ,,=Z.rn-I Both, culprit and victim, are considered as distributed circuits. The equations governing the phenomenon may be written as: a2 3 [VI = $& a2 lil = CBl ~2 CA1 s Cvl R=LU where [il where: ll(cl+cl2)+ml2cl2 -llcl2-ml2(c2+cl2) CA1 = 1 L-12c12-m12(cl+c12) m12c12+12(c2+c12)_l = a/sb6 6 is the skeen depth in metal = G, As for common-mode voltage, the any ground-loop rejection ratio is applied to the common-impedance coupling to find the differential-mode voltage that appears acros8 the victim input and constitutes the Potential EMI threat. - v- 330 - sIMuLAT1OW Package characteristics for research and industrial A package applications for computer aided design in the area of BMC has been realized on the basis of the different coupling modes developed in It covers most problems related to the IIV. Two-Box modelr . interconnection. we can consider a simple wire-pair, a twisted wire-pair or a we can also consider shielded wire-pair: shielded and twisted wire-pair or a coaxial line with the shield grounded or not. l boxe grounding. we can choose solid grounding, isolated boxes or inductive grounding with or without cabinet bond. We can also consider optical isolators. As offending signal the three forms developed in 5x11 are used. A real signal f(t) can be entered directly on a computer terminal. The package is especially designed for users who are not particularly expert in EMC. This has required the development of special procedures to facilitate the input of data and the output of results. The modularity of the package enables it to be useful even when the initial hypothesis are inadapted; in this case it is easy to replace or to modify the corresponding routines. This full interactive package, written in Fortran 77, is implemented on a mini-computer (HP 1000) and is thus easily transportable, Computed results As illustration, we present spectral simulations that show the effect of the grounding and the interconnection nature on the ground-loop rejection (figure 1). 0 . -- -figure lA) coupling by the wire-pair B) coupling by the shield isolated t boxes inductive C) coupling by the wire-pair grounding with D) coupling by the shield I cabinet bond E) coupling by the wire-pair with solid grounded boxes We notice that for low and especially for medium frequencies, use of inductive grounding in addition to a cabinet bond is particularly efficient compared to solid grounding (-4 to -25 da), also we notice that coupling by the shield is negligible compared to the coupling by the interconnection, In the time domain, the simulation of a system with two configurations is presented. The offending signal is considered as a current source and we simulate the response of a ground plane with the ground-loop rejection (figures 2-3). I; t(/ksec.) figure 2- solid grounded boxes _. - 331 60~5 - I t(psec.1 figure 3- isolated boxes wo conclusions are obvious: with the ungrounded boxes the peak voltage is 13 times than with leas grounded ones. Elsewhere, oscillations appear around 4.5 MHZ which is the maximal susceptibility frequency of the system. Experiments were made in spectral domain. They were carried out in a Faraday cage. As illustration, we present the calculated and mesured ground-loop rejection for two lengths of wire-pair (figures 4-5). figure 5- cable length = 113.5 meters agreement up to We notice loogood 30ckBz. Beyond this, oscillations, due to the propagation in the wire-pair, are noticed. VII-aNzLoBIoN figure 4- cable length = 40 meters We have presented a methodology to study the susceptibility of an electronic equipment to electromagnetic perturbations. An associated interactive package has been presented which operates in both the spectral and time domains. The different starting hypothesis were always incorporated with the aim to realize a versatile tool adapted to minicomputers and corresponding to an investment easily justified by the quality gain on the equipment to design. - 332 - EM1 may now With numerical simulations, somewhat predictable. as being be regarded The package is not limited to the computer aided design but could also be used for a posteriori analytical study of a system with Simulation its immunity. improve a view to coupling a improved by quality could be station to the calculator so that testing real perturbing signals could be directly analysed. C63 technique for CLYTON R. PAUL - A simple - CH 1936-2/t33/0000estimating crosstalk 0430, 1983, IEEE, p.430. 173 HENRY JASIK, Editor - first ring handbook Bill Book Company, Inc. I81 EUGENE D. KNOWLES Cable effectiveness testing - IEEE Electromagnetic Compatibility, 16, No. 1, Feb. 1974, ~~~16-23. and C.D. TAYLOR HARRISON, J.R. Cl1 C.W. terminated transmission Response of a line excited by a plane wave field for IEEE incidence arbitrary angles of on Electromagn. compatibility, vol. Trans. EMC-15, No. 3, August 1973. 193 M.A. DINALLO, L.0, HOEFT, J.S. HOFSTRA, effectiveness of D. THOMAS - Shielding typical cables from lMBz to 1000 MHz The BDM Corporation, 1601 Randolph Road, N. M. 67106, pp.499-493. SE Albuquerque, C21 L.J. GREENSTEIN, H.G. MBIN - Analysis of cable-coupled interference - IEEE Trans. Interference - vol. on Radio Frequency RFl-5, NO. 1, March 1963. 233 Donald R.J. White, MSEE/PE - A handbook on electromagnetic shielding materials and 1960, second copyright performance edition, sec. 1.3, pp. 1.8-1.14. C41 J. LIFERMANN - Theorie la transformation de Masson 1977. et applications Fourier rapide de - C51 V.M. TCJRESIN - Electromagnetic compatibility guide for design engineers - IEEE on Electromagnetic Compatibility Trans. vol. EKC-9, No. 3, Dec. 1967. Antenna edition, engineeMcGraw- shielding Trans. on vol. EWC- simulation des AZRAK Etude et Cl01 G. subies perturbations electromagnetiques systemes electroniques dans les par These de l'appareillage electrique a 1'Ecole Oocteur Ingenieur presentee Centrale de Lyon (France) le 16 Oct. 1994. Cl11 G. AZRAK, Ph. AURIOL - Calcul de la susceptibilite electromagnetique des systemes electroniques industriels - S.E.E. Journees d'Etudes 9, 10 et 11 Mai 1984, Grenoble, France. - 333 61JS - ELECTROMAGNETIC WAVE SCATTERING OF THIN CYLINDRICAL ANTENNAS LOADED BY NONLINEAR W. Krzysztof Technical University Wroclaw A procedure for treating a nonlinearly loaded thin cylindrical antenna as a scatterer is described and illugtrated. Following the procedure the use of computer to solving nonlinear magnetic-field-type (Hallen-type.1 integral equation for a current distrlbution on the antenna has been shown. The equation IS solved numerically by method of moments and nonlinear problem is treated by applying the Bairistov numerical procedure. Numerical and experimental examples for several loads are presented. Introduction A number of man-made objects and many physical phenomena are inherently nonlinear in nature. In the area of nonlinearitles may electromagnetics, be unforeseen and undesirable In the design of a particular system. For example, in radar applications it may exhibit nonlinear effects which In turn result in new frequency components appearing in the backscattered field. Whereas others nonlinearitiee may be essential in functioning of the system i harmonic radar detecting system). One example of this nonlinearity involves metal-to-metal contacte were nonlinear properties are belived to be caused by discontlnultiss or joints of imperefect contacting junctions and oxide layers or films in the component. Other important example of such man-made nonlinearities are antenna systems containing devices involving semiconductor junctions (diodes, integrated circuits, voltage limiters, etc. I . This nonlinearlties are especially present at high RF levels (transmitters, lighting strike, EMP 1 . The objective of this paper is to extend the formulation of nonlinearly loaded antenna or scatterer. The behaviour of antenna or scatterer when loaded with nonlinear element can be than ed greatly from that observed under 9 inear conditions. In some cases, the nonlinearity causes effects such as the harmonic ( Intermodulation) products. Various methods of calculating IMPEDANCES ik of Wroclaw , POLAND of an antenna with a nonlinear load have been addressed in literature. Alternate integral equation have been develaped by Schuman [ll , Liu and Tesche 121 and Sarkar and Weiner 131 to study influence of a nonlinear loads on the behaviour of a linear Wire Network loading of the wire antenna. antenna or scatterer was reported by an alternate techniLandt [41 . Here, For the present dique is presented. scusion of an antenna with a nonlinear load, which is localized at a single point on the antenna, the magnetic-field integral equation (MFIE)is used for obtainfng a solution. The electromagnetic-field problem is reduced to a network problem by appllcation of the method of moments. Sol ving MFIE numerically by this method, reduces It to a linear system of nonlinear algebraic equations. The study of behaviour of nonlinearly loaded antennas or scatterers can next be described as an electric circuit analysie problem with characteristics of nonlfnearity and periodic steady-state conditions. The formulation is cast for the stead -state response, and solution is oz tained by extrapolation method, based on the harmonic balance method. The nonlinear problem is solved by applying the Balristov algorithm. The numerical approach is outlined, and the frequency-dependent reponses Of several examples are presented. Formulation Throughout this paper, we shall consider the antenna configuration depicted In Fig.1 . Most of numerical results will be for the scattering problem. The wire antenna (cylinder) with radfus a and length 2H is assumed to be perfectly conducting. The axis of the wire is taken to be parallel to the z-axis of Cartesian oo-ordfnate sytem, ae shown in Fig.1. It is placed in fbee space and is excited by electric e field, tangential to the wire. A current i Is induced on the antenna by incident wave and this induced current in turn produces, a acat- 334 - L2k] =_{ t+z;tl K(H,z*l , dz’ Medium Kit,z*~= KiH,zj- = [(z-z) Rin Flg.1: Conducting cylinder with central nonlinear load, illuminated by slnusoidel time-dependent plane wave tered electromagnetic field. If an impedance is added at the center of the the induced current Is modiantenna, fied and likewise the scattered field. The technique of impedance loading has been applied to a linear antenna to modify Its radiation and scattering characteristics before. The nonlinear resistive element contained in the load, which is located at z-0 has a votage-current i v-i) characteristic defined by vLiO,t) - F C iL( 0,t)l ii) FL.3 is, in general, a known zewhere ro-memory nonlinear function and v are the instantenuous voltbge and IL across and current through the load respectively. The formulation element, of solution of the harmonic responses of thin-wire loaded antennas as scatterers is covered in great detail by invokes the usuKrrysztof lk C51 , it al thin-wire approximations to a space-time domain MFIE. Th? advantages of using the MFIE of Hallen-type for solving antenna problems have been discussed in a previous paper C61. We shall use this formulation for the solution of the induced current on the antenna with a nonlinear load. The analysis was limited to a thin-wire antenna in the Interest of simplicity in developing the theory. As such, the treat ment can be applied to the large class of objects modeled by wires. For conf iguratlon illustrated in Fig .l solving the boundary? value problem, the MFIE for the total axial current i(r,t) Is given, in operator form, by LEI it,tll -- FCi (0, t 11 C3(zI - C,eJz)Ep,owt (21 where LCliz,t)l- LIC i(s,t)lI,( z,jt)Klr,z‘) Cd(Z) L2Ciiz,t)l, dz , R -’ 11 R 4 expi-jkoRZ) , , expi-jk RH) + a 2 ,I/9 RH quantity z - C,(z) =Lcos C3 (z) - sin CJzl - cost kozl/cos( H (k,z)/cosi k,i H- , , k,W~I/60 11/30k,, , cos ikoH) k,H) , iiz,t)is the unknown current along the thin straight wire surface, FliiO,tll relating cuIs a known function ill rrent to the total tanget&ial E-field along the wire sIrface, k Is the waand E is the tangential ve number, component of the’incident E-field along the wire surface. For x=0 i iz,tl = i(t) as can be observed in Fi’.l. The first term on the right side o $ $\now contains the unknown quantity . The relation between load voltage VL is in geneand the load current i ral described by a non&near function. Several methods can be used to treat nonlinear elements. A method that has been found to give correct results Is outlined here. Recent works indicates that an appropriate approach Is to model of nonlinear v-i characteristics by a Taylor-type power series expansion, which is intrinlsically polynomial in nature. Both,analytical and experimental studies indicate that such an expansion is a valid representation and, furthemore, the backscattered response is characterized by harmonics and intermodulation products of the input frequencies, as predicted by the polynomial model. Let as assume that Fc.1 is adequatly characterized by the power series l vLit) = 2 A ikit) (3) k=i k L are theconstants represenwhere Ak ting the parameters of the nonlinear load. By putting this nonlinear relation in 12) we observe that the only unknown quantity is the current through the load. Hence, the induced current on the surface of antenna linearly related to the incident eld. Numerical Treatment Is nonE-fi- Nonlinear MFIE (21 may be reduced to a linear system of nonlinear algebraic equations through the method of moment8. A PopovlE’s polynomials C6,61 expansion procedure is employed in space-domain for the current basis functions. The wire structure is approximated by straight segments of finite radius, and point matching is employed at the centers of these segments. At a point 2*0 and time t, the numerical approximation of (21 yields 61~6 335 the following ctlymonotically eguat ion increasing. Numerical Li; 2 P;i) I~%osWtl}. (4) 151 r-l=1 N = 43 (z) &Ak( I +lPncos@t) I k -c.$’ REcosmt L (-1 is the value resulting where from numerical approximat$on of the integral of kernel K( z,z 1, are the Popovid* s polynomial expansioP are unknown constans of current, nts of the Popovi~*s polynomial expansions, and E$ , Cg), C,(z) are the quantity as in(Z) . After the expansion of power series (31, the application of the extrapolation method based on harmonic balance method, for different harmonics f requallows the equation (4) to be encies, rewritten as a set of nonlineer algebraic equations [51 . Such a mathematical developement leads to formulawhich can be interpreted in tions, terms of the equivalent networks as shown in Fig.2 . Fig.2: Equivalent circuit representations of cylindrical antenna with nonlinear load for three harmonic frequencies As seen in Fig.2, nonlinear load have been replaced by series current-cont&led voltage sources A I (01. The nonlinear portion of condbct’o 5 A manifests itself as a dependent vo f tage sources, driving the linearized circuit, different for other harmonic f requencles . The solution of nonlinear algebraic eguations (4) may be obtained by using a standard numerical procedures, such the Newton-Raphson or Bairlstow as metl!rods. It Is sufficient to state that for the configuration under study in this paper the solution exist and is unique if the nonlinear load’s v-l. characteristic function F1.l la stri- Results Following the procedure outlined above the method under study is now applied to the analysis of center-lodipole antenna irradiated aded , thin by a 330 MHz E plane wave parallel to the dipole. Specifically, the analysed system consists of a center-loaded dipole of length 0.57 m and raload con.002 m . The nonlinear dius sist of single diode or two parallel Schottky-barrier HP 2800 diocircuit des. To continue the analysis of a diode loaded system, it is necessary to have a reasonably accurate v-l relationship for these type of load.The data needed to specify the v-i characteristics of the diodes were measured and then fit in to the analytic Ak expansions ( 3 1 . The COeff ICientS of series expansions of the single diode and two parallel diodes are chosen as load for circuit 1-11;:’ ;:;Q@rn 2tw .. (11 flz fq3 R4 fl? .::y$,~ ..JIII,t7 :+# :$I1 :s $r~~‘~,~..~t_f.fl :$X:23 (rljf~..‘l.!.~ :!#:f:g‘I. +:rf~‘~).+.A:.$:#4) 4pYdll.,l.fl :#t$!~) 3.277x Ql ‘- ,/G975E 9@ .“, 82 1SF-W3 =, -’ I 2@64E-l$4 2 Y61 ?ME-88 2888 4. %I41.9E 68 5 I/3!?22E-85 -I 8264E-04 - 1ZVE!E,-@8 5.%E!lE-I.0 n respectively. The scattered field from the nonllnearly loaded antenna structure contains many components at frequencies, such as 330 MHz (f,) , 660 MHz (2f I MHz (3f I, etc.. To obtain th& i9.z:: strengths at different harmonic frequencies the current distributions &nduted on the antenna surface are determined at frequencies of interest .6ince the voltages across the nonlinear loads are known at different harmonic frequencies the current distribution on antenna structure Is obtained by premultiplying the voltage vector eveluated at a certain frequency by the admitance matrix of the antenna structure evaluated at the same frequency. Once the current distribution is known the problem is reduced to a conventional antenne analysis problem. The antenna is then excited by these currents at different frequencies to give the scattered field. The harmonic backscattering cross sections G be used to quantitavely descrkbe !tx scattering characteristics of nonlinearly loaded antenna structure as a scatterer. It is clear that the knowledge of higher order 6, could provide addStiona1 signifficant information useful in classifying and identifying the scattertng antenna. Also, the knowledge of should allove to predict the powk r spectrum of the received signal once the power spectrum of the trensmitted signal is specified. The &k Is def lned, as - 2 Ep 1 = 10 log 1 E;i)& ‘k where the E ( k1 re-t-ad E-field are [d6 m21 amplitudes the reradiated st 336 k-t-h of shows the computed and meas function of the broadE-f iald at frequency side inc k dent f,330 MHz. -80 1 v -20 -30 EP ;‘0[d13”/rn I0 k* kre$l experiment 20-H ’ $ 000 xxx ‘2 6 3 ___ ber (five) of terms in the series exThe discrepancies can be pansion (31 further red&ed if a more elaborate model of the nonlinearly loaded antenna will be used. fre- ~%~~~%‘F3 -40 - theory Conclusions In general, the analysis of scattering by nonlinearly loaded wire antenna is an extremly difficult pro kkreces demonstrated in this paper, the special case of ainusoida1 excitation, for a nonlinearly loaded antenna can be described as an electric circuit analysis problem involving it*s nonlinearities and respoas well as, preassumed steady nses, - state conditions. In such applications of the procedure can be adopted that do not require a detailed knowledge of the transient phenomena. In this paper the use of the extrapolation method, based on numerical approximation of nonlinear MFIE by means of method of moments and the harmonic balance method has been investigated. In any har-c, particular, the responses at manic frequencies are obtained by solving of the linearized equivalent network in which the nonlinearities sppear as known excitations. To validate the analysis the redictsd responses of the system fol e owing known incidens have been compared with those measured by the author. The two sets of results agree favorably . Acknowledgments The author wish to express his thanks to Prof. 0.3. Bern of Wroclaw Technical University forhis discussion and valuable comments. References (11 -30 40 -io . 0 EP ) [dBV/ml backscattering cross Fig .3 t Harmonic sections of cylinder (2H no.67 a- .002 m ) with central m, nonlinear load : A/ by single Schottky-barrier HP 2800 diode B/ by two parallel circuit similar Schottky-barrier HP 2800 diodes, a5 function of 330 MHz E-field strength of broadside incident wave -40 The compairison shows the relative agreement between the measured and theoretically predicted harmonic beckecsttering cross sections. The exsting discrepancies may be attributed to the simplicity of the assumed model, especially to nonzero junction capacitance of the diodes and other parasite components. Moreover, the nonlinear v-i characteristics of the loads were approximated by a relatively small num- 121 [31 141 [51 Schuman, H.: Time-Domain Scattering from Nonlinearly Loaded Wire. IEEE Trans. Antennas Propagat . , vol .AP22 611-613, July (1974) Liu, T.K., Tesche, F .M. r Analysis of Antennaa and Scatterers with Nonlinear Loads. IEEE Trans. Antennas Propagat . , vol.AP-24, No.2, 131 -139, March (1976) Sarkar, T.K., Weiner, D.D.r Scat tering Analysis of Nonlinearly Loaded Antennas. IEEE Trans. Antennas Propagat., vol. AP-24, No.2, 125131, March (1976) Landt, D.A.: Nstwork Loading of Thin-Wire Antennas and Scatterers in the Time Domain. Radio Science, ~01.16, No.6, 1241-1247 , ( 19811 Krtysrtofik, W.: Electroms netlc Wave Scattering of Thin Cy 9 indrical Antennas with Nonlinear Loaded ImPh.D. Dissertation, Wropedances. claw Technical University, (1983) [61 Bern, D.J., Wa1kowiak.M.: Polynomial Approximation of Current Dlstributlon . . . . Archives of Elect romagnatiques, t .XxX, 463-476, (1981) - DETERMINING EMI 337 62 - Kl IN MICROELECTRONICS - A REVIEW OF THE PAST DECADE James J. Whalen Department of Electrical and Computer Engineering State University of New York at Buffalo Amherst, New York Summary During the past decade there has been a considerable effort to determine the effects of EMI upon microelectronic circuits. Some efforts have been essentially experimental. The microelectronic circuits investigated experimentally have included analog small-scale integrated (SSI) circuits such as broadband amplifiers and operational amplifiers, digital ss1 circuits such as NAND gates and line driver/receivers, digital medium-scale integrated (MSI) circuits such as lk memory devices, and some very preliminar work on very large scale integrated (VLSIr circuits such as microprocessors. Other efforts have focused on developing models for computeraided analysis and prediction. The Nonlinear Circuit Analysis Program NCAP has been shown to be useful for predicting low-level RF1 effects in both bipolar and FET analog SSI circuits. The computer program SPICE has been used to predict RFI-induced upset in both bipolar and FET digital SSI circuits but not in MS1 nor LSl nor VLSI circuits. Prediction of EMI effects in MSI, LSI, and VLSI circuits which have thousands of transistors must await the development of macromodels for this purpose. Such a development is occurring, but at a slow pace. More recently the emphasis of several investigations has shifted toward being able to determine the statistical variations of EMI in microelectronic circuits and toward developing procedures for coping with these variations. The review will attempt to place in perspective what success has been achieved during the last decade and what problems remain for future investigators to solve. 1. 14260, U.S.A. bility criteria for analog and digital Circuits which are valid today [2]. UnfOrtUnately, there is not enough space to review the earlier papers on EM1 in microelectronics, and this paper will concentrate on the work done during the last decade. Reviews on the status of determining EMI in microelectronics were given in 1981 at Zurich [3] and at Boulder [4] and were updated in 1983 at Zurich [B]. The purpose of this paper is to consolidate and to extend the previous reviews, to describe on-going investigations, and to indicate the additional efforts needed in the future. The emphasis continues to be on how microelectronic circuits respond to conducted EMI signals. It is convenient to organize this paper into the following sections: the basic approach used to determine EM1 in electronic systems containing microelectronic Circuits; predicting EM1 caused by UHF and microwave signals; predicting EM1 caused by MF, HF, and VHF signals; statistical investigations and probabilistic approaches; future needs and problems. 2. The basic approach used to predict RF1 in electronic systems containing microelectronic circuits is to partition the problem into four parts [6]: (1) (2) Introduction During the most recent decade the author has been extensively involved in several major efforts to determine EM1 in microelectronics. He has also observed closely other similar investigations. However, efforts to determine EM1 in microelectronics started long before this author's involvement. Two decades Schulz and Clapsaddle wrote a set of pa ers on the EMC aspects of microelectronics E11. In 1968 Cowles and Showers wrote a paper on "A General Model for Integrated Circuit Susceptibility Prediction" in which they proposed suscepti- Basic Approach (3) (4) to predict the electromagnetic fields inside a system enclosure for a known electromagnetic environment (field-aperture penetration); to predict the pickup of the internal fields by the cables and wires inside the system enclosure (field-to-wire coupling); to predict the resultant EM1 effects in the microelectronic circuits connected to the cables and wires; to predict the behavior of the electronic system which results from the EM1 effects induced in the microelectronic circuits in the system. All four parts of -the - EM1 prediction problem were reviewed in J-61. The reader interested in items (1) and (2) should examine the references given in 161. In this section Item (3) - 338 - on predicting conducted EM1 effects in microelectronic circuits will be emphasized. The choice of techniques used to predict conducted EM1 effects in microelectronic circuits is strongly influenced by the frequency of the interfering signals [7]. For EM1 frequency less than 100 MHZ, electronic circuit analysis programs can be used for EM1 analysis [8-IO]. These computer programs contain models which have been developed to predict semiconductor device performance at frequencies up to the device cutoff frequency. In other words the models are appropriate for the frequency range for which the semiconductor device is used to perform its intended circuit functions. At EM1 frequencies greater than the semiconductor device cutoff frequency, additional factors must be considered. Thus it is convenient to partition the prediction problem for conducted EM1 in microelectronic circuits into frequency ranges below and above a nominal semiconductor device cutoff frequency of 300 MHz. In Section 3 efforts to determine EM1 in microelectronic circuits above 300 MHz will be reviewed. 3. Predicting EM1 Caused By UHF and Microwave Signals A major program to determine Integrated Circuit (IC) Electromagnetic Susceptibility (ICES) which was initiated approximately one decade ago has been reported upon in considerable detail [II-133. In its initial phase the effort was focused on determining experimentally the permanent damage and degradation caused by conducted EM1 injected into IC terminals at the following frequencies: 200 MHZ; 900 MHz; 2 GHz; 5 GHz; 9 GHz. In its intermediate phase the effort was focused on determining experimentally at the same set of frequencies interference effects which existed only when the interfering signal was present. In its final phase the effort was focused primarily on developing models which could be used to predict the main effect observed experimentally which was interference caused by rectification of the UHF and microwaves signals injected into the IC. In this section the prediction models developed for EMI rectification effects will be reviewed. Before discussing these relatively new models, it is useful to review two obstacles to making accurate predictions of EM1 in low frequency circuits caused by UHF and microwave signsls: (I) parasitic effects of passive components and (2) semiconductor device behavior above its cutoff frequency. Predictions made with computer programs such as NCAP, SCEPTRE, SPICE, etc., are only as accurate as the circuit and device models used. Accurate predictions require a detailed and accurate description of the entire circuit, including the parasitic effects associated with passive components, such as resistors, capacitors, and inductors, and with the wiring among components 1145. Modeling parasitic effects at radio frequencies greater than 10 MHz is a difficult task, but one which is receiving attention 115-161. Even if success is achieved in modeling the parasitic effects mentioned, there is still another factor to consider. The semiconcutor device models used in electronic circuit analysis programs, such as NCAP, SCEPTRE, SPICE, etc., were developed to account for device performance at frequencies where the device is normally used. At frequencies greater than the semiconductor device cutoff frequency, the validity of the semiconductor device models used in these electronic circuit analysis programs should be questioned [17]. An important contribution during the last decade has been the development of new semiconductor device models for predicting the main EM1 effect caused by UHF signals in low frequency ICs [13]. These models account for rectification effects in transistors at radio frequencies greater than the transistor cutoff frequency. One model is a small-signal model for bipolar junction transistors (BJTs) [18]. It is useful at low RF power levels where EM1 effects first manifest themselves. It has been applied to linear circuits (including op amp circuits) at RF frequencies well above 100 MHz. Similar models have been used to account for EM1 rectification effects in single-stage BJT circuits [I91 and in field-effect-transistor (FET) circuits [20]. Another model for the BJT, called the modified Ebers-Moll model, is useful at both low and high RF power levels [21]. It has been applied to bipolar linear and digital integrated circuits [22-231. A characteristic of these models is that they are low-frequency models; the RF-induced rectifications effects are accounted for by adding low-frequency generators whose excitation levels are controlled by the RF power absorbed by the semiconductor device. The important assumption is usually made that abosrbed RF power equals the RF power incicent upon the device. This assumption is viewed as a worst case assumption, which causes EM1 rectification effects in bipolar semiconductor devices to be overpredicted. The main practical advantage of this assumption is that it eliminates the need for the development of an accurate model at radio frequencies for the electronic circuit which accounts for the parasitic effects mentioned previously. The new models can be used with existing electronic circuit analysis programs, such as SCEPTRE and SPICE, and the EM1 analysis can be carried out at dc for CW EM1 [22-231 or at the AM-modulation frequency for AM-modulated RFI [24]. In essence, the new models shift a major part of the EM1 analysis from the RF region to the AMfrequency region. Clearly, this procedure cannot answer all questions about how semiconductor devices respond to UHF/microwave signals. However, it does provide useful information about how such signals are rectified to cause undesired low-frequency responses in bipolar discrete transistors and ICS. 4. Predicting EM1 Caused By MF, HF, and VHF Signals in Analog Microelectronics For EM1 signals with frequencies below 300 MHz, it is possible to use general-purpose electronic circuit analysis programs, such as SCEPTRE and SPICE, to predict EM1 - effects in analog integrated circuits F-91. BothSCEPTRE and SPICE COntain large-Signal semiconductor device models and time-domain (transient) analysis routines which can be used to analyze analog ICS. However, there are two numerical difficulties which may occur singly or jointly depending upon the microelectronic circuit being analyzed and the EM1 signal conducted into it. USUallY, the dc bias voltages are in the 1 to 20 V range. If the desired and the EMI-derived signal levels are in the pV range, the node voltages must be calculated to UV precision. This is usually not done - e.g., SPICE2 node voltages are usually calculated to a 50 uV precision. To increase the precision from a 50 pV limit to a 1 pV limit would undoubtably caused a considerable increase in computation time. Another numerical problem occurs when the EM1 signal contains both short and long time constants. Calculations must be made at time intervals (time steps) short compared to smallest time constant and must be continued for a time interval long compared to the longest time constant associated with the EM1 signal. For example, consider one of the worst cases for time domain analysis: the AM-modulated RF carrier. The size of the time step must be a fraction of the RF period, and the time interval over which the analysis is carried out must contain several periods of the AM-modulation signal. The minimum number of time steps required exceeds ten times the ratio of the radio frequency to the AM-modulation frequency. Even for radio frequencies less than 100 MHz, the number of time steps required can approach one million. As a result, the cost of a single computer simulation will be high, and the cost of many different simulations may be too high for most EMC analysis. A limited effort was carried out by the author and co-workers to demonstrate the feasibility of using the transient analysis routine of SPICE to predict EM1 in analog microelectronics. The EM1 was injected into the input of a CA3026 broadband cascade amplifier circuit. The circuit used is the same as that described in [27). A 50% AM-modulated signal was used. The AM-modulation frequency was 1 kHz. The RF carrier frequency was only 50 kHz. The signal generator available power output was -35 dBm into a 50 ohm load which resulted in an EM1 signal level of approximately 5 mV at the amplifier input. The EMI response at the amplifier output at the 1 kHz AM-modulation was measured to be 1 mV. The SPICE2 prediction was 1.35 mV. The number of time steps was 1000. If the RF frequency was increased from 50 kHz to 50 MHz, One million time steps would be needed. The cost of the computer simulation would increase greatly. The effort demonstrated that the SPICE (or SCEPTRE) transient analysis routine could be used to predict demodulation effects in linear integrated circuit amplifiers. However, the cost of using these programs needs also to be considered. Time-domain analysis techniques are very expensive to use when steady-state responses of electronic circuits to AM-modulated RF signals are desired. 339 - 62 HI Another computer program exists which is well-suited for calculating the steady-state responses of linear SSI and MS1 microelectronics [lo]. The Nonlinear Circuit Analysis Program NCAP contains nonlinear incremental models and uses frequency-domain analysis techniques to analyze nonlinear interference effects in weakly nonlinear circuits [25]. It should be emphasized that NCAP cannot be used for EMI analysis of digital circuits because the NCAP models are nonlinear incremental models. However, NCAP is especially Wellsuited for predicting very low-level EM1 effects in linear circuits caused by sinusoidal EMI. It has been used successfully to predict how AM-modulated RF signals with radio frequencies up to 100 MHz are demodulated in bipolar ICs to cause undesired low frequency responses [26-301. The ICs analyzed included a CA3026 dual-differential-pair used in a broadband cascade amplifier circuit and a ~A741 operational amplifier (op amp) used in a unity gain voltage-follower amplifier circuit. The predicted and measured EM1 results for radio frequencies in the 0.050-100 MHz range were in good agreement. It is interesting to note the use of macromodels containing just two transistors to predict EM1 effects in bipolar op amps which contain 25 transistors 1301. Although macromodels are not essential for EM1 calculations in smallscale integrated circuits (SSI), macromodels will be essential for EM1 calculations in very large scale integrated circuits (VLSI) which contain more than 10,000 transistors. The author believes the use of macromodels to predict successfully EM1 effects in bipolar OP amps is one of the most encouraging results of the past decade., The program NCAP does not contain a model for Metal-Oxide-Semiconductor Field-Effect Transistors (MOSFETS) and cannot be Used to predict EM1 in electronic circuits containing either MOSFETs or MOS ICs. For this reason a MOSFET model was developed that could be incorporated into NCAP or used with other electronic Circuit analysis programs to make EM1 predictions for electronic circuits containing MOSFETs and MOS ICs [31]. Initially, a CMOS amplifier circuit was investigated to determine how well the model could predict how amplitude-modulated (AM) RF signals are demodulated in MOS circuits to cause undesired 1OW frequency responses. The experimental and calculated values for demodulation EM1 agreed within 3 dB over the RF frequency range .l to 100 MHz [32]. Next, Bi-MOS operational amplifier (op amp) was investigated which had MOSFET input transistors followed by bipolar interior and output stages. Again the experimental and predicted values for demodulation EM1 were in good agreement over the RF frequency range .l to 100 MHz [33]. The MOSFET model developed has also been Used successfully to predict third-order intermodulation products in a CMOS amplifier [31]. It iS anticipated that the MOSFET model developed will be incorporated into NCAP at some future time. - 340 Since the program NCAP contains models for junction field-effect transistors (JFETs), it can also be used to predict demodulation RFI effects in op amps with JFET input transistor [34-351. The NCAP predictions indicated that demodulation EM1 effects are weaker in JFETbipolar op amps than in bipolar op amps. Experiments performed subsequently were in agreement with the predictions [36]. Again macromodels were used to represent the op amps, and the values for the macromodel parameters were determined from information provided in data sheets published by the op amp manufacturers. Predicting demodulation RFI effects in op amp circuits is one thing. Doing something about suppressing demodulation RF1 effects in op amp circuits is something else. Fortunately, a very valuable paper on suppressing these effects has been published recently [37]. By adding small RF1 suppression capacitors in several locations, it is possible to significantly reduce demodulation RF1 effects in bipolar op amp circuits. The same techniques have been used successfully also on JFET-bipolar op amp circuits t-381. 5. Predicting EM1 Caused By MF, HF, and VHF Signals in Digital Microelectronics The computer program SPICE (Simulation Program with Integrated Circuit Emphasis) is well-suited for transient analysis of digital microelectronics circuits [9]. SPICE simulations of EM1 effects in a 7400 TTL NAND gate have been described [39]. Sinusoidal EMT with an amplitude in the range 1 to 20 V and an RF frequency in the range 0.1 to 100 MHz was used. The EM1 signal was injected into different terminals of an input NAND gate in a cascade of two-NAND-gate inverters. Waveforms at the output of the second NAND gate were predicted and compared to waveforms corresponding to no EMI. Parameters were defined to characterize degradation in voltage and current waveforms. The investigation demonstrates that SPICE (or SCEPTRE) large-signal transient analyses can yield useful information on EMI performance criteria for digital SSI microelectronics. A similar set of SPICE simulations of sinusoidal interference in a bipolar Differential Line Receiver have also been performed 14Cl]. The CW interference was injected at the two inputs of the IC as either common-mode or differential-mode voltages. Representative output waveforms were described. EM1 effects such as change of output logic state and propagation delay were described as a function of the sinusoidal interference amplitude and frequency (dc to 500 MHz). RF1 effects in MOSFET ICs have also been simulated using SPICE [41]. Sinusoidal RF1 is injected at the output of MOS output driver stage which is connected to a buffer stage which is loaded by another buffer stage. Values of risetime delay and fall-time delay as a function of RF voltage amplitude and frequency are presented. It was noted that for a large enough RF voltage stuck-at faults occurred. In particular a stuck-at zero logic state was observed. The computer simulation results were used to analyze the RF1 susceptibility of the input/output (I/O) section of a microprocessor. - Many papers have been presented on determining conducted EM1 in Small-Scale Integrated Circuits (SSI) which contain less than 100 active devices. Few papers have been presented on determining EM1 in Large-Scale Integrated Circuits (LSI) which contain over 1000 devices. An exception is a recent paper on measuring the EM1 susceptibility of a lk NMOS memory IC [42]. In that paper the problems associated with measuring the susceptibility of digital ICs were discussed. A method suitable for determining the susceptibility of a lk NMOS memory and its susceptibility levels was presented. Sinusoidal RF signals with frequencies in the range 1 to 500 MHz were injected onto one line. That line was connected to an address pin or the chip enable pin or a data-in pin. An important observation was that stored data were not altered by the application of RF during a Read Cycle. All RF-induced errors occurred when information was being transferred in or out of memory, The errors were caused primarily by the wrong memory cell being read, by the wrong IC being enabled, or by the wrong information being written into memory. Plots of incident RF power or RF volts (peak-to-peak) vs RF frequency are given. It was noted that some of the RF-induced errors (e.g., the wrong cell being read) cannot be detected by error checking codes such as a parity check. The use of repeated operations to detect errors may not be successful if the RF is present during all repeated operations. It is doubtful that an IC as large as a lk NMOS can be simulated in its entirety using a complete model. Even computer simulations of EM1 effects in a single line receiver using a complete model in which every resistor, diode, and transistor is modeled completely, as in 1401, require much computer time and are expensive. When a digital system consisting of line drivers, transmission lines, and line receivers is to be simulated on the computer, a complete model for each integrated circuit is undesirable because computer time and expense may be too great. Both computer time and expense can be reduced by using macromodels. Recently, the EMC design of digital systems containing line drivers, transmission lines, and line receivers usin macromodeling procedures has been described il 431. The paper describes the macromodel topology for the ICs and their interconnections, macromodel parameters and procedures for determining their values, and computer simulation results. The development of macromodels for digital microelectronics that will facilitate the EM1 analysis of large digital ICs is one of our most important needs. Initial efforts have begun for small-scale digital ICs such as TTL NAND gates [44]. However, much more remains to be done before LSI and VLSI microelectronic circuits can be simulated for EMI. 62 - 341 6. Probabilistic Approaches and Statistical Data Bases - A procedure for determining how an electronic system which contains microelectrOniC circuits functions in a specified electromagnetic environment can be called a methodology for EMC in microelectronics. A methodology based upon two well-established reliability procedures known as Fault Tree Analysis and Failure Modes and Effects Criticality Analysis has been reported upon recently [45-471. In a parallel development which is also based upon well-established reliability procedures, a probabilistic approach to EMC modeling and analysis has been developed [48] and refined [49]. Performance criteria, acceptable performance, EM1 performance curves and performance thresholds are concepts related to EM1 susceptibility levels in a probabilistic manner. In addition, the interactions at different levels such as system, subsystem, equipment, and component are also discussed. Because large portions of systems are being replaced by complex ICs such as VLSI microelectronic circuits and because the EM environment and equipment susceptibility are random in nature, a probabilistic approach may enable one to develop a statistical macromodel. In such an approach, detailed circuit models and functions are replaced by statistical models where probability density functions are used to evaluate probabilities and statistical averages associated with responses at various operational levels. Probabilistic approaches cannot be used by EMC engineers until data bases are acquired. Unfortunately, statistical data bases on EM1 in microelectronics are very limited. An effort to acquire statistics on the EM1 susceptibility of 7400 TTL NAND gates has been described recently [50]. A large number of NAND gates were subjected to CW sinusoidal EM1 at VHF frequencies. DC rectification effects and waveform distortion measures were tabulated. 'The author and his co-workers have long been interested in determining demodulation RF1 effects in analog 113. The specific RF1 effects investigated is how amplitude-modulated (AM) RF signals are demodulated in operational amplifiers (op amps) to reduce undesired low frequency responses at t Re AM-modulation frequency [26-301. The undesired demodulated response may then be processed as a desired low frequency signal by the low-frequency components that follow the op amp. Initially, the emphasis was placed on comparing predicted RF1 effects to RF1 measurements made on a few op amps of each type. Now an experimental investigation to determine the statistical variations of RF1 demodulation effects in op amps is being conducted. The op amps being investigated are the 741 bipolar op amp which has conventional npn input transistors, the LMlO bipolar op amp which has less conventional pnp input transistors, the LF355 JFET-bipolar op amp which has JFET input transistors and the CA081 MOS-bipolar op amp which has MOSFET input transistors. Mean values and standard deviations for demodulation RF1 for 30 units of each type have been presented [5l]. One lmPor- Kl tant observation was that in the frequency range 1 to 20 MHz where the demodulation RFI effects are largest, the mean values are 10 to 20 dB lower for FET-bipol ar op amps than for bipolar op amps. New results will be presented at this session [52]. The experimental investigation to measure RF1 demodulation effects in op amps will produce statistical EMI susceptibility data for four op amp types. It is anticipated that the op amp statistical data will also be used to evaluate the usefulness of probabilistic approaches to EMC analysis and prediction. 7. Conclusion This paper has attempted to review investigations to determine EMI in microelectronics conducted during the last decade. Almost all the investigations reviewed were on how one or a cascade of two small-scale integrated circuits (SSI) responded to conducted EM1 signals. In the future it is anticipated that the emphasis will be on developing probabilistic approaches for EM1 in microelectronics and the statistical data bases required to use such approaches. The emphasis will shift from determining EM1 in SSI and MS1 microelectronics to determining EM1 in LSI and VLSI microelectronics. Such efforts have already begun [53]. Acknowledgments I thank the many who have worked extensively with me during the past decade: Dr. Don We iner of Syracuse University; Mr. Carmen Paludi of RADC; Mr. Curt Larson and Mr. Jim Roe of MDAC. Several others have also aided whenev er requested to do so: Mr. Johh Spina and Dr. Gerry Capraro of RADC; Dr. Bob Richardson and Dr. Vince Puglielli of NSWC. I am especially grateful to my students from SUNY/ Buffalo: Dr. Ta-Fang Fang; Dr. Gordon Chen; Dr. Kun-Nau Chen; Dr. Yue-Hong Sutu; and Dr. Joe Tront (now at Virginia Tech.). All should know that they have done the work and that I was allowed to tell the story. References [II R.B. Schulz and R.L. Clapsaddle, "Electrocompatibility Aspects of Microelectronics:' IEEE Trans. Electromagnetic Compatibility: vol. EMC-6, pp. 37-46, January 1964. Also "Environmental Effects of RF1 on Microelectronics," Electrotechnology, vol. 75, pp. 46-50, June 1965. RI W.W. Cowles and R.M.Showers, "A General Model for Inteqrated Circuit Susceotibilitv Prediction," 1968 IEEE Intern'1 Elkctro- " magnetic Compatibility Symposium Record, pp. 280-290, Seattle, Wash., July 1968. c31 J.J. Whalen, "Current Status of Determininu EM1 in Microelectronics," A~Proc. 4th S-m . & Tech. Exhibition on Electromagne;i! Compatibility, pp. 141-145, Zurich, March 10-12, 1981. 342 [41 [51 - I.J. Whalen, "Determining EM1 in Micro?lectronics& Overview,' 1981 IEEE Intern'1 Electromagnetic Compatibility jymposium Record, pp. 75-78, Boulder, t:olorado, Auqust 1981, IEEE Pub. 3lCHl675i8. _ Whalen and C.A. Paludi, "Computer[I41 J.J. Aided Analysis of Electronic Circuits-The Need to Include Parasitic Elements." Int. ji Electron., Vol. 43, pp. 5oi-511, Nov. 1977. I.J. Whalen. "EM1 in Microelectronics-An Jpdate," Proc. 5th Symp. & Tech. Exhibition on Electromagnetic Compatibility," 3~. 455-458, Zurich, Switzerland, March 3-10, 1983. "Modeling [I51 J.A. Woody and C.A. Paludi, Jr., Techniques for Discrete Passive Components to Include Parasitic Effects in EMC Analysis and Design," Rec. IEEE 1980 Internat. Symp. Electrzagnetic Compatibility, pp. 39-45, Baltimore, MD, Oct. 7-9, 1980, IEEE Pub. 80CHI538-8 EMC. I61 V.G. Puglielli, "Current Status of RF1 Iredictions in Electronic Systems Containing Semiconductor Devices," Proc. 3rd Symp. & Tech. Exhibition on Electronagnetic Compatibility_, pp. 275-279, Rotterdam, The Netherlands, May l-3, 1979. c71 J.J. Whalen, "Predicting RF1 Effects in Semiconductor Devices at Frequencies Above 100 MHz," Guest Editorial for Special Issue of RF Interference Effects in Semiconductor Discrete Devices and Integrated Circuits, IEEE Trans. Electromagnetic vol. EMC-21, pp. 281-282, C81 J.C. Bowers and S.R. Sedore, SCEPTRE: A Computer Program of Circuit and Systems Analysis, Englewood Cliffs, NJ: PrenticeHall, 1971. [91 L.W. Nagel and D.O. Pederson, "SPICE: Simulation Program with Integrated Circuit Emphasis," Electronics Research Laboratory Univ. California, Berkeley, CA 94720, Tech. Memo. ERL-M382, Apr. 12, 1973. [lOI "Nonlinear Circuit Analysis Program Documentation," Tech. Rep. RADC-TR-79-245, ~01s. I-III, Rome Air Development Center, Griffiss AFB, NY 13441, Sept. 1979. J.F. Spina, C.A. Paludi, Jr., D.D. Weiner, and J.J. Whalen, Engineering Manual, vol. 1, J. Valente and S. Stratakos, User's Manual, vol. II, J.B. Valente and S. Stratakos, Programmer's Manual, vol. III. till "Integrated Circuit Electromagnetic Susceptibility Handbook," Final Version, Report MDC E1929, 1 August 1978, McDonnell Douglas Astronautics Co., St. Louis, Missouri 63166. This handbook contains a bibliography which includes approximately 20 titles of technical reports published on the MDAC Integrated Circuit Electromagnetic Susceptibility Investigation conducted over the period 1972-1978. Copies of the handbook (NTIS N79-14312) can be obtained from the National Technical Information Service, Springfield, VA, 22161, U.S.A. Woody, "Modeling of Parasitic Effects in Discrete Passive Components, "Technical Report RADC-TR-83-32, RADC, GAFB, NY 13441, February 1983. [I61 J.A. Cl71 C.A. Paludi, Jr. and J.J. Whalen, "The NCAP Nonlinear T Model for Bioolar Junction Transistors at UHF Frequencies," REC. IEEE 1979 Internat. Symp. Electromagnetic Compatibility, pp. 112-117, San Diego, CA, Oct. 9-11, 1979, IEEE Pub. 79CHI383-9 EMC. "Modeling of LowCl81 R.E. Richardson, Jr., Level Rectification RF1 in Bipolar Circuitry," IEEE Trans. Electromagnetic Compatibility, vol. EMC-21, pp. 307-311, Nov. 1979. Also "Small Signal Rectification Effects in Linear Transistor Circuitry," Proc. 3rd. Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 281-285, Rotterdam, May l-3, 1979. [I91 M. Elliott, "The Susceptibility of Analogue Circuits to Radio Frequency Interference-prediction and Measurement," Proc. Conf. on Electromagnetic Compatibility, pp. 103-112, Univ. of Surrey, September 21-23, 1982, IERE Publ. No. 56. c201 M.L. Forcier and R.E. Richardson, Jr., Microwave-Rectification RF1 Response in Field-Effect Transistors," IEEE Trans. Electromagnetic Compatibility, vol. EMC-21, pp. 312-315, November 1979. C2ll C.E. Larson and J.M. Roe, "A Modified Ebers-Moll Transistor Model for RF Interference Analysis," IEEE Trans. Electromagnetic Compatibility, vol. EMC-21, pp. 283-290, Nov. 1979. Also in Proc. 3rd Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 257-262, Rotterdam, May l-3, 1980. Whalen, J.G. Tront, C.E. Larson, and c221 J.J. J.M. Roe, "Computer-Aided Analysis of RF1 Effects in Digital Integrated Circuits," IEEE Trans. Electromagnetic Compatibility, vol. EMC-21, pp. 291-297, Nov. 1979. [l21 Session on Predicting RF1 Effects in InteWhalen, C,E. Larson, and [231 J.G. Tront, J.J. grated Circuits, Proc. 3rd Symp. & Tech. J.M. Roe, "Computer-Aided Analysis of RF1 Exhibition on Electromagnetic Compatibl'lity, Effects in Operational Amplifiers," IEEE pp. 251-285, Rotterdam, May l-3, 1979. Trans. Electromaqnetic &mpatibility,. EMC-21, pp. 297-306, Nov. 1979. Also in Cl31 Special Issue on RF Interference Effects Proc. 3rd Symp. & Tech. Exhibition on in Semiconductor Discrete Devices and ,Electromaqnetic Compatibilu pp. 269Integrated Circuits, IEEE Trans. Electro_I__ 274, Rotterdam, May l-3, 1979. magnetic Compatibility, vol. t'-21, pp. 281-315, Nov. 1979. - 343 Kl Tront, "Some Results from Using the Modified Ebers-Moll Model to Predict EM1 in Active Filters," Proc. 4th Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 147-150, Zurich, March 10-12, 1981. Whalen, "MOSFET Nonc331 K.N. Chen and J.J. Linear Incremental Model for NCAP," 1982 IEEE Intern'1 Electromagnetic Compatlbility Symposium Record, pp. 66-73, Santa Clara, California, Sept. 8-10, 1982, IEEE Pub. 82CH1718-6. 0.0. Weiner and J.F. Spina, 'Sinusoidal Analysis and Modeling of Weakly Nonlinear Circuits with Application to Nonlinear Interference Effects," New York: Van Nostrand Reinhold Co., 1980 (ISBN 0-442- Whalen, c341 K.N. Chen, G.K.C. Chen and J.J. "Using Macromodels to Compare RF1 in Bipolar and FET-Bipolar Operational Amplifiers," Proc. 4th Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 157162, Zurich, Switzerland, March 10-12, 1981. II241 J.G. cm 62 - 26093-8). Fang, "Nonlinear System Analysis in Bipolar Integrated Circuits," Ph.D. Dissertation, State Univ. of New York at Buffalo, Amherst, NY 14226, Feb. 1979 (copies of the dissertation can be obtained from University Microfilms, 300 N. Zeeb Road, Ann Arbor, MI 48106); also published as Tech. Rep. RADC-TR-79-324, RADC, Griffiss AFB, NY 13441, Jan. 1980. [261 T.F. t-271T.F. Fang and J.J. Whalen, "Application of the Nonlinear Circuit Analysis Program NCAP to Predict RF1 Effects in Linear Bipolar Integrated Circuits," Proc. 3rd Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 263-268, Rotterdam, May l-3, 1979. C281T.F. Fang, J.J. Whalen, and G.K.C. Chen, "Using NCAP to Predict RF1 Effects in Operational Amplifiers," 1979 IEEE Int. Ep, pp. 96-103, San Diego, CA, Oct. 9-11, 1979. II291T.F. Fang, 3.5. Whalen, and G.K.C. Chen, "Using NCAP to Predict RF1 Effects in Linear Bipolar Integrated Circuits," IEEE Trans. Electromagnetic Compatibility, vol. EMC-22, pp. 256-262, Nov. 1980. [301 G.K.C. Chen and J.J. Whalen, "Macromodel Predictions for EM1 in Bioolar Ooerational Amplifiers, IEEE Trans. Electromagnetic Compatibility, vol. EMC-22, pp. 262-265, Nov. 1980. Also see Proc. Conf. on Electromagnetic Compatibilitv, Univ. Southampton, 16-18 Sept. 1980, IERE Conf. Proc. No, 47, pp. 363-375. [311 K.N. Chen, "Nonlinear Modeling of Metal- Oxide Semiconductor Field-Effect Transistor with Application to Radio Frequency Interference Analysis," Ph.D. Dissertation, State Univ. of New York at Buffalo, Amherst, NY 14260, February 1982. Available from University Microfilms, 300 N. Zeeb Road, Ann Arbor, Michigan 48106, USA. [321 K.N. Chen and J.J. Whalen, "A Nonlinear Incremental Model for Predictins EM1 in MOS Transistors," Proc. Conf. on Electromagnetic Compatibility, pp. 113-129, Univ. of Surrey 21-23 September 1982, IERE Pub. No. 56. Whalen, "Comparative c351 G.K. Chen and J.J. RF1 Performance of Bipolar Operational Amplifiers," 1981 IEEE Intern'1 Electromagnetic Compatibility Symposium Record, Boulder, Colorado, August 1981, IEEE Pub. 81CHI675-8, pp. 91-95. C361 Y.H. Sutu and 3.5. Whalen, "A Comparison of RF1 in Operational Amplifiers," Proc. 5th Symp. & Tech. Exhibition on Elexmagnetic Compatibility, pp. 477-482, Zurich, Switzerland, March 8-10, 1983. Goedbloed, K. Riemans and A.J. Stienstra, "Increasing the RF1 Immunity of Amplifiers with Negative Feedback," Proc. 5th Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 471476, Zurich, Switzerland, March 8-10, 1983. [371 J.J. [381 Y.H. Sutu, "Demodulation Radio Frequency Interference Effects in Operational Amplifier Circuits," Ph.D. Dissertation, State Univ. of New York at Buffalo, Sept. 1984. Available from University Microfilms, 300 N. Zeeb Road, Ann Arbor, Michigan 48106. Alkalay and D. Weiner, "Computer Simulation of EM1 Effects in a 7400 TTL NAND Gate," Proc. 4th Symp. & TecQxhibition on Electromagnetic Compatibility, pp. 151156, Zurich, Switzerland, March 10-12, 1981. Ii391 J. c401 T. Dave and D.D. Weiner, "Computer Simu- lation of EMI Effects in a Differential Line Receiver," 1981 IEEE Intern'1 Electromagnetic Compatibility Symposium Record, Boulder, Colorado, August 1981, IEEE Pub. 81CH1675-8, pp. 96-98. c411 J.G. Tront and D.W. Royster, "RF1 Effect in MOSFET Integrated Circuits," Proc. 5th Symp. & Tech. Exhibition on Electromagnetic Compatlblllty, pp. 459-464 Zurich, Switzerland, March 8-10, 1683. Roach, "The Susceptibility of a 1K NMOS Memory to Conducted Electromagnetic Interference," 1981 IEEE Intern'1 ElectroImagnetic Compatibility Symposium Record, Boulder. Colorado. Auuust 1981. IEEE Pub. 81CH1675-8, pp. 85-90: [421 J. - 344 [431 S. Caniggia, "EMC Design of Digital Systems Using Macromodeling Procedures for Integrated Circuits and Their Interconnections," Proc. 5th Symp. & Tech. Exhibition on Electromagnetic Compatibility pp. 465410, Zurich, Switzerland, Margh 8-10, 1983. [441 J.C. Bowers and R.S. Vogelsong, "Basic EMC Technology Advancement For C3 Systems, Macromodeling of Digital Circuits," Technical Report TR-82-286, vol. IIB, RADC, GAFB, NY 13441, April 1984. c451 C.A. Paludi, Jr., R. Bossart and J. Shekleton, "A Methodoloav for EMC in Microelectronics," Procy4th Symp. & Tech. Exhibition on Electromagnetic Compatibility, pp. 163-168, Zurich, Switzerland, March 10-12, 1981. [461 R. Bossart, J. Shekleton and B. Lessard, "EMC in Microelectronics-A Methodoloov," 1981 IEEE Intern'1 Electromagnetic Co% patibility Symposium Record, Boulder, Colorado, IEEE Pub. 8lCHl675-8, pp. 79-84. r471 R. Bossart, J. Shekleton, and B. Lessard, "EMC in Microelectronics, "Technical Report RADC-TR-83-30, RADC, GAFB NY 13441, February 1983. C481 A. Ephreth, D.D. Weiner, G. Capraro and C.A. Paludi, Jr., "A Probabilistic Approach to EMC Modeling and Analysis," 1982 IEEE Intern'1 Electromagnetic Compat%?lity Symposium Record, Santa Clara, California, September 8-10, 1982, IEEE Pub. 82CHl718-6, pp. 81-84. - c491 A. Ephreth and D.D. Weiner, "EMC Modeling and Analysis-A Probabilistic Approach," Technical Report RADC-TR-83-102, Rome Air Development Center, Griffiss AFB, New York 13441, U.S.A., April 1983. c501 D.D. Weiner, J. Gormady, G. Capraro, C.A. "Random Susceptibility of an Paludi, Jr., IC 7400 TTL-NAND Gate," Proc. 1982. Government Microcircuit Application Conference, Orlando, Florida, Nov. 2-4 1982, pp. 348-351. Also 1983 Intern'1 Electromagnetic Compatibility Symposium Record, Washington, D.C., August 23-25, 1983. IEEE Publication No. 83CH1838-2, pp. 21-24. Whalen, "Statistics for [511 Y.H. Sutu and J.J. Demodulation RF1 in Operational Amplifiers," 1983 IEEE Intern'1 Elkctromagnetic'Compatibility Symposium Record, pp. 220-225, Washington, D.C., August 23-25, 1983. (IEEE Catalog No. 83CHl838-2) Whalen, "Demodulation [521 Y.H. Sutu and J.J. RF1 in Inverting and Non-Inverting Operational Amplifier Circuits," in this issue of Proc. 6th Symp. & Tech. Exhibition on Electromagnetic Compatibility, Zurich, Switzerland, March 1985. t-531A Dad/NBS Conference on Microelectronic Electromagnetic Susceptibility is scheduled for March 12-13, 1985 at Washington, D.C. (by invitation). The main topics are the characterization and control of upset and damage in large digital microelectronic circuits. - 345 63~2 - COMPARISON OF THE RF1 SUSCEPTIBILITY OF SEVERAL TYPICAL IC PIN DRIVERS/RECEIVERS Joseph G. Tront Department of Electrical Engineering Virginia Polytechnic Institute and State University Blacksburg, VA 24061 USA (703) 961-5067 ABSTRACT VDD P PADDUT4 Several typical input and output stages for a digital IC have been simulated using the electronic circuit analysis program SPICE2. An RF1 signal is injected into the circuit at the point where the output stage is connected to an input stage. The effects of this RF1 are exhibited as a changes in the dc transfer characteristics of the circuit, as well as, a changes in the transient behavior of the circuit. A comparison of the relative RF1 susceptibility of several different types of integrated circuit pin drivers and pin receivers is made. tl A-+ PAD Figure 1. Schematic diagram driver PADOUT4. for the pad INTRODUCTION Digital integrated circuits are being used in a wide variety of applications. Many of these applications leave the integrated circuit exposed to a harsh external environment. In some . applications there exists the possibility for the introduction of radio frequency interference (RFI) into the digital circuitry. The effect of this interference could be detrimental to the digital circuit. Thus, it is necessary to determine how various levels and types of RF1 will effect a digital circuit. Since many digital circuits are fabricated as MOSFET integrated circuits, an investigation has been undertaken to determine the susceptibility of a MOSFET IC to being 'upset" by the effects of RFI. The term "upset" means that the digital circuit will operate uncharacteristically during all or part of an RF1 event, but will return to normal operation after the event is ended. This excludes the condition where any part of the circuit is permanently disabled by the RF1 event. Several researchers have used simulation techniques to investigate the problems caused by RF1 in bipolar analog as well as digital circuits [l-3]. RF1 effects in MOS circuits have also been studied [4,5]. Typical simulation tools used to conduct these studies include the computer programs SPICE [6] and NCAP [7]. These programs perform a simulation of the circuit operation at the electronics level. This paper reports on the continuation of the work performed by the author and reported on at Zurich in 1983 [8]. In that work a single driver/receiver pair was analyzed. Here, a comparison will be made of the RF1 susceptibility of several different types of pin drivers and pin receivers. NMOS IC DESIGNS The majority of the ICs being produced today are fabricated as NMOS custom designs. This study is concerned with the effects of RF1 in NMOS circuits. The term custom design means that the internal circuits of the IC are comprised of a variety of individually designed sizes and shapes of depletion and enhancement mode transistors. However, the circuit stage which is used to drive the pins of an IC is usually somewhat standard. This standardization is motivated by the desire for an IC to be TIL compatible. Thus, standard libraries of pin drivers and pin receivers have been generated and are used by IC designers [9].. The susceptibility of several of these standardized circuits will be analyzed here. Pad drivers are usually large transistors driven from internal logic by a sequence of increasingly larger transistor stages. There is. usually three levels of buffering separating internal logic-performing transistors from an IC pin. A schematic representation for atypical pad driver called PADOUTO is shown in Fig. 1. Pad receivers can be as simple as a wire interconnecting the pad with a logic circuit, or as complex as a configuration of transistors which allows the pin to be multiplexed as either an input or an output. Receivers usually contain an electrostatic discharge device to protect the gates of the MOS transistors which are directly connected to a pin. - 346 - o VDD 1 L PADOUT Figure 2. I PADIN I with Load Diagram of the RFI-injection scheme with the transistors for the circuits PADOUT and PADIN shown. RF1 INJECTION The sources of EM1 and the mechanisms by which it is coupled into an IC are not within the scope of this paper. It is assumed that RF1 is picked up by any of the conductive elements of a circuit. This includes the conductive traces of a pc board on which ICs are mounted, as well as, the conductive materials within the IC. However, for a given EM1 strength, the RF1 voltages and currents induced in leads external to an IC will be stronger than those induced internally because of the physical size of the conductive materials involved. Hence, the injection of RF1 into the circuit is modeled as an effect seen at the pins of the IC. We have chosen to model the RF1 as a simple voltage source (VRF) in series with a capacitor: This source is connected to the wire which joins an output pin of one chip with the input pin of another chip. The effects of injecting RFI'into the power supply pins is not considered here since it is assumed that the power supply pins capacitively bypassed. A appropriately are diagram of this injection scheme is shown in Fig. 2. Figure 2 further provides .a view of the electronics internal to the PADOUT and PADIN , .circuits. The magnitude (VMAG) and the frequency (f) of the source VRF are varied to simulate different sources and different coupling of RFI into the circuit. The effects of the RF1 on the chip are measured at the output of the transistor stage which follows the pin receiver. This type of RF1 injection causes a change in both the dc and the transient behavior of the IC [8]. For high levels of RFI, the circuit exhibits a stuck-at zero behavior. At lower RF1 levels the circuit's transient operation is affected. Specifically, there is a change in the delay characteristics of the circuit. The nominal output fall time delay TDF is the time it takes for an input stimulus to cause the output to go low. Likewise, the output rise time delay TDR is the time it takes from when an input stimulus is applied to when the output rises. The nominal values for TDF and TDR are measured with VRF=O. When RF1 is injected into the circuit, the value of TDF decreases while the value of TDR increases. These changes as a funtion of both VMAG and f were reported on in [8]. This paper reports on further data collected driver/receiver different pin several from configurations. SIMULATION EXPERIMENTS Figure 3. Schematic circuit. diagram of the PADMUX The pad drivers/receivers used in the simulation experiments performed here are taken from a library of IC components [9]. Three pad receivers and two pad drivers are used. (The PADMUX circuit is used as both a driver and as a receiver.) Various combinations of these circuits five different experiment provided Schematics for the PADOUT4, configurations. PADMUX, and PADIN circuits are shown in Figs. l-3. The PADIN circuit, which is used in one of the experiments, is basically a sizing variation of the PADINC circuit. Simulation runs were made for each configuration with the frequency of VRF set to 100 MHZ, 150 MHz, and 200 MHz. These frequencies were from results of earlier chosen the experiments. Ideally, more frequency points would 63~2 - 341 - tDRT tDR 10 9 t PADllUT4- PADIN 10 9 8 7 6 5 4 3 2 1 PADClUT4- PADIN 8 7 6 5 '4 3 2 1 1-1 1 1 I2 I3 4 ; 11; 6 7 8 5 f ‘I 9 1 I > 10VMAc Figure 4. A plot of the change in TDR(nsec.) as a function of VMAG(volts) and f PADOUT4-PADIN the for configuration. be used, and future plans call for these simulation runs to be made. The range of voltages used for VMAG (and correspondingly the value of the coupling capacitor CRF) was designed so that at the largest value of VMAG, the circuit would be driven very near the point where stuck-at behavior is observed. This range of voltages and the value of CRF are the same for each configuration so that a relative comparison of configurations can be made. Eleven different voltage values are used for VMAG. Each configuration was analyzed at three different frequencies. Five configurations were simulated. Thus, 165 simulation runs are reported on here. Each run consumed about $2.00 in ’ computer time. I I I I I I I I I I > 1 1 2 3 4 5 6 7 8 9 10VMAG Figure 6. A plot of the change in TDR(nsec.) as a function of VMAG(volts) and f the PADOUT4-PADIN for configuration. SIMULATION RESULTS The results of the simulation experiments are shown as plots of the change in the value of TDR and TDF as a function of the VMAG. Each plot is a composite and shows the change in TDR and TDF values for three different frequencies of RFI. The changes in TDR and TDF taken together tend to cause negativeigoing pulses to become wider and conversely, positive-going pulses to become narrower. In the limit, this is what produces the stuck-at zero behavior. However, even before the stuck-at situation occurs, a change in the width of a pulse can cause an upset to occur in a digital system. tDF T -c 10 I--T1 PADllUT4- PADIN I I I I I I I I > 1 1 2 3 4 5 6 7 8 9 lOV,,, Figure 5. A plot of the change in TDF(nsec.) as a function of VMAG(volts) and f PADOUT4-PADINB the for configuration. Figure 7. A plot of the change in TDF(nsec.) as a function of VMAG(volts) and f for the PADOUT4-PADIN configuration. Changes in TDR and TDF for the PADOUT connected to PADIN are shown in Figs. 4 and 5. Notable for this configuration are the larger changes in TDF at lower values of VMAG for lower frequency values. However, the curves for each value of frequency meet at high values of VMAG. The results of the PADOUT - PADIN circuit simulations are shown Figs. 6 and 7. It can be seen that the behavior of this configuration is similar to that of the PADOUT - PADIN circuit. Because of this similarity in behavior, the PADIN receiver was not used combination with the other pin drivers to form further configurations. It assumed that it would behave in the same manner as the PADIN receiver. tDR - PADMUX The behavior of the PADOUT configuration is shown in Figs. 8 and 9. This configuration exhibits a slightly lower level of than do the other two. In susceptibility particular, the value of TDF at the high values of VMAG is noticeably lower than is the case for the previous two simulations. The PADMUX - PADMUX simulation results, shown in Figs. 10 and 1'1, indicate that this configuration is also less' PADIN PADOUT the than susceptible slightly more but, it is configuration susceptible than the PADOUT PADMUX configuration in terms of changes in TDF. tDR 1‘ 10 T PADMUX - PADMUX 9 8 7 6 Figure 8. A plot of the change in TDR(nsec.) as a function of VMAG(volts) and f the for PADOUT4-PADMUX configuration. PADOUT - PADMUX Figure 10. A plot of the change in TDR(nsec.) as a function of VMAG(volts) and f for the PADMUX-PADMUX configuration. PADMUX - PADMUX 8 7 6 “MAC Figure 9. A plot of the change in TDF(nsec.) as a function of VMAG(volts) and f for the PADOUT4-PADMUX configuration. Figure. 11. A plot of the change in TDF(nsec.) as a functicm of VMAG(volts) and f for the PADMUX-PADMUX configuration. . - 349 63~2 - TDF are taken together, ti;Di;;;t susceptible PADMUX the configuration is combination. Use of the analysis technique demonstrated IC designer to make allow an will here of drivers/receivers when choices appropriate designing an IC which may be exposed to RF1 in its environment. REFERENCES PADMUX - PADIN 1. J.Tront, J. J. Whalen, Ci E. Larson, J. M. Roe, "Computer-Aided Analysis of RF1 Effects' in Operational Amplifiers", IEEE Trans. on EMC. vol. EMC-21, pp. 297-306, Nov. 1979. change in plot of the A TDR(nsec.) as a function of for VMAG(volts) and f the PADMUX-PADIN configuration. Figure 12. The most susceptible circuit combination found in those tested is the PADMUX - PADING configuration. The simulation results for this configuration are presented in Figs. 12 and 13. Examination of these figures shows that the values of TDR and TDF increase rapidly with increasing VMAG. In fact, most of the data points for these two plots are off the scale used here. 2. J. J. Whalen, J. G. Tront, C. E. Larson, J. M. Roe, "Computer-Aided Analysis of RF1 Effects in Digital Integrated Circuits", IEEE Trans. on EMC, vol., EMC-21, pp. 291-297, Nov. 1979. 3. "Computer Weiner; J. Alkalay, D. D. Simulation of EM1 Effects in a 7400 TTL NAND Gaten, Proc. 4th Svmw. Tech. Exibition on EMC. Zurich, March, 1981. 4. K. N. Chen, G. K. C. Chen, J. J. Whalen, "Using Macromodels to Compare RF1 in Bipolar and FET-Bipolar Operational Amplifiers", Proc. 4th Svmw. Tech. Exhibition on EMC, Zurich, March, 1981. 5. J. N. Roach, "The Susceptibility of a 1K NM& Memory to Conducted Electromagnetic Interferencd', Proc. Conf. on EMC, Southamnton. UK. Sent. 1980. 6. CONCLUSIONS This study shows how several standard pairs of drivers/receivers are more susceptible than others, giving an indication of the relative susceptibility of the configuration pairings. The most susceptible combination is the PADMUX PADIN configuration. When the change in TDR and 10 9 0100 MHz 0150 MHz A200 MHz El Y I I I I, I I I,, I , , , , , , , > 1 2 3 4 5 6 7 8 9 10VMAc Figure 13. w. Nagel, D. 0. Pederson, "SPICE: Emphasis", Electronic Research Laboratory, Univ. Calf., Berkeley, CA, Tech. Memo ERL-M382, April 1973. 7. nNonlinear Circuit Documentation", Tech. vols. I-II, Rome Air Griffiss AFB, NY, 1979. 8. J. G. Tront, D. W. Royster, "RF1 Effects in MOSFET IntegratedCircuits", Proc. 5th Svmn, Tech. Exibition on EMC, Zurich, March, 1983. 9. J. A. Newkirk, R. Mathews, The VLSI Designer'sLibrarv, Addison-Wesley, 1983. PADMUX - PADIN 8 7 6 5 4 3 2 L. Simulation Program with Integrated Circuit A plot of the change in TDF(nsec.) as a function of VMAG(volts) and f for the PADMUX-PADIN configuration. Analysis Program Rep. RADC-TR-79245, Development Center, - 351 64~3 - DEMODULATION RF1 IN INVERTING AND NON-INVERTING OPERATIONAL AMPLIFIER CIRCUITS - Yue-Hong Sutu and James J. Whalen Department of Electrical and Computer Engineering State University of New York at Buffalo Amherst, New York Summary The paper describes an investigation to determine statistical variations for RF1 demodulation responses in operational amplifier (op amp) circuits. Amplitude-Modulated (AM) RF signals were injected into the op amp signal input terminals to produce undesired demodulated responses at the 1 kHz AM-frequency. The RF frequency was varied over the range 0.1 to 400 MHz. Previously 30 to 35 741 op amps were tested in a non-inv$ing circuit with voltage gain Av = +l. same 741 op amps were tested in an inverting amplifier circuit (Av = -10) without and with RF1 suppression capacitors (27 pF). Mean values and standard deviations were determined for the demodulation RF1 which was characterized by a nonlinear transfer function H2. When the RF1 suppression capacitors were included, the experimental mean values for H2 were suppressed from 10 to 35 dB over the RF frequency range 0.1 to 150 MHz except at 0.15 MHz where only 3.5 dB suppression was observed. A method of comparing values for RF1 demodulation responses (H2) for a non-inverting circuit (Av = +l) and an inverting circuit (Av = -10) is described; the mean values for H2 for the two circuits were found to be surprisingly similar with both sets of data having a 10 dB peak value. Experimental and NCAP computer simulation results are compared for a 3-stage op amp circuit without and with RF1 suppression capacitors. Macromodels were used for the 741 op amps in the NCAP simulations. 1. Introduction, Over an extended period of time, a series of investigations has been carried out to determine conducted RF1 effects in analog microelectronic circuits [l-lo]. The conducted RF1 effect investigated is how AmplitudeModulated (AM) RF signals are demodulated in operational amplifiers (op amps) to produce low frequency responses at the AM-modulation frequency. Subsequently, the undesired demodulation response may be processed as a desired low frequency signal by the low frequency components that follow the op amp. 14260, U.S.A. Initial investigations were focused on the.741 bipolar op amp which has conventional npn lnput transistors l-l-41. The next investigation considered also the LMlO bipolar OP amp which has less conventional pnp input transistors [5-61. This was followed by an investigation on the LF355 JFET-Bipolar op amps which have junction field-effect transistor (JFET) input transistors [7]. Next to be investigated were CA081 MOS-Bipolar op amps which have Metal-Oxide-Semiconductor FieldEffect-Transistors (MOSFET in ut transistors [B-9]. In investigations t 11- If91 the objective was to determine how well predicted RF1 effects compared to measured RF1 results made on a small number of op amps of each type. More recently the emphasis of our investigations has shifted to measuring the statistical variations of RF1 demodulation effects in 30 or more op amps of each of the four types investigated previously [lo]. In all the previous investigations the op amp circuit configuration was kept constant [l-lo]. The specific op amp circuit used was the unity voltage gain circuit called a voltage follower which is shown in Fig. 1. The circuit configuration is also called non-inverting because both the intended signal and the RFI signal are injected into the non-inverting input denoted by (t). We are now investigating the inverting op amp configuration shown in Fig. 2. The intended signal voltage gain Av = -R2+Rl and the intended signal input impedance is Rl. Our ultimate goal is to investigate the combinations of Rl and R2 listed in Table 1 for 30 or more of the four op amp types (741, LMlO, LF355 and CA081). We plan to measure demodulation RF1 using the experimental configuration shown in Fig. 3 for RF frequencies in the range of 100 kHz to 400 MHz. The statistical variations of RF1 for each op amp type for the resistor combinations listed in Table 1 will be determined. Statistical parameters such as means and standard deviations will provide quantitative measures on variations in op amp RF1 susceptibility similar to that reported upon in [lo]. NCAP simulations corresponding to the experiments will be performed to determine how well the mean values for the RFI susceptibility - 352 - I I , T 4 L__--___-_-___--_-__A Fig.1: Experimental set-up for demodulation RF1 response measurement of non-inverting unity gain op amp circuit. The 687 ohm resistor is often replaced by a short-circuit in voltage follower circuits of this type. can be predicted. The NCAP simulations*are known to be sensitive to the values of lnternal transistor capacitances and to external parasitic circuit capacitances. This sensitivity is also being studied. Finally, RF1 suppression capacitors, especially in the feedback path, are known to yield positive results [ll]. Experiments and NCAP simulations illustrating the usefulness of RF1 suppression capacitors are also currently being conducted. The total amount of data to be measured is very great, and it will take many man-hours to accumulate. We plan to report upon what we have accomplished at several stages during our investigation. This paper reports our first set of demodulation RF1 data for op amps in an inverting circuit configuration similar to that shown in Fig. 2. An input resistor Rl = 10 kohm and a feedback resistor R2 = 100 kohm were used. Data were obtained both with the RF1 suppression capacitor omitted (C4 = 0) and included (C4 = 27 pF). Information on the 35 741 op amps reported upon is given in Table 2. TABLE 1 COMBINATIONS OF Rl, R2 and C4 FOR INVERTING OP AMP CIRCUIT CONFIGURATION Input Resistor Feedback Resistor: Voltage Gain RF1 Suppression Capacitor Rl (Q) R2 (@:A, C4 (F) 10 k 100 k:-10 0 10 k 100 k:-10 27 P 10 k 1 M:-100 0 10 k 1 M:-100 27 P 100 k 1 M:-10 a 100 k 1 M:-10 27 P This paper is organized in the following manner. The experimental procedures are described in Section ‘2. Experimental results for demodulation RF1 for the 35 741 op amps Fig.2: W;r;ing V op amp circuit with voltage = -RZ/Rl. The capacitor C4, when included, provides RF1 suppression Ill]. msig.(irn. lwKl Attewotor HP6061 *nplifin undn Test Lo*Paw - TmM wtfnrter Filter WOSSiil - - Gft(600 f&RF’“emcr ft”RF f&tF-b fA@ Modulation: !NMb AM at fw fnquUrY yig.3: Block diagram of experimental system for measuring the demodulation RF1 response. tested are given in Section 3. Statistical results (means and standard deviations) are given in Section 4. In Section 5 experimental mean values are compared to mean values predicted using the Nonlinear Circuit Analysis Program NCAP. Section 6 is the conclusion. 2. Experimental Procedures The actual circuit used to measure demodulation RFI effects in inverting op amp circuits is shown in Fig. 4. The circuit is called the 3-stage op amp LED circuit. The circuit was designed to correspond to a circuit in a system in which a transducer generates a signal which must be amplified to turn on a warning light such as an LED (Light Emitting Diode). A complete description of the circuit is given in [12]. The RFI is injected into the inverting input of the first stage. The RF1 signal has an RF frequency fRF in the range 0.1 to 400 MHz and is 50% AM-modulated at an audio frequency = 1 kHz. Because of nonlinearities in fAF the 1st op amp, a demodulated RF1 response at ‘I kHz exists at the output of the 1st op amp stage. The demodulated RF1 signal generated within the 1st stage is amplified by the 2nd stage and by the 3rd stage which drives the LED. The demodulated RF1 signal at 1 kHz when amplified to 1.1 V (rms) can light the LED to produce a visual display of the existence of RFI. - 353 3-Stage Op Amp LED Experiment Fig.4: Three-stage op amp LED experiment. J.JF unless otherwise specified. Resistor values are in Q. In all measurements of demodulation RFI responses, the 741 op amps for the 2nd and 3rd stages of the 3-stage op amp LED circuit shown in Fig. 4 were the same. The 1st stage op amp was changed. Each of the 3.5 units of 741 op amps listed in Table 2 was used in the 1st stage. The input RF signal at a specific RF frequency was adjusted so that the demodulated 1 kHz Audio Frequency (AF) component at output V3 of the 3rd stage reached rms values of vi; = 1.0 V and 0.4 V sequentially. At the same time, the meter readings of the RF generator and RF voltmeter were recorded. This procedure was used to verify the square-law response region described in [4] and [lo]. To understand this, note that the two AF voltages (1.0 V and 0.4 V) at output V3 correspond to a ratio of -8 dB. Therefore, the two corresponding meter readings of the RF voltmeter or RF generator should give a ratio of -4 dB. If otherwise, the data taken do not reflect the characteristics of second-order nonlinearities. If the data correspond to the square-law response region, the secondorder nonlinear transfer function H2 can be used to characterize the demodulation RFI response at 1 kHz. See Cl], 141 or [lo]. AF The 1 kHz AF rms voltage Vo3 at the output is related to the RF rms voltage Vi' from the RF generator b the second-order transfer function H2(fl,-f2Y by the expression mti = m(m!F)21H2(fl,-f2)/ 64~3 - (1) Capacitor values are in where m is the modulation index, fl the RF carrier frequency fRF, and f2 the lower sideband frequency fRF - fAF. In obtaining Eq. (l), it was assumed that the lower and upper sidebands contributed equally. The factor &!arises because amplitudes are used in nonlinear transfer function expressions. From Fig. 4, we note the voltage relationship between the RF RF generator voltage ViF and the voltage V, indicated on the RF voltmeter is given by VRF = 4vFF g Substitute Eq. (2) into Eq. (1) and solve for H2(fl,-f2) with m = 0.5. The result is IH2(f,,-f2)1 = fiV;;/(4V;F)2 AF RF where Vo3 and V, are in rms volts. (3) Equation (3) can be expressed in dB as 2010g101H2(f,,-f2) 1 = 201wlo$~ - 4ologlovllfF- 21.0 (4) The second-order transfer function Hi of the 1st stage alone, i.e., between the output node of the 1st stage and the node connecting Cl and Rl, can be related to the second-order transfer function H2 of the complete circuit. Accounting for the linear gains of the 2nd stage (100) and the 3rd stage (10) and the attenuation factor (0.5) of the input attenua- $54 - 74 1 Invortlng Ampllflor 35 Dovlcoa R, = 10 kfl R. = 100 f,, RF FrcqwnC~ In MHS Fig.6,:Measured values of the second-order transfer function H2(f,,-f2) of the Fig.5: Measured values of the second-order transfer function H2(f,,-f2) of the 3-stage op amp LED circuit vs RF frequency for 35 741 op amps. RF1 suppression capacitor included. 3-stage op amp LED circuit vs RF frequency for 35 741 op amps. RF1 suppression capacitor omitted. for H2(fl,-f2) which are denoted by E2 and the tor, we obtain the results IH;(f,.-f2) 1 = IH2(f,,-f2) 1(2)2/(100*10) (5) = (H2(f,,-f2) 11250 or 2010g,0/H;(fl.-f2) ( = 2010q01H2(f,d2) - 48.0 3. 1 (6) Experimental Results As just discussed, the demodulation RF1 at 1 kHz in the square-law response region can be characterized by the one parameter H2(fl,-f2). Using Eq. (4), experimental values were determined for H2(f,,-f2) of the 3-stage op amp LED circuit for RF frequencies in the range 0.1 to 150 MHz for 35 741 op amps. The experimental values for H2 are plotted in Figs. 5 and 6. For Fig. 5, the resistor and capacitor combination was Rl = 10 kfi,R2 = 100 kQ, and C4 = C5 = 0. The H2(f,,-f2) values at a specific RF frequency vary + 3 to + 11 dB. For Fig. 6, the resistor and capacitor combination was Rl = 10 kQ, R2 = 100 kn, and C4 = C5 = 27 pF. The H2(f,,-f2) values at a specific RF frequency vary + 4 to + 15 dB. Whereas H2 values in Figs. 5 and 6 are for the complete 3-stage op amp LED circuit, the H2 values for the 1st stage alone can be obtained easily by subtracting 48 dB from H2 values for the complete circuit. See Eq. (6). The value of the ordinate in Figs. 5,and 6 is reduced by 48 dB. The variations of H2 values in dB at a specific frequency are idential to those of H2 values. 4. kfl Statistical Result2 Shown in Figs. 7 and 8 are the mean values standard deviation, (I,for 35 units of the 741 op amp tested in the 3-stage op amp LED circuit. One of the two sets of data in each figure corresponds to ,Rl = 10 kfi,R2 = 100 kn, and C4 = C5 = 0. Another set corresponds to Rl = 10 kQ, R2 = 100 kR, and C4 = C5 = 27 pF. The mean values for H2(f,, -f2) indicate clearly the effect of the RF1 suppression capacitor C4 when it is connected in the feedback path of the 1st stage. The suppression of demodulation RF1 effects caused by a 27 pF capacitor results in a reduction in K2 from 3.5 dB to as much as 36.5 dB at a specific RF frequency. Another effect of the RFI suppression capacitor C4 is indicated by the two sets of standard deviation data in Fig. 8. That plot shows that the spreading of H2 values is wider in general by including C4 = 27 pF. An increase as much as 5 dB in standard deviation is observed at certain RF frequencies. In Fig. 9, RF1 demodulation responses in two 741 op amp circuits are shown. One set of data are mean values of measured H2 for the 741 unity gain buffer shown in Fig. 1 which were reported upon previously [13]. Another set of data are mean values of measured H2 for the 741 inverting amplifier with a voltage gain of 10 and no RF1 suppression capacitor (C4 = 0). For comparison purposes, the H2 values for the 741 unity gain buffer from [13] have to be adjusted. We want an Hi for the unity gain buffer that relates the audio-frequency voltage at the op amp output to the RF voltage at the noninverting input terminal. We must account for the voltage division at the output which contributes a factor of (2) and the voltage division at the input which contributes a factor of (2)2. See Fig. 7 in Chapter 4 in rl2]. The relationships between Hi and H2 for the unity gain buffer are given - 355 64~3 - 0 I$’ I 30 2 ^. f 2 . . . $ . 1 t* : : .* . A .* I 0. . . . I .l . lhlty t Buffor aoht 0 fnr*rtln# A~llflW .I (&I=-10, Vduo~ hwr bwa wwted I I 1 2 ,n( gd w H& raw,(* 10 UW.) I I 6 . 1 20 60 100 I 200 600 In WI= RF frequency for two 741 op amp circuits. The H2 values for the unity gain buffer were obtained by adding 18 dB to the H2 values reported preThe H2 values for the inverting amplifier were obtained by subtracting 48 dB from the H2 values shown in Fig. 7. -1.1 741 kbvatkw Am66fku 35 Dwkn lo- R, = 10 kn oC.=Cs=O IPI R, = 100 WI nC.=C,=27pF justed mean values for H2 are surprisingly L, similar in that both sets of data points have a peak value near 10 dB. The comparison shows that the 741 unity gain buffer has an average lower RF1 response for RF frequencies below 8 MHz. Above 8 MHz, the average RFI demodulation response in the 741 inverting amplifier is lower. II - o 6; 4- 2- 5. Fig.8: Standard deviation, a, of the measured H2(f,,-f2) of the 3-stage op amp LED circuit for two feedback capacitor values, C4 = 0 and C4 = 27 pF. by * lttmwtla,t.w In wwlrmt viously in [13 1. Stmdudw b . C.=O)’ Fig.9: Measured mean values of the secondorder transfer function H2(fl, -f2) vs 14- 12- . . f,, RF Fr.4Wnc7 the 3-stage op amp LED circuit for two feedback capacitor values, C4 = 0 and C4 = 27 pF. This plot shows the suppression of the demodulation RF1 effects by the feedback capacitor. . . “* .2 l. . .* ,~fk.tlm ’ l. . I 3 s- Fig.7: Mean values of measured H2(fl,-f2) of . . . . . . I t”:: -/ . . . IH;(fl,-f2)( = 81H2(fls-f2)I (7) Previously, we have described the NCAP procedure for determining the nonlinear transfer functions which characterize many nonlinear effects including demodulation in weakly nonlinear circuits [4]. We have also described the macromodeling procedures for a 741 (bipolar) op amp [51. Now in this section we shall present NCAP simulations of demodulation RF1 effects in the 3-stage op amp circuit used in our experiments. The NCAP predictions are compared with measurements. The NCAP values for the second-order transfer function H2(fl,-f2) will be compared to the mean measured values of H2(fl,-f2) since most NCAP 2010g,01H;(fl>-f2) t = 2010g10~H2(fl~-f2)~ + 18 NCAP Simulations of Demodulation RF1 (8) The change in nodes at which voltages are measured results in an increase in 18 dB in the H2 values. Also H2 values for the 741 inverting amplifier from Fig. 5 have been adjusted using Eq. (6). The result is a decrease of 48 dB in the values plotted in Fig. 5. The adjusted H2 values in Fig. 9, therefore, exclude the linear amplification and attenuation stages in the experimental setups. The ad- parameter values were determined from typical values given by manufacturers. These typical values are believed to correspond to mean values. The incremental circuit of the J-stage op amp LED circuit for NCAP simulations is shown in Fig. 10. All three 741 op amps have been replaced by an equivalent 741 macromodel. The values for the 14 linear macromodel elements and the 16 nonlinear BJT parameters are not significantly different from the values given in 151. The values actually used are - 356 &waemtrl Macromhl Of lb 3-2tap - 0, Amp LED Clrcuit Fw NCAP Slmuhtlon I)* lOOh r----------r -vvv-I---1 Fig.11: NCAP simulation and measured mean values of the second-order transfer function H2(fl, -f2) for the 3-stage op amp LED circuit without and with RF1 suppression capacitors. TABLE 2 -INFORMATION ON 741 OP AMPS TESTED 1 Fig.10: Incremental macromodel of the 3-stage op amp LED circuit for NCAP simulation. given in Tables 3-2 and 3-5 of [12]. Values for the 14 linear macromodel elements were determined from typical values of 741 op amp characteristics using the procedures developed by Boyle et al. L-141. Most of the 16 nonlinear BJT parameters were obtained from Fang's measurements of the 741 input stage [4]. The NCAP input data for the circuit shown in Fig. 10 are listed in Appendix A in [12]. First we verified that the second and third stages of the 3-stage op amp LED circuit indeed did provide linear gain for the demodulationAF component produced by the first stage. Using NCAP we calculated values for the magnitudes of H2(fl,-f2) at nodes 17 and 42 of the circuit shown in Fig. 10. Throughout the RF frequency region of interest, the ratio of the two H2 magnitudes at nodes 17 and 42 is 72.95 2 0.05 dB, which corresponds to a linear gain of about 4440. The experimental value for the linear voltage gain at 1 kHz is $5 + v;; = 4000 (72 dB). Shown in Fig. 11 are NCAP simu1atIon.s of H2 of the 3-stage op amp LED circuit with Rl = 10 kS2and R2 = 100 kQ for two values of the RF1 suppression capacitor C4 (0 and 27 pF). The corresponding measured mean values are Maker - Year Week 5 RCA 1981 27 2 5 Fairchild 1976 36 3 10 Fairchilda 1981 38 4 5 Fairchilda 1981 35 5 5 Fairchildb 1982 49 6 5 National 1982 - Group -- Units _ 1 a Manufactured in Hong Kong b Manufactured in Korea. also shown. Let us first discuss the two sets of experimental results obtained with C4 = C5 = 0 and C4 = C5 = 27 pF. Note that the RFI suppression capacitors reduce H2(fl,-f2) at every RF frequency. The reduction in H2 varies ,froma low of 3 dB at .15 MHz to a high of 36 dB at 10 MHz and exceed; 15 dB at all RF frequencies above 10 MHz. Now let us compare NCAP simulation results to experimental results with the RFI suppression capacitors omitted (C4 = C5 = 0). Below 1.0 MHz the NCAP simulation values exceed the measured values by 10 to 17 dB. From 1.0 to 6.0 MHz the NCAP values and the measured values agree within 1 to 4 dB. Above 6 MHz i6h;oNf;Pd;alues exceed the measured values by Finally let us compare NCAP simulation results and experimental results with the RF1 su pression capacitors included (C4 = C5 q 27 pFP. Below 1.0 MHz the NCAP simulation values exceed the measured values by 23 I - 357 to 42 dB. Note that below 1 MHz NCAP Predicts an increase in H2 when the RFI suPPression capacitors are added and that this is contradictory to what is observed experimentally. Above 1 MHZ NCAP predicts a decrease in H2 when the RF1 suppression capacitors are added; this is in agreement with what is observed experimentally. From 2 MHz to 20 MHz the NCAP values and measured Values agree quite well with differences in the range 0 to 7 dB. Above 20 MHz the NCAP values decrease much more rapidly than the experimental values which exceed the NCAP values by 20 to 40 dB. The NCAP predictions leave something to be desired, and this indicates that additional modeling effort is required. It would be worthwhile to include in the NCAP simulations the appropriate parasitic elements associated with all the passive components, but especially RI, R2 and C4 in the first stage and the capacitance associated with the Printed Circuit Board wiring. 5 Conclusion and Recommendations Based upon the results obtained to date there are three major recommendations for future investigations. Each recommendation will now be presented. (1) Additional Measurements of Demodulation RF1 in Inverting Amplifiers Additional measurements of demodulation RF1 in inverting operational amplifier circuits are needed. The sets of values for the input resistor RI and the feedback resistor R2 listed in Table 1 are suggested. Data should be obtained for four op amp types: the bipolar 741; the bipolar LMIO; the Bi-MOS CA081; the JFET-bipolar LF355. The measurements should be made without and with RFI suppression capacitors in the locations suggested by Goedblood et al. [ll]. (2) Automatic Measurement System The data generated for this paper was obtained with a manual measurement system. The volume of data was large and almost unmanageable. The volume of data that would.be generated by implementing the first recommendation requires an automatic measurement system. The automatic measurement system would need to set the RF frequency and amplitude, turn on the 1 kHz AM-modulation, measure RF and AF voltages, store the data, compute secondorder transfer functions such as H2(fl,-f2), and plot the results for each op amp tested in each circuit configuration. The data stored for all op amps of one type could then be used to calculate mean values, standard deviations, etc. which also would be plotted. (3) Additional NCAP Simulations In Which Parasitic Effects Are Accounted For More Thoroughly It would be desirable to model more completely the external circuit wiring on the printed circuit board to account for small capacitances from wire to wire and wire to 64~3 - ground. It would also be desirable to model the external circuit component parasitics. Of particular importance are parasitic effects in the op amp input resistor RI, the feedback resistor R2, and the RF1 suppression capacitors. References 111 T. F. Fang, "Nonlinear System Analysis in Bipolar Integrated Circuits," Ph.D. Dissertation, State Univ. of New York at Buffalo, Amherst, NY 14226, Feb. 1979. Also published as Technical Report RADCTR-79-324 by Rome Air Development Center, Griffiss Air Force Base, NY 13441, USA, January 1980. I21 T. F. Fang and J. J. Whalen, "Application of the Nonlinear Circuit Analysis Program NCAP to Predict RFI Effects in Linear Bipolar Integrated Circuits," in Proc. 3rd. Symp. Tech. Exhibition on Electromagnetic Compatibility, Rotterdam, May l-3, 1979, pp. 263-268. L31 T. F. Fang, J. J. Whalen and G. K. C. Chen, "Using NCAP to Predict RF1 Effects in Operational Amplifiers," in 1979 IEEE Int. Electromagnetic Compatibility Symp. Rec., San Diego, CA, Oct. 9-11, 1979, pp. 96-103. L.41T. F. Fang, 3. J. Whalen and G. K. C. Chen, "Using NCAP to Predict RF1 Effects in Linear Bipolar Inteurated Circuits." IEEE Trans. Electromagnetic Compatibility, Vol. EMC-22, pp. 256-262, Nov. 1980. 151 G. K. C. Chen and 3. 3. Whalen "Macromodel Predictions for EM1 in Bipolar Operational Amplifiers," IEEE Trans. Electromagnetic Compatibility, Vol EMC-22, pp. 262-265, Nov. 1980. Also see Proc. Conf. on Electromagnetic Compatibim, Univ. Southampton, 16-18 Sept. 1980, IERE Conf. Proc. No. 47, pp. 363-375. L61 K. N. Chen, J. J. Whalen and G. K. C. Chen, "Using Macromodels to Compare RF1 in Bipolar and FET-Bipolar Operational Amplifiers," Proceedings of the 4th Symposium and Technical Exhibition on Electromagnetic Compatibility, pp. 157162, Zurich, Switzerland, March 10-12, 1981. r71 G. K. C. Chen and 3. 3. Whalen, "Comparative RF1 Performance of Bipolar Opera, tional Amplifiers," 1981 IEEE International Electromagnetic Compatibility Symposium Record, pp. 91-95, Boulder, Colorado, August 18-20, 1981. - PI I31 II01 358 K. N. Chen, "Nonlinear Modeling of MetalOxide Semiconductor Field-Effect Transsistor with Application to Radio Frequency Interference Analysis," Ph.D. Dissertation, State Univ. of New York at Buffalo, Amherst, NY 14226, Feb. 1982. (Available from University Microfilms, 300 N. Zeeb Road, Ann Arbor, MI 48106, U.S.A.) K. N. Chen and J. J. Whalen, "MOSFET Nonlinear Incremental Model for NCAP," 1982 IEEE International Electromagnetic Compatibility Symposium Record, pp. 66-73, Santa Clara, CA, September 8-10, 1982. Y. H. Sutu and J. J. Whalen, "A Comparison of RFI in Operational Amplifiers," Proceedings of the 5th Symposium and Technical Exhibition on Electromagnetic Compatibility, pp. 477-482, Zurich, Switzerland, March 8-10, 1983. [llJ J. J. Goedbloed, K. Riemens and A. 3. Stienstra, "Increasing the RF Immunity of Amplifiers with Negative Feedback," Proceedings of the 5th Symposium and - Technical Exhibition on Electromagnetic Compatibility, pp. 471-476, Zurich, Switzerland, March 8-10, 1983. II21 Y. H. Sutu, "Demodulation Radio Frequency Interference Effects in Operational Amplifier Circuits," Ph.D. Dissertation, State Univ. of New York at Buffalo, Amherst, NY 14260, U.S.A., Sept. 1984. Also to be published as Technical Report RADC-TR-82-281, by Rome Air Development Center, Griffiss Air Force Base, NY 13441, U.S.A. Cl32 Y. H. Sutu and J. J. for Demodulation RFI Amplifiers," 1983~It ElectromagneticaE &cord, pp. 220-225, _--August 23-25, 1983. No. 83CH1838-:?). Whalen, "Statistics in Operational International ibility Symposium ..,L,'..,C. . WUXIIIIYLO~, U-L., (---lttt catalog - Cl41 G. R. Boyle, B. M. Cohn, D. 0. Pederson and 3. E. Solomon, "Macromodeling of Integrated Circuit Operational Amolifiers." IEEE-J. Solid-State Circuits, Vol'. SC-g, . pp. 353-363, December 1974. - TRANSIENT 359 FIELD DISTRIBUTION 65 - IN A TRANSMISSION LI LINE SIMULATOR 0. DAFIF, C. BARDET and B. GECKO L.C.O.M. - U.E.R. des Sciences University of Limoges 87060 LIMOGES CEDEX (FRANCE) determines ABSTRACT The purpose theoretically bution is the in an sections wire structures. space-time The is transient by electric the conical plates nuniformity of of problem is determined the field amplitude, components the rise-time the The between area. The the with no- presence nonnegligible variations are of the current structure solving Secondly field is area a space- the deduced conical induced transient everywhere, and in the working area. II - INTEGRAL EQUATION The of radiation is the induced is problem wires [2] APPROACH of transient conducting used these analyzed. transient equation. in the equation techniques. wire thin integral particularly thinin integral and in the working transversal amplitude, treated numerical field by time the electromagnetic transmission is an distri- simulator approximated using solved The flat-plate are domain field simulator. witch on any is to determine electric three of line paper transient E.M.P. composed which of this firstly time involves currents currents waveforms A [I!. which formulated on are from technique computation wires. obtained terms perfectly domain the the in of In free solving the space integral [21 : equation I - INTRODUCTION The purpose is to provide ximates the at sion in the to plate The of the modes. This technique a guided wave sloping of a central or meshes wires to be perfectly and are the ground conducting. this response I) consisting to opposite parallel simulator are Our rigourous s and c s q and the are 9 s J (s-so)* the are 0 wire + a2 ’ observation the unite contour, a is the incident plane (x o y on figure and tangent is the source points, vectors, wire radius c and is O zi field. The parallel- the joined plane two space-time approximated the R domain a (Fig. rectangular because treated presents plates plates this frequency determine triangular conducting have simulator : where to+ of important TEM and higher paper be explo- that between via to from it is very authors of the to in a transmission differencies in terms of two grids the Then (1) appro- a nuclear different wave. that wavefront The field quite simulator field from theoretically domain The is Most generally sides atmosphere. determine problem E.M.P. traveling distance traveling fields. of plane some simulator a plane an an electromagnetic expected line of plate. by is wire large. of the wire image obstacle account the considerations used in function rnethod (the which size effect (cf. of the wave In the of a perfectly I) introduces is of the 5 III), our treatment assumed approach presence following this form conditions : usely twice) to For the application modifies ground. to operator conducting the image modifie the take into numerical method the is Green [3] [4] . the integral takes - p(s,t) 360 - a~ (s,,r) o = +f0 3.& Z.-d*a1 _-J.& mar (So’$)- c R*2 5 (SoJo*) - c* R” &e,t) & - 4% q C aI R z (soJo) 0 + c $* $-(so,to) . $0 0 - using R* t*_t_0 C so* Ea(s,t) of source, and zo* in vector is the sould the be introduced are deduced field related For example the magnetic knowledge electric III - NUMERICAL 1) - reducing the scattered fields of the field is and the effect by ters are analysed. given (3). of step is wires constitued into basis Similar is in the segment in terms bouring segments. representation choosed to (2) and and the solution procedure rical equation means. spatial time current in A and and the thin for A these one for the and the Finally is obtained using a time accuracy how by are equation account the of presence a (2) perfeclty in the computer substantially to take into conducting plane. Morever allowed 3081 by IBM) the being the storage computer (5500 limited line for every user, on time on the middle ted an some is by : T N 2.19/b, excited on the and the by : A 2 0.69101. and field _____ distribution dependance wire Figure of near the 2. The induced current generator shape is presen- of the transient is conserved. Figure the 3 currents for different a shows space pulses along values - The distribution along of the shape the the Ex(P,t) are presented (x observed on as (Fig. = the shown Ey(P,t) for values middle 6. wire, 4) observed m) metallic region, shape the current line conical spherical snapshots of time. transverse typical a I) parame- - exp (-I%)] Current and front has Figure EZ(P,t) P (x = 6 m, the [S]. wave in has plates 5 where components y = 2 m, z = 1 m). - On the given in Figure the sphericity requirement Koctets IV-l. of Secondly of some (Fig. conical : is defined - The - In numerical the requirement numeintegral rates domains, [31, [51 using stepping the sampling reduced form of the well approximated storage a in the presented. simulator at half-height - The the matrix results are voltage The rise-time voltage interpolation a on neigh- polynomial [21. the time in the take temporal to : A = 1.05 V with time. and order Lagrangian give matrix modification V(t) = A rexp(-at) width scheme space values the segments. applied second of expressing equations to set interpolation used been order of of current depends has In division also an [2]. Clearly, solution first of uses be (3) the each [61 is (1) (3), segments. is step -to express (Z), defined, segmentation second the elementary Ns current unknown is by functions The equations typical The CONSIDERATION solve field areas working currents. by a transient To (x o Z IV - RESULTS and electric the the program: of simulator to the voltage. from into symmetryeproperties Some The (3) & } dso matrix. applied transient (+ 0 , Fig. plane is the imagepoint the unite “$- (S0’ t 0 *) - C* & modifications ’ - s *)’ + a2 (2) 0 0 : R* = J(s (so,‘) dr 1 dso as 0 0 - GK * ai at +(S0’0t *) - C&R” where O* ar R”S 0 0 dt g0 line A slight be the electric longitudinal of the can region 6 the wave observed. wave inclinaison is field plane in the Morever practically of the (x conical o lines are z). The transmission in the a Ev component working plane one. appears - 361 65 - -2 -1.5 -I L G’ ” 0-z Simulator Fig.l-Schematic diagram of Simulator.Wires radius:lcm. the Fig.b-Transverse current on the LI 10 wires distribution transverse simulator of line x=6m. 10 wires V(t):z=Sns,A=150nd. P:x=6m,y=2m,z=lm Fig.2-Induced wire segment current near on one the middle generator. Fig.5-The time-dependent components in the E field conical region. (V/m): Simulator 10 wires Fig.G-Electric regionsfor Fig.3-Current along the space middle snapshots wire. field lines t=190ns.Simulator V(t):t=lOns,A=200ns. in 0.0626 c--) the two 6 wires. - 362 - pI \ \ ‘\“ / I I \\\“ 1 I \\\” j 1 \ \ 2m i *lmI x=53m I I 1 \ ' 1 1 I I I Simulator:...20 Fig.?-Transverse in the electric working Simulateur region for Fig.lO-Effect lines transverse t=212ns. the 6 wires.V(t):r=lOns,A=2OOns. . 2-r wires ---22 fields of t=46ns \ electric conical I p 4 wires I added wires field on the lines in region. IhlW~mFl) x=6.3m Simulator: \ x=6m mT$v,m WY ground 0734Vfm - 0 \ t=3*.5ns 'f t 1 -2 I -1 y(m) I 6 1 2 Yll, ,....I....1 -2 . -1.1 ..I... -I -.a cr:constant Fig.8-Wires transverse 6 “‘.“‘.‘I”“*““’ .a 6 1 current number Simulator:(l)with 2 on transverses wires. (6 wires) transverses [(2) without Fig.ll-Transverses wires effects region effect current 1.5 the trically connected to the wires. elec- longitudinal distribution. one,on the transverse current distrib- ution. I tz I(&) Added Simulator:20 x=6m 22 wires==Fig.9-Effect the transverse of added current wires on 6 wires I wires wires+++ Simulator ! I ,! 1 \ ’ \ I- I I I distribution. - with ---- without Fig.lZ-Transverses field homogeneity I transverses l 1 ground ll.61 I",m wires transverses Y wires wires effects on the in the conical region. 363 on the Figure tinuity of the simulator _ The electric field vertical the 6, witch plane Figure observed (x 7. near provides shape = the (France) discon- prOjeCtiOfls m) are in a presented inhomogeneities be Dl DAFIF can dependent on on Figure bution [8] the of increasing effect of the is analysed. the as slight inhomogeneity on the field - The near The more can be on seams is observed But de a (Fig.10) inside electrically ones; in the Figures the the conical 11 and is a field effects working 12. slightly uniformity on the region field remain V - CONCLUSION been in developed a treated wave using application in the of describe EM field is results problem solved directely program distribution in was and techniques Computer has propagation The formulation numerical which approach pulse simulator. domain. transient experimental domain a rigorous time simulator time to guided by the space good gives inside agreement the with [7]. REFERENCES [II “Transient ed EM field” : LB FELSEN. New-York 121 MILLER G.J. for Topics Vol. E.K., time [3] DAFIF POGGIO domain Journ. Verlag. equation analysis filaires and A.J. of thin of Camp. Physics 0. and JECKO des structures ~011. nat. Physics (1976). “An integrodifferential trures” in Applied 10. Springer B. “Diffraction en presence sur la Compatibilitk BURKE technique wire I2-24-48( struc1973) d’IEM par du sol” 2e EM.Trigastel de d’ondes des presence de “Aspects regime cycle structures en Universite 0. en 3eme electromagne- par quelconques DAFIF Ann. pp.215-225(1983). doctorat (1983) impulsion- mitalliques” transitoire Nationales des d’IEM Limoges numiriques transitoire” Microondes. Jour- Toulouse, juin contrat champs a ligne no413 EM c&s de dans transmission” 103 01 un simulaRapport DGA/DRET/CEG (1984). CSI MEIXNER J. “The fields edges” Antennas to be studied. A diffraction la teur quite 9. distribution But B., de [7] “Etude was modification effects the [61 JECKO EM 1982. inter- wire, Figure introduce large. homogeneity the distribution the current and region of middle longitudinal shown transverse modified 8), de formes n”8-83 sol”. &es The on wires to the is region, (France). wires. transverse connected A current shown du number. -These de shown (Fig. obstacles de la diffraction distri- uniform. simulator. transverse important, wire region modification to a 20 wires as d’ondes Vol. 38 n05-6, en regime filaires quite plate’s the is almost iS wires, current spacing added of to metallic constant field 0. tiques distribution number with the electric wire the 8, and tends Within - The current des Telecomm. “Etude wires. transverse par des IV.2. Main parameters effects - The B. “Diffraction nelles on LI juin 1983. [41 JECKO (x = 33 m). vector 53 Some the from 65 - of behavior I.E.E.E. and Propagation, for elecromagnetic Transactions july 1972. on - 365 CALCULATION AND MEASUREMENTS ELECTROMAGNETIC FIELDS 66 - L2 OF TRANSIENT IN EMP SIMULATORS H.-D. Briins, D. Kijnigstein Hochschule Hamburg, der Bundeswehr West Germany Summary The method of moments was applied to calculate transient effects-on two different types of EMP-simulators, which are located above a plane of high conductivity. The simulators are composed of a great number of thin rods. A computer program was written, based on an integral equation, which can be derived from Maxwell's equations. In order to find a numerical solution triangle functions were used to approximate the currents along the wires. A new method has been employed to treat arbitrary thin wire juncfields tions. First electromagnetic were determined in the frequency domain. Then a Fourier transform technique was used to get the desired transient responses in the time domain. 2. Calculation of Transient Fields and Currents 2.1 Basic Theory In order to fully understand the following sections it is necessary to give a short outline of the mathematical background, that is the basis for all computer-calculations. More details can be found at [Il. Starting point is the following boundary condition A lot of measurements were carried out with different kinds of EMP-antenna arrangements. So it was possible to compare measured and calculated results. The electric and magnetic fields were measured by means of Dand'B-dot sensors respectively. A Tektronix waveform-processing system R 7912, combined with a desktop computer, allowed measurements and storage with a system-risetime of 0.7 nanoseconds. It is shown that there is a good agreement between measured and calculated results. We assume that all conductors of the structure are wirelike with a circular cross-section, which is very small compared to the wavelength of the highest frequency applied by the excitation system (thin wire approximation). Each wire of the system is divided into a certain number of segments or subdivisions, as shown in Fig. 1. (I) where St,, field, region. is the impressed only acting -+S electric in the generating E tan is the tangential compo- nent of the field, caused by all the charges and currents of the system under consideration. 1. Introduction EMP-simulators nowadays are widespreadly used to test the shielding effectiveness of electronic equipment. A complete system can be divided into several components. In detail we are able to distinguish an exciting voltage source, an antenna, which, as far as possible, has a matched termination and a measuring system to record the responses of the system under test. To have a check of the measured results or to optimize the antenna termination it sometimes is desirable to calculate the electromagnetic fields, currents and voltages at discrete points below the simulator arrangement and on the simulator rods respectively. Fig. 1: Location of subdivisions m and n in Cartesian coordinates Using equation (I) we are able to derive an expression for the voltage over segment m due to the current in segment n of the length AZ': - Um~=E;*h&*"em lJ = j,, 366 In this equation the same integral types as in (2) can be found. Thus, no further numerical solutions have to be developed. The observation point is the centre of a small segment of length 2*Al pointing in radial direction. The derivative of the magnetic vector potential with respect to z has been replaced by a finite difference approximation. Applying inverse coordinate transformations and summing up all portions we get the H-field of the whole antenna. = I(z')$(ro)dz'~~n*~m*ALm+ OAzl - ,‘,, dI(z') --J, dz' with (r_)dz’ I 6(r) =1/4n*exp(-jkr)/r. In order to simplify the mathematical effort it is assumed that all the currents, whose influence on other portions of the wire structure shall be calculated, are z-directed. Therefore, it is necessary to carry out coordinate transformations to treat the general three-dimensional case. The method proceeds by approximating the current distribution by a linear combination of elementary expansion functions with unknown amplitudes. Each expansion function exists only over one interval of the structure (method of subsections). For example, a contiguous set of pulses covering each wire have can be chosen. Investigations shown that triangular functions deliver more accurate results with fewer functions per wave length. Inserting the corresponding equations, we obtain a short form of equation (2) urn = Zmn - In. Zmn has the dimension - Generally, the investigated structures consist of arbitrarily oriented and partially interconnected wires. In order to treat wire junctions a special technique has been developed [21. The method is based on the continuity equations, which can be written as K C Ik=- jw k", C A{ X(b')de' K=l (7) K For w=O we simply have the Kirchhoff current law. Using equation (7) it can be proved that the expansion functions have to be located in the proximity of a junction as depicted in Figure 2. To demonstrate the basic principle, an arrangement of three wires (K=3) meeting in a common point was adopted. (3) of an impedance and I, is the complex amplitude of the nth triangular basis function. Furthermore, the resulting voltage at segment m is a consequence of all N current elements on the antenna: urn = N c Zmn'In' n=l ' (4) Finally, all mutual couplings are taken into account, giving a set of N equations, which can be written in matrix form: [VI = [zl l [II (5) The known current distribution, easily obtained by means of matrix inversion routines, is the basis of further computations such as electromagnetic fields at fixed observation points or voltages at lumped loads. If we make use of a circular coordinate system, we are able to calculate the cp-component of the magnetic field from a z-directed current element: aAz 1 - $1 J I(z')$(r)dz'l M H/--'-=pO ap AZ’ F+&[ J I(z’N(p+rz,zWz’ - AZ' -J AZ’ I(z’)q(p_,z,z’)dz’l (6) Fig. 2: Junction wires geometry of connected Every current portion, which contacts the junction, is split on the other wires in the shown manner. An advantage of treating the junction problem in this way is that no assumptions on the amplitudes of the involved currents or other additional constraints are needed. Thus the method is ideal for programming purposes. Up to the present all the considerations were made with regard to the frequency domain. In order to get the desired transient responses in the time domain, it is necessary to compute all currents, voltages and fields for a sequence of increasing discrete frequency values G(w) up to a limiting frequency UC. All the complex system responses are multiplied by the corresponding values of the spectrum of the exciting voltage H(w). Subsequently, an inverse Fourier transform is performed by means of the following equations - f(t) 1 UC = y Jo (R(u)cosut F(W) = H(W) - G(O) - X(W).sinot)dw = R(W) +jX(w) 361 (8) (9) 2.2 Application to EMP-simulators As the intention was to calculate transients with rise times tr of only a few nanoseconds frequencies up to 300 MHZ must be considered. On the other hand one has to choose a certain fundamental frequency, so that all muliple reflections during one half period have been damped to zero. The described relationship between fo and tr is the reason for the fact that often several hundreds of system responses have to be determined. Thus some methods to reduce the computational effort were introduced. First an interpolation scheme to obtain the responses in the frequency domain has been employed with great success. Investigations have shown that there was no necessity to compute the values of the fields and currents for all frequency steps, but only for every second or third step. The lacking system responses can be calculated subsequently by spline interpolation taking only a comparatively small amount of computer time. Most EMP-simulators consist of rods or wires, which carry the same current distribution. Therefore, we can utilize the symmetry involved in many arrangements to increase the speed of computations. Furthermore, there is no need to calculate all mutual interactions with the same degree of accuracy. If the distance between two subdivisions increases, comparatively simple and time-saving formulas can be used. It is a well known fact that a certain number of current basis functions per wave length is needed to achieve a stable numerical solution. As a consequence a frequency-dependent automatic, computer-controlled segmentation of all simulator rods was introduced. Usually, in the lower frequency range the number of current amplitudes is small compared with the situation at the highest frequency, which is taken into account. For every step the matrix [Z] has the minimum order. 2.3 Determination of transient signals Measurements were used to verify the validity of the theory outlined above. The transient measurements of nanosecond pulses were made by broadband probes, each of which had an electrically small size. All probes were mounted on a highly conductive ground plane made of copper and were connected to 50Ginstrumentation via low-loss coaxial cables. For physical reasons the probes are only able to measure the time derivative of the fields and are commonly referred to as D- or Bdot sensors. Due to the mentioned fact 66 - L2 it would be a great advantage to compute the sensor output voltages in a direct way. To reach this aim the geometries of the two mentioned sensor types were modelled by rods with small radius. The D-dot sensor only consists of a short vertical cylindrical monopol. A 50R-resistor located at the basis represents the impedance of a standard 5OR-cable, to which the measuring device is attached. The B-dot sensor was formed of 6 rods in the shape of an inner and an outer loop (see Fig. 8). The outer loop is a relatively thick rod with a narrow gap at one upper side. The thick conductors operate as a shield to minimize the influence of the transient electric field. For all parts the thin wire assumption is fulfilled. 3. Measurement setup Fig. 4 shows the configuration of the measurement setup which was used for the practical measurements of the fast transient pulses in the different simulator arrangements. It consists of two main components. First the R 7912 Transient Digitizer, capable of capturing fast transients with a minimum duration of 5 ns with a resolution of 512 by 512 points. Second, a desktop computer hp 9825 for digital signal storage and processing. The equipment was located in a shielded cabin, approx. 10 m from the simulator. The measuring signal from the sensors is transmitted to the 7 A 29 amplifier by 15 m of low-damping coaxial cable. This results in an overall stepresponse time of the system of 730 ps as shown in Fig. 3. A separate triggering cable from the source enables precise time determination between the loo+---+-----+&m------ II 2 3 4 time (ns) 5 Fig. 3: Measured step-response of measuring system including 15 m signal cables (source: 250 ps) 1: picture of oscilloscope trace 2: presentation after signal processing - 368 - signal-cables trigger-cable *----;Source firing Shielded Enclosure ._ .-.-. Plotter Time Base e 7 B 92 7 A 29 hp 7470 1 . L l R 7912 Transient Controller Digitizer z ) Memory Display hp 9825 A 16 Bit I DMA Unit Interface Magnetic Tape (100 Signals Floppy Fig. 4: Configuration of fast transient pulse measurement setup different measuring positions. The Memory Display Unit allows repetitive display of the once captured transient signal and visual display on an oscilloscope screen. Voltage, pressure and firing of the impulse voltage source of the simulator is controlled by a pneumatic control and can thus be handled from the shielded cabin. The once captured transients are transferred to the computer together with the zero-line of the trace and processed to get the middle value of the original trace. The curves can be plotted as shown in Fig. 3 or stored for later evaluation. A floppy disk supplies a variety of programs for presentation, integration, differentiation, comparison, stretching, and other kinds of signal processing. 4. Examples of investigated EMP-simulators A graphical presentation of two investigated arrangements is given in Fig. 5. Both simulators (further on called type I and type II respectively) are pulse excited by means of a pressurized spark gap connec,ted to a charged coaxial cable (20 kV). The total extension of this source is fitted to the requirements, outlined in section 2. The diameter of the simulator rods were 8 mm (type I) and 6 mm (type II). Simulator II consists of rods which branch out at one point 12 cm above the ground plane and slightly rise until they reach a height of 80 cm. Then they parallely descend to the ground plane. At the end, each of the rods carries a resulting resistance of 600 ohms as depicted in detail in ,* I Disk 9885 Fig. 5. The lower resistor may be replaced by a 50 ohm cable in order to measure the current. The other simulator is terminated by only one resistor. The working volume is bounded at the upper side by 5 horizontal rods of 125 cm length. At different positions below the antennas, sensors are located to analyze the transient performance. - _I-,--- I---- Fig. \- 320cmM 5: Dimensions of presented lator arrangements simu- - 369 5. Comparison of measured and calculated results 5.1 Simulator type I Some interesting examples of measured and calculated signals are presented in this chapter. Generally one can say that the bandwidth of the measured signals (500 MHz) is higher than the maximum frequency used for the calculations, which was up to 300 MHz to save computation time. To compute the true risetime of the measuring system (.7 ns) one would have to use frequencies around 1 GHz for Fourier synthesis. On the other hand, the measured signals have the disadvantage to show any noise or crosstalk caused by weak signal coupling to the simulator and not by the field signals. So one has to compare and interpret the curves carefully to find out the true influences. On the other hand, control of measurement by computation and vice versa is a very helpful tool. 66 - L2 4 3 z "0 lo Fig. 20 30 40 50 60 time 70 80 (ns) 7: Measured D-dot signal (1) and integrated electric field strength (2) at position x = 425 cm (simulator type I, source: 20 kV, 2 ns) Fig. 8 proves the possibility of modelling and computing the output signal of a B-dot probe consisting of short cylindrical rods. V Fig. .-. a (2) time (ns) 6: Calculated output of D-dot sensor at position x = 425 cm (1) and principle of sensor modelling (2) (simulator type I, source: 1 V, 3 ns) Fig. 6 as a first example shows the calculated output signal of a simulated D-dot sensor which was modelled in the computer program as a short cylindrical rod terminated by 50 ohms. One can see the typical reflections of a not terminated simulator and the travelling time of the wave to the termination and back to the measuring position. Fig. 7 shows a pretty good agreement of the measured signals with small differences that can be noise or the influence of elements 'not considered in the computation. The D-dot sensors that were used have the shape of small conical monopoles matched to the cable impedance as described by King [31. In the considered frequency region no significant difference to a short cylindrical rod could be observed. The integrated curve presents the electric field strength at the measuring position. The method of numerically integrating the D-dot signal proved to be much better than using hardware integrators, as the D-dot signal level is more insensitive to noise and no integrator discharging time constant has to be considered. r=50R A i _ 150 120 90 time (ns) ‘, 180 Fig. 8: Calculated output of B-dot sensor at position x = 200 cm (I) and principle of sensor modelling (2) (simulator type I, source: 1 V, 3 ns) The measurements with this type of probe and comparison with a commercially available B-dot sensor showed a good agreement between both types and the calculation. 5.2 Simulator type II For this type of simulator some computed signals are selected for presentation which can be derived directly from the computed electric and magnetic field or the currents in the used wire elements. Fig. 9 shows the magnetic and the electric field strength at a position 438 cm from the source of the simulator. It is obvious that the choosen termination resistance of the simulator is nearly equal to the intrinsic impedance of all simulator wires as the first peak of the field strength has nearly the same amplitude as the final value. On the other hand, reflections - 370 - do occur from the termination resistors which are mounted at a distance of 174 cm from the considered point. This results in a length of the first peak of approx. 10 ns. The conclusion is that the single wires which are terminated with equal resistors are not matched individually in this case. This will be shown better in Fig. 11. The initial relation between E- and H-field is the intrinsic impedance of the free field but this changes after the first reflection, when the current in the inner rods rises and the voltage drops respectively. Fig. IO shows the actually measured D-dot and E signals that are in pretty good agreement with the computed values and show the same characteristics. 5 5.. :' 5 o 'i '34 ’ 68 ’ 102 ’ 136 '170 '204, B CI-1 w 0 0 4 ’ ’ 34 ’ 68 Fig. 9: Computed tric (2) sition x type II, ’ ’ 102 ’ ’ ’ ’ ’ ’ 136 170 204 time (ns) magnetic (1) and elecfield strength at po= 438 cm (simulator source: 1 V, 3 ns) In the next figure the currents in the individual termination resistors of the separate rods are shown derived from computation and measurement. For the outer rod the value of 600 ohms is good matching to its individual intrinsic impedance, while the middle and the inner conductor need a higher value for correct matching. From the relation between first value and next sten of amnlitude the correct matching resistor _ can easily be determined. 160 2 0 time (ns) Fig. 11: Computed (A) and measured (B) currents in termination resistors at outer (I), middle (2) and inner (3) conductor (simulator type II, source: 1 V, 4 ns (A)/ 50 V, 250 ps (B)) 0 20 40 60 80 100120 6. Conclusion The method of moments combined with triangular functions of currents and a new method of wire junctions treatment was used for the computational analysis of EMP-simulator arrangements. Two types of field probes were simulated in the computer program and manufactured. Using a fast transient pulse measuring and processing system, the magnetic and electric field strength and their time derivatives and the current in the termination resistors were determined. Control of measurements by computation and vice versa proved to be a helpful tool for the interpretation of the true transient behavior of a system under test. This will ease the design of simulator arrangements, impulse voltage sources, termination resistor values and positions as well as the construction of simple field probes. 7. References 111 Harrington, R.F.: Field Computation by Moment Methods. The Macmillan Company, New York 1968 1 -'O [21 Briins,H.D.: Berechnung der StromII 10 2Ov 30 40 50 6@ 70 80 time (ns) 90 100 Fig. IO: Measured D-dot signal (1) and integrated electric field strength (2) at position x=438 cm (simulator type II) belegung auf dreidimensionalen Stabantennenstrukturen. AEU, vol. 38, No. 4, 219-226, (1984) i-31 King, W.P.: The conical Antenna as a Sensor or Probe. IEEE Trans. Electromagn. Compat., vol. EMC-25, 8-13, (1983) - 371 EMP SIMULATION 67~3 - BY PULSE INJECTION bY Torbjijrn Karlsson, GiSran Und& and Mats Gylemo FOA Box 1165, S-581 11 Linkgping, Abstract In order to meet the demand for EMP ,hardness validation, FOA have within a low budget program developed a number of test methods and test facilities. With relevant analytical support simple pulse injection technique have proved to be an effective instrument for validation tests. In the paper a number of illustrative experiments are presented. Tests on feed-thru capacitors, filters and a power plant have been carried out to demonstrate the applicability of the pulse injection equipment. The presentation of each experiment is concluded by a detailed discussion of the results. Introduction Current injection in order to simulate EMP effects is important as part of the EMP validation process. While being the predominant test method during the development phase or the construction period of a system, pulse injection has not been considered useful for hardening verification tests. However, if it is used in conjunction with free field simulation and well designed supporting analyses, pulse injection techniques promise to emerge as suitable tools for system testing during the production and the deployment phases as well. With the objective of attaining a useful technique for pulse injection into buried structures and other large objects not suited for tests in EMP simulators, FOA have developed a number of test methods. It is important to remember that the assessment of EMP hardness generically always include analysis, even if only to scope the test. We want to show how an integration of simple empirical and analytical approaches emerges as a really useful technique for EMP hardness validation. Test of cable entries The hardening of electronic equipment is preferably accomplished by the method of controlled EM topology. Long lines and cables need penetration Sweden treatment at the boundaries in order to prevent the comparatively large induced EMP current from damaging internal electronic equipment. In order to keep the EMP hardening costs to a minimum level the use of standardized hardening components is desirable. The evaluation of cable penetrations for approval requires a reliable test facility which FOA have accomplished by assembling ordinary standard components. The pulse generators used were simple laboratory models. Pulse injection using a coaxial line In order to evaluate the quality of EMP hardened cable entry vaults, we use a simple injection line consisting of a steel tube in the center of which the injection cable is located. This configuration constitutes a coaxial line with a characteristic impedance in the 100 R range, depending on the diameter of the inner conductor. At one end of the tube, a pulse is injected into the cable while at the other end the tube is circumferentially connected to the shield around the cable entry. Experiments show that this device is useful for testing both hardening design and components used in the installation. Steel cabinet Steel cabinet L Fig.1: Pulse Measurements injection facility. acquired: 0 pulse generator voltage o current on the cable close to the cable entry o residual voltage inside the shielded cabinet "g I "r - To avoid parasitic oscillations the coaxial line is backterminated at the generator. Validation tests of hardening components Simple pulse injection experiments combined with relevant analysis have proved to be an effective tool for validation tests in a low budget program. FOA have carried out a series of experiments which demonstrate the importance of a topological good hardening design. In this chapter, we present two experiments that will illustrate high performance protection using standard components. At the same time the versatility of the pulse injection equipment is demonstrated. Hardening by using coaxial feed-thru capacitors Using ordinary feed-thru capacitors installed in the topological barrier is an excellent method of controlling low frequency cable penetrations, particularly power entries. The important feature is the virtually nonexisting high frequency coupling between the conductors on different sides of the shield, which is impossible to achieve by using pigtails. Boundary--D: 372 - In a real EMP hardening situation a power cable is always connected to three or four capacitors, one on each conductor. Supposing a balanced cable, the induced current from an EMP is almost equally distributed between the different conductors, which means that the capacitors are coupled essentially in parallel. In our case, a total current according to figure 4 was injected into the cable. All conductors were assembled together at the cable entry and connected to one capacitor only. Fig.4: Incident current I. A comparison between our injected current and a current in one conductor induced from an EMP estimated by Vance [ll indicates that our injected pulse is not far from threat level, The residual voltage calculated from the theoretical model given by the equivalent circuit (fig.3) has the shape of a double exponential pulse. As expected, measured values turned out to be in good agreement with those calculated. The results are assembled in the figure below. I Fig.2: Coaxial f eed-thru capacitor. Two different coaxial capacitors conformable to the illustration were tested in the above mentioned pulse injection line. A lumped element model of the whole test set up is demonstrated by the equivalent circuit diagram in figure 3. X . Calculated values Microseconds ---------7 I R.3 Ir---k-T Fig.3: ; - 1 ‘i I I r-!---l Test set up, equivalent circuit. Fig.5: Residual only). voltage V,. (Capacitors The risetimes of the residual voltage are significantly longer than those of the incident pulse. Risetimes measured in the order of microseconds demonstrate that the feed-thru capacitor has the important quality of being impervious to the high frequency components that are characteristic of EMP spectrum. - In order to avoid isolation break down in an open-ended capacitor caused for example by a very long pulse, a diversive element has to be coupled parallel to the capacitor. The effects caused by the diversion of a varistor or a surge arrestor, respectively, are shown as a reduction of the residual voltage in figure 6. Fig.6: Residual voltage value 0.5 vF). 373 67~3 - adapted telephone cable using the pulse injection facility. After analysis of the filter was modified deficiencies, accordingly. V, (Capacitor Eig.7: The late time behaviour reveals a noticeable difference between the performance of the two nonlinear components. When using the varistor, the decay time constant is not significantly changed because of the comparatively high impedance in the varistor. On the other hand, an activated surge arrestor decreases the effective load impedance, thence decreasing the decay time. Since this will result in a reduction of the penetrating energy, the combination of a feed-thru capacitor and a surge arrestor appears as a penetration treatment to be preferred. An elucidative measurement was carried out which increased our understanding of the physics behind the activation of the surge arrestor. The voltage of the capacitor was measured on both sides of the shield. At early times, the voltage on the outside could reach several kilovolts while the voltage at the same time on the inside was still in the 10 volt range. This high voltage transient due to inductance in the bolt caused the surge arrestor to break down, contributing to the diversion of the current, although the voltage on the inside never reached the static break down level of the arrestor. The unique ability pertaining to coaxial feed-thru capacitors of breaking the high frequency coupling was thus demonstrated in an illustrative way. Hardening by using filter modules for ten telephone pairs The same protection princinle is applicable to all kinds-of low frequency penetrations including telephone cables. In the hardening design, though, a practical approach implies the integration of protection devices in a filter module. A new design of such a module (fig. 7) was validated by injecting a surge current into an Filter module pairs. for ten telephone The EMP protective device is of a hybrid type construction consisting of a combination of rare gas tubes and filters. A design objective was to achieve a convenient method of installation with several modules stacked together. The module was the object of two, approximately double exponential current pulses with different risetimes, 5 ns and 25 ns respectively. In both cases, the peak value was 1 kA and the half time pulse length was 0.5 vs., The residual voltage is shown in figure 8. A remarkable result is the decrease in amplitude for shorter risetimes of the incident pulse, the reason being the decrease of energy coupled into the filter before the striking of the gas tubes. Fig.8: Residual voltage V,. Oscillation frequency is a filter characteristic and, consequently, independent of the pulse risetime. An iterative procedure of modification alternating with validation tests has been carried out and further modifications will follow. The most recent result in the process of development will be presented at the conference. - 314 - Pulse injection on a power plant A pulse injection experiment was carried out by FOA being the first in a series of planned tests in order to discover the EMP vulnerability of certain power systems. The tests serve several objectives, including support and validation of the analysis, actually determining the system response and providing system response data. The particular purpose was to examine the pulse transfer characteristics of an ordinary 84/7 kV transformer. The generator used, delivered a low level current pulse with a broad frequency spectrum. The current was injected into three suspended copper wires connected to the aerial power line junction of the power plant. As a reference plane and return path for the pulse generator, we used a wire netting. Pulse propagation was studied both inside and outside the power plant. Power Pulse injection on a power plant . The power switch gear and distribution conduits which in the building run all the way down to the power transformer have the same effect on the propagated pulse as a low pass filter. Registrations of the current on both sides of the transformer are presented in figure 10. Incidentcurrentil r;~ Transferred currenti, ~~~ 0 12 3 4 s 6 7 8 910 Microseconds Fig.10: Conclusions The integration of analysis and simple test methods utilizing basic experimental equipment has turned out to be a highly valuable resource for EMP simulation. Simple laboratory models built up within a low budget program constitute the necessary foundation for the test program. Fairly accurate results have been obtained using a skilled staff familiar with the special technique pertaining to theese non-standarized procedures which altogether appear as a Cost effective way of EMP-hardening validation. station Fig.9: The short incident pulse gives a direct response seen as a spike at the very beginning of the transferred current before the low frequency oscillation becomes dominating. For early times the transformer obviously attenuates the pulse one order of magnitude which demonstrates the mainly capacitive, high frequency coupling thru the transformer. The dominating current in the response consists of a natural frequency oscillation built up by energy translated to the internal distribution system by several coupling mechanisms. Being the first one of a series of tests, this experiment has given valuable insight in the special properties of large systems. It has also indicated ways of procedure for analysing the EMP vulnerability of power systems. 012345678910 Microseconds Currents on both sides of the transformer. Relative measurements. Reference [ll Vance, Ed.: Coupling to Shielded cables, John Wiley & Sons, Inc. 1978, ISBN o-471-04107-6 - 375 - 68L4 A PORTABLE PROGRAMMABLE PULSER AN’D HIGH-SPEED, LOG-WEIGHTED PEAK-LEVEL RECORDER FOR DIRECT-DRIVE TESTING M. E. Gruchalla, A. J. Bonham, J. Gibson, P. G. Johnson EG&G Washington Analytical Services Center, Inc. Albuquerque, New Mexico, USA Traditionally, there have been two major deterof direct-drive testing: applications mination of damage level and susceptibility to in the majority of These tests, upset. have been conducted in a laboratory programs, environment on individual system modules, or In such a test line replaceable units (LRU). simulate the it is necessary to environment, object at least to the extent that host test the unit under test is effectively interfaced This can be in a proper operational manner. and time a very costly task both in material Further, in each LRU is quite unique. since many direct-drive tests, as in verification the data required is testing for example, Detailed limited to the peak signal levels. waveform information provided with traditional requires data acquisition systems data reduction procedures to reduce the acquired information to that desired. This often causes a serious delay in the availability of reduced test data and further increases cost. The application of the Portable Programmable allows Pulser the direct-drive source to be taken to the test object where the various LRU’s may be tested in-situ in their ideal environments. The additional operational the application of Peak-Level Recorder the provides immediate real-time data on waveform peaks allowing rapid assessment of accurate response and instrumentation adjustment to optimum test and acquisition parameters. The Portable Programmable Pulser Traditional direct-drive testing is genand often reerally a laboratory procedure special power conditioning, generation quires of very specific signals and responses, total environmental control (cooling, etc.), and in mechanical environments. some cases specific It is quite difficult to provide these special requirements using conventional laboratory test equipment for any but the simplest Test fixtures are therefore utilized devices. provide the critical elements of into terfacing to the test unit. Each unit of a test object that is to be tested is typical generally quite unique and requires its unique This requires a special support environment. test fixture for each unit to be tested. Since there can be, and very often is, a large test number of units to test in a typical the fabrication of the required test ‘object, fixtures can represent a major expense in the test program as well as a major time consuming element in the schedule. Once the data have been successfully collected in a typical laboratory-type test environment, the question of correlation of the measured response to the response in the actual test object arises. test The environment is obviously much different from that of the actual host environment in which the test unit normally operate. would Therefore, some type of extrapolation must be used to predict, from the test results, the actual response to be expected in a real operational environment. This is relatively simple in some cases, but quite of ten significant controversy arises as to the actual best method of correlation of the test data to actual in-system response. Another method of direct-drive testing is the testing of the various units in-situ in their actual operational environment. With in-situ testing, all the operational requirements of the individual units to be tested are satisfied without the need of any special test fixtures. Further, there is no question as to the correlation of the test response of a unit to that in the actual operational environment of the host system. However, the question still remains as to perturbation caused by the test system encroachment and the overall validity of the test in general. Those questions will always be present. It is the purpose of the test system designers to develop instrumentation that is as cost effective as possible and as closely satisfies the needs of the various tests as possible. The unique operational requirements of the individual elements of a test object are perfectly satisfied with in-situ operation, but the requirements of the test system must also be considered. If the test system must be significantly more complex than that required for laboratory testing, the overall impact on the test could be negative. This is not the case in the majority of applications. In direct-drive testing, it is the intent to effect testing by means of either direct injection of current or direct application of potential to the various identified test points. Certain special apparatus, such as break-out boxes, are of ten needed to allow access to the test points, but these are relatively inexpensive compared to the special 376 test fixtures needed in laboratory testing and be fabricated as needed. can often Some method of driving the test points and some method of data collection is needed. A great many data collection systems, such as the Peak-Level Recorder presented below, are available. from various sources typically including EG&G. The Portable Programmable Pulser (PPP) system was developed by EG&Gto address the test point driving problem. Since the various drive requirements are generally similar from one test object to another, very a relatively universal drive system can be configured. The basic limitation on the drive its drive capability. system is This sets a maximum limit on the peak current, peak potential and total output power available. If the drive system is made too large, it to be convenient and cost effective to ceases convey to the test object. However, for total drive powers on the order of several a reasonably convenient, portable kilowatts, system can be configured that can provi.de a very cost effective approach to in-situ direct-drive test.ing. The development of the Portable Programmable Pulser system at EC&G was directed specifically to the needs of in-situ direct-drive testing. System _l_Elements _The EG&GPPP system, model number PPP-2, consists of a Programmable Pulse Generator unit, PPG-1, two 1 kW power amplifier units, APA-3A, and an optional Sequencer, SCU-2. The the system is approximately total weight of 1500 lbs.: 600 lbs. each for the power amplifiers, 250 lbs. for the Sequencer. Protective covers are included to provide adequate protection for shipment via padded van or other similar The basic carrier. system is shown in Figure 1. Portable Pulse Generator. The Pro---II -Pulse Generatorunit, PPC-1 , congrammable sists of a commercial frequency synthesizer, Figure 1. Portable a high speed zero crossing detector, a damped computer for and a dedicated sine generator, This unit and communication. control system panel may be controlled either by front unit, or by a Sequencer controls, by the all functions remote computer system. Also, the power amplifiers, except for switching of from the PPG of mains power, are controlled All programmed parameters are displayed unit. Also, all on convenient front panel displays. whether programmed from the these parameters, front panel or by remote means, may be interrogated by the Sequencer or a remote Other features such as damped sine computer. over run warning as well as various indicators The communication link to also included. are computer is a the Sequencer or remote The use of the fiber-optic RS-232 link. communication path reduces the fiber-optic causing upset or risk of unwanted coup1 ing even damage to the test system and test Although designed as an integral object. element of the PPP-2 system, the PPG-1 may be as a stand-alone unit with user-supplied used power amplifiers and sequencing system. In the power application, however, such an by the amplifier control cannot be provided Major features of the PPG-1 are shown PPG- 1 . in Table 1. Programmable Pulse Table 1 Generator Major Features *Damped sine generation - 10 kHz to 150 MHz fundamental frequency - 100 Hz resolution - Q’s from 1 to 19,999 depending on frequency - Q accuracy of +/- 5% *Burst sine - 10 kHz to 150 MHz - 100 Hz resolution - Interval on of 2.5 ~3 to 25 seconds - Duty cycle of 50s Programmable Pulser System - Table 1 (continued) *Repetition rate - 0.02 Hz to 200 kHz - Accuracy of +/- 2% *Psuedo random repetition rate average repetition rate -0.6 x selected reoetition rate *Pulse count - 1 to 65,536 *Polarity reversal *External trigger *Pulse delay *Cain - -60 dB *Cain control - -80 dB range The PPP-2 system is Power Amp1if ier . designed to support two APA- power amplifier The APA- power amplifier Unit iS a units. 1 kW unit designed to deliver 1 conservative The output impedance, kW to a 50-ohm load. is 100 ohms to allow very convenient however, parallel or push-pull operation into 50 ohms. is designed to drive any VSWR load The unit the full Also, without damage to the unit. of one APA- may be reverse applied to output the output of another without damage simplifying combined power amplifier configurations. The maximum current available into a short circuit is 7A peak and the maximum potential available to an open circuit is approximately Either of these values may be roughly 7oov. parallel or series operation doubled by depending on the drive desired. The push-pull provides the best distortion operation performance for a given output power and provides a total output power of 2 kW into 50 ohms. The APAmay also be used as a stand-alone unit and may be operated manually from its front panel controls or, with suitable interfacing, remotely from a user supplied system. Sequencer. The SCU-2 Sequencer is a comme~-computer repackaged for portable use. It includes the general software for basic system control and monitoring, but does not include any application software. This unit allows total control of the entire PPP-2 system, except for mains power, and allows interrogation of the various programmed parameters for accurate logging of front panel entries. Accessories Various accessories are generally required for the use of the PPP-2 system. As mentioned earlier, such devices as break-out boxes and other similar units are test object dependent and must be identified for the particular test object and test points concerned. Current drivers are convenient for direct injection of current into conductors without the need of breaking the conductor. Often, only the harness restraints need be removed to install current drivers. However, the physical size of the driver increases with the lower limit of drive frequency and maximum power desired, and for frequencies near 10 kHz , the size required may be too great for convenient installation. Other Applications A general purpose system such as the PPP-2 is not limited to applications in direct-drive testing. It can essentially be used in any application requiring a versatile programmable RF source. One such application 377 68L4 - The high output is in the area of cw testing. power coupled with the protection against poor load matching results in an excellent driver the Further, for the illumination antenna. and the wide range of gain programmability convenient means of allow a very control tailoring of the response to specific needs. Log-Weighted Peak-Level -- Recorder is probably the most testing Pulse popular method of performing elecwomametic the response of various characterization of test objects and systems. However, the actual the desired information data collection of environments has always been a from these Countless data acquisition systems problem. and data links have been developed - each with The majority of merits and shortcomings. its these systems have addressed the problem of the desired data and sensing of accurate faithfully communicating it out of the hostile In this respect, the major EM environment. concern has been with data quality and not particularly with data throughput. Obviously the primary importance. data quality is of However, data throughput can be a serious concern due to of testing, the cost availability of the test object, and simulator scheduling . Further, in many types of testing such as verification direct-drive applications, only peak information is needed and that information must normally be reduced from the collected data. The Peak-Level Recorder was developed to provide rapid and accurate peak-level information. In applications requiring only peak data, no further reduction is required for access to needed data. In conventional data acquisition systems, the peak-level information allows accurate setup of the acquisition equipment with a minimum of trials thereby improving data throughput. Each data collection environment is unique in its operation and associated problems. However, there are classically several things that limit the overall data collection rate that many tests have in common. One major contributing factor to test duration is the time required to reconfigure the test object after each test shot. There are a number of instrumentation techniques which could be applied here that could improve the data rate. A typical example would be a single data link with multiple multiplexed inputs such as the EC&GODS optical data links which provide eight such inputs. Since only one input at a time can be active, one pulser shot (at least) is required for each input instrumented as for any single channel data link. The real advantage is in the capability of instrumenting a number of data points with a single entry into the test object. This eliminates the need to open the test object for each data point that a specific link is to access. This can be a significant advantage with test objects in which a significant amount of time is required for reconfiguring the data links. For example, the ODS-68 system can access 64 data points in a single intrusion into the test object. The reconfiguration time, although somewhat equipment dependent, is also highly a function of test management and the type of test object. A generally more serious limitation on the data rate is the number of - each test needed to acquire shots pulser This not only affects the data rate point. site cost since the pulser is but al so the lifetime device or at general ,ly a limited one with a capability of a limited least service major between pulses number of the Reduction of procedures or failures. each data number of pulser shots to acquire reduces test time and significantly point pulser degradation during a given test. There result in are two basic problems which shots being needed for each multiple pulser variability and limited pul ser test point: of the data collection system. dynamic range With contemporary pulsers, the PUlSer output variability is low enough that it is within tolerable limits for good data collection. Limited dynamic range of the data system is The range almost always a limiting factor. which typical data can vary, from the over test point least excited to that most excited, magnitude is generally several orders of greater than the dynamic range available in Therefore, typical data collection systems. the signal level must be adjusted to fall the data system. within the usable range of This is generally accomplished with input either manual or attenuators, remote Some limited success has been controlled. obtained with logarithmic amplifiers but the and the dynamic range is stability is poor generally below the data range so attenuators are still required. It is not unusual to use several pulser shots to get an attenuator properly adjusted, and if it is necessary to collect several data points in the same shot for correlation purposes, a reasonably large number of shots, perhaps five to ten, could be used. It would be ideal if the number of shots needed to set the attenuators could be reduced to zero. That would require that the peak value of the data be known reasonably accurately. Indeed, if the data were known that accurately, many tests need not be done. Another approach is to use one pulser shot to assess the data peak of each instrumented point. That peak information can then be used to set the corresponding attenuators allowing good data quality on a second pulser shot. This is possible due to the good pulseto-pulse performance of modern pulses, as mentioned earlier. This then reduces the number of pulser shots required to acquire a set of simultaneous data points to two, and almost all sets of data in a single instrumentation session can be simultaneous. There are of course anomalies which on occasion require more than two shots to get all the desired data, but this peak level technique has proven to be a very valuable tool in improving data throughput. Peak Level Detecting -There are several methods that could be considered for peak-level detection but first, a brief review of the actual performance needed would be valuable. A key parameter is the dynamic range. A value of 40 dB is a reasonable minimum value since only the peak value is of interest and not waveform structure. Also, both polarities must be sensed although the sign generally need not be preserved. Sign of a peak, however, is one additional piece of information that may have some value and could easily be obtained from 378 - Speed is a a peak detecting instrument. very and a parameter critical second A nominal 100 MHz capability subjective one. adequate for the range Of generally is interest in the larger experimental Sites. The resolution needed is a function of the A resolution of 1 dB to specific data system. 3 dB will generally satisfy most requirements. Finally, the question of linear or logarithmic Almost without response should be considered. exception, the data system scaling is in some way keyed to the dB parameter, the attenuator settings for example, so it seems only logical done on a that the peak detection be Some reference level must logarithmic basis. logarithmic chosen for the then be An initial value of 10 mV was measurement. chosen to compliment the data links with which These the prototype was to first be used. values are certainly subjective and subject to discussion, but they do represent a reasonable starting point for a peak detection system. These parameters are tabulated in Table 2. Initial Table 2 Peak-Level Recorder Design *40 dB dynamic range for both positive negative polarities, sign preserved “100 MHz response *l dB resolution *Logarithmic response, 10 mV reference Goals and level Traditional means of peak capture are generally based on some form of analog peak capture or some type of A/D flash converter. The analog methods suffer from both speed limitations and poor response flatness with frequency. The conventional flash converter is less than desirable due to its need of a trigger to capture the desired data. A high speed memory could be used to store a large amount of data from some trigger point and the peak extracted after the shot. However, a reasonably large memory would be required as well as a high sample rate, perhaps 500 megasamples per second for reasonable 100 MHz information. There is quite a bit of work being done in the industry on such A/D systems and at present they are not available commercially. In the development of the PLR system, we chose to use a variation of the flash converter. A conventional flash converter generally incorporates a single comparator for each discrete bit value in its operating range, a 4 bit converter requires 16 comparators, a 12 bit 4096. The desired 40 dB dynamic range with 1 dB resolution woul’d require 512 comparators for each polarity, or a total of 1024 comparators. If the divider were logarithmically weighted only 80 comparators would be required. However, the ladder structure of the conventional flash converter is not one consistent with good high frequency performance. A transmission line conf iguration provides superior performance. A better high frequency configuration of the flash converter is the distribution of the comparators along a terminated transmission line separating each comparator from the preceding one with a 163. pad. This configuration can provide very good high - 379 frequency performance as well as good accuracy. One problem encountered was that the input impedance of the comparators selected was poorly behaved with frequency. This required a reasonably complex compensation network and even then the upper response was somewhat less than that desired Each comparator but still quite functional. is arranged in a latching configuration. This makes maximum use of the comparator speed and minimizes the total number of high speed With this configuration, components needed. comparator is tripped, in order, such each that all those below the highest tripped will Thus, only the highest also be tripped. determine comparator tripped need be found to This configuration will retain the the peak. highest peak applied since the last reset higher occur successively peaks whether nanoseconds or hours apart. The 80 comparators are arranged on four individual flash converter PC boards each containing 10 positive sensing and 10 negative An RF amplifier is also sensing devices. included to provide a convenient means of frequency compensation on each flash converter These are then series connected with card. lengths of coaxial cable to form a dual 40 comparator long converter. signals from output the 80 The are applied to an interface card comparators where a dedicated microcomputer is used to format the data to drive a front panel display and for transmission to a dedicated host computer in the system chassis. The four flask converter cards and the interface card are housed in a subchassis Figure 7. Peak-Level 68~4 - which results in a convenient complete flask converter module. Each module has a front panel with displays for both positive and negative peaks and a control for manual reset. Eight of the flash converter modules a.re housed in a chassis unit, four in a master chassis, four in a slave chassis, to provide an eight channel system capability. The individual module front panels form the individual channel data dispiays. A dedicated computer in the master chassis handles communication protocol to both a front panel display and to external data collection equipment. Either RS-232 or IEEE-488 communications are available for external data transfer . The system may be controlled either from front panel controls or remotely via the communication channel. The dedicated computer contains a software module which also calculates the attenuation needed in each channel for a given channel compression level and detected peak. That information is provided at the chassis front panel display for immediate use and to the remote communication link for automatic control or logging. The complete PLR system is shown in Figure 2. Field Test Results TheirstR system was fielded and after a troubleshooting and training period, it proved to be very valuable in improving data throughput. Data throughput improvement of as much as a factor of five and in some cases even higher were recorded. The completed system met all the design goals Recorder System 380 except for speed. The speed of response is limited by the comparators. A 70 MHz half sine pulse is detected 3 dB low which corresponds to a. 70 MHz, 3 dB band-pass. The 70 MHz limitation did not present any noticeable limitations. the dynamic range, Also, resolution, and reference level proved to be very good choices for the particular first application. The concept of logarithmic response interfaced excellently with the setting of the attenuators. In all respects, the concept of Peak-Level Recording provided the performance expected and the prototype was considered to be a complete success. Other Applications described here The PLR system was basically developed as an aid to conventional data acquisition. However, in many cases the final quantity of interest at the various data points is only the absolute peak signal. In those applications, a PLR system and data link would be the only data acquisition tools needed for complete peak data recording. Further, in those applications where total waveform structure is desired, a PLR system could be used in a pretest evaluation to prioritize the data points in order of their response peaks. This information could then be used both to rank the data points in or.der of importance as well as to set the attenuator values for the specific data points as they are instrumented for test. Conclusion A very cost effective solution to directin-situ testing where the drive testing is unique operational requirements of the various Further, system units can easily be provided. remains as to the little or no question correlation between the test response of a response in its actual unit and its operational environment since the two This environments are virtually the same. type of a versatile drive testing requires unit which is both portable and flexible in its We feel that the PPP-2 capabilities. system as outlined here satisfies these requirements and offers significant advantages to direct-drive testing in terms of cost, flexibility, and test time. Similar1 y, a peak-level recording instrument can prove an effective addition to a This type typical EM data acquisition system. of instrument, when effectively used, can provide the information necessary to collect high quality data with the minimum of pulser firings. This improves both data quality and throughput reducing test time and subsequently test cost. Also, since a minimum of pulser firings are needed to acquire a given set of data, pulser lifetime is extended. Furthermore, in applications where only the peak data are desired, a PLR system with a suitable data link is all that is required for complete data acquisition further reducing test cost and complexity. - 381 - 69L5 BLACK BOX BOUNDS Carl E. Baum Air Force Weapons Laboratory Albuquerque, New Mexico USA This paper applies time-domain norm concepts to bound the failure of a black box to multiport excitation in terms of the failure responses to single port excitation. Appropriate assumptions concerning the nature of black box response made and the are discussed. 1. Introduction In characterizing the interaction of electromagnetic fields with complex systems one can make the problem more tractable if, instead of trying to obtain the actual signals at various positions in the system, one settles for something less detailed, in parRecently ticular, bounds on these signals. several papers have addressed this approach [4,5,6,7], from the points of view of both transmission-line network theory and the scattering equations encountered in (quantitative) electromagnetic topology. It is becoming clear that the concept of norms plays a central role in bounding the electromagnetic response of complex systems CZJ31. The general interaction questions [3,4] are conveniently cast in forms involving supermatrices which appropriate norms can reduce to scalars. The electromagnetic signals of concern propagate "down" to the circuit level where various undesirable effects can occur. These effects are usually divided into two sets designated upset and (permanent) damage. In this note we take some set of such circuits which are physically grouped together into what are often termed "black boxes" which are in turn typically interconnected by signal transmission lines (wires, waveguides, etc.). Characterizing such a black box as an Nport network the N signals (considered independent) are cast in the form of voltages and currents, or equivalent voltages and currents for cases that the signals are in the form of more general electromagnetic waves (modes). In this form black-box terminals are put in a form compatible with the equivalent voltages and currents presented to them by the rest of the system in the format of transmission-line network theory or electromagnetic topology. 2. Black-Box Characteristics For our purposes the common "black box" is considered to be a network with N input ports as indicated in figure 2.1. There are also M internal "failure ports' Cl]. These are indexed as n = 1,2,...,N (input ports) m = 1,2,...,M (failure ports) (2.1) The N input ports are assumed to be known a priori. However, the number of failure ports (M) may be a priori unknown as may be the location of any or all of the individual failure ports. A failure port is defined as any port (with two terminals) inside the black box where some signal at this port can cause failure. This is interpreted in the sense of any change in the black box function or capability to function resulting from some signal there. This includes any transient upset (change of logic state) as well as permanent damage attributable to the signal driving the failure port. Let the input signals be F,(t) = a,f,(t) (2.2) where fn(t) is some appropriately normalized waveform and a is an arbitrary (real) ampliHere tie F"(t) can be interpreted as tude. voltages, currents, or some linear combination of the two (such as combined voltages C31). Let the response at the mth failure port be given by G,(t) = (2.3) ,5, angm,n(t) ~~li~ee~il+i(t) is the response due to fn(t) nth input port. Of course, this type of response assumes linearity, at least for times of interest. Stated in vector/ matrix form the input is (F,(t)) giving a response (G,(t)) = (g,,,(t)) l (a,,) (2.4) Now f,(t) might be any kind of waveform, including a 6 function such as - 382 - input port t- 5 1 "1 failure ports 2 1 0 -- + t I1 t 2 . . . . . . L B B "2 0 -I2 . . . . M . I3 N "2 -I3 Fig. 2.1. fnH Black box representation = s(t - tn) (2.5) If t, is allowed to vary then gY,,(t) may vary as a function of t, in a camp ex way If If it is the system is not time invariant. time invariant then f,(t - tn) produces gm Jt - t,) and the two are related by a Perhaps we might betcofivolution operator. ter assume a certain set of piecewise timeinvariant states so that f,(t - tn) produces gh'A(t - tn) during system state T. Wave- fokms are not "allowed" to cross state boundaries (particular times). This can be stated by (for causal functions) gl\‘A(tj , (I$) (t)) = T,!,',!,(t) 0 fn(t) , = (4(1)(t)) m,n l (an 1 o E convolution with respect to time x(t) 0 y(t) 3 IL x(t - t')y(t')dt' -m Thus far we have not specified whether f,(t) and gm ,(t) represent voltages or currents or so?ne linear combination of the this is purposes For present two. - 383 Table 2.1. Terminations at input ports for different kinds of singlewaveform excitation Input-Port Waveform fn(t) Input-Port Termination for n' f n voltage V short circuit current I open circuit combined voltage V + ZI (incoming wave) (measure both V and I at nth input port) I__ impedance Z taken as a frequencyindependent resistance (assures only outgoing waves) uGm(t)n => failure at mth failure port (3.1) i L< 'rn=> II G (t )II b 0 otherwise llG(t)Itl E j- -cQ fG(t)\dt (3.4) llG(t) Ila, 5 max G(t)( t' More generally the p norm is Failure Norms -- Now we need some measure of G (t) to determine if failure occurs at the m& failure port. Remember Gm(t) could be a voltage, current, or combined voltage; whichever it is may not be important for present purposes. Let us define some failure measure as a norm 3 ' rm (3.3) = 0 iff G(t) s or has zero "measure" per the particular norm Examples of norms might be one will unnecessary. However, for the f,.,(t) eventually have to choose some form to perform the experiments involving sequential single-port excitation. The remaining input ports will then have to be properly terminated as indicated in table 2.1. 3. 69m - non-fa lure at mth failure port Let us take system fa lure as system failure <=> at least one failureport failure (3.2) system nonfailure <=> no failures at any failure port It will be further assumed that such a system failure, whether upset or permanent damage, will be observable, even if the particular failure port or ports which fail are not This observation might take the observable. form of a check of system logic states and/or functional performance after the test of interest. Some questions are: What is an appropriate form? Do all failure ports have the same norm? Fortunately, if there are such a norm or norms, these answers are not essential since all norms (vector norms) have certain properties [8] uuG(t)ll= Ial llG(t)ll C(: scalar llG(t)up= [ i_, )W)lP 1I/P dt 1 (3.5) Here integration is actually over times for which G(t) is significant. In particular, integration needs to be limited to times in the Tth time-invariant state of the black box. Note that only real G(t) are considered since we are dealing with physical timedomain sisnals. If energy is the failure mechanism then the 2 norm might be appropriate. However, suppose the failure mechanism is peak voltThen the failure mechanism may not be age. bipolar, failure requires +l volt or e.g., -10 volts This difficulty can be overcome by defining the experiment so that both G (t) and -G (t) are produced (different tests7 from (FnTt)) and -(F,(t)) with failure in either polarity defining failure-port failure. Such norms then apply to bipolar experiments. Of course, if G,(t) has equal positive and negaThis tive peaks only one test is needed. might be the case if the F,(t) were sinusoids (of a common frequency) making the G,(t) Practically this would require sinusoids. slowly and smoothly turning the exciting Damped sinusoids are sinusoids on and off. more problematical. Let us define a special kind of norm as a time-invariant norm iff iG(t - t,)u f function of tQ (3.6) Of course this is only meant to apply within Note a time-invariant state of the system. norms are all time that the above p invariant. The p norms in (3.4) and (3.5) have the property that if the integration is truncated one obtains a lesser value for the norm since the integrands are positive semi-definite. 384 vary an (real) until a failure occurs at some (perhaps unknown) mth failure port. Determine the maximum (positive) a,, and minimum (negative) a, for failure. Define Stated another way, we can define tf IiG(t )Ii PJf f 1 I__ (G(t))P - dt I”” max Z monotone non-decreasing function of tf (3.7) which also applies to (3.4) by restricting -w r, (3.9) if failure occurs (from (3.1)). can m be interpreted as a failure time for the mth failure port. In section 2 transfer convolution functions were defined relating Gm(t) to the This involves fundamentally the fn(t)* For our failure assumption of linearity. norms to apply it is only necessary for a failure port (and the signal transport to it) to be linear for times up to tf . After this m time the failure port will have failed, which by assumption is detected as a box failure. is irrelevant and Stated another way t > tf m linearity for such times is not needed to Even more generally insure the result. define tf = Thus tf min tf 1tmcM m Single-Port Tests Suppose now that we apply F,(t) at the nth input with all other inputs zero and terminated per table 2.1. Noting that F,.,(t) = a,f,(t) (4.1) F,,(t) cz0 for n' f n min 1 (4.2) > 0 so no failure occurs without an input so that A, = minimum IanI causing failure anywhere within the black box (4.3) Note then that for all Gm(t) under the above condition IIGm(t.)lI = Ia,g,,,(t)I! = Ia,1 llng,,,(t)ll < r, for all m with equality for at least one failure port (giving the blackbox failure) (4.4) with failure norm (and r per (3.1)) being that appropriate for each &h failure port. The point is that for all a, individually with 0 c IanI < A, , n = 1,2,...,N (4.5) there is no black-box failure. each case of Furthermore (3.10) giving the first failure at any failure port, which of course gives black box failure. Times greater than this tf are unimportant and linearity is not required for such times. Thus we do not need a completely linear system for our results to hold. Define this linearity as required kind of lesser linearity to failure. 4. ,-an IanI < A, with an, = 0 for all n'fn (4.6) gives IIG,(t)ll < rm for all m = 1,2,...,Y which is failure. 5. the requirement for no system Bounds on Failure under Multi-Port Excitation Now let there be signals on all N input ports. The failure port signals are (G,(t)) = (g,,,(t)) with failure norms l (a,) (5.1) 385 IiGm(t)H= &$A11 1"4, gm,n(t)anl n also N < 69Ls - I1 IanI “g, , “(t)l (5.2) This last result is interpretable as the 1 norm of a vector whose elements are failure norms of the signals from each nth input port, i.e., (5.3) with (F,(t)) = (a,f,(t)) (5.4) assures > rm for any m = l,?,...,M => box failure IlG,(t)~l < r, for all m = 1,2,...,M => box non-failure (5.5) Non-failure of the box is then assured if Il(lanl lig,,(t)ti)lll < r, for all m = , 1,2 ,*a*,M From the single-input-port tests 4) one has (5.6) (section A,,llg, ,(t)s < rm for all m = 1,2,...,M with , equality for at least one m and for all n = 1,2,...,N (5.7) non-failure. component vector ($) n . Consider the N Then an < 1 => box non-failure (A;;)1 II I (5.11) This is then a bound on multi-port excitation (the set {a 1 of input-port excitations) to assure non-fgilure in terms of the results of single-port excitations (the levels la I = A, for black-box failure due to sing1 e-port excitation). Note that the index m does not appear in (5.11) so that the location (m) of a failure in the black box is not needed in determining Our task here is to find conditions under which the black box will not experience a failure. This is based on - (5.10) ( this bound. Note that i(; for all A, > 0 is a valid norm for arbitrary One might call this norm a weighted 1 ~~~~m)(Similarly one could define a weighted . Note that the above norm is a tight one since for all an = 0 except for n = nc (5.12) this norm in (5.11) reduces to Ian-l -A,,‘l (5.13) which is exactly the result from an experiment concerning non-failure for single-port excitation at the n'th input port. A looser but simpler bound can also be obtained. Write an (5 1 = (k ln,mj (a,) l (5.14) A, > 0 for all n = 1,2,...,N so that or rm 119,,(t)e < 7i- for all m = 1,2,...,M , n n = 1,2 ,***,N Replacing rg,,,(t)w by (5.8) rm/An only increases the sum (1 norm) in (5.6) so that requiring (5.9) or (a,,) = (Anln,m) l (2) n Il(a,)u< ll(A,l,m , (5.15) (5.16) Now for diagonal matrices for any associated matrix norm (see C61) ~(A,l,,,)ll= max IA,.,~ = max A, n n (5.17) - Also we have fO,)fl @I > G,* and - 0.452)2t1.852 in passing we may write as C a," _ 0.5 Q -1nS We now attempt to extend these results to nonperfect conductors. Such an extension requires that we approximate the field at the surface of the conductor, and we write I J=uE- . . . (67) .*. (59) 2na6 - 0.168 where which compares directly with and is more accurate than the result of Lee and Leung the skin depth [4]. We can virtually recover the result of Lee and Leung by ignoring the logarithmic term in (61), when the estimate for a becomes by /- (+o= We then have 0.5 that )I= . . . (60) Q -1np is given 2 6= a-" CONDUCTIVITY the I0 field at z = 0 is e(i*a)t . . . (69) 2naSu - 1.03 Lee and Leungs formula having 0.46 in place of 0.5. An approximation for the frequency can be obtained from (31) but the correction to w, appear0 to be insufficient. We can obtain a more accurate first approximation by noting that the non-zero value of p is the source of a and 00-w and that G,* is insensitive to 6. Accordingly we elrpect G, - G,"r and this implies Differentiating with respect to time and introducing the result into (17) we obtain after cancelling colmaon factors iw-a -_ = u. 2na6u ;(i@-a)Do'/c 4~ "0' I [[;]2t _ n q-J-&l _002% a vc2 (i*a)2]i';'i~)Do~ccosv~ dn' . . . (61) or -R (70) Introducing f/fo"l- This compares with Or vafo Lee and . . . (62) .*. (71) a6uh Leung's result rewritten in this form 2~s i/fo - 1 - " Separating into real and imaginary Parts and approximating as before we have the estimate ... (63) Sfo Both these results can be converted cylinder approximationeby dividing to closed by (1 + 0) a16 ” C *G+ *b% Fc*G2* *' + G2"2 IG, + s/Zc GI* I + 9/2w G,* I . . . (72) - Writing e explicity and noting that region where the correction becomes 395 71 - L7 in the significant *.. (77) we G, >>G, find that . . . (73) we can write _EXPERIWEWTAL oI6"+ This approximation will cease the skin depth approaches the ie when to be valid when cylinder radius, lX&-$l a > __ aa . . . (75) 4 when this obtains, J= we write I . . . (76) rraL and PRGGRAWME [1+2+&l the estimate for al6 becomes The experimental procedure was to illuminate 2 m dipoles in a bounded wave IMP simulator and monitor the centre current using broad-band current probes. The signals were led into a microwave transmitter coupled into a wave guide with a dielectric break and thence to a remote screened recording room. The current pulses were recorded on film and analyaed graphically to obtain the attenuation constant and w was determinea directly from the late-time response. Two conductors, copper and nichrome were used, the nichrome to enhance the resistance effect and so make it observable with the short lengths of antenna used in the experiment. 1-o EXPERIMENT 0.9 ) (NICHROME ? : SSEM 0.8 MOM (ROBERTS) SEM ITESCHE I 1 MOM 0.7 --- APPROXIMATION (771 USING 166) ’ O-6 0.5 u = 1~1.1065/m 0.L o-3 (T = 6.25.107 - 0.2 * o=oo 0.1 0.8 EXPERIMENT SSEM 0.7 0.6 0 MOM A MOM (ROBERTS 1 t Fig 2 ‘l/3 -----e slm r\ - - 396 - CONCLUSIONS Fig 3 4-.- G/f” been singularity expansion method has The applied to the determination of the resonance cylinder. The parameters of a resistive estimates refines reproduces and approach obtained by other techniques leading to close agreement with the results obtained by a more sophisticated application of numerical methods together with an encouraging agreement with results. The analytical experimental approximations are useful for aspect ratios as low as 5 where the geometric assumptions become questionable out to values beyond lo5 where resistivity dominates even for good conductors. Combining these results with the early time enable presented in predictions Cl1 will accurate predictions of antenna currents over the entire time scale. extreme simplicity of The adequate approximations indicates their utility in threat assessments of larger scale systems. WoM PREDICTIONS An existing Method of Moments programme was used to predict the impulse response of 2 m cylinders, the resonant frequency being determined b y Fourier analysis of the results while the attenuation constant was determined graphically. The progrannne allows for the finite conductivity of the cylinder. RESULTS The results of the various computations for the decay constant, together with the experimental data are presented in Fig 1 while the correponding curves for the resonant frequency appear in Fig. 2. The close agreement of the various computations is to be noted, together with the acceptable agreement with experiment, particularly with regard to the effect of resistivity. The resulting pole trajectory is presented in Fig.3. Cl1 EWP Coupling I L Gallon to Long Cables EMC'79 c21 On the Singularity Expansion Method the solution of Electromagnetic Interaction Problems CE Baum IN 98 AFWL. for Expansion Method as c31 On the Singularity applied to Electromagnetic Scattering from thin wires IN 102 AFWL. F M Tesche Frequency of a c41 The Natural Resonance thin Cylinder and its Application to EMI7 Studies. SWLeeandBLeung IN96AFWL. [S]f&,&ularity scattering Expansion from thin of Electromagnetic and thick closed cylinders. J L Roberts L W Pearson IN 431 AFWL - V&l_II)&TION USING THE OF EMP RESPONSE LXGHTNING 397 72 - CALCUL&TION 6JN CIEFtIf%L OF STROKE METHODG CABLE TO ricr emr7 of t.he 14 pairs The that tishield of the two buried connrcted at each end to r-resi~.tivc; ground. L8 FI cable cablfzli a 30 and are ohms On the! 14 pairs overhead cable, the the di+‘fcrential modes arm c*mmo17 and 600 matched at the far end through a a 120 ohms resistances, ohms and Resi st i vE3 val tag!2 rccspmcti vrl y* arc conn~~tad at the near end dividers in order to match the and the common dif+crrsn.tial circuits and give - 398 WREGISTREMENT DU 23mva3 StPRIVOT MESURE HF -Voie: 6 -Amplitude: 250 Seuil: 300 Gamma -Pente: 2500 Gamma/ms Date seuil: 3 h. 30 mn. 19 sec. 365 ms. U(P) = J$",(P,x)us(x)dx + /eG (P,x)Ie(x)dx 0 v2 .(P) = fiII(P,x)Us(x)dx + P G12(P,x)Is(x)dx 0 - 399 72 - The f 01 :I.owi ng (fig. is ‘-* The height is neglicJibIc length. pcrfextly of the c:abIs + 0 i(t-F/c) + 7 the ’ the (35) 1 di(t-r/c) dt = whmre c:urrcnt (6) I is l.(t) , c: the df kstanco thr measured speed of 1 ightning t hI light, r be?tw~en the 1 ightning fal 1 point and the dititt-ibutsd on the gable, and h the height return stt-ok@ channel. stroke source af thr both I ground dipole above;? to 1 ;5 Jti(t-r/c)dt 2h = 4nEo 3 One of made were c:onductivE). compared 1 H(t) on9 4). I*. T h w ciai1 E(t) a3sumpti L8 problem is h R Thi% to can 1.n j exted the dextermine br done+ by cxrrent the value mcasuriny and the magnetic field at a particular point af the the line. If mceasur- i ng point is near rnuugh from the injection paint, the field icsr al most magnetic proportioncirl find h z 2nr to the 2 H(t) i(t) currtz+nt and one can (7) (5) and (6) we tranepaserd in frequency domain and ueed to czompute the va3 tage and current distributed swurC:‘es CEil whi rh are nfwded for the integrals (3) and (4). Equations the E(w) = & H(U) = g (8) 0 1 [ F2 + $]W) (9) 0 -10 measurement -20/ - CHARACTERISTICS 401 OF UNUSUAL 73Ml - POWER MAIN TRANSIENTS William T. Rhoades Xerox Corporation El Segundo, California 90245 INTRODUCTION For nearly thirty years, researchers in the field of Electromagnetic Compatibility (EMC) have been puzzling over a multipartite question about power main susceptibility: what amplitude value, waveshape or frequency spectrum and time duration must equipment be designed for? The propagation modes, classes and stress levels for transients have been defined (1). A large number of common occurring transients have been characterized for each class of transient ( 2 ). These transients include the classical contact arc having an inductive load, lightning and the inrush current from tungsten lamps and motors. However, in order to have error free operation and avoid any serious detrimental effects, the manufacturer must design and test products for all, not just the common, power main transients, This paper will describe the characteristics and measurement techniques for three of the many unusual power main transients. with no AC/DC/AC conversions. The load is connected to the secondary winding of the CVT. With a power failure, a battery supplied AC waveform is applied to a primary winding of the CVT. This low power UPS design is called an off-line UPS whereas the UPS in Figure 1 is called an on-line UPS.) Manufacturers of UPS have defined the output tran-sient, output harmonic distortion and, sometimes, the Electra-Magnetic Interference conduction output (EMI). Unlike many power main transients ( 2)) online UPS transients are continuous and vary in characteristics. These power main transients are a strong function of the design of the UPS and the load on the UPS. UPS The main use of an Uninterruptible Power System (UPS) is for protection against power failures lasting more than a few power main cycles up to When protection is needed for outages minutes. approaching a half hour or longer, usually an auxiliary power source is used or the computer/office equipment is placed in a non-operating, low-power standby mode. Because more and more computers/office equipments are being used in critical or real-time applications, the use of UPS equipment is steadily growing. It has been estimated that 5% of business sites have UPS and 30% of all large mainframe computer sites have UPS back-up equipment. The use of UPS is a function of the application, not the quantity of power used. Nearly all UPS designs work on the principle shown in Figure 1. In the normal continuous operation, the AC voltage is converted to a DC voltage which supplies the necessary battery charging current and the inverter current as shown in Figure la. The inverter converts the DC voltage back to an AC source. When a power outage occurs, the battery supplies the inverter as depicted in Figure lb. If an overload occurs or the UPS fails, an automatic transfer switch (like thyristors or FETs at low power levels) usually phase synchronized to the input AC mains; connects the load directly to the AC mains as illustrated in Figure lc. (Another UPS design conthe utility power to the primary of a nects ferroresonant Constant Voltage Transformer (CVT) T (b) I w I Figure 1: BANK The three operating modes of an ON-LINE UPS - a) with AC power present, b) during an outage, and c) with an overload on the UPS. Synopsis of UPS hverter Designs How the DC to AC inverter is designed has a major effect on the input and output harmonics; therefore, a short review will be given. - The simplest DC to AC inverter drives a bipolar square wave into a transformer. The bipolar (positive and negative) square wave is created by using four thyristors in a bridge. This rarely used method produces an unregulated square wave output voltage with very limited overload protection. By using a CVT, the fixed voltage output is quasi-sinusoidal (up to 8% distortion) at a frequency which is determined by the resonance of the transformer leakage inductance and an external capacitor. The CVT method is automatically current-limiting and voltage regulating at a constant square wave drive frequency. However, due to the low efficiency of CVT, the design becomes uneconomical above 10 kVA per phase. The most serious problem of the CVT method is that an output transient is created with load changes. Most CVTs have a 100% overvoltage transient with a 50% load reduction. This overvoltage transient can last for 20 to 30 cycles of the output frequency. Due to the self-limiting CVT characteristics, the applications of a full load creates a droop in excess of 25% lasting 2 to 3 cycles. Thus, the CVT method cannot be used with large motors or switching power supplies having no inrush control (2 ) . To improve the output regulation with DC input voltaee changes, a quasi-square wave could be used. At e&h end-of. the-primary winding, a power stage (like a 4 thvristor bridge) creates a square wave. The two square waves buck each other and,can create a dead time in the pulse drive to the CVT between the change of polarity. Maximum voltage at the CV$’ output occurs when the two square waves are 180 out of phase and vice versa. Regulation is achieved by changing the effective pulse width and amplitude, however, the dual drive CVT method still suffers from load transients, efficiency, non-adjustable output voltage and output harmonic distortions. To overcome the CVT disadvantages, the transformer is made normal and electronic feedback (voltage and current) from the load is used. The undesirable output harmonics from the quasi-square wave drive are filtered by large tuned tank series/parallel filters. the transformer Since the efficiency becomes resonance is removed, insensitive to inverter control frequency. With electronic control, the short term overload capability for inrush loads can be achieved. Transient performance due to load changes is usually 240% due to a major trade-off: the greater the number of output filters - the better the harmonic reduction but the poorer the transient response. To maintain good transient response and still reduce the output harmonic at the desired output frequency, the quasi-square wave is replaced with a high frequency pulse train which is modulated at the That is, a series of desired output frequency. positive pulses of varying duration followed by a series of negative pulses of varying duration are applied to the transformer. The desired output frequency is the frequency of change from positive to negative pulses. By making the duration of the high frequency pulse proportional to the amplitude of the desired output waveform (like 50 or 60 Hz), an integrated output waveform is created. At zero crossing of the output waveform (50 or 60 Hz), the pulse duration is made as small as possible. At the crest of the output waveform, the pulse duration is the maximum. Although the amplitude of each high frequency pulse to the primary winding is constant, as the duration of each pulse is increased, more energy is applied to the transformer. As the pulse train frequency is increased, the transient control is 402 - better and the harmonic distortion at the desired output frequency becomes lower. The main advantage of this pulse width modulation method is that the output filter need only remove the carrier and the For example, if the pulse sideband fre-quencies. train frequency is 1200 Hz, filtering need only to be from 800 to 1600 Hz, not 120 to 1600 Hz. The upper frequency limit is the thyristor component limitation and hence the upper power limit is about 100 kVA per phase. Above 10 kVA per phase and nearly always for very large three phase inverters, the synthesized stepped converter is used. The output waveform is broken up into either six or 12 portions or steps. The higher the number of portions, the better the tranand the lower the harmonic sient performance distortion becomes. Thus, the 12 steg inverter has Each step each step spaced at 360°/12 or 30 . employs a thyristor bridge feeding power to a portion of the output transformer from the DC input voltage. Each step is a square wave at the desired output frequency that is phase shifted away from the normal 30’ (en) position to regulate the ultimate output voltage. Since this method inherently reduces the harmonics below the 12th, the output filter is usually a simple low pass filter. Power Main Transients From An UPS The maximum distortion of the input current is sometimes specified rather than voltage distortion to avoid the changes of input power line impedance. Typical distortion values- are 10% of the maximum input current for 3 phase and 15% for single phase. Note that this specification is % RMS distortion, not peak voltage or current, pulse duration or the value of any harmonic. The power main transients are not controlled by the power utility, by safety requirements (NEMA, NEC, OSHA, UL, IEC, etc.), by EM1 regulations (FCC, FTZ, BSI, etc.), or by industry standards (IEEE, etc.). Therefore, an understanding of these transients and protection against these transients must be provided by the manufacturers of equipment that shares the power mains. UPS Transient Example The larger the load on the UPS, the larger the main transient as the transient source power impedance is reduced. Unfortunately, the lighter (or smaller power) the load sharing the power mains with the UPS, the larger the UPS transient will be applied to the light load. Consider a full load on a three phase 208 V on-line synthesized 12 step 80 kW UPS. The measured transients on a shared 120 V line are shown in Figure 2. One should not conclude that this type of UPS is the worst case transient penerator, rather it is a typical UPS design. A 100 V spike at 400 US width is shown in Fieure 2. As nointed out in the ‘UPS synopsis, the UPS- steps are hot at 30’ or 1.39 ms apart since the UPS is under full load. The point to be made is that the power main transient can be at any phase position on the shared power main waveform. Since transients have a Fourier spectrum, we can analyze the transient by the harmonic content. The harmonic amplitudes up to the 13th harmonic are shown in Figure 3 for a 50% load on this UPS. The 11th harmonic is about 4 V peak with the others ranging from 1.4 to 2.6 V (fifth harmonic). The amplitudes of the harmonics are not reduced at higher frequencies. There is a define band of harmonics that repeat and last until 35 kHz as shown - in Figure 4. The sum of all these harmonics creates a transient of 50 V amplitude and 120 ps duration on the shared power mains. 403 73Ml - Effects of UPS Transients Nearly all power main monitors will not detect UPS transients, yet circuits in commercial digital equipment are susceptible to these transients. The EM1 filters commonly used to meet EM1 regulations (10 kHz and up) give no protection. Although it is known that the addition of inductors and tuned circuits at the inputs to the UPS rectifier reduces the feedback of these transients to the input AC mains, rarely can the shared power main commercial manufacturer request this added feature due to cost and efficiency. The best UPS transient protection is by internal equipment cost effective designs because tracking power main filters or CVTs have limitations and most peak voltage limiting surge suppressors such as MOVs semiconductor clamps, etc., give no protection against these low amplitude disturbances. CVT TRANSIENTS Figure Input power main transients from a 12step UPS at full load. Note that the steps are not exactly 30° apart. 2: 10 DBIDIV It was pointed out before that CVT transformers are used in the design of UPS, for the control of UPS transients and are excellent suppressor surges ( 3). As in all transformers having loads (4 ), there is an inrush current with CVT that can be up to 30 times the normal peak current with rise times the range of microseconds. Observe in Figure 5 that the maximum inrush current (10 times normal) did not occur at the peak of the input waveform, but rather near zero volts. In fact, the inrush current is only limited by the impedance of the service and the DC resistance of the CVT primary winding if the inrush current adds flux to the CVT. Thus, if there are power outages lasting for only a few cycles, one should be very careful when using CTV’s. Power outages exceeding minutes do not create such transients as the trapped flux in the CVT is removed. t AC MAINS CURRENT I I I I 1 Figure 3: 3 5 I 7 _ HARMONIC NUMBER Input power main spectrum from a 12step UPS at 50% load. Of the odd harmonics, the 11th is the largest. AC MAINS VOLTAGE IZOV (RMS) Figure 5: Inrush current of a CVT. tbe large current transient zero volt crossing. Observe occurred that near 10 DSlDlV CRT FLASHOVER 5 KHZ/W! i- Figure 4: The 12-step UPS has spectrum to 35 lcHz with the harmonics shown in Figure 3 repeating every 720 hertz. Many products need high voltage for their operation: copiers, illuminators, Cathode Ray Tubes (CRT) in TV sets and monitors, etc. One theory that explains the CRT internal flashover is that the high potential on the inside of the neckglass creates field emissions from the metal structure of the CRT gun (5 ). Then there is an electron avalanche along the neckglass that makes gas accumulated on the glass surfaces due to electron stimulated desorption. - The ionization in the desorbed gas and ion feedback causes a run away condition. This condition forms a plasma that creates breakdown, followed by a cleanup that restores normal conditions. The occurrence of high voltage flashovers in CRT has been known for despite rigid quality control many years yet, procedures and new inventive designs, the flashover has a finite probability. The flashovers occur at random intervals well separated in time. Because the intervals between flashovers become larger as the life of the CRT increases, the flashovers are rarely noticed by the viewer until the repeated flashovers create a failure. With almost a billion of CRT equipservice with other sharing the power ments electronic equipment, a major concern is the power mains transients induced by CRT flashover which can There can also be be as high as 2500 volts (1). radiated and conducted interference to equipment interfacing with the CRT circuitry, for example, computers interfacing with CRT monitors. Even with arc protection in the CRT circuitry, there can be castastrophic failures in the interfacing equipment. Model of Main Arc Current Path The positive high voltage is connected by the picture tube ultar or anode button to the internal coating as shown in Figure 6. The high voltage return is connected to an external tube coating called AQUADAG. The capacitance (Cl) formed between the internal and external coatings on the funnel with the glass serving as the dielectric, is about 1 to 2.5 mF. The distributed capacitance between the yoke mounted on the tube neck and the internal coating (C,) is about 10% of Cl. With the yoke return common to the AQUADAG, the total capacitance increases with the size of the CRT. The high voltage (15 to 20 kV for monochrome and 25 to 35 kV for color) creates a charge Q=CV between 20 to 63 PC and an energy (J = l/2 CQ) as high as 65 joules. The distributed resistance of the internal coating is about 20 ohms over the entire length so one can assume the effective resistance in the arc model is 10 ohms. Many newer designs add a 400 ohm series resistance (discrete or distributed in the coating) to limit the peak and rate of the rise of the current in the arc. 404 - When an arc occurs in the CRT, the current flows through the elements of the CRT gun to the pins of the CRT. Then arc current leaves the pins hunting a return path to the AQUADAG. The inductance of the path inside the CRT is about 0.4/.1H. The inductance of the arc from the CRT pins to the AQUADAG con-nection depends upon the length of the arc return path which is longer than the length of the neck of the CRT. A typical inductance for the external path is about 0.3 PH. The arc loop area will determine the radiation but the unavoidable external arc return will create a major conduction source. The main arc path can be modeled as a series RLC circuit with the primary sparkgap G. The primary sparkgap voltage drop becomes a low value after breakdown and has little effect on the arc current. As the energy stored in the capacitor CT (the sum of Cl and C2) is dissipated in the loop series resistance, a large current flows as shown in Figure 7. This arc current (100 to 800 A), as in all sparkgaps, will continue to flow until the current is below the extinction value. By the time extinction occurs, the original stored energy in capacitor C is reduced to almost zero. With no high voltage on 3 he CRT, the CRT becomes blank and acts like a CRT during initial power turn on. With a large arc current having a nanosecond rise time (50 ns in Figure 7), large voltages are developed across portions of the loop inductance. With large voltages being developed around the arc loop, secondary sparkgaps are encountered. If these secondary sparkgaps are incidental, equipment failure with repe.ited discharges is most likely. Thus, controlled secondary sparkgaps having standoff voltages \rmsh less than the CRT high voltage are used for predicable performance. These secondary sparkgaps must handle large currents. For example, with a cathode arc, the beam current normally 60 mA peak, jumps to 480 A for about 200 ns and decays to the normal value in about 15 c(s. This effect creates CRT Bower supply transients 60 AMPSlDlV Figure 7: The CRT arc current begins to ring 240 A peak with a 50 ns rise time. at YOKE ANODE ANODE BUTTON H.V. SOURCE Figure 6: Dispersion BUTTON 7 ELECTRON GUN R 2 R 3 1 Circuit model of the during CRT flashover. main current path of The Arc Current With the controller video, sync and other low level signals referenced to logic ground, normal CRT connections tie the controller ground to the CRT drive ground. With high bandwidth requirements for the video signal, CRT cathode drive as shown in Figure 8 is the commonly used connection. We see that a suppressor for secondary arcs is used to protect the cathode drive circuit when the main arc current (I ) returns to the AQUADAG connection. The differynce in voltage between point A and point B even with a wire or briad having very low inductance, during the flow of the main arc current flow is about 800 V. This voltage allows about 10% of the - 405 400 A main arc current to escape. There are many paths for this escape current (I,& to get back to the AQUADAG. Normally, shielde cables are used to control radiated emissions and the cable shield is connected to the enclosure of the CRT or frame. If the reference or ground of the CRT circuitry is connected to the CRT frame, then some arc current (ISH) flows through the shields of the cable. The stray capacitance of the cable to the CRT frame or controller frame creates another current path (ICl or IC2). With well filtered controller voltages, all controller voltage lines are a path for the escaped arc current. Often an EM1 filter on the DC power lines is used to control emission and this filter is a major path for the arc current. Stray or leakage capacitance creates another arc current path through the AC mains. This arc current (IS) can return through the AC safety wire as shown m Figure 8 or through the DC power return if the power supply of the CRT is not part of the circuitry. These multiple current paths create problems as escape current divides. In a typical case somewhere between 10 to 100 V is developed across a critical part which is upsetted and sometimes fails. Downstream current blocks are needed to control the escape current to less than l/Z% of the main arc current. Reducing The Escaping Current Since the main problem is flow of current, rerouting or the placement of cables has little effect. Furthermore, a single point system ground does not block the multiple shared arc current loops. The escape path impedance is complex with typical values between 20 to 100 ohms at the ringing period near AQUADAG 73Ml - 300 ns. Low capacitance diode clamps and series resistance can be used to block the current in the signal lines in the interconnect cables leaving the cable shields, DC power leads and logic grounds to be controlled. One begins to reduce the escape current by noting that the inductance of the cables between the CRT circuitry and the CRT interface (like a controller) has a major influence. The larger this inductance is, the smaller is the escaping current. Hence, baluns and ferrite beads for the cables are very helpful as any series inductance in the ground or the shield degrades the system performance. To be an effective block, the series impedance must be near 500 ohms, not the typical 1 to 5 ohm impedance of the cable shield inductance at the ringing frequency. As the escaped arc current leaves the logic ground, shields or DC power leads, the path can be modeled as a current divider with the complex impedances in each branch, The DC power and AC safety leads can be blocked with a series inductance in each branch which does not resonate below 25 MHz. Large inductors like 500 PH having stray capacitance of 20 to 50 pF convert the arc current ringing in the low MHz frequencies to problems at 5 to 15 MHz frequencies. Long leads on the inductors or poor placement of the inductors allow the escaping arc current to radiate to the critical interface circuitry. As each low impedance path is blocked, a side effect takes place - the escape path impedance is being raised in the current dividers. This means that more applied voltage will be presented to the impedance paths and hence more arc medium AC FRpME CONNECTION FILTER / \ / I ~,-j+‘-• B ISH f I,@ f-1'""'""";" CRT \ STRAY CAPACITANCE IN POWER SUPPLY +--- i 4-v RF, FILTER IF USED i I SPARK GAP, DRIVE CIRCUITS IEX A * I I lls ( -- LOGIC GROUND * 1 CRT MONITOR AC SAFETY WIRE IF USED ) Figure 8: f-DC The complex arc current escape path. Blockage of each path requires different solutions. CONTROLLER BYPASS - current. Changing the layout by impedance between the arc current critical circuitry is very helpful. creating more path and the 406 - REFERENCES W.T. Rhoades, mercial Power Standard”, pp. posium. 2. W.T. Rhoades, “Development of Power Main Transient Protection for Commercial Equipment”, pp. 235-244, 1980 IEEE EMC Symposium. 3. “The Propagation and AttenuaF.D. Martzoff, tion of Surge Voltages and Surge Currents in Low Voltage AC Circuits”, pp. 1163-1170, May 1983 IEEE Transactions on Power Apparatus and Systems, Vol PAS-102, No. 5. 4. A.C. Franklin and D.P. Franklin, “The Book”, pp. Transformer 548-562, Butterworth, Boston, Mass. CONCLUSIONS The characteristics of three unusual power main transients have been given. On-line UPS transients are continous; can occur at any phase angle of the power main waveform, have up to 100 V amplitude, last up to 600~s and have spectrum up to 35 kHz. The inrush current of CVTs is only limited by the power service; and has /,& rise/fall times at 10 to 100 times the normal current. The final transient described was the CRT flashover; not only can a 2500 volt transient be injected in the power mains, there can be major transients from the arc in the CRT How the arc current is interfacing equipment. dispersed and some unique ways to reduce the escaping arc current are given. “The Ratiocination of a ComMain Conducted Susceptibility 269-276, 1981 IEEE EMC Sym- 1. 5. J and P 1983, K.G. Hernquist, “Studies of Flashovers and Preventive Measures for Kinescope Guns”, pp. 117128, IEEE Transactions on Consumer Electronics, May 1981, Vol. CE-27, No. 2. - 407 7h2 - THE DEVELOPMENT OF AN IEEE GUIDE ON SURGE TESTING FOR EQUIPMENT CONNECTED TO LOW-VOLTAGE AC POWER CIRCUITS Fraqois D. Martzloff Corporate Research and Development General Electric Company Schenectady, NY 12345 INTRODUCTION IEEE Std 587-1980, Guide on Surge Voltages in LowVoltage AC Power Circuits, now designated ANSI/IEEE Std C62.41-1980, was published after several years of preparation by a Working Group of the Surge Protective Devices Committee of the IEEE [1,2]. The same Working Group is now preparing a new Guide on Surge Testing. This document has been refined and is now close to achieving consensus; only a few items remain to be finalized. However, final approval and publication by IEEE are not expected for some time. Since the “Guide on Surge Testing” is based on information contained in IEEE Std 587-1980 and IEC Report 664-1980 131, familiarity with these two documents is almost a prerequisite to complete appreciation of the issues discussed in this paper. However, the overview of the issues presented here will be useful to individuals concerned with the various aspects of surge protection and surge testing. BACKGROUND Standards published before IEEE Std 587-1980 were generally influenced by traditional dielectric test concepts or special equipment requirements. For instance, some military standards [4,5] required voltage tests without specifying source impedance; the widely overused SWC test [61 specified realistic test conditions and procedures, but only for the limited field of solid-state relays in high-voltage substations; IEC Report 664-1980 addressed only the voltage aspects of surge occurrences because it is primarily concerned with insulation coordination. The publication of IEEE Std 587-1980 introduced several new concepts in the field of surge occurrences in low-voltage ac power circuits. The development and wide acceptance of silicon avalanche diodes and metal oxide varistors as low-voltage surge protective devices had made the surge current an important new factor in the design of surge protection schemes. Their increasing use motivated the writers of IEEE Std 587-1980 to describe in a systematic way current surges in the environment of ac power systems. Hence, two circuit conditions were defined in IEEE Std 587-1980: for lowimpedance circuits, a current surge is the relevant parameter; for high-impedance circuits, a voltage surge is the relevant parameter. Furthermore, IEEE Std 5871980 introduced the definition of three location categories. In these categories, the amplitudes of current surges decrease from the outdoor environment toward the interior points of the wiring system within a building, in contrast to the amplitudes of voltage surges, which do not decrease in the case of a low load on the system. The primary purpose of IEEE Std 587-1980 is to provide a description of the environment, including the statistical and exposure-dependent nature of surge occurrences. Warnings are provided on the pitfalls of “worst case” approaches in defining the environment. Furthermore, a specification for any required level of withstand capability of hardware is carefully avoided. Nevertheless, equipment specifications are appearing in the trade, where such statements are made as “. . . meets the requirements of IEEE Std 587” (italics added). Such misuse of IEEE Std 587-1980 was one of the factors that motivated the writing of a guide on surge testing, in anticipation of publication by IEEE of a comprehensive document on low-voltage surge protective devices and their application, in the late 1980s. CONTROVERSIAL ISSUES While the general approach and contents .of the “Guide on Surge Testing” were readily accepted by committee members and reviewers, three issues were somewhat controversial during the process of writing the text: uncontrolled 1. The concept of the unprotected, environment given in IEEE Std 587-1980 and the concept of the orderly voltage staircase of IEC 6641980 cannot be interchanged or used concurrently; however, reconciliation can be accomplished by coordinated expansion and clarification of the texts in the two documents. The “Guide on Surge Testing” can be a vehicle for such reconciliation. In contrast with conventional high-voltage practice, where separate tests are conducted for voltage impulses and current impulses, electronic equipment is best evaluated with a surge test system inherently capable of delivering either a voltage wave or a current wave, or both in succession (but the upon depending simultaneously), not impedance of the test piece. Guidance on how to select appropriate withstand levels for generic types of equipment was difficult ” ,. THE ANS,,,EEE STD C62.41-1960 CONCEPT OF LOCATION CATEGORIES . . LOCATlON CATEGORY C “Ou,s,de and Service Entrance” . : : wa,or Feeders and Short Branch Clrouits” . - IN UNPROTECTED : . . LOCATlON;ATEGORY l 408 CIRCUITS LOCATION CATEGORY A “Outlets B”d Lang Branch Cl,C”IW : VOLTAGES 3 kA CURRENTS 2. TYPICAL : B kV lmp”lS* 0, fling 10 kA 0, more EXAMPLES OF INDUSTRIAL Impulse500 A Ring OR RESIDENTIAL 6 k” Ring . CIRCUITS Alternate: ““dergraund cable S*r”iC* LEGEND: t&v: Arc welding supply M: Watt-ho”, mete, FA: Fixedappliance PI: aurge B,,BS,B, P2 IO: fv: MB: t? SE: (secondary rating) surge a,,es,8, (secondary rating, Wl3,: 3. THE IEC REPORT 664-1960 CONCEPT industrial drive system D,i”B” motor Main breaker Transbent p,otec.to, Service entrance (may take many forms depending on specific ca2.B Of syste”?) VW receptacle WithoUt atte”“atio” OF CONTROLLED CA: Co,d.co”“ected appliance + COMP:computerWithbufferedinput KS: industrial control system Lc: Line power conditioner P3: surge ponector WI32 wall ,eceptkx With attenuation provided by: Z-swies impedance C-shunt impedance Hi=:Co”s”me,&C1,0”iC9 P4. surge p,otector PC: PerSOnal compute, TV: T,a”sfo,me,-isolated &%t,O”iCS VOLT+GES Ilnmntmlld Independently from the location of a device or equipment in the above figure, it should remain safe (no fires, no personnel hazard) over the full range of available surges at any point within the installation. It may also be desirable, under particular circumstances and for specific devices, to proscribe damage as a result of testing at higher levels than might be suggested by its typical location. Notes: (1)The Controlled Voltage Situation of IEC Report 664 requires the presence of interfaces; these can be surge protective devices such as PI, P2, P3 or P4, or the existence of well-defined impedance networks such as Z and C shown in the circuit diagram upstream of WR2. (2)Voltage levels following the designation of Installation Category (IV, III, II or I) are shown in parentheses for a system with 300 V phase-to-ground voltage, and next for 150 V phase-to-ground voltage. The voltages shown are implied as 1.2/50 p.s impulses. (3)This diagram makes no allowance for the possibility of surges associated with ground potential differences that may occur, for instance, with a sensor connection to the KS control system, a cable TV connection to the line-isolated TV set, etc., or the flow of ground current in the impedance of the grounding conductors. Figure 1. Similarities C62.41-1980 and differences and Installation between the concepts of Location Categories in IEC Report 664-1980 Categories in ANSI/IEEE - to define. The controversy was as much the question of whether or not guidance is desirable at all as the question of what format would discourage such guidance from being lifted out of context and considered to be specifications. In the final draft, a table showing examples of withstand levels was deleted because of the concern for misapplication of the table. Concern was also expressed that specific reference to ANSI/IEEE C62.41 (formerly IEEE Std 587) and IEC 664 documents might imply a restriction of the “Guide’s” applicability to a limited set of surge waveforms. The final version of the “Guide” is expected to include sufficient information to avoid this limitation. The first two of these issues are discussed below in more detail. 1. IEEE 587-1980 and IEC 664-1980 IEEE Std 587-1980 defines “Location Categories” to acknowledge different current, voltage, and energy levels in a low-voltage system, while IEC 664-1980 defines “Installation Categories”* to acknowledge different voltage levels in such a system, since its prime concern is insulation coordination. There are both similarities and differences between the IEEE Std 587-1980 concept and the IEC 664-1980 concept. Both recognize the decreasing severity of the surge environment from outdoors to the inner recesses of a building. The differences originate from the fact that IEEE Std 587-1980 describes a condition where existing, uncoordinated clearances and solid insulation determine the maximum overvoltages that can occur but where the voltage breakdown levels of these clearances and solid insulation are unknown and uncontrolled. IEC Report 664-1980 addresses primarily a coordinated insulation system where the voltage breakdown levels of clearances and solid insulation are known and controlled. Figure 1, excerpted from the “Guide on Surge Testing,” provides a graphic illustration of the similarities and differences. 2. Current/Voltage Testing Versus Separate Tests The separation of surge tests has long been used in the utility industry, where the need for surge testing was first identified. For instance, evaluation of the dielectric strength of transformer insulation can be performed satisfactorily with tests involving only voltage impulses. Valve-type arresters, on the other hand, require two tests, a voltage test to characterize the front-of-wave sparkover and a current test to characterize the discharge voltage. These two tests are performed separately, and two distinct surge generators can be used to apply the test waveform. The situation is different when tests are performed on electronic equipment where the outcome of the test is not merely’ pass or fail, as in the case of insulation test of a transformer or a simple description of the two characteristics of a surge arrester. 409 - 74~2 is unknown, two separate tests might seem appropriate and sufficient: first, a voltage surge, then a current surge. However, the separation of the two tests could yield misleading results. If, for example, tests are conducted on equipment whose behavior is not completely defined, the outcome of the test can depend on changes of impedance occurring during the test, so that a prior commitment to performing either a voltage or a current test has the risk of missing the actual outcome. Two examples are given below, and the literature [7,81 provides further insight into this pitfall. As a first example, consider the testing of a protection scheme that uses a spark gap combined with a varistor. If the test plan is to apply a separate voltage surge - quite possibly from a high-impedance surge generator - then the gap sparkover will be observed and the test operator might be satisfied that he has characterized the gap behavior. However, the effect of the gap’s abrupt switching to high current upon sparkover will not have been observed. Moreover, a highimpedance generator will produce a low rate of current rise following gap sparkover. The preferred lowimpedance generator, on the other hand, will produce a high rate of current rise. Such a high rate of current rise can induce additional voltages that will be impressed on the downstream circuit. Only a test conducted with a generator inherently capable of delivering a high current as soon as the test piece changes from high to low impedance will disclose this subtle behavior. As another example, consider the case of a test where the failure mode and effect are the desired test criteria. If insulation breakdown occurs during a voltage test applied from a high-impedance surge generator, the subsequent current will be too low to produce the effect that would result from an actual lightning surge impinging on the equipment. Thus, generators ing either test rather there is a need to perform surge testing with that have the inherent capability of delivervoltage or current during the surge, in one than in separate tests. MAJOR TOPICS DISCUSSED Planning of Surge Testing Surge testing is generally performed to determine the surge withstand capability of specific equipment. In such a case, the first decision to be made is to reach agreement on the nature of the surge environment for this equipment. The major concerns in the “Guide” are switching surges and lightning-induced surges. Surges associated with nuclear electromagnetic pulse and high-frequency noise are the subject of other documents. IEEE Std 587-1980 points out the duality of current versus voltage surge testing because of dependency on the impedance exhibited by the equipment connected to the power circuit. Consequently, when a test is to be performed in the laboratory with a surge generator, one might think of selecting either a voltage surge or a current surge, depending on prior knowledge of the equipment to be tested; or, where the equipment impedance It is important to differentiate between design tests and diagnostic tests, on the one hand, and production and qualification tests, on the other hand. In the first, the limits of withstand are sought and tests are conducted until failure; in the latter, a specified level is applied with the expectation that no failure will occur. Thus, experience has shown that the expected results or consequences of a test need to be defined before the test is performed. * The term “Installation Categories” used in the 1980 issue of Report 664 will be supplanted by “Overvoltage Categories” in later IEC publications. Figure 2 illustrates the considerations leading to the selection of the appropriate surge generator and monitoring instrumentation. - 410 - r Purpose of Test Nature of EUT l l l , l l l l Surge AC Interface Surge Other Points’ Surge Component Terminals Environment Selected Failure Criterion (6) Possible Outcomes Powered Unknown EUT Impedance High EUT Impedance Low EUT Impedance EUT With Impedance that Changes Duiing the Test - I /I Figure 2. I I \ Type of 1 I Surge Generator Monitoring the EUT Relationships between the Equipment Under Test (EUT) requirements and the selection of the test equipment Powered Versus Unpowered Testing In powered testing the surge is applied to the test piece while the test piece is connected to its normal industrial-frequency power supply. This powered testing is more complicated than unpowered testing, where the test piece is subjected only to the surge, as an isolated component. However, in some cases useful measurements cannot be gained without full simulation of the actual power conditions with the test surges superimposed: 1. If the testing goal is to evaluate the effect of the surge on the normal operation of the equipment. For instance, a disturbance caused by a surge test on data processing equipment can only be ascertained if this equipment is in fact engaged in data processing at the time of surge testing (connection to its normal power supply is implied). 2. If the testing goal is to evaluate the effects of powerfollow current. Power-follow current is the current flowing in the test piece from the industrial-frequency power source, after the test surge has initiated a low-impedance path in the test piece. Examples are the intended sparkover of a gap valve arrester or the failure of a component. A power-follow current equal to that which would occur in an actual application can only be obtained if the test piece is powered by an industrial-frequency source capable of delivering that current. 3. If the testing goal is to evaluate the effect of energy being deposited in the device. For instance, a surge of large amplitude and long duration might deposit enough energy in a varistor to heat it up to the point of initiating thermal runaway if the steady-state voltage rating of the varistor had been selected too low. An unpowered test cannot disclose that condition. Back Filters Superimposition of the test surge to the normal power supply can be accomplished by a series connection or a shunt connection of the surge generator (Figure 3). Because the design of the transformer used in series coupling limits flexibility in specifying the test surges, the shunt connection is generally preferred. However, in a typical laboratory application, two limitations are imposed on the application of a surge in shunt to the power system: 1. The surge must not affect other equipment connected to the same power supply. Decoupling is required between the test circuit and the industrial-frequency power source. 2. The generator cannot be connected directly in shunt across the power supply. Because the industrial-frequency power source has a low impedance, this low impedance would load the surge generator to the point that it could not deliver the specified voltage at the terminals of the test piece. Both of these conditions can be achieved by the insertion of a filter between the industrial-frequency power source and the test circuit. This filter is referred to as - 411 - 74~2 Its presence in the test circuit has “back filter.” definite effects on the overall behavior of the circuit and on the waveforms that can be obtained in feasible test circuits 191. -T Monitoring the Equipment Under Test :yi=JEuT T = Surge Coupling BP = Bypass Filter (A) Monitoring the input surge is required to verify the characteristics of the applied surge. Monitoring within the equipment during surge application can clarify failure mechanisms. Finally, monitoring at the output, if applicable, can provide data on the risks of passing on a surge to downstream equipment. At the minimum, voltages should be monitored. Monitoring for current is an effective method of detecting sparkovers, clamping effects, or other circuit behaviors which might not be obvious from an inspection limited to recordings of voltages. Figure 4 shows one example of the monitoring method discussed in the “Guide.” Transformer Series Coupling Example Coupling the Surge C = Surge Coupling BF = Back Filter (B) Figure 3. Capacitor Shunt Coupling Example Series coupling and shunt coupling of the surge generator to the Equipment Under Test (EUT) For power supply lines which typically involve more than two conductors (line, neutral, and grounding conductor on single-phase circuits, for instance), two different coupling modes are defined. Those where the grounding inductor is connected to the surge generator are referred to as “common mode”; those where it is not, as “normal mode.” Depending on the nature of the equipment, one may be a greater threat than the other, so that testing for both is a prudent approach. Table 1 shows an example of the variety of possible coupling modes for a single-phase circuit. The “Guide” also addresses polyphase circuits. Safety Because the voltage and energy levels involved in surge testing are inherently hazardous, the “Guide” places considerable emphasis on the safety aspects of test procedures. Minimize area cross-hatched in order to avoid contaminating measurement by voltage induced in probe loop \ CR0 frame should not touch any other frames Line AC Power Neutral Equipment Grounding Conductor Grounding conductor of generator Figure 4. Isolate or disconnect All other connections to EUT of oscilloscope *Or suitable differential probe/amplifier Monitoring within surged equipment with voltage probes in differential connection - 412 - Table 1 SELECTED COUPLING MODES FOR SINGLE-PHASE (One line and neutral with grounding conductor) Test ( Connection Type Ground of Generator Normal L HG HN HH = = = = L Common L Common L Common HG Common HG Common connection connection connection connection to to to to surge surge surge surge generator generator generator generator Diagram for Normal Coupling Mode Basic Diagnostic Example of Connection Coupling Line Mode .n/ low (Lo) high (Hi) by coupling capacitor CG high (Hi) by coupling capacitor CN high (Hi) by coupling capacitor CL For each test type shown horizontally in the table, the surge generator is to be connected as indicated in the three “Connection of Generator” columns. The connection diagram in the table shows as an example the jumpers required to obtain the “normal” coupling mode. When several H’s appear on one horizontal row of the table, the coupling requires several coupling capacitors, shown as CG, CN, CL, between each of the conductors indicated and the surge generator high, in order to apply the surge simultaneously to the conductors shown. CONCLUSION When published, the “Guide” will provide valuable guidance for performing surge tests on a technically sound basis, with repeatable results, and in a manner reflecting the actual exposure of equipment to surges occurring in service. The advanced information and perspective provided here should be helpful for application of the guidance compiled in the forthcoming “Guide on Surge Testing.” If a one-liner may be used to summarize the goals of the “Guide,” it is: Don’t kid youme& don’t kill yourself! ACKNOWLEDGMENTS The concepts presented in this paper result from a 3-year effort of the IEEE Working Group on Surge Characterization, composed of over 30 members who contributed their technical and practical experience in planning, performing and evaluating surge tests. C.L. Fisher provided advice in clarifying and unifying the presentation of the concepts, in the “Guide” as well as in this paper. REFERENCES 1. Martzloff, F.D., “Guideline on Surge Voltages in AC Power Circuits Rated up to 600 V,” Proceedings of the Third Symposium on EMC, Rotterdam, 1979, pp. 449-454. Std C62.41-1980, Guide on Surge Voltages in Low- Voltage AC Power Circuits. 2. ANSI/IEEE Report 664-1980, Insulation Coordination Within Low- Voltage Systems, Including Clearances and Creepage Distances for Equipment. 3. IEC 4. MIL STD 461B, Electromagnetic ceptibility Requirements netic Znterference. Emission and Susfor the Control of Electromag- 5. MIL STD 1399, Interface Systems, Section 103. 6. ANSI/IEEE Standard C37.90a-1974, stand Capability @WC) Guide for for Shipboard Surge With- Tests. Voltage and Current 7. Richman, P., “Single-Output Generation for Testing Electronic Systems,” Proceedings of the ZqEE EMC Symposium, August 1983, pp. 47-51. ftir die Isolation8. Wiesinger, J., “Hybrid-Generator ET2 104 (21), 1983, pp. 1102skoordination,” 1105. 9. Richman, P., “Changes to Classic Surge Test Waves Required by Back-Filters and for Testing Powered Equipment,” Proceedings of the Sixth Symposium on EMC, Zurich, 1985. - 413 75 M3 - CHANGES TO CLASSIC SURGE-TEST WAVES REQUIRED BY BACK-FILTERS USED FOR TESTING POWERED EQUIPMENT P. Richman KeyTek Instrument Corporation Burlington, Massachusetts, U.S.A. Realistic surge testing of modern electronic equipment supplied from the ac power mains requires two significant departures from prior surge test tradition. The first, already described elsewhere (1, 2, 3), is the change to a Combination Voltage/Current impulse to replace separate voltage and current impulse-wave testing. The second change results from the back-filter required when doing powered testing, both with normal equipment power-line current, and with the high, so-called follow current that flows from the power line following either flashover, or equally likely, the crowbarring action of a protector. In either case, the back-filter can bring about changes in the character of the classic, 1.2/50 and 8/20 waves used for surge testing powered equipment. These changes have been accepted for some years on a de facto, yet not fully quantified basis, in laboratories doing such tests. Results of surge testing can,therefore vary. Modified surge test waves are pi?OPOSed, based on the same logic that underlies the derivation of the Combination V/I wave. That is, the new waves are designed to fully meet the needs of modern, powered surge testing, while remaining as faithful as possible to traditional waves for test Situations not requiring line power. Results of testing with the newer waves on unpowered circuits should be comparable with results obtained using traditional waves. The crucial point is that this should hold true for all relevant loads including even the newest, most complex protectors; incorporating as they may, crowbar devices, clamps, and most recently, inductors and/or capacitors. Introduction It has already been shown to be far preferable to perform modern surge tests on electronic equipment using a Combination Voltage/Current surge test impulse wave, with the inherent capability to deliver a 1.2/50 voltage wave across a high-impedance test piece, and an 8/20 current wave into a low-impedance test piece (1, 2, 3). The input impedance of the Equipment Under Test (EUT) is generally unknown; it may include one or more simple or complex, clamping and/or crowbarring surge arresters, and it may or may not incorporate auxiliary L-C filters. In addition, there is always the everpresent possibility of flashover. The complexity of this input impedance has been the justification for changing from separate V and I testing to the new, Combination V/I wave. Of greatest importance is the wave's inherent capability to simulate nature. It does so by changing automatically and instantaneously from 1.2/50 voltage to B/20 current and even, if necessary, back again, as required by the instantaneous input impedance found by the test wave at the input of‘the equipment being surged. In addition to the change from separate V and I tests to use of the Combination V/I wave as described above, modern surge testing has introduced the need for a second major departure from the tradition of the last half century. The new problem which traditional surge test waves cannot adequately address without alteration, is the need to surge equipment powered by the ac line. The Surge Back-Filter The requirement for powered surge testing leads in turn to the need for a surge back-filter between the power line and the EUT, One purpose of the back-filter is to isolate the test surge from other equipment that may be powered by the same line. Another is to provide the surge generator an impedance to drive. Without the filter, the test generator would have to develop a surge across the extremely low impedance of the unfiltered line. Note that even if such a powerful surge - 414 - generator were available, one capable of surging an isolated power source like an MG set without a back-filter, for example, it would still not be satisfactory. The reason is that such a super generator would inevitably supply tens of KA in short-circuit mode; far greater than the hundreds or just a few thousands of amperes of short-circuit surge current required by modern Guides and Standards (1). The back-filter usually consists of, or at least includes, inductors in series with two or more of the input power lines. Fig. 1 shows the basic configuration. The inductors are difficult loads across which to develop the exact, classic, unidirectional l.Z/5O open-circuit voltage impulse, with or without surge coupling capacitors. The primary reason is that the inductors must pass normal ac line current to operate the EUT, with minimum voltage drop. In some cases, they must pass current due to power follow as well. These requirements imply low inductance values. However, while it is possible to deliver long, high-energy, essentially unidirectional impulse waves across lowvalue inductors, it is difficult to do so and still deliver some of the lowervalued (3kA, IkA, etc.) 8/20 short-circuit currents. The powerful, low-impedance circuits necessary to supply long-duration voltage impulse waves to low-value inductors, will tend to furnish short-circuit currents that are far too large and too long. This paradox and a proposed solution are dealt with later. The need to do powered testing has followed closely the recent recognition already referenced, that a Combination V/I wave is necessary for testing mod- ern protectors. That recognition is based on the sensitivity of protectors, unpremeditated flashover, and surge remnant or coupling that reaches downstream equipment, to: 1. dV/dt of the leading edge of the open-circuit voltage surge, and 2. energy and general waveshape of the short-circuit current surge. Fig. 2 shows the open-circuit voltage (Fig. 2a) and short-circuit current (Fig. 2b) supplied by an ideal test surge generator which delivers the Combination V/I Wave (2), for un-powered testing. Fig. 2a: Classic 1.2/50 Open-Circuit Voltage, lkV/.5cm, lOus/.5cm Fig. 2b: Classic 8/20 Open-C rcuit Current, 500A/.5cm, 5ps/.5cm SIMPLIFIED POWER-LINE SURGE BACK-FILTER I I HIGH LF AC LINE nci CF Fig. 1: TYPICAL SURGECOUPLING CAPACITOR I CF- I BACK-FILTER INDUCTORS / 1 1 EUT SURGE GENERATOR I ’ Typical Surge-Generator Connection to an EUT Driven by an AC Power Line 75 415 - M3 Open-Circuit Voltage (OCV) It is worth noting that "open-circuit voltage", as the expression is conventionally used in surge testing, refers to the fact that the surge generator is unloaded, i.e. its output is opencircuited. In other words, the EUT (Equipment Under Test) is disconnected. The back-filter needed for powered surge testing, however, cannot be disconnected from the surge generator. From the standpoint --Lof the EUT the back-filter --_is part of the surge generator; so it must remain connectedwhen discussing the "open-circuit" surge voltage waves. Figs. 3a and 3b show a typical "opencircuit" voltage wave obtained when using a back-filter, when surging powered equipment. The open-circuit voltage wave duration happens to be somewhat longer than 50 us. More important, it exhibits significant "undershoot" also termed "ringback" or "backswing ?I; i.e., it starts to ring below the baseline. The wave is, in effect, underdamped. It can be referred to as a differentiated impulse, using nomenclature already introduced elsewhere (4). Fig. 3c shows a typical short-circuit current wave that results from the same test configuration. It is the same as the classic 8/20 impulse, except for a decrease in duration due to the coupling capacitor, and a small undershoot due both to the coupling capacitor and to added inductance in the cable required to reach the filter. Realistic test waves in laboratory surging of powered equipment have for years looked far more like the waves of Fig. 3 than those of Fig. 2. The key question is whether such waves are satisfactory for testing complex modern protectors and the electronic equipment they are designed to protect. In this connection, it is particularly important to note that for such protectors, the back-filters will seldom be completely shorted, as they are in short-circuit current mode. Before proceeding, it is first necessary to recognize that for any backfilter application, the open-circuit voltage impulse must undershoot after return to the zero baseline. This follows from the fact that capacitance can't couple dc -- and inductance can't support dc. The total area of the wave above the zero baseline must therefore be equalled by the total area below the zero baseline, in order for the average to be zero. The only remaining question is the character of the undershoot. The two possible extremes are low amplitude/long duration, and high amplitude/short duration. Fig. 3a: --- Typical Open-Circuit Voltage Wave When Using Back-Filter; Approximately 1.2/5O;n-+30% Undershoot. oCv=6kv lkV/,5cm, lOus/.5cm Fig. 3b: Same Wave as Fig. 3a Except on Different Time Scale to Show- 30% Undershoot Caused by the Back-Filter. OCV=6kV lkV/.5cm, 50us/.5cm Fig. 3c: Short-Circuit Current Resulting From the Wave of Fig. 3a. oCV=6kv 500A/.5cm, 5ps/.5cm Clamp-Voltage and Clamp-Circuit Current With suitably-rated single clamps -metal-oxide varistors (MOV's) or appropriate strings of avalanche devices for example -- there may be no problem with even fairly large undershoots in opencircuit voltage. Coupler/filters 0r the kind shown in Fig. 1 yield acceptable short-circuit currents. The Combination V/I wave test surge generator that supplied the open-circuit voltage wave of Figs. 3a and 3b, and the short-circuit current wave of Fig. 3c, was used to test a 13OV MOV clamping arrester's performance with undershoots. The resulting approxi- - 416 mately 460V MOV clamping voltage wave is shown in Pig. 4a, while the resulting MOV current wave appears in Fig. 4b. Fig. 4c shows the same MOV current wave on a different time scale, to more fully exhibit the backswing. Note that the MOV voltage “re-engages” the clamp, in Fig. 4a, for a time that is much longer than duration of forward-current conduction. But current magnitude during this reverse-clamping interval is extremely small, as shown in Fig. 4c. Fig. 4d displays the same current wave as Fig. 4c, but enlarged to show how small the ring'back current actually is. As a result of the low amplitude of the ringback current, in spite of its duration the reverse energy, or current integral, is only a fraction of forward energy or current integral. Parenthetically, the total MOV clamp current wave is termed "clamp-circuit current", or CCI, to distinguish it from shortcircuit current, or SCI. It differs from the no-undershoot current that the ideal surge generator's waves of Fig. 2 would apply into a clamping arrester, because of the surge coupler-filter. For testing these simple arresters then, conventional surge coupler/filter designs are satisfactory. Problems can arise, however, when surge protectors or other circuits being surge tested involve the combination of one or more crowbars or clamps with one or more components that store energy: i.e., capacitors or inductors. For these more sophisticated combinations (and also for simple surge arresters that don't have on the order of 50% energy safety factor versus the "pure" test impulses of Figs. 2a and 2b), a coupler/ filter that can be used successfully to surge a single clamp, can supply significant and even destructive undershoot energy into the test piece. This can be particularly true for an arrester "buried" within a sophisticated, combination protector network. For testing such a protector network, the only tolerable voltage undershoot may be one whose peak is less than the minimum protector clamping voltage -- i.e., only slightly greater than peak ac line voltage, to insure small relative current undershoot energy. - Fig. 4a: Voltage Wave Across a Clamping Protector Clamping at 460~, for the Open-Circuit Voltage Wave of Fig. 3a and Short-Circuit Current Wave of Fig. 3c. oCv=6kV 200V/.5cm, 200 us/.5cm Fig. 4b: Current wave into the Protector of Fig. 4a. OCV=6kV 500A/.5cm, 5 ys/.5cm Fig. 4~: Current Wave into the Protector of Fig. 4a. OCV=6kV 500A/.5cm, 50 us/.5cm Thus for a 6kV surge, as required for high exposure areas (l), acceptable undershoot for a single 300V to 500V clamp may be as great as 1.5 to 2kV. However for more sophisticated, combination arresters, it may be only 200V or so, for a 120V rms line application. The principal issue is current undershoot into a possibly "buried" arrester, within the protector assembly's circuits. Forward and reverse surge energies for the clamp in Figs. 4a, 4b and 4d, integrate to about 17 and 5 joules respectively. Reverse/forward ratio is ~30%. Fig. 4d: Current Wave of Fig. 4c (partially offscreen). OCV=6kV 50A/.5cm, 50Fts/.5cm - 417 75M3 - Protectors and Undershoots Figs. 5a and 5b show the input and output voltage waves, respectively, for a more complex, four-terminal, combination protector, surge tested with the same surge generator-coupler/filter combination used to surge the MOV of Figs. 4a through 4d. Note the undershoots. It was necessary to run this test at significantly-reduced surge voltage (1.5 kV instead of 6kV) to avoid destroying an internal surge arrester. Note re-engaging of the output clamp in the reverse direction. Duration of the re-engagement is already several times the duration of clamp engagement in the forward direction, even at the 1.5 kV open-circuit setting that was employed. More important, conduction of the internal, shunt output clamp in the reverse direction is almost certainly at a high current level, unlike the MOV reverse conduction shown in Figs. 4b, 4c and 4d. This output clamp current can't be directly measured, since the clamp is "buried" within the protector assembly. However, the fact that the internal clamp fails when the input surge is raised to 6kV,which it doesn't do when surged on a generator without coupler/filter and therefore without backswing, would seem to imply that the reverse conduction current is indeed high with the coupler/filter in place. Fig. 5a: Input Voltage Across a Specific Four-Terminal Protector, Surged Using a Conventional Surge Coupler/Filter. OpenCiXUit Set Voltage=1.5kV lOOV/.5cm, 50us/.5cm Fig. 5b: Output Voltage from the FourTerminal Protector of Fig. 5a, Using the same Conventional Surge Coupler/Filter. OpenCircuit Set Voltage=1.5kV lOOV/.5cm, 50us/.5cm A New Coupler/Filter Design A new coupler/filter has been designed to significantly reduce undershoot. It is intended to be used with the same basic surge generator employed in all tests reported in the foregoing sections. Fig. 6a and 6b show the results of using the original surge generator and new coupler/filter combination to surge the same MOV whose waveforms using the original coupler/filter appear in Figs. 4a, 4b, 4c, and 4d. Fig. 6a, for voltage, is to be compared with Fig. 4a. Undershoot is negligible. Of even greater interest, the current undershoot displayed in Fig. 6b, to be compared with Fig. 4c, is negligible as well. Figs. 7a and 7b show input and output voltages, respectively, for the same complex protector whose performance was shown in Figs. 5a and 5b. However for Figs. 7a and 7b, the original surge generator was also combined with the new coupler/filter. This time, full 6kV/3kA stress application presented no problems; even at this level, the output clamp is not re-engaged in the reverse direction. In addition, even when testing with a positive-going surge at 270 degrees on the ac line, or Fig. 6a: Voltage Wave Across a Clamping Protector Clamping at 460~, for the Open-Circuit Voltage Wave of Fig. 3a and Short-Circuit Current Wave of Fig. 3c. NOTE: USES NEW COUPLER/FILTER DESIGN. oCv=6kV 200V/.5cm, 50us/.5cm Fig. 6b: Current Wave into the Protector of Fig. 6a. NOTE: USES NEW COUPLER/FILTER DESIGN. OCV =6kV 500V/.5cm, 50ps/.5cm 418 a negative-going surge at 90 degrees, the output clamp still didn't re-engage in the reverse direction. Thus it is possible, at least at the 3kA level, to suppress undershoot into even complex, combination protectors. Modern Surge Waves for Testing Powered Equipment Two major alternatives exist for applying traditional surge test techniques to the requirements for surging linepowered ac systems. The first involves refining, as far as possible, surge generator and coupler/filter design, so as to facilitate delivering Combination V/I waves with absolutely minimum voltage and current wave undershoots. For voltage in particular, backswing should surely be less than several hundred volts, even with 6kV peak impulses. This approach is exemplified in Fig. 6 and Fig. 7, at least for 3kA short-circuit current. The second alternative for powered surging is based on recognizing that both for surge currents that are less than a few kA, and also for power follow currents over one or two hundred amperes, super-low-undershoot open-circuit voltage waves are difficult if not impossible to deliver. For this reason it may be desirable to consider lowerfrequency (5kHz to 20kHz, for example) damped cosine waves for open-circuit voltages for higher energy, as well as for the lower-energy applications in which they are already specified (1). Such waves are widely found in actual on-site surge measurements (5). Front time can be maintained at 1.2us, to provide correspondence with present test waves. Short-circuit, clamp-circuit and protector-circuit currents can all be specified as damped oscillatory waves as well. Waves can be defined -frequencies, Q'S, etc. -- so that total delivered energy can be the same as, or carefully related to, energy transferred by the more conventional 1.2/50 and/or 8120 impulse waves. Conclusions Traditional impulses have already been replaced by Combination V/I waves for surging powered equipment. Requirements for doing powered surge testing over a range of short-circuit currents, with more sophisticated protectors and with medium and high follow currents, now lead to consideration of undershoot clamping energy versus forward clamping energy. These factors seem to be pointing in turn to the increased desirability of using damped cosine waves for newer surge standards. These oscillatory waves appear to offer not only greater correspondence with real-world surges, but more practicable test configurations as well. Undershoot energy into clamping protectors for such waves can be part of total wave energy as specified, instead of representing unspecified additional stress. Acknowledgement Fig. 7a: Input Voltage Across a FourTerminal Protector, for the Open-Circuit Voltage Wave of Fig. 3a and Short-Circuit Current Wave of Fig. 3c. NOTE: USES NEW COUPLER/FILTER DESIGN oCV=6kV lOOV/.5cm, 50us/.5cm Fig. 7b: Output Voltage from the Protector of Fig. 7a. NOTE: USES NEW COUPLER/FILTER DESIGN oCv=GkV lOOV/.5cm, 50us/.5cm Gregory J. Senko made important design contributions to the low-undershoot coupler/filter, and also performed a thorough technical review of the manuscript. His inputs are gratefully acknowledged. References Cl1 ANSI/IEEE Std C62.41-1980, Guidean Surge Voltages in Low-Voltage AC Power Circuits, p. 23. c21 Richman, P., "Single-Output, Voltage and Current Generation for Testing Electronic Systems", Proceedings of the IEEE EMC Symposium, August 1983, pp. 47-51. c31 Wiesinger, J., "Hybrid-Generator fur die Isolations-koordination," ETZ 104 (21), pp. 1102-1105 (1983). c41 SAE AE4L-81-2, Test Waveforms and Techniques for Assessing the Effects of Lightning-Induced Transients, 15 December 1981. c51 Ref. (11, pp. 17-19. - 419 - 76~4 PERFORMANCE DETERIORATION OF METAL OXIDE VARISTORS BY CURRENT SURGES V. Scuka Institute of High Voltage Research Uppsala, Sweden Summary The conduction mechanisms of MOvs are Still not fully understood. A new mechanism (I) may be responsible for formation of potential barriers in ZnO grain interfase. This, if so, may influence future composition of additives in MOV and give an improved aging characteristics. The experimental results obtained by the present author suggest a modified physical design of varistor discs for high duty performance applications. Referring to the experimental results it is also stressed the importance of adoption of varistor parameters to the actual operational and environmental conditions. Introduction The surge limiting efficience of MOV is superiortothe efficienceofthe conventional gap arrestors primarily due to a well defined threshold level, high power dissipation ability and fast surge pulse response. These characteristics of the MOV are well known and make MOV very suitable for applications in sensitive electronic systems which may be exposed to sub microsecond surges (which may be) generated in the system itself or those entering the system through its power supply network. Often it is not stressed enough that the surge limiting performance of MOV is in some extent a matter of their exposure to electric surges, working temperature, physical construction and the bulk material microstructure and a matter of some other environmental conditions. It is therefore of primary importance to select carefully the elements to obtain an optimized performance in a particular application. In this paper are discussed some characteristical features and parameters which influence electrical characteristics of MOV and their useful life time. The intention of this discussion is to make everybody aware of the performance limits of MOV and to help the engineer to avoid unproper design of surge protective circuits. To understand the complex mutual relationsofdifferent phys- ical and electrical parameters, it is useful to review briefly the main features in the electric conduction mechanisms and the main steps in the manufacturing process of MOV. The bulk conduction in ZnO - varistors The bulk conduction in the low current density region, the leaka e current region of approximately 10 A/m9 or less, is partly due to the conduction through the high resistive intergranular material, filling the space between the ZnOgrains, and partly to the conduction through the randomly distributed thin junction barriers between the individual ZnO-grains which lie in touch to each other. Bi203, diffused into the surface of ZnO-grains, is believed to be primarily responsible for the formation of the intergranular barriers in the bulk material of the varistor (6). These barriers create a complex network of current paths with randomly distributed cross-section areas of varying current transmission abilities. The resistivity of ZnO is low (0.01 ohm.m), so the interface areas between the touching ZnO-grains are at higher field strengths surely the predominating contributors to the electric properties of the bulk material. At low field strength, however, the electric properties of MOV are influenced even by the electric propertiesofthe intergranular material. In general the leakage current is temperature dependent and is nonlinear with applied electric field. It has been found experimentally (3) that varistors obey in the surge limiting current region the follwoing current density "j " - electric field "E" relation: ... (I) J = jo exp tcE] The electron potential energy configuration of the intergranular ZnOboundary region and the current breakdown mechanism in MOV may be as follows. Free electron from ZnO are trapped in an about10011 thick trapping layer, thereby forming positive space charge donor layers which extend into thegreins of ZnO on either sideof the trapping region. - 420 - Because of the small and negative temperature coefficient of the breakedown, which has been experimentally videnced, it is reasonable to assume that tunneling through the forward biased space charge barrier occurs at the breakdown current. To explain for the high values of the unlinearity coefficient a voltage dependent tunnel length was considered (3). The space charge density in the grain barrier may be influenced by the metal oxide additives (2) and is assumed to be in the range of lO24 mm3 (6). SO, the voltage drop across a single interface junction between two grains is in the breakdown current region nearly constant - about 3 V and does not vary for grains of different sizes. The conduction through these space charge regions is responsible for the strong nonlinear current-voltage characteristic of the MOV: i= kua Observations An ideal component should consist of grains of equal shape and size with uniform intergranular space. This is, however, in practice not the case. The grain size may vary throughout the bulk and the grain size distribution may be inhomogeneous. The same is valid for the intergranular space. In some materials the cavity formation in the intergranular space or/and the porosity of the grains may be observed. Measuring the size of the grains the following general observations have been made: . ..(2) where the nonlinearity exponent a is in the range of 20 to 70. However, it should be remembered, that this valuedropsdramatically at current densitiesofthe order of 1 MA/m2. So, it should be avoided ith current surges to load MOV-discs exceeding 0.1 MA/m Y if the surge limiting efficiencehastobe fully preserved. A new approach in explaining the conduction mechanisms in MOV has been reported by Einzinger (1). The formation of the potential barrier at the ZnOgrainsinterphaseis explained by the effects of the oxygen deficienceof ZnO, creating so temperature dependent concentrations of oxygen and Zn vacances. Referring to this model Bi203 is not an unavoidable composite additive in MOV for the creation of varistor function. Even more, Bi203 has been pointed out as responsible for the aging of the varistors. This due to a presumed uncontrolled phase transformation of Bi203 from a high resistive fior y -phase to a low resistive b-phase. Permutations may be initiated locally in the microstructure of the varistor bulk material by high current densities. This could well explain the pulse aging characteristics reported by Scuka (7) and later in this paper. Varistor and the way of controlling the cavity and flow formations in the grains and the intergranular space. In general, the nominal voltage of a MOV-disc is given by the thickness of the disc and the maximum power rating by the diameter of the disc. fabrication and construction The starting material is a powdercom position of ZnO (90.0 v.%), Bi203 (4.0 v.%), Sb203 (3.0 v.%) and other additive oxides. The composition is dispersed in a liquid, mixed and milled, dried, pressed into discs of predetermined thickness and diameter. After pressing the discs are sintered above 1200 OC. The Bi203 is molten above 825 OC, at higher temperatures grain growth occurs forming a structure with controlled grain size. The process of grain growth and the intergranular space material formation is strongly influenced by the raw material, the rate of temperaturechange, pressure change Fig. - 1: Electron micrograph of a broken surface of a 220 V MOV disc. Nonuniform size of ZnO grains may easily be recognized The grains in the middle of a disc have larger mean effective diameter withahigher standard deviation as the grains at the upper and lower disc surface. - This observation is very pronounced for discs with a nominal voltage below 60 V and is not so pronounced in discs for higher nominal voltages. - The porosity of\the grains has been observed only in some discs for low nominal voltages; - The cavities in the intergranular space are more often observed in the low voltage discs than in the discs for higher voltages. After sintering the varistor discs are equipped with a metallized electrode layer. The electrode layer, the electrode termination and the rest of the mechanical assembly of the MOV are very important parts in a component. They have to be considered in each specific kind of application where the demands of high duty performance are of importance. - Fig. 2: - Electron micrograph of a broken surface low voltage MOV disc. The grain porosity may be observed The points of consideration are: the electrode layer thickness and its interface to the MOV-disc the mechanical and electrical characteristics of the terminal connections the electrode symmetry the mechanical assembly. Some observations Microcracks and flaws below the metallized electrode layer may cause irregular current density loading of the disc. Similar effects may develop at irregular current outlines of the electrode surface or at centric displacement of electrode surfaces. The resistance of the electrodes must be low and the thickness of the electrode layers must be large to prevent radial current tracking on the electrode layers caused by high current pulse loading. The interface area between the terminal and ,the electrode surface must be as large as possible, without any flaws. The mechanical assembly of the component must fit the environmental conditions and the physical circuit layout. The mechanical assembly is in large extent responsible for the failure mode when the component is overloaded. A non-destructive mechanical inspection of a component may be done by sonic waves at about 5 MHz. During this procedure the components are dopped into the water. A flaw of about 1 mm size in the bulk may be discovered by this method. The effect of elevated working temperature The useful life of MOV is strongly influenced by the working temperature. This is due to the electrochemical processes in the varistor bulk, the'atomic migration of additives and the positive electric temperature coefficient of the component. Higher temperature causes resistance decrease and higher power 421 76 M4 - dissipation. This may result under certain loading circumstances in thermal runaway of the component. So, for example, when the working temperature for a certain varistor with a characteristic constant D = 13130°K and MTBF (25OC) = 5 . 108 hours (5.7 ’ 104 g ears) is increased from 25OC to 130 C, the mean time between failure will be reduced according to the Arrhenius law (ref. 1 equation 2) to about 7 months. The rate at which the component is aging is given by the expression sP(t)/Gt = -K exp C-D/T] ...(3) and the reliability of satisfactoral operation during the time period of 7 months will consequently decrease at the elevated temp. from the expected value of 99.999 % to 37 % as it may be proved by the following expression: H(tlT2) = [H(t,T,)I exp [(D/T,) * (a-1)/a-J ...(4) High energy pulses It has been established experimentally that each component can withstand only a certain number N(j)max of current suxges j at certain pulse shape, e.g. 8/20 us. If we assume N(j)max to be 10 we found that j is varying between the average values of 10 MA/m2 and 40 MA/m2. The lower value is applicable to the varistors below 60 V nominal voltage and the higher value to the varistors for higher nominal voltages. This observation supportsthegeneral physical model of MOV asitwas discussed in the previous sections. Ifweassume the creationofhot spots onthe critical intergranular barrierswemay calculate by theuseof thermal analysisof Carslaw and Jaeger the critical density: 1 . ..(5) JHs = TX/ [UBDfHS with parameters as follows: the thermal conductivity of the x Zn-grain ,-I W/m°K the melting temperature of the T Bi203 intergranular layer z 820°C UBD the breakdown voltage of the intergranular barrier M 3 V rHS the equivalent radius of a hot Spot rHSL 1 I.lm. A contact area of an equivalent radius Of 'HS = 1 urnwill cause melting at a junction current density of 270 MA/m2 which is about 10 times higher than the average maximum allowed impulse current loading of the component, when a 10 pulses life of the component is assumed. It should also be remembe ed that at a current density of 200 MA/m 5 electromigration of metalls in semiconductors is very pronounced. High power pulses When MOV are exposedtocurrentpulses of nanosecond rise time, however of a relatively low total energy, no simple - 422 between the pulse energy and the deterioration of the component has been observed. The deterioration is also poorly related to the number of applied pulses. A stepwise deterioration of the electric characteristic of the components has, however, been observed, the deterioration being related to the increased crest value of the current at the preserved current shape. Similar results, however at the current shape 8/20 us, have also been observed by other authors (4). At a rate of rise of current density in the range of lOI A/m2s, only a very limited volume of the MOV disc is initially loaded by the current pulse. Consequently, the current density in this region of the bulk material is extremely high. Intergranular barriers of successively larger equivalent cross sections (rHS) will burn out according to equation (5) when the pulse current amplitude at the preserved current shape is successively increased. The most exposed region with highest current density is the outermost part of the MOV-disc. It is therefore of vital importance that the outer boundary of the electrode lies well inside both the disc areas and that the length of the electrode circumference is large enough to prevent critical current density at sub-microsecond current pulses. Unfortunately these aspects have until now not influenced the construction technology of varistor discs. It would therefore be of great value if further experimental investigations could be performed with MOV-discs of the construction shown in Fig. 3. relation terminals varistor disc + electrodes Fig.3: Varistor disc design which might resist high sub-microsecond current surges Functional reliability of MOV The functional reliability of MOV may be expressed in a general way as R(t) = exp[ -idt/MTBF(T)] ...(6) and may be simplified for small failure rates as follows: R(t) M 1 = ; ti/MTBF(Ti) ... i=l Index i in the equation (7) is related to the time intervals with particular working temperature or time intervals with particular type with an other environmental condition, e.g. environment with high energy current surges. By loading the component with current surges the MTBF(T) is shortened by a factor Nmaxmn Nmax where n is the number of applied pulses and Nmax is the maximum number of specified pulses of a certain magnitude which can be applied during the useful life of the component: N -n (8) MTBF(t, Tn) = MTBF(t, T) F... max Selecting a suitable MOV-disc When selecting a MOV-disc the recommendations given by the manufacturer have to be followed. In addition to what is stated above, the following has to be carefully considered: The disc diameter has to fitthetechnical design requirements related to the maximum expected surge currents of 8/20 1~spulse shape. Itiswell known that the nonlinearity exponent a decreases dramaticallyathigher current densities. To preserve the surge limiting efficience of the component, the maximum surge current loading should not exceed 0.1 MA/m2. The disc diameter and physical design has also to fit the requirements related to the unhomogeneous loading of the bulk material during sub-microsecond current surges. References (I) Einzinger, Richard: Evolution of Physical Models for ZnO-Varistors - A Review, Proc. of Int. School of Mat. Sci. and Technology, July 1984, Springer Series in Sol. St. Sci. (2) Kazno, Eda: Transient conduction phenomena in non-Ohmic Zinc Oxide ceramics, J. Appl. Phys. 50(6), p 4436-4442, 1979. (3) Karner, Hermann: Alterungsvorgange bei Zinkoxid-Uberspannungsableitern, Institut fttrHochspannungstechnik, Technischer Universitat Braunschweig. (4) Levine, Jules D.: Theory of varistor electronic properties, CRC Critical Reviews in Solid State, Sciences, p 597-608, November 1975. (5) Mahan, G.D., Lionel M. Levinson, and H.R. Philipp: Theory of conduction in ZnO varistors, J. Appl. Phys. 50(4), pp 2799-2812, April 1979. (6) Morris, G. William: Physical properties of the electrical barriers in varistors, J. Vat. Sci. Technol., Vol. 13, No. 4, p 926-931, July/ Aug. 1976. (7) Scuka, V.: Lasting effects of transients on equipment performance, UURIE:147-83, Uppsala University, 1983. - 423 - 77 M5 CHARACTERIZATION OF DISTURBING TRANSIENT WAVEFORMS ON COMPUTER DATA COMMUNICATION LINES Maurice Tetreault Digital Equipent Corporation Stow, Massachusetts c)1775 Franc;ois D. Martzloff General Electric Company Schenectady, New York 12345 SUMMARY CIRCUIT‘IYPES As parat of a systematic effor*t aimed at char*acter*ization of disturbances on data cormmlnication lines, measur*ementswerae per*formed at a distr*ibuted computer* site durding lightning storms. The recordings ar’e analyzed for* statistical par*ameter*sIn or*der*to obtain generaic data. The jnformation Is presented as an invitation to share data on the subject. CWcuits and cabling methods used In LANand DCS systems are an extension of standard pr*actlce In room-size systems. However., they do not involve the much longer, lines encountered In telephone cirdcults. The RS-232 inter-connecting standard Is the most commonly used technique at the present time; with the demand for* faster, tr+ansmisslon reates, RS-423, RS-422, haseband, and hroa.dband ciracuits ar’e becoming m0r.e prdevalent . IMJKXXJCTION Increasing use of computer corrmunication techniques for J,ocal Area Networgks (LAN) and Distrelbuted Computer*Systems (DCS) requlr,es rAelIable er*r*or-fr*ee data transmission in hostile enviraonments. Many such systems have experaienced damage f r*omenvir*onmentally induced In addition to the cost voltage transients. aspects, this prIoblem also diminishes the va.lue of the J,ANand JXX installation. Inter*fer#ence irmnunity of these systems may be achieved by the application of praotective devices at the data cable point of entry, by the pr’opera install.ation of cable shielding, by utilization of inherently lrtmune pr*otocols - 01’ by application of all of these remedies. Hardware pr*otectlon against major threats can only be achieved by diver%ing the unwanted ener’gy away from sensitive components/devices. In or*der*to apply pr*otection devices or* techniques cordrectly and economically, more knowledge about the char*acter*lstlcs of the distur%ances is required. The present paper* reporats the results of mea.surements and analyses frlom a distr*ibuted computing site In The centrsal Floral da dur*lng l.ightnlng storms. data is offeraed her*e as a contr+bution to establishing a better. charaacter,lzatlon of the environment and as an invitation to other* wor*ker+sIn this field to come forward and shar*e Information. * fi The Sur*ge Pr*otective Devices Comnlttee of the Im is curerently sponsor*lng such an effore at character*lzlng the envlr*onment . Additional inputs to the data base are invited; contact the pr*lncipal author* of this paper’, M. Tetr-ault, Digital Fqulpment Corpor*atlon, 40 Old Bolton Road, Stow, Massachusetts 01775. I?ach of these cir*cult types has differ*ent sour*ce and load impedance char*acterlstics, and the line dr*iver*s and receIverIs have differ*ent overvoltage withstand capabllltles. Cables used in these systems may be of the twisted-pair4 (shi.el ded 01’ unshl elded) type, or* The RS-232, 423, and 422 links coaxial.. normally use twisted pairs, while baseband and brdoadband links typically use coaxial cables. Cable shielding methods r%nge fr*om nonexlstent to single 01’ multiple shields. %ounding of the shield Is var*iously found at one end OL* both ends, scmetlmes with deliber*ate 01. accidental gr*oundlng at multiple intermediate points. Tunis diver*sity of clr*cuj t types might, at fir*st glance, make sny measurement that uses a single example for4 character*lzlng a complex situation What is rvporated here Is seem to be smbltious. Indeed only one example: the Intent Is to descr*lbe a measurement prQcedur4e and the method applied to reduce the raw data to useful However,, we believe engineering parameters. that the same methods can be applied to other* types of installations (LAN or ES) impacted by overvoltage sour*ces other, than lightning, such as Induced disturebances from adjacent power’ cir*cults or* electrQstatlc dlscharage. SITE FACILITY The site facility consisted of one centraal computer camnvllcating via RS-232 with various distrolbuted video display and hardcopy terminals (Fig. 1). The cable on which the measurements were made was 650 m long, containing 32 twisted palr*s of #20 AWTr wire, ea.& pair* having an individual shield of thin aluminum foil tied to grvxlnd at one end. Routing of the cable Included dir-ct bur*lal fat* - 424 conductor* of its power cored. With these pr,obes connected to the 1 M ohm input impedance of the oscilloscope, an effective impedance of 10 M ohm was thus connected between the points beine: monitored and local ground - negl,igihle compared to the circuit impedance. The oscillograms were t*ecot*ded with the oscilloscope operating at a sampling rate of 20 MHZ- that is, 50 nanoseconds per’ point. The tseceptacle for the oscilloscope power. cord was about 1 m from the building service entt*ance, so that the voltage measured by the oscilloscope was essentially the potential with respect to the building ground potential. Fig. 3 shows a.n example of the recor*dings obtained during the measurement pt*oject . Oscillogtam 3A contains two waveforms syncht*onously and simultaneously t*ecorded by Channel channel 1 and channel 2, respective1.y. 1 monitored a transmit line (pin 2) and channel 2, the associated signal refetaence line (pin 7) - r*efet* to fig. 4. Oscillogt~sm 3R shows the al.gebtaic difference pt*ocessed fkom the simultaneous taecor*dings of the two channe1.s. L Fig. 1 Site plan indicating cable routing (Measurements performed at cable entrance to building containiny computer) _ . 46 m at a depth of about 1 m, aluminum conduit for 380 m al.ong va.t*ious metallic racks, and pipes 2 to 5 m above graound, the taest of the ca.ble being strung overhead on utility poles. 3 shows a si.mpl.ifi ed l_ogic drawing of a.n Fig. RS-232 cormunieation circuit. The line dt*lvets ate 1488 devices: the receivers are 1489 that is, they devkes . !They arae single-ended; operate in t*efet~ence to signal common rather than in a dlfferaential mode. These signal commons are then bonded to the chassis of each equipment cabinet, which in turn are bonded to their respective ac power grounding conductor. This type of arrangement may result in a dlfferdence In the potential of the signal In an conmon at the two ends of the cable. attempt to tdemedythis situation by equalizing the potential, signal refetdence wires (pin 7, EIA RS-232 Standard) are run along and twisted with the respective tt*ansmi.t and receive signal of equalizing the wi tss . The effectiveness potential of two chassis, each bonded to a local. ac gt*ounding conductor,, can be expected to decraease as the ohysical dktance and/or the difference between the two chassis p’.otential i: ncrease. Fig. 200 3 3A example of field recorded waveform 160 120 s --T_-J I__- -- 2 6 80 40 > Fig. 2 Simplified circuit logic drawing of a RS-232 I~WlJHJMF~ATION AhID RECORDING MFTHCECJ~~Y A Nicolet Model 2090 digital oscilloscope was ’ used for monitoring and recording tra.nsi.ent voltage waveforms appearaing on the the data Compensated pt*obes with 1.0:1 cable. attenuation wet&eused; the shl.elds of these pteohes we1.e not grounded to the c-lrcuit signal common, but only to the oscl lloscope chass-i s, the latter* being t?.eA to the ac grmmtllng ii” g 40 ;1 E 80 :: 2120 160 200 3R exist@ normal mode voltage - 77M5 42.5 - !The waveform shown in Oscillogram 3P illustr*ates that a norinal. mode voltage can exist between between the conductors of a twisted pair,. This normal mode voltage is in fact the vol.tage that till he 1mpraessedacr’oss the output pins of the line dr*iver*/r*eceiver* circuits and their, rdespectlve signal commons. Note the offset exceeding -12 V and the transient voltage exceedj.ng -4o V, with a dur*atlon extending beyond the tj.me window of the recording (102.4 JW). The diffenence tn voltage between simultaneous traansients occur*rlng on the twisted pair. consisting of the signal and r*eferance conductor*s is due to the differdent terminating impedances of the conductors, as illustr.ated in Fig. 4. RFmT,Ts The raw field data wer’e trgeated by numer*ica1 methods which ar*e descr+bed in the Appendix. Table 1 shows the r#ange of frequencies contained in the transients, their, r*ise times and their* dureations. Fig. 5 shows a r#epr*esentative waveform that may be cited as ‘typical of what a near*by lightning stroke produced In the cable at this site. It must be understood that the author*s are pr*esenting figur*e 5 as r*epr*esentative of only the data obtained at the patWcular+ site descr*ibed in this paper’. It is not possible to extrapolate a universal r*epresentative waveform given the small size of this sample. As stated the intention is to pr*esent a method of measur*ing, r*ecordlng and reducing the data to useful engineering parsmeteras. 1BOVOLTS 250KHz Fig. Fig. 4 Partial schematic of driver receiver (1489) pair 5 (1488) and Representative waveform derived statistical study from TABLE1 4 is a partial circuit schematic of a Fig. RS-232 dr*iver*/r*eceiver~ pair*. Note that the draiverahas a 300 ohm output impedance and the r*ecelver* has a 4000 ohm input Impedance. !IWS differaence in terminating impedance means th.at a balanced syrrmetr~ical transient tr*aveling along the twisted pair* (signal and reefer*ence wirees) will cr*eate a diffenent voltage dr*op acr’oss the diffenent terminating impedances, in other, words a difference of potential between the signal and common pl.ns of the terminating That differaence, which is only caused devices. by the unbalanced devices input/output Impedances, converts an Impinging commonmode tr*ansj.ent into a normal mode traansient at the termlrmting devices. Over* the period of observation at this site, r*ecordings such as that of Fig. 3 were These r*awdata were subsequently obtained. analyzed by statistical methods in order to obtain a “typical” descr*iption of the effect the envkonment on the cable. 64 of’ To place these r*esults Into perspective, test rdesults on the withstand voltage of commonly used m-232 dr+iver*and receiver. integraated circuits indicate that fai1ur.e occurs in the rdange of 40 V to 90 V, wl.th times to failune ranging fr*om 3 ,us to 35 JB . Failune of the 1488 dralver+is gener*al.ly due to burnout of the 300 ohm output r#esistor,, an enetegy-dependent mechanism wher*e duration of the transient Failurae of the 1489 recelver. is dominates. genersally by Internal arolng, a voltage level-dependent mechanism where dur*ation of the transient has minimal effect. DURATION (note 3) total events 4.8 ).“s to graeater than 102.4 us single events 0.1 ns to 75 E.ls 52.6 pa 4.5 ps Note 1 Upper, frequency instrzunentatlon appendix. ll.mlt Is 7 MHz due to limitations. Refer+ to Note 2 Lower* limit of Cl.1 ,us was arhitr*ar4ily imposed due to Instrumentation limits. Refer, to append1x . Note 3 Two ranges ar’e pr,esentad since natuna.1 transient events tend to occur’ in bursts of related voltage excursions (fig. 3). Values listed as “total event” raeprdesentthe duration of the entir*e event as rdecorded. Values listed as %ingle event” represent the dunatlons of the individual voltage excursions occur*r*lng durWg the entir*e event. . - 426 - INSTALI,ATION EF'JXCTS ON TRANSIFW CRARAcTeRISTICS The field raecorded waveformsIn this study repreesent the effectsof transientsgenereated by lightningdischangeson a specificcable plantand its terminating impedances.That Is the electricalchar*acteristlcs of the complex circuitmade up of the data cable,CPU, terminals,the variousac power* feeds,and the earth r*eferaences for*both power and cable shields,etc., will determinethe responseto a given transientstimulus. The specific re;ponseto a hypothetical impingingtransient of knownDatameterswould not likelybe the same for,any two LlANand/or* DCS Installations. To further. complicatethe effortto char*acterize the electricalenvir*onment of LAN and DCS installations thereare a varietyof possiblesourcesfrom which the initial transientsimpingingupon the systemsmay be generated. The transientgener*atlng sourcesto whichL#AN and DCS systemsar'eexposedincludeFSD, lightning,trdansients enteringthe systemsvia the power' mains,transientscoupledover*from near*by conductorssuch as power OL* contrIo lines (evenstnmturs.1steel),and transients gener*ated due to likelyearthpotential differ*ences momentarily existingbetweenthe variousearth referaences of the system. The char*acteristlcs of the Initialimpinging transientsonto thesesystemswill of course exhibita varietyof waveformsexisting individually and in combination, consistingof both oscillatory wavesvaryingover,a wide r*ange of possiblefrequencies and unidinectional impulseshavinga wide rangeof possibleratesof rise and dunations. The Snhenentcircuitchar~acteristlcs of the particular* installation on which r*ecor*ding ar*e beingmade will have a gneat effecton the electricalchar*acteristics of any transient tnavelingalong the installation.The existenceof voltageand cunrentstandingwaves and reflections can be expected,whichwill certainlycontributeto the char+acter*istics of any observedtrsnsientwaveformson the cable. The physicallengthand the velocityof. pr*opagation of any cablewill Influencethe r*ates of rise that can be expected. Typicalwith&and levelsof data cable Insulation are in the rangeof 5 kV. Any transientinducedon the cableabove that value would be limitedby insulationfailur*e, so that an upper* limitcan be expected. Damaging levelsof the terminating equipmentis nnuch lower*, so that the rangeof interestincludes all voltagesbetweenthe withstandlevelof terminalequlpm&t and the upper'limit,should it be reached,set by the cable insulation. The highestvoltagemagnitudesrecor*ded during the measurement peniodexceeded200 V. The peak magnitudesof this eventwas estimated, basedon the observableportionof .ltswaveform char+acter&rtics of dunation,rise and fall times,to have been in the neighborhood of 3oov. l- BernardRicker,, PrincipalFngineer', Corporaate ComponentEvaluation Laboratory, DigitalFquipment Corpor*ation, Northbor*o, MA. 2- DonaldGauss, PrMiuctSupportRngineer*, DistrictField ServiceSupport,nfgital EquipmentCorporation, Onlando,FL. APPENDIX- NUMERICALMFTRODS All data were analyzedusing DigitalFqUipment Corponationsoftwarae. The pr*lmary pr*ogrems used are containedin Y'he Laboratory SubroutinePackage"r11, other, utilitypr*ograms were writtenas requir*ed. The specific subroutines used were the Fast Fourier Transform@!?I?) and the PowerSpectrum on the Nicolet (PCWRSP).Each waveformstoraed disk was changedto ASCII formatand storedas an individual. file on a VAX 11/780. Since the fielddata were digitizedcertain limitations in pr*ecision will occur' duningthe necor*ding pr*ocess.The Nicolet2090 stonesthe r*ecor$ed data in a digitizedformatof up to 4096points,sampledat up to a 20 MHZ rate with 8 bit precision.Each of the waveformsin this studywas recordedusing 2048 sampling pointsat a 20 MHz samplingr*ate(50 nanoseconds per+point). Voltagemeasunementaccuraacy is a functionof verticalamplifier@ linearityand bit 0.4% resolution.8 bit nesolutionpr*ovldes That is, the value of the Least rdesolution. Significant Bit as a functionof the total numberof bits determinesthe smallestpossible subdivision of the measunement unit. So for,a 8 bit r*esolution devicethe ISR Is l/256of the unit of measunement. Resolutioncapability combinedwith the amplifierslineanlyaccuracy rendersa statedspecification forathe Nicolet 2090 r2i of 0.5% full scaleaccunacy. 'Ihefrequencyrecor*ding limitIs determinedby both the samplingraate and ampl.lfier~ bandwidth. Samplingat 20 MHz can accunatelyrecord frequencies up to slightlyless than 10 MHz. Samplingtheoryindicatesthat a given fr*equency (F) must be sampledat more than 2F in order* to be accurgtelyreconded,any frequencyequalor'greeaten than l/2 the samplingrate will not be accuratelyrecorded. The amplifier1 bandwidthimposesan overall7 MHz limit. Any frequencyinputto the device above 7 MHz will be recordedwith enrotsin amplitude. FlguneAl is an exampleof the outputof the analysisprogrem,usinga 100 kHz sine wave. AlA Is simplya graaphic of the waveformbeing analyzed. AlB 1s the outputof the power* spectrumroutine. !l'his describesthe proportional relationship of power. magnitudes betweenthe varies frequencycomponentsof the waveformbeinganalyzed. 77M5 - 421 - .., I- 12. Data pertainingto transientraates of nise and dunationswere not gr*aphlcally depicted,since statistical data adequatelypr*ovide the Information required. ?d 4. ;r d > -4. -12. _*o.;: : j i I : : : j 44 ( i : : ; 88 ( : : : : 132 MICRO , : :: 176 220 SECONDS Fig. AlA 100 kH.zwaveforminputto La.bor*atory Subr*outine Package. The remainingfiguresare examplesof the transientsr*ecor+ded durl.ng the pr,oject.There is a tendencyof the transienteventsto occur* in burst of nelatedvoltageexcursions.Figur*e A3A and A3B are examples. The entireeventhad a dur*ation of gr*eater+ than 102.4us -the limit of the Nicoletrecor*dlng windowusing a 20 MHz sampling[*ate and storaing 2048 points (50 nanoseconds X 2048). Containedwithinthe eventwer*enumer*ous voltageexcursionsof varyingamplitudesand dur*ations. There is no way to pnacticallydescr*ibe such eventsusing one set of statistical InformatIon.So the decisionwas made to determinetwo sets of parsmeters.One descniblngthe eventsas an entiretyand the other+ descr*lbing significant voltageexcursionsoccur*r*ing withinthe events. 20. :001 .Ol .l 1 FREQUENCY 100 IO ,I : : : , : i’i I , : : : I , : : : : , : : 8 : 12. (MHz) AlB Outputof power* spectrumr,outine. g 4. 2 d > -4. Figurde A2 Is pr*ovided for*compar*ison, it depictsthe resultsof the analysisusinga 100 -12. is a kHz squar'e wave input. As expectedther*e significant differencebetweenthe pr*opor*tional.m power. magnitudesof the fundamental. and the 0 the subsequentharmonics. Fig. A3 lllustrlates rdesults of the analysisroutineon one of the field r*ecorded waveforms. 'Phlsexample 1 indicatesthe gneatestpower'contentin the raange of 30 kHz to 80 kHz;. Then two other, pr4cxninent peaks occur* at appnoximately 110 kHz to 180 kHz and the other. at appr*oximately 240 kHz to 260 kHz. _“. _20,;, : i4: / j i8 i :,A2: j / :,& : 72 MICRO , :001 Pig. 48 24 : ,:, : .Ol , l:, 96 120 SECONDS : I .l FREQUENCY “, I :>t, 1 (MHzl : 10 :‘: 100 A3A and Fig. A3B ar'eexamplesof transientwaveformanalysisr*esults. “, 220 MICRO A2A Fig. 1 ,, : SECONDS 100 kHz squarewave inputto LaboratorySubr80utine Package. : : : : : : : : : : : : :_, Fig. A4 Typicalrecordingsof normaland c0mMlr-r mode voltages. ( FOLLOWING PAGE ) .8 + = .S ._ 5 I! :: 4: .2 f o” :, ,001 .Ol .i, / .l FREQUECNY .,:,:., : 1 $ ,/ 10 / ir 100 (MHzl A2B Outputof power' spectrumroutine. Suh*outinePr*ogrammer*'s 1. Labor*atory ReferenceManual,DigitalFquipment Corpor*ation, Manlboro,Massachusetts, 1980. 2090 Digital 2. ServiceManualforeSerlies Oscilloscopes, NicoletInstrument CorporAatlon, Madison,Wisconsin,1981. - ” -MICiiO 10 0 20 30M.&O 428 - MICRO SECONDS SECONDS !&O~DS 70 80 YU 10 0 - 20 30 MI$O 80 90 100 S%O;DSz 200 160 120 0 10 20 30 M&l J !&O~D 0 10 20 3”Mlci?0 &Q&t 70 80 90 100 78 - 429 - THRESHOLD SIGNAL AND PARAMETER IN NON-GAUSSIAN Nl ESTIMATION EMC ENVIRONMENTS* David Middleton** New York, N.Y., Abstract Based on the author's recently developed results for canonical optimum and suboptimum threshold detection theory in nongaussian noise and interference [ll-[3], the present paper provides a concise summary of the extension of these concepts and methods to the parallel problem of the canonical threshold estimation of signal waveforms and parameters in nonqaussian electromagnetic interference (EMI) environments. The critical problem of terminating properly the resultant (weak-signal) series approximations for the canonical estimator algorithms is solved, preserving the required locally optimum Bayes estimator (LOBE) structure and insuring asymptotic optimality (AO) of performance. As in the corresponding detection situations [l]-141, it is essential to include a proper bias term in the estimator. It is shown that the desired LOBE's are linear or nonlinear func,tionals of an associated locally optimum Bayes detector (LOBD) and A0 structure, with proper bias. A sufficient condition for this LOB and A0 property is that the associated LOBD's are asymptotically normal, with appropriate means and variances under the particular hypothesis states involved. General examples are based on: (i) the "simple cost function (SCF)," [5], [61, from which are generated unconditional maximum likelihood estimators (UMLE's), which, in turn, are simple linear (i.e., derivative) functionals of the associated LOBD's; (ii) the *This paper is based on current studies done in part for the Institute ill-[31, of Telecommunication Sciences (ITS) of the National Telecommunications and Information Administration (NTIA), U.S. Department of Commerce, Boulder, Colorado 80303. **The author is at 127 E. 91 St., New York, N.Y. 10128, USA. USA quadratic cost function (QCF), for which the corresponding LOBE's are found to be nonlinear (i.e., integralexponential) functionals of the associated LOBD's in this case. Explicit classes of (threshold) estimator are obtained for the SCF and QCF, corresponding to the (1) coherent, (2) incoherent, and (3) composite modes of observation [ll, II+&, for both optimum and suboptimum algorithms. The principal types of nongaussian EM1 here are described by the author's canonical Class A and B noise mechanisms [7], developed during the last decade. The general results are illustrated by a specific example. 1. Introduction As is well-known, in the Bayes the[6] estimation is the "twin" ory [51, of detection: the former aims to determine the particular characteristics of a desired signal, e.g., amplitude, waveform, phase, frequency, etc., once the signal has been detected, while the latter is concerned with the basic question of the desired signal's presence or absence. In a very broad sense "estimation" can be regarded as an extension of detection: both subsume appropriate cost functions for the derivation of optimal algorithms and for the measures and comparisons of performance. In fact, in a more general Bayesian sense detection and estimation are the two, coupled component elements of the composite process of signal extraction: detection and estimation joined together by a suitable cost function, reflecting the frequent situation where it is not completely certain, for the estimation process, that the desired signal is present, which, in turn, results in biased estimators [81. Here, however, we shall assume that the desired signal is known a priori to be present, and that certain waveform and parameter features of the signal are to be estimated, when the accompanying EM1 is highly nongaussian (the "classical" situation of gaussian EM1 is a,special case of our general model). Now the estimators are un- - 430 biased. In any case, we may expect the estimation process, particularly optimum estimation (in the Bayes theory 151I [6]), to be closely related to the detection process, since both are derived from appropriate likelihood functions. This will be seen explicitly in the canonical* threshold theory outlined below (cf. Sec. 3). From the practical EMC viewpoint, signal estimation is a common communication requirement: signal amplitude (or level), waveform, frequency, epoch, modulation are each important elements of the reception process, whether it be telecommunications, radar, television (and their analogues in acoustical and optical regimes). Threshold reception (here estimation) plays a critical r81e in EMC applications, because it permits one to help maximize performance under the constraints of limited spectrum OCcupancy, against competing signals in the presence of non-intelligent EM1 and noise backgrounds. Designing receivers for effective threshold performance usually ensures better (but not necessarily optimum) performance at strongsignal levels. As in the detection cases 111, an optimal theory provides limiting estimation algorithms and performance, which are models to be approximated in practice. Also, as in detection, a canonical theory is possible in the weak-signal cases (as long as reasonably large time-bandwidth products are permitted, of course: the desired signal must be "extractable" under the observational constraints). Such a canonical theory also provides standards against which practical, suboptimum (i.e., approximate) algorithms can be evaluated and compared 111, 121. From the EM1 scenario, namely, the set of quantitative conditions on the interfering signal sources (e.g., source distribution in space and time, propagation conditions, waveform structures, etc.) which permit the a priori calculation of the parameters of the probability distributions of the interference [3], [7], we can determine the quantitative analytic form of these distributions, and consequently, the explicit structure of the desired estimation algorithms and their expected performance. These interference parameters are also directly obtainable from empirical data [73, and are physically meaningful elements of the basic statistical physical models of the dominating nongaussian interference here. *BY "canonical" is meant an analytic structure where the form of the processing algorithms and their performance is independent of the particular waveform of the signal and statistics of the noise and interference, i.e., of the specific physics of the problem in question. As in threshold detection theory t11 a fundamental problem now in developing effective threshold estimation algorithms is to obtain expressions of limited complexity, which retain their optimum nature when sample size (n) becomes increasingly large. The latter is the case, of course, when the signal is weak, since large effective data samples are required for small expected errors in the resulting estimates-analogous to small probabilities of decision error in optimum (threshold) detection. However, just as in detection [l], without a suitable "bias" term in the extraction algorithm this algorithm demands progressively more terms in its approximative form. This rapidly defeats the key requirement of processing feasibility, particularly for signals and interference of practical use, and destroys analytic tractability, as well. Again as in (threshold) detection theory, a major difficulty with much of the earlier work (cf. discussions in [3], [4]) has been the lack of correct "bias" component in the LOBE forms, so that they in turn can be very suboptimum at all (small) signal levels and sample sizes. Accordingly, from the viewpoint of the above, and consistent with the severe limitations on space, we present here a concise summary of optimum and suboptimum threshold signal estimation, for the three qeneric modes nf reception: (1) coherent estimation, 72) incoherent estimation, and (3) composite estimation, where a linear combination of the coherent and incoherent algorithms is employed. Because of spatial limitations we shall confine our explicit illustrations to the simplest situation of amplitude (or scale) estimation, where signal waveform is otherwise known at the receiver. However, the effects of fading, doppler, and other propagation conditions can be included. [The general formalism outlined here, nevertheless, is quite capable of providing explicit algorithms in the more complex situations involving waveform, frequency, phase estimation, etc.] The EM1 is additive, generally nongaussian, and is either Class A or Class B interference (i.e., respectively coherently or incoherently received in the (linear) front-end stages of a typical narrow band receiver, accompanied by (additive) external and internal gaussian noise [13-[4], [71) * These interference models account for most of the EM1 encountered in practice. Among the new results cited here are: (1) LOB and A0 estimator structures for threshold signals; (2) composite estimator algorithms; (3) useful suboptimum algorithms; (4) estimator variances; and (5) explicit functional relations between the LOBE's and their associated LOBDs, with the accompanying A0 condi- - 431 tions emphasizing the critical rale of the "bias" term. This overview is accomplished in Sections 2-5: Section 2 gives a short summary of the relevant results of estimator theory; Section 3 is devoted to the associated LOBDs and their resultant optimum estimators, while Section 4 considers an illustrative example of amplitude estimation. Section 5 concludes the presentation with a brief discussion of the results and their EMC implications. 2. Analytical Background: Summary Here we briefly summarize the needed main elements of the decision-theoretic formulation of signal estimation theory [21, [51, [61. For an estimator we write (2.1) 2 = &($j&I,X)= g&), where J&=(81,...,e,) is a set of signal parameters (or waveform samples, etc.) to be estimated; &=(Xl,...,Xn) is the set of received data samples in which the estimate is to be based; the "decisions," or estimates made are denoted by ~,=(y~,...,yM), and u in Y0 indicates the a priori probability (density) governing the parameters $, e.g., a=~@), or J+ in 0(h). [We note that xc is an estimator for all permitted 5, while for a particular set X=X', yc becomes an estimate.] The de"cizion rule 6 is here .&(Y(X)=$,(y-x0(&(;)),which is an (M-dime&Tonal7 delta function. When xc(=gE) is given, i.e., the estimating receiver gE($) is specified, the average error (or risk) is determined from R(a,6) s or e= I, a@_ or lJ’, ~,(~(~:(g..W($j_or Q)& (or i,-yo)dz a) l (2.2) where C is an appropriately chosen cost function, naturally proportional to the measure of error selected. The estimator y. is a point estimate, and F is the conditional pdf of 3, given S($$. Another useful type of estimation procedure is interval estimation, defined by the probability P that a particular point estimate, Yc (for given E) falls within (l+h) 100% of the true value of the quantity [g=waveform, or @parameters in s(i)] being estimated, v1z.: ~{(l-h) ES or Ql Q Y,(a or SIX_)Q 8 < h < 1 (2.3) (1+X)[k or a>, where h is a measure of the prechosen confidence interval selected, -X8 to +hg for example. In the case of a single parameter 8, for instance, we can write PI (1-A)e 6y,(e\z) (l+h) (1-A) ; Wl(y, G (l+h)el (8) )dy,, = (2.4) 78 - Nl where Wl(Yc(8)=p(Yle) is the conditional pdf of the estimator y, conditional on 6, formed here from p(Y]e)= I r Fn(~l~(8)s(yel~)d~, with obvious formal extensions to the multidimensional cases, cf. 111, [51, [61. The unconditional pdfs of the estimators themselves are obtained from p(l)= gor 8' We emphasize thatr, is a point estimator, embodying the specific structure, gE, (2.1), of the receiver performing the estimation. On the other hand, interval estimators, as expressed by P, (2.31, (2.4), yield a probability which is a measure of the efficiency of the point estimator for any particular application (i.e., choice of &). The average error (or risk) (2.2) measures the expected cost or average error in using x0, considered over all possible ($1 received. For optimum, or Bayes estimation we seek estimators YX which minimize the average error or?isk R(o,6), (2.2). The general form of the resulting y* depends, of course, on the choice o B "Cost“ or error function C&_ or Q,y,). For example, for the quadratic cost function (QCF) c(~,&J)=colo-$J12 = ? (2.5) ‘0 ,i, (em-Y$m) Lf the associated optimum estimator is found to be (cf. Chapter 3, [51, Chap- (2.6) with A$, o(Q)+cr(s),a-+@ in the case of estimating signal waveforms. Note that 1% is generally a nonlinear operator on the received data, 8. Another cost function of considerable interest is the simple (or rectangular) cost function (SCF) given by M" (2.7) C(~,~c)=Co mil [Am'G(Ym-em)l with appropriate choices of Co, Am to ensure meaningful results (e.g., positive errors, etc.). Minimization of average risk here leads directly to (cf. Sec. 21.2-1 of [6]) the following relations determining yg SCF: I a log ae m c4 0,) w, (4 I.e,) 1 I em=eg=y* all m=l,...,M, where WJ&l = 0, m (2.8) ~,=e,)" , and )>p,=,_,_ sn /a(v,=e,) a(Y,=f3,)=,, with Q'=all Q except 8 The con& ition (2.8) deter92. mining x={y&] is precisely that determining the unconditional maximum like &,,and the proper r3iisdsti for A0 behavior are 0(21)* B(21)*,_ n;coh n-cob L(2) n =-2 C [a oj'j>& (3.11) (21): B(21)*_ On-inc n-inc--2 (SCF) =- $ - y $_lRj I (SCF) F (aiRj+R:Gij)i j 3.3 Suboptimum Estimators We parallel the treatment in [l], 131, where g: is replaced by the suboptimum g,, now appropriately adapted to the estimation structures (i.e., hypothesis test) required by the SCF, QCF, etc. We illustrate the procedure with the case of SCF and the use of simple (cross- and auto-) correlation receivers, well-known to be threshold NOW optimum in gauss interference. (3.5)-(3.7) reduce to this case direct&i=-xi (:.R:=-l), and ly on settin Similarly, for =l, L 8 &iss=2. "L&&s the QCF cases (3.9), (3.10), we make the same substitutions, in both the bias and data portions of giCF. 8, (3.6) Reception (SCF) g*n-camp =g*n-cob +g;2_inc' (3.7) where, as before [l], a .=a (ti) is a (normalized) signal amp??tuSie, s.= s(ti, 0,$') is a normalized signa i waveform, such that =1, and Ri= d logwi(xi)/dxi, R!=dR'/dXi, where wl(xi)=lst order p&f ot the interference (nongauss+gauss) above, in the The proper bias terms, usual way. for weak-signal AO, are speci+0*2, ficg?ly here (all SCF): B;_coh?~c~~_co&L(2) fCaojsj>i i =B* B* n-cob +"A-inc; n-camp 2 B;: inc:- 1,* 2 on-inc=-8 l f l(L(4)-2L(2)2) 6ij+2L(2)2iARij, :-: 'J (3.12) 2 2 where AR..-) is suffic$ently small, thereOexists an ,,,(< max the threshold optimal character of gfi breaks down and becomes suboptimum. At larger input signals gfi (+gn now) and hence Y&-V,, etc., may or may not be monotonically better, in absousulute terms, than gf, qIq_,,x. for sufficiently strong signals, ally, gg (now g,) is absolutely better, unless the information-bearing portion of the signal is destroyed by the algorithm itself, e.g., clipping destroys - 434 - waveform detection/estimation, but not phase parameter extraction, for example. _ 4. An Example: Amplitude with the SCF: Coherent Estimation Reception The simplest useful example is that of estimating the scale (or amplitude) of an otherwise fully known signal -0 waveform received in (generally) nonThis is also gaussian interference. the pEoblem of estimating signal intensity 1;=&;2. Here we have em=0 =ao, with 2' all With other relevant signa 1 parameters. coherent reception, signal epoch E is precisely known, e.g., E=E~, and we se=a, with independent lect ,=Sj_max Accordingly we write noise s2 mples, Physically, our a=smax= /2 here. present example can represent slow fading, whose changes are negligible over the data acquisition period. Accordingly, to (3.5) reduces j=l this to o1 laoI i('4.1) gives R(xi) ‘: (5) = - i 9,j/nfiL(2), z<H aO* 1 (=l)=-JZTL(2P,1an8 we can show that G&l) (4.5) (ifj); &R2+7Lw) 0 from g2a2L(2)2 ijl 0 (i=j). Thus a* is vnconditionally unbiased, e.g.,fr~~a:t:=9 (=Aa,/2) and $fz a;+ao, (>O). -(4.6) Here, from al(a )=l/Aa, we have a$ (Aao)2/%in (4.5) to establish an-upper bound (a )max( <2 7 1a 0=&* 0=o, (3.3), 8m=81=ao, - L(2)ain-aoJzz 2 a <<_1 varOR _= 0 2 2 varOR directly (s*n-coh)SCF=-L (2)azn-aoJZ y R(x ) Applying al( 2 The "smallness" condition on a iS (from Sec. 6.4, Eq. (6.71) of [31? like (4.2), if a(~)=m&lul(6m); otherwise one has M-coupled equations to be solved for Bm=Sm. Also, if the QCF is used, the structure of the optimum estimator is generally much more complex, due to the nonlinear functional relation between g& and ~8, cf. (3.2). If the parameters to be estimated ap-, pear functionally in the waveform S(t,i) then one obtains functional solutions for the estimators. The important general result here (for all signal levels) is the explicit functional relation between the (optimum) detection algorithms and the resultant optimum estimators, cf. (3.1), (3.2): detector structure provides the initial analytic relation, from which the estimator is the appropriate functional relation, the form of which depends on the choice of cost function. Another result of importance is the COLL responding threshold approximation, which when carried out properly in the associated detector is then A0 as well for the LOBE. Thus, to carry out the derivation of the desired estimator, we must start with the appropriate detector algorithm [l]. The evaluation of estimator performance 's provided, by '( *) , in the exusing the estimator, Y, pressions for the average error (or risk), cf. (2.2). Finally, we remark that our presentation here is basically a detailed summary of methodologies, which need further development in the form of specific algorithms and various numerical examples involving real-world Class A and B environments, including - 435 - the extension to the important situations (mentioned in Sec. 1) of joint detection and estimation (i.e., extraction), when it is not known precisely that the desired signal is present [8]. The fundamental importance of these procedures for EMC is explicit in the requirement for weak-signal capabilities in reception: the weaker the extractable signal, the greater the effectiveness of the telecommunication process generally. References D. Middleton, "Threshold Detection in Non-Gaussian Interference Environments: Exposition and Interpretation of New Results for EMC Applications," IEEE Trans. on Electormagnetic Compatability, Vol. EMC 26, Feb. 1984, p. 19-28. A somewhat more compressed version of this paper was presented orisinallv in the Proceedings of the 5th Electromagnetic Compatability Symposium and Exhibition, Zurich (Switz.), March 8-10, 1983. , "Threshold Signal Reception [21 , in Electromagnetic Interference Environment: Part III. An Introduction to Canonical Threshold Signal and Parameter Estimation," Report NTIA-83-21, Jan. 1983, U.S. Dept. of Commerce, Washington, D.C. [3] D. Middleton and A.D. Spaulding, "Optimum Reception in Non-Gaussian Electromagnetic Interference EnviOptimum and Suboptironments, II. mum Threshold Signal Detection in Class A and B Noise," Report NTIA83-120, May 1983, U.S. Dept. of commerce, Washington, D.C. Ill 78 Nl A.D. Spaulding, "Locally Optimum and Suboptimum Detector Performance in a Non-Gaussian Interference Environment," NTIA Report 84-142, Jan. 1984, U.S. Dept. of Commerce, Washington, D.C. r51 D. Middleton, Topics in Communication Theory, McGraw-Hill (New York), 1965; Chapter 3. Introduction to Statistical 161 -ication Theory, McGraw-Hill (New York), 1960, Chapter 21. "Canonical Non-Gaussian [71 -Models: Their Implication for Measurement and for Prediction of Receiver Performance," IEEE Trans. on Electromagn. Compat., Vol. EMC21, no. 3, p. 209-220, Aug. 1979. Also in Proc. 3d Symposium on EMC (Rotterdam, Holland), May l-3, 1977, paper 71-L3. t81 D. Middleton and R. Esposito, "Simultaneous Optimum Detection and Estimation of Signals in Noise," IEEE Trans. on Information Theory, Vol. IT-14, No. 3, May 1968, p. 434-444. [91 L. LeCam, "On the Asymptotic Theory of Estimatation and Testing Hypotheses," Proc. 3d Berkeley Symposium in Mathematical Statistics, Univ. of California Press, Berkeley, 1956, and "Locally Asymptotically Normal Families of Distributions," Univ. of California Press, 1960, Vol. 3, No. 1, p. 3798, Publications in Statistics. [lo] B.R. Levin, Theoretical Principles of Statistical Radio Engineering, Vol. 3, "Soviet Radio," (TransMoscow, cf. Chapter 3. lation) Ii41 - 431 - 79N2 Locally Optimumand Sub-Optimum DetectorPerformancein Non-Gaussian "Broadband" and "Narrowband" Interference Environments A. D. Spaulding National Telecommunications and Information Administration Institute for Telecommunication Sciences Boulder, Colorado 80303, U.S.A. Since the normally assumed white Gaussian interference is the most destructive in terms of minimizing channel capacity, substantial involvement in system performance can usually be obtained if the real-world interference environment (non-Gaussian) is properly taken into account. In this paper, the performance of the locally optimum Bayes detector (LOBD) for CPSK signaling is compared with the performance of various ad-hoc nonlinear detection schemes. Monte Carlo simulation results are given for both broadband and narrowband "impulsive" noise. The simulation results are compared with the LOBD theoretical performance results (which are valid only for sufficiently small signal level and for the number of independent samples N increasing without limit). It is demonstrated that these theoretical results can be misleading in actual operational use (large signal level, small N) of LOBD detectors. Introduction The real-world noise environment is almost never Gaussian in character, yet receiving systems in general use are those that are optimum for white Gaussian noise (i.e., linear matched filter or correlation detectors). It is well-known that Gaussian noise is the worst kind of noise in terms of minimizing channel capacity or in its information destroying capability. This means that very large improvements in the performance of systems can be achieved if the actual statistical characteristics of the noise and interference are properly taken into account, and there have been various significant efforts in the last few years in this area [1,2]. When confronted with real-world noise, the earlier and usual approach was to precede the "Gaussian receiver" by various ad hoc nonlinearities (e.g., clipper, hole punchers, hard limiters, etc.) in order to make the noise look "more Gaussian" to the given receiver. Later, optimum systems were derived using models of the actual noise (e.g., [1,3]). These systems are adaptive in nature and usually very difficult to realize physically. If, however, the following two assumptions are made: 1. the desired signal becomes "sufficiently" small ('sufficiently" small is precisely defined in [Z]) and 2. the time-bandwidth product is large, so that a large number, N, of independent samples from the interfering noise process can be used in the detection decision process, then a "locally optimum" detector, generally termed a "locally optimum Bayes detector" or LOBD, can be obtained. Under some rather strict conditions, these LOBD detectors approach true optimality (asymptotically) as the above two assumptions are met, and usually take the form of the "normal" Gaussian receiver preceded by one or more particular nonlinearities. In actual use, the desired signal may not be always "sufficiently small" and the time bandwidth product may not be particularly large. It is the purpose of this paper to present Monte Carlo computer simulations results of the LOBD and various sub-optimum detector performance and to compare these results with the previously derived theoretical results (based on the above two assumptions). Results for both Class A (narrowband) and Class B (broadband) interference for the CPSK system will be presented. Note that developing these results involves the actual implementation (via software) of the LOBD (and other) detectors heretofore treated only theoretically for realistic interference environments. LOCALLY OPTIMUM DETECTION The techniques for deriving the locally optimum detector for various signaling situations are well known and covered in detail in [1,2]. Here, we simply review some of the results for binary CPSK that we need to refer to later. The problem for binary CPSK is to decide optimally between the two hypotheses: Hl H2 : X(t) = Sl(t) + Z(t) o$T : X(t) = S2(t) + Z(t) octEo) = e l- (9) XI'lt~ ( > 1 OiE< - - 03, For Class A narrowband "impulsive" noise: 439 - 79N2 For actual implementation of the required nonlinearity R(x) (2), or for generation of random samples from the interference process, the models (8) and (IO) are much too complex, and much simpler, approximate models must be used. Some time ago Hall (3) developed an ad-hoc mathematically simple model for atmospheric noise (Class B), and Middleton has shown (5) that his Class B model reduces, approximately, for special parameter values, to expressions of the Hall type. The Hall model has two parameters, e and Y, and is given by r Am co P,(z) = e-A e-z2/2oi , c m=O m! p “p ; P,(Z) = (10) F(.!$)j L2 + Y"] e'2 (13) 2~ru~ and where P[E>E,] = m/A t I" ltr' (1’) ’ The closest match between the Middleton model (8) and the Hall model (13) are for the Middleton parameters o=l, A,=l, and n=4~10-~ with the and, for the envelope, 00 P(E'Eo) = esA c m=O Am fie corresponding Hall parameters 0=2 and -E2/02 om (12) The Class A model has two parameters: A and r'. A is termed the overlap index, and as A becomes large (%lO), the noise approaches Gaussian (still narrowband) and r' is the ratio of the energy in the Gaussian portion of the noise to the energy in the non-Gaussian component. An example of the theoretical performance results that we want to check by measuring (via computer simulation) the performance of actual LOBD detectors is that for Class B noise, the LOBD detector is always better than the hard-limiter detector, but the performance of the hard limiter is always within 2 dB of the LOBD performance [8]. The corresponding results for Class A are shown on Fig.2. The "degradation" is the difference between the LOBD performance (asymptotically optimum) and the hard-limiter performance. ForA=l,I-'=K)-? L=3299 ’ Figure 2. (14) 1 Comparison of the LOBD nonlinearity for Class A noise with the hard-limiter. y= J&10-". For both the Middleton model (cl=l) and the Hall model (f3=2)the 2nd moment does not exist, so the normalizing parameters R and y and set to match measured data. For f3=4, however, the first 3 moments exist. For the Hall model, the required nonlinearity is simply given by (Fig.l), BXi (15) yi=r. xi + Y2 In actual systems, the various nonlinearities operate on the magnitude of the complex received waveform sample; that is, the phasor sum of the signal vector and the noise vector. For the noise (Hall) random samples from the envelope pdf are efficiently obtained from E =Y ( -2 UT=_, ) l/2 , (‘6) where U is uniformly distributed on [O,l]; and the noise phase angle is uniformly distributed on [0,271]. Fi .3 shows the two examples of Class B noise 9 Hall) for which we will present simulation results. The case 8=2 results can be compared against the theoretical results for the Middleton model, a=1 and Ao=l. While we can use the Hall model for Class B noise, the only model developed to date for Class A interference is the Middleton model (13,14). Some time ago Spaulding and Middleton [9,Appendix] developed directly from the characteristic function for the Class A model a simple approximation to the pdf of the instantaneous amplitude. This approximation was given by the sum of two exponentials and the corresponding envelope pdf is cumbersome and not suitable for random sample generation. The following approximation, based on the above, however, - 440 can be used. Without going into detail, it turns out that the Class A APD can be closely approximated by PIE>Eo] = -& l+A where I e - where, 2 2 2 2 -EB/oo + A e-EBIol -E;/of + A e-Ez/"; (17) I o: = r'/ltr' , The parameter EB must be determined "experimentally." For our Class A case above and -3 A=0.35)(A=0.4) E =O.l. Fig.4 (r’=.5xlo shows the A;0 for our Midd!e&on Class A example and the corresponding APD (17) along with points generated via 10,000 random samples using (18). 20 ProbCE > EmI = a 0 I t\ 1 HALL. B = 4 ! -Ll I uF 1 A -20 % 0 - SIMULATION 10,000 POINTS 2 -40 -60 I ., 10-l \ (37 dB) I \ Figure 4. \ _40LliI I IIll 10-6 10-4 .Ol .1 .2 .4 P&[E Figure 3. .6 .8 .9 .95 I, .98 .2 .4 .6 II I I I .a .9 .95 .9G .99 ProbEE > Eel An example of Middleton's Class A distribution and corresponding simple approximation along with values obtained via computer simulation using the simple approximation. .99 > EJ The Hall Class B noise model for 8=2 and 8=4, the APD is normlized to the rms envelope level. Extensive use has been made of the particular Middleton Class A example, r'=.5~lO-~,A=O.35. To match this with (17), the appropriate parameters are r'=.5x10m3 and A=O.40. Using (17), random samples for the pdf of the 'Class A envelope are obtained from: The above gives the Class A approximation for the envelope. We also need the corresponding result for the pdf of the instantaneous amplitude in order to implement the LOBD nonlinearity, K(x). Since the phase angle is uniformly distributed, this pdf can be obtained from co P,(X) = x 71 d+ p(E)dE , (19) I E= owl1 -- (18) (-o:,, The , and E = where p(E) is the pdf of the envelope. result is 3 [$ (l-U)l)l/2S U1 is a coefficient which takes accou&of dimensioality; ec i,M % P ia a procesa'in the 1-th state; f(l)(z) i0 a vector function which meets Lipshits's conditions; T d X= X,)... x, i T is a.eymbol of transposition. (t,r) A standard test stimulus % uted has a form of normally diertri white noise with a(Bfffe;tf+8matrix of g-pulses with a flux intensity 3 and pulse-Vamplitudefq distribution P(A). When fixation of spatial coordinates is effected, it follows form (1) that (2) - 444 The EMIs are described by equation (*)It'z! Shown in [I] that SDEs (1) and (2) are equivalent from the point iew of one-dimensional densities $0 (&) of their solutions. Therefore, being interested only in onedimensional statistical characteristics of the EMB, a "concentrated" analog of an SDE (1) in the form of an SDE (2) should be used. It is ehown in [I] and [2] how one can construct models of EME and EM1 on the basis of systems with random structure in the form of SDES (1) and (2), when one prescribes one-dimensional density of EME components and their correlation functions in each of 1 states, the number of these states M and the intensity of conversions from the state 1 to the state r and vice versa, An example of of these models for construction of a program-controlled complex for modeling of interference is described in 131 . In the present Report the main results of [I 21 are deveio ed in the direction o? EM'Scontrol. guch problems arise in control of the modeling devices described in well as in operational EME with predetermined statistical characteristics such as a variance of interference or spectral-and-corf-elation properties. Solution of the latter problems can be useful in, organizing a frequency-dispatcherservice, controlling the spectrum and radiation power in a system of radio facilities, and the like. Let us confine ourselves to a search for control of models in the form of concentrated SDEs of type (2): because a theory of control of spatially distributed stochastic systems is developed insufficiently [4] . Here u(t) is a,dsterminate controlVo, i.e. a control is vector and u limited. While directinally varying the EXE, a control u(t) should be selected on the basis of optimization according to some criterion. Let us establish criteria, accordfng to which a control in SDE (3) should be optimized. When solving the EMC problems, it seems expedient to prescrib.e the following criteria for optimality of control: - a minimum of a variance of interference for all the statest - 1, = m2 62&J,g,u,t) ; (4a) - a minimum of difference of the interference spectrum in all the states from the required one: a minimum of a mean frequency (probability) of exceeding of some predetermined level C by an interference &[5] : P (t(),lL,t ,x XJ I, =fkn u (4c) Here G(w) and G (w) are a real radiation energy-@&ctrum and a required one, respectively; teQ?:tt; P(e) is a probability of the pact that-x exceeds a level& It can be shown that an optimization according to criterion (4a) provides for optimization cording to 8 Thus, opcriteria (4b) and (413). timization of control can be performed according to a quadratic criterion of type (Ya), which enables to use extensively the main results of a theory of optimum control [4, 61. As it is shown in [4, 63 order to find a control veotk in(l) U (t) it is necessary to solve a Bellman's equatfon of the following form: X)With G (w)wO; if this is not the case th@ eatisfying the criterion (4a) means only evaluation from be[4, 6] and majorises'criterion I in contrast to criterion I which 2s not quadratic in a genei: al case. 80~3 Here Sign Z I 1 when Z>O -1 when Z,LO i In order to obtain simple analytical results let us confine ourselves to a case when nml because taking a multidimensional case will not change anythQng in principle but will require more complicated calculations and designatieyq. ) in Let us search for S(t,x the form ,...x ) S(t A')) Here x(l) ctionnwh!.chs&ves is a BelTman*; for determination of a vector of optimum control 14 6) Equations (5a) and*(gb) should be solved with a pre@)ermined finite condition S(t , & >. From now 0k we shall assume that the control is'terminal [4,6] and, just as in [1,2], we shall accept that 'tf&t, &ccI. From a phys cal point of view, the latter assuJtion means that a change of states occurs relatively seldom in comparison with a "life time" of each EME state. In order to solve the Bellman's equations constructively it is expedient to use a method of statistical linearization 163 or a solution procedure which was described for the first time in [7] . In so doing it should be taken into account that it is sufficient to confine oneself to solution of (5a) because a vector of optimum control practtfqlly depends only on the type of f (z, u1,i.e. depend on the character can be shown that a minimization of the right-hand part of (5a) leads to obtaining an expression for optimum control, which has the following form: Let us use a statistical linearization of nonlinearities F(x) and S&gn(*) 2nd set the coefficients for X and X e ual to each other. As a result, it ss easy to show that in steady-state conditions )(it should be reminaded that ++$.~,$~%I) we have u P - u. Sign x . (8) Thus, an optimum control is a relay control and is invaria under a stimulus of the form @)(t) In order to apply technique described in [7] let ue take into account that the Bellman*s is nonlinear relative to In order to linearize use approximation by replacing as& x(& stochastic process U,Si8n.w by a random tele raph signal. Then equation (7'5will take on the following form Substituting (6) into C3a) we have: x> Index "1" ie omitted for simplicity. xx)Here only equation (7) is linearized, not the initial SBE (3) and equation (7) as it was done in the previous case. - 446 - Then, using a technique described in it can be shown that the first [71 appr&imation of solution of equation (9) in stead -state conditions . coincides with (8;Y Taking account of a sufficiently rapid convergence of a procedure of sequential approximations described in [7] it can be accepted that in each of ICI3 states an optimum control is (with practically sufficient precision) a relay control. As it was stated before, this conclusion can be broadened to cover a case of more complicated EMl2models for n31. Here one should especially single out a control of narrowband stochastic processes which can be used as models of radiation from a number of radio facilities. These models are described by secondorder SDEs of the following form: By resentin..a solution of the SDE ( 0 in the Porm of oscillations x1p t =A (t) cosq)(t) - where A(t) and q(t) are slowly varying random functions, we obtain a relay control of the form (8) but this time for the case of an envelope A(t). Thus, from the above-stated, the following important conclusion should be drawn: independent of a character of interference which are EME components, an optimum control of these interferences is a relay control in most of the problems of EMC guaranteeing. In practical realization of a relay pgrjtrol,an evaluation of a state x (t) is necessary but algorithms of obtaining such evaluations are known [4, 61 and, in principle, change nothing in beyond the this conclusion. It framework of the preXlt Report to present the principles of realization of such algorithms of evaluations of EME components, by means of electronic computers. A question has remained not discussed substantially, which concerns an o timum control of an EME when G PW) L 0 (criterion (4b)). It is o%%ous that a criterion of o timality of control is not quadra!ic in this case, which essentially complicates a theoretical investigation of this question. However, the possibilities of control of interference spectrum are very great. The following example may convince one that this is really so. Suppose we have that the 1-th component is a normally distributed lowfrequency process which is described by an SDE of type (3) when n-1 d, 0) &l,then U(t), cos c,+,t and & that a correlation ::n:~;o~eo$F!) (t) will have a form eosO$,i.e. close to a close to &-.({) correlation function of a sinusoid with a random phase. This example demonstrates great possibilities of changing an interference spectrum by means of control and indicates that it is expedient to continue to work at solution of problems of an optimum control according to a criterion (4b). If References 1. N.Buga, V.Kontorovich,Y.Polozok "A theory of systems with random structure as a constructive method of description of electromagnetic environmental models.-International Symposium on EMC, Tokyo, October 16-18, 1984. H,H.Eiyra,~.ii.~onTopoBTnu,~.~.~~0~030K 2. &'sviCtWls 3lIeKTpOMarHvlTHHX BOM3X Ha OCHO30 TQOKMM CMCT(SBdCJ'fYYatiHoa 3. A.~,~p~ce~~o3,~,H.~oHTopoaMY,B.3. ~~HJJpeC,3.~,~OJIO3OK npOrpaMMHOMO ‘3JIMpOBaHvlfi yI'IpaBJIf@MHR KOMlXI0KC pyJiH HMM ClHAyCT kll&JIbHHX nOMdX.- 5! Pqao,!982, W5,c.37-41 4. I'.E.I~OJIOCOBCMHTe3 OnTMMaJIbHHX aBTOMTMWcKMX CMCTBM npM CJIqaf4HHX 803 BWCTBMAX. -M. ,HayKa, 1954, 5. c.2 8 5. 3.~.~onosoa,~.~.~asec~~~,~.~.~o~- TO~OBMY ic sdopy napatdeTpoB '61~~ycTpi4aJIbHHXnOM%X,Onp~A0JI~D~X MX wuawe)% ~qemc~scleHa annapaTypy waJioro3Hx in ~W$possdx ClrlcTe~ napeAWL-iYiUiCJyHap0 Hfir@CklMno3M M no 3MC BpounaB 1978 c 2&S-29$ On&MM3a6. ~.Elaaa~o3,if,M.Ap~~~b~B UMR. AklHaldWR9CKMXCMCTBM CnyYafiHoa 8~ y~yp~.-M. HFyFa,19t30,c.d81. I o!i ~o~%HKo,~.~.~oHTo~oB~Y Me~08 akiwm3a MapKoBcKux ~0Aeaeti ~enaH&iHHX IIMHtlMWEJCKklXcwc~e~.-l/la~.Aii CCCP” T~~xH~W~CK~~~IKwSepHeTMKa, 1975: N.5, c.5558 l - 441 8101 - SPREAD SPECTRUM COMMUNICATIONS - INTERFERENCE CONSIDERATIONS A Tutorial Overview Phillip M. Hopkins and Donald N. Cravey Lockheed Engineering and Management Services Company, Houston, Texas Overview Of Spread Spectrum Techniques Spread spectrum (SS) communication systems utilize the wideband, noise-like properties of pseudorandom binary code sequences to provide several useful properties. These include reduction of power spectral density, multiple access to the channel, and a degree of immunity from interference. The two most widely used SS techniques are Direct Sequence Spread Spectrum (DSSS) and Frequency Hop Spread Spectrum (FHSS). Figure 1 shows the basic configuration of a pgigq.J Lpjijq Fig. 1: Functional diagram - DSSS DSSS system. A pseudorandom sequence (code) of keying rate (chip rate) much higher than the modulated bandwidth of the data signal is used to directly modulate the data carrier. The resulting signal occupies a radio-frequency bandwidth proportional to the code keying rate, hence the name "spread spectrum." The process of removing the pseudorandom code (despreading) is accomplished in the DSSS receiver by mixing the received signal with a synchronized replica of the code. The FHSS system, as depicted in figure 2, uses the pseudorandom code to select the frequency of a frequency synthesizer, which is Fig. 2: Functional diagram - FHSS then used to translate the data signal to its radio frequency (rf). The rf carrier then hops about in a seemingly random manner. At the NOTICE: Inc. receiver, another synthesizer, driven by a synchronized version of the pseudorandom code, translates the FHSS signal to a fixed intermediate frequency where demodulation of the data signal can be accomplished. The rf bandwidth of the FHSS signal is determined by the minimum and maximum frequency selections Of the synthesizer, and the distribution of the signal power over that bandwidth is determined by the number of hops, the frequency spacing between hops, the properties of the pseudorandom code, and (to a lesser extent) the modulation process of the signal before spreading. In both cases, the SS process results in an rf signal with spectral properties determined primarily by the pseudorandom code rather than by the original modulation signal. Therefore, the convenient properties of the code can be exploited by the system designer to achieve a desired systematic result without placing unnecessary constraints on the design of the information-bearing signal. Pseudorandom codes for SS systems are usually linear feedback shift register sequences. The most well-known of these sequences fall in two families, the maximum-length sequences cl,21 and the Gold codes [31. Maximum-length sequences are useful in single-access communication systems wherein interference between authorized channel users is not a major concern. The sequences are less desirable for multiple access systems because of the fact that in a set of maximum-length sequences of a given length there is a limited subset among which the crosscorrelation properties will permit separation of multiple signals competing for the same channel. Most practical SS systems, particularly multiple-access communication systems, use Gold codes for the pseudorandom spreading codes. A Gold code is formed by the linear combination of two maximum-length sequences. It has approximately the same properties as the maximum-length sequence, but has the added advantage that a large set of Gold codes of a given length can be formed with good crosscorrelation properties. These properties are summarized in figure 3. The autocorrelation function of a pseudorandom code determines the system's ability to accomplish synchronization TECHNICAL DATA AUTHORIZED FOR EXPORT UNDER LICENSE TO ALL DESTINATIONS - 440 (a) Typical autocorrelation function (b) Typical power spectrum Fig. 3: Characteristics of the Gold codes in the despreader; a desirable autocorrelation function has a single large correlation peak at each multiple of the code period and only small (ideally zero) minor correlation peaks at other delays. The Gold codes have minor peaks, called correlation sidelobes, which have been proven to be of small enough magnitude to be of relatively negligible importance. The power spectrum of the code is significant in that the rf power spectrum of a OSSS signal is determined by the power spectrum of the code. Figure 3 shows that most of the power of the Gold code is distributed over a bandwidth of twice the code keying rate, RC (called the chip rate). As a point of reference, the sinc() function shown in figure 3b represents the envelope of the power spectrum of a maximum-length sequence. SS Susceptibility To Interference An SS communication system is resistant to interference because the SS receiver is very nearly optimized for a wideband signal with particular pseudorandom properties. In order to determine the degree of resistance to interference that an SS system will exhibit, analysis is most often performed by modeling the interference source as either broadband noise or a single-tone signal. Performance analyses for both modeling techniques are provided in references [4] and [Sl for DSSS systems and in references [61 and [7] for FHSS systems. The effects of the despreading (SS detection) process are illustrated in figure 4 for the DSSS case. The desired signal, spread to appears at the receiver input bandwidth 28 accompanied 8' y an interfering signal of bandwidth much less than B . This example includes the familiar case of ti; e single-tone jammer, which represents an unmodulated carrier Within the SS bandwidth. The despreading process restores the desired signal to a bandwidth proportional to the information data rate Bd, and the same process spreads the interfering signal to a bandwidth 28 (the SS bandwidth). Consequently, the demodu 7 ator in the receiver, which accepts only the information signal bandwidth, receives only a portion of the interfering power. If the signal power is P,, the interthe information bandwidth is 25 andwidth is fering power is Pi, and the SS E' then the signal-to-interference ratio B (%R) is approximately given by this equation: SIR = (Ps/Pi) X (Bs/Bd) (1) In equation (l), the factor B /Bd is known as the processing gain (P.G.) ana is a parameter of key importance in the design and evaluation of SS communication systems. P.G. = Rs/Bi and SIR = (Ps/Pi) X (2) P.G. (3) P.G. is a measure of the resistance of an SS system (in this example, a DSSS system) to narrowband interference. In the case of an FHSS signal, the end effect of a narrowband interferer is much the same as for a DSSS signal. As shown in figure 5a, the power spectrum of the desired signal at the receiver input is spread to a bandwidth of i_B, about f . The instantaneous bandwidth, of course, on?y occupies the data bandwidth, B for binary frequency shift keying (FSK) mo !' ulation, which is much less than B Thus, the hopping of the carrier within B, p&ides the greatly increased effective bandwidth and the interference rejection property of FHSS systems. Figure 5b reveals this property by showing the effective power spectrum ‘0-~iu~Yk-0s ‘0 0 s (a) Power spectrum at receiver input ,NTEiVERENCC POWER P,B& clPSiP,llRSisDi lNTERFER,NGSlGNIL SIR (a) Power spectrum at receiver input w I s (b) Power spectrum a;tir despreading Fig. 5: Narrowband interference to FHSS system (b) Power spectrum after despreading Fig. 4: Narrowband interference to DSSS system of the demodulator input of an FHSS system. The receiver has translated the pseudorandomly hopped carrier to the center frequency at f, while translating the interferer to many frequencies in the range +B, about fi. Due to - 449 this translation, the total power of the interfering signal within the data bandwidth about f is reduced to approximately Pi(Bd/Bs). TRus, equations (1) and (3) again apply. The effect of interference on OSSS system performance using coherent binary phase shift keying (BPSK) modulation is shown in figure 6. 8101 - power becomes much greater than the desired signal power, then (even for unlimited SNR) the P approaches 1/2N. An increase in K will natureally increase this asymptotic value for This leads to the conclusion that wideband P o:'multitone jamming does cause degradation of the FHSS system, but the degradation is limited by the total available power of the jammer since the effect of each tone is a function of SIR. These conclusions .are illustrated in figure 7, which shows the effect of wideband fig. 6: Performance of binary DSSS channel with interference The P, is plotted versus Eb/N,, for several different cases of P.G., for a fixed SIR of -20 dB. The equation governing these curves is: P, = Q((2/(1/SNR + l/SIR))**0.5) (4) where Q() is the normal complementary error function; i.e., the noise is assumed to be additive, white, and Gaussian noise (AWGN). It is further assumed that the interference to the demodulator is also Gaussian, because of the filtering effect of the demodulator filter. For an FHSS system using binary FSK modulation and noncoherent detection at the receiver, there are several cases of interest with multitone jamming. If there are K jamming tones transmitted and N total hopping slots used by the system, then the probabilities of "hitting" the binary MARK, the binary SPACE, both, or neither can easily be derived. Furthermore, the probability of bit error can be calculated, given the various possible "hit" conditions, signal-to-noise ratio (SNR), and the SIR, as has been shown in references [Sl and [9] to be given by: P, = ((N-K)(N-K-I)/~~N(N-~)~) EXP~-SNR/ZI + IK(N-K)/IN(N-~)II ExpI-(z/sNR+l/sIR)-'I + [K(K-~)/[zN(N-~)~~ Ex~~-(Z/SNR+Z/SIR)-~] (5) In equation (5), if K = 0, P, is determined by the first term. If K = 1, P, is determined by the first two terms. The first term is the P over the unjammed frequencies as a funct?on of the SNR. The second term'introduces the effect of interference on the P, at the jammed frequency (K = 1). If the interference Fig. 7: Ferformance of binary FHSS channel -with interference interference on FHSS system performance, assuming binary FSK modulation. For multitone jamming, the greatest degradation occurs when the interference bandwidth is such that the interference power per slot is just slightly greater than the signal power. This power condition is assumed for the cases in figure 7, which shows P versus E /N for various ratios of jammed slo!? s to tota !? s ots in the FHSS systern. These performance curves suggest the need in FHSS systems for some form of error correction coding, which most FHSS systems employ. SS Interference To Other Systems SS signals, being wideband, may cause rf interference to conventional systems which operate in the spectral vicinity of the SS system. Scenarios for this type of interference include the cases of SS interference to conventional systems and SS interference to other SS systems. These scenarios include the familiar problems of Code-Division Multiple Access (COMA) interference and "splatter." Interference To Conventional Systems Interference to conventional systems by SS systems includes two cases of particular interest. First is the case of a conventional communication system of bandwidth much narrower than the SS signal, falling within the SS bandwidth. Figure 8 depicts the power spectrum of such a situation. The conventional signal in figure 8 is shown to be centered at frequency f,, which is approximately midway - 450 approximately 10 times the chip rate, the magnitude of the SS is approximately 30 dB less than the maximum in-band power density. Fig. 8: SS interference to conventional channel between the center of the spread spectrum bandwidth and the first spectral null. The ratio of bandwidths (conventional-to-spread) is 1:20, and the power ratio is PC/P,. When the conventional signal is demodulated, the bandwidth of the demodulator is typically constrained to include only the primary lobe of the power spectrum which is approximately equal to twice the data rate. In this example, the ratio of the demodulator bandwidth to the bandwidth of the interfering signal (SS) is Bd/B,. Using the simplifying assumption that the SS power, P,, is uniformly spread over a bandwidth 5, of twice the chip rate, then the interfering power in the conventional demodulator is: Pi,d = (Ps/Bs) X Bd = Ps/P.G. (6) and the SIR is: SIR = Pc/Pi,d = (Pc/Ps) X P-G. (7) where P.G. is the P.G. defined in part 2, equation (2). It is interesting to note that the P.G. is of mutual benefit to both the narrowband user and the SS user. In the second case of interest. the user experiencing interference from the SS signal is not located within the orimarv bandwidth of the SS signal. This interference-is often referred to as "splatter." The term is descriptive of the splatter of power from the SS signal outside its intended bandwidth due to the high keying rates typical of SS signals. Splatter is particularly prevalent in fast-hop FHSS systems. Figure 9 shows the magnitude of Fig. 9: Out-of-channel power (splatter) from an SS signal the envelope of splatter as a function of frequency separation from the center frequency of the SS signal. This plot is the magnitude of the relative maxima of the (SIN x/x) function for large values of x. Figure 9 shows, for example, that for a frequency separation of Thus, a conventional communications channel, separated from the center of the SS channel by 10 times the chip rate will experience interference (splatter) from the SS channel at a level of -30 dB relative to the center of the band. The resulting SIR depends on the relati,vepower levels of the desired and interfering signals at the receiver experiencing the interference. One scenario which is affected by splatter is shown in figure 10 Fig. 10: The near-far problem which depicts what is known as the near-far problem, discussed in more detail in the context of interference between SS users. Interference Between SS Systems SS communication systems operating in a shared range of frequencies also interfere with each other to varying degrees. Indeed, one of the most common applications of SS is to provide multiple access to a common communications channel. Code Division Multiple Access. When several users OCCUDY a CDMA communications channel simultaneously, they experience interference from each other, the degree of which is determined by the parameters of the particular CDMA scheme and by the relative powers of the various users. When all users have equal power (e.g., when they all receive from the same satellite, are separated far enough to eliminate a near-far problem, and occupy the same bandwidth) then the interference can be analyzed conveniently. For the case of a digital data transmission system using CDMA (DSSS) with up to 502 users, the probability of error for the signal experiencing the interference is shown in figure 11. The figure shows the effect of "graceful degradation"; i.e., as the number of users of a DSSS CDMA system increases beyond design limits, the performance of the channel for each user degrades incrementally rather than catastrophically. In figure 12, a similar situation is shown for an FHSS CDMA system, using binary FSK as the basic data modulation scheme. This figure shows more pronounced degradation with an increasing number of CDMA users than in the case of DSSS, in addition to the generally wurs(? performance due to noncoherent FSK modulation _ ~~- 451 81 OI - If the interfering transmitter is nearer to the receiver than is the desired transmitter, then the P.G. is diminished. As an example, if the interfering signal is at center frequency, which is common in CDMA systems, and if SIR > 10 is required for reliable communications, then it is required that: If this power ratio cannot be achieved, then the interfering transmitter will jam the receiver. Self-Interference In SS Systems Fig. 11: ,P;;P;;;1ance of a binary DSSS CDMA In the preceding sections of this paper, emphasis has been on the interference between an SS communications channel and other user channels. There are also interference effects that involve only the SS receiver; these are self-interference effects. The most significant of these effects is the self-noise due to partial correlation of the pseudonoise code. Partial Correlation Partial correlation of the pseudonoise code occurs in the despreader when one or both of the following occur: 1. The correlation operation is over less than a full period of the code 2. The correlation process is nonlinear, as when a square-law detector is used. Figure 13 shows an example of partial correlation. In this example, the autocorrelation Fig. 12: Performance of a binary FHSS CDMA channel and demodulation. In each case shown in figure 12, the probability of error approaches an asymptote defined by the following equation: 'e = K/2L (8) where K is the number of interfering users and L is the length of the spreading code in chips. The Near-Far Problem. When the receiver (location-1 in figure 10) is nearer to the interfering transmitter (location 2) than to the desired signal transmitter (location 3), then the interfering signal has a natural power advantage over the desired signal. In such a situation, the SIR is: SIR = CP3/P2(fd)l(d12/d,3)2 x P.G. (9) where P2(fd) iS the power density of the interfering signal at location 1 at a frequency separation fd from the center frequency of the interfering signal. Fig. 13: ,PiP;ial correlation of a pseudonoise function contains a relative maximum at a phase (tau) delay of jTc, of amplitude k. If the despreader synchronizes to the partial correlation peak rather than to the true (zerodelay) peak, then the despreading operation will be degraded by a factor of k2. Synchronization The analyses in the preceding sections of this paper have all been for steady-state conditions; i.e., under the assumption that the despreader in the SS receiver is synchronized to the incoming pseudonoise code. In fact, synchronization of the despreader in the presence of interference is no small problem. Obviously, if proper synchronization cannot be achieved, then the P.G. of the SS system can-not be realized. - 452 A particular case of interest is the case of false synchronization due to partial correlation at high SNR's. When the partial correlation peak of figure 13 is of sufficient magnitude that it is detectable in the correlation detector, then the despreader may synchronize to the partial correlation peak. This can occur if: k*(SNR) > SNRmin (II) where SNRmin is the SNR at the threshold of acquisition. Figure 14 is an example of acquisition teSt data from the Electronic Systems Test Laboratory at the National Aeronautics and Space P.G. of the system. A significant example is the case of the nearby transmitter, either SS or narrowband, which overcomes the SS P.G. by sheer power. Similarly, SS channels cause interference with conventional channels near the frequency of the SS channel, but the inherently low power spectral density of the SS signal mitigates the problem. Again, the near-far problem represents a significant example of interference. Acknowledgements Portions of the work on which this paper is based were performed for the NASA Lyndon B. Johnson Space Center on contract NAS 9-15800. In particular, the authors wish to acknowledge the cooperation of NASA in providing the test data presented in this paper. References Cl1 Hopkins, P. M.: A Unified Analysis of Pseudonoise Synchronization by Envelope Correlation. IEEE Trans. Commun., vol. COM-25, Aug. 1977, pp. 770-777. r.21Pickholz, R. L.; Schilling, 0. L.; and Milstein, L. B.: Theory of SpreadSpectrum Communications - A Tutorial. IEEE Trans. Commun., vol. COM-30, May 1982, pp. 855-884. c31 Gold, R.: Optimal Binary Sequences for Spread Spectrum Multiplexing. IEEE Trans. Inform. Theory, vol. IT-13, 1967, pp. 619-621. Fig. 14: Typical test results showing false synchronization effect Administration (NASA) Johnson Space Center in Houston, Texas. The curves show acquisition time versus SNR in the despreader bandwidth for the S-band transponder receiver of the Space Shuttle. It is clear in this figure that at very low SNR (noise limited region), acquisition time increases dramatically as SNR decreases. This is an expected result because the detector is operating below its noise threshold. In the region of 0 to 10 dB SNR, the synchronization time is well-behaved and conforms to predictions. The interesting effect in figure 14 is that for SNR > 10 dB, the acquisition time increases rapidly because of false locks during the acquisition search mode. This result was predictable from an examination of the partial correlation properties of the despreader. Summary and Conclusions SS communication systems, although inherently resistant to most types of interference, are susceptible to interference if the interfering power is sufficient to overcome the c41 Simon, M. K.: The Performance of M-ary FH-DPSK in the Presence of Partial-Band Multitone Jamming, IEEE Trans. Commun., vol. COM-30, May 1982, pp. 953-958. c51 Smith, I. R.: Trade off Between Proc- essing Gain and Interference Immunity in Co-Site Multichannel Spread-Spectrum Communications. IEEE Trans. Commun., vol. COM-30, May 1982, pp. 959-965. C61 Solomon, G.: Optimal Frequency Hopping Sequences for Multiple Access. AD915892, Proc. 1973 Symp. Spread Commun., vol. 1, pp. 33-35. r.71 Spellman, M.: A Comparison Between Frequency Hopping and Direct Spread PN>as Antijam Techniques. IEEE Commun. Mag., vol. 21, no. 3, Mar. 1983. C81 Torrieri, 0. J.: Principles of Military Communication Systems. Artech House (Dedham, Massachusetts), 1981. c93 Viterbi, A. J.: Spread Spectrum Communications - Myths and Realities. IEEE Commun. Mag., May 1979, pp. 11-18. - 453 8202 - EMC PROBLEMS IN DATA TRANSMISSION OVER INDOOR POWER LINES USING SPREAD SPECTRUM TECHNIQUES Lehrstuhl Postfach K.Dostert, Member Universitat Kaiserslautern fi.ir Grundlagen 3049, D-6750 Summary Practicallv every building contains an electrical- power “distribution network, which can additionally be a valuable communications medium for data and control signals. AS indoor power lines are heavily stressed with electrocommunications special interference, magnetic techniques will be needed to provide secure information transmission. Furthermore electromagnetic compatibility with equal or similar communications systems operating at the same network is paramount. The use of spread spectrum techniques, which are generally applied in the range of high frequencies, can help to meet the mentioned requirements /I/. With spread spectrum techniques some interesting features become additionally available, such as multiple access and selective calling /3,4/. This paper shows, that EMC prowhich are involved by application of blems, spread spectrum techniques for data transmission over indoor power lines, can be overcome by careful system design. Especially using the line voltage as global reference will help to solve EMC problems and at the same time significantly reduces receiver costs. Measurements with an experimental set-up show negligible synchronization errors and confirm the feasibility of the global reference concept. Introduction Numerous measurement, control and supervision tasks in private and commercial buildings can advantageously be performed autamatically, e.g. by a personal computer or a microcomputer installed anywhere: Control and supervision of central heating, or remote and illumination air conditioning control and supervision of various consumers of proelectrical energy, e.g. by clock-controlled grams or supervision of fire and burglar alarm systems. For performing those tasks, it will be necessary to transmit and receive control signals, time information as well as temperature and consumer data to and from many different places. So realizing such systems will first of all call for large scale wiring to connect the devices which are involved. In existing buildings additional wiring is almost impracticable. A data network which was planned communications during the construction phase of a building often turns out to be inflexible afterwards. The electrical power distribution network however IEEE der Elektrotechnik Kaiserslautern, W.Germany reaches almost every place in a building and existing thus represents a val uabl e, already medium for communication purposes. The power line channel The power distribution network of a building may be heavily stressed with electromagnetic interference from various sources. Narrowband and broadband continuous wave interferers may appear as well as spikes or hazards caused by switching, e.g. through devices such as thyristors and triacs. Measurements revealed, that this inherent interference power density is especially high in the frequency range up to about 40 kHz. So it will be advisable to fix the lower limit of the frequency range for data transmission at 40 kHz to keep most of the inherent interference off. Some interferers, such as dimmers, incorporating triacs or thyristors, may cause a 30dB enhancement of interference power density in a frequency range up to several hundred kHz. Furthermore interference power density significantly differs in one-family houses and industrial buildings or laboratories. The upper limit of the transmission band is given at 146 kHz, determined by the start of the longwave broadcast band. The transmission signal amplitude must not exceed I.1 V (peak to peak) in the frequency range 30...15OkHz, according to the rules of the “Deutsche Bundesposttt. Above 150 kHz the rules VDE 087516.77 apply, allowing amplitudes of only some millivolts. This means, that the spectrum of a data transmission signal must be cut off at 150 kHz sharp. Fig.1 shows four spectra for comparison. The records 1, 2 and 3 were made in a onefamily house; record 4 was made in a factory building. Trace 1 belongs to a transmitted signal with an amplitude of 1.1 V which is swept from 40 kHz to 180 kHz. TRg’ lower trace (2) shows the interference power density spectrum in a one-family house. It can be seen, that the interference is especially strong below 35 kHz. In the range of the transmitted signal the interference power density is generally more than 50 dB below the power density of the transmitted signal. So data transmission over power lines would be no problem, if this performance could always be expected. Trace 3 in Fig.1 shows the interference power density spectrum caused by a dimmer with a load of 100 W. At 45 kHz the inter- - lO_OkHz Horiz.scale: Fig. 1: FnFstra 20kHz recorded Vrqt. scale: IOdB at mdoor power ference power density is now only about 15 dB below the power density of the transmitted signal and less than 30 dB up to 150 kHz. If the transmitted signal is attenuated on the way to the receiver more than lOdB, severe detection problems may occur. In a factory building the situation may be even worse, as trace 4 indicates. In one-family houses measurements show attenuation values of less than 1OdB in the frequency range from 40 kHz to 150 kHz even with cable lengths close to 100m. In factory and laboratory buildings however attenuation values of 20dB and more can be measured at cable lengths below 20 m. This is due to large capacitors installed for reactive power compensation. In connection with attenuation, the impedance of indoor power lines in the frequency range from 40 kHz to 150 kHz is of importance. Measurements revealed, that an approximative value of 33 Ohms, which is given e.g. in /S/, does not apply in most cases. The measured impedance values varied from 15 Ohms to 60 Ohms, and significant amplitude and frequency dependent nonlinearities could be registered. Furthermore considerable impedance variations occurred with time. The impedance variations make proper matching of the transmitter output to the power line network impossible; so varying attenuation will be the consequence. Amplitude and frequency dependent nonlinearities can distort the transmitted signal severely. In the following the application of spread spectrum techniques to the hostile environment of indoor power lines is examined with respect to EMC, interference immunity, synchronization problems and bandwidth efficiency. Dealing with spread spectrum systems will first of all call for careful consideration of the receiver synchronization. Synchronization concepts Receiver synchronization in spread spectrum systems generally causes high effort /2,3,4/. As in the proposed applications numerous receivers will be needed, many of the known synchrornzation concepts /4/ are ruled out by their costs. Merely concepts with transmitted reference promise acceptable solutions: A reference signal could be fed into the power distribution network being globally present at each 454 receiver and transmitter, delivering all necessary Such an clock and synchronization impulses. additional signal however cannot be recommended for the following reasons: An extra channel within the transmission bandwidth would have to be reserved for the reference. Selective attenuation may separate some receivers and transmitters from the reference temporarily or completely. Similar spread spectrum systems with such a synchronization concept are not generally electromagnetically compatible. Therefore careful blocking of the reference signal towards neighbor buildings is necessary, whereas spread spectrum signals normally must not be blocked, due to their low spectral power density. In this paper the power line voltage is proposed as reference signal. This reference is of course available everywhere in a power distribution network, practically unaffected by interference and with high stability of frequency and its use causes no EMC problems. The power line frequency can be multiplied appropriately by phase-locked loop (PLL) circuits to generate carrier and clock frequencies. Initial synchronization may be performed at each transmitter and receiver periodically at the zero-crossing instants of the line voltage. The proposed synchronization concept involves two difficulties: During power failure communication will be interrupted, and a phase shift of the power line voltage due to cable inductance and load causes synchronization errors. The second point is especially severe, because the phase shift between the line voltages at two different places is multiplied by a factor, which gives the ratio of spread spectrum clock frequency to power line frequency. Let us for example consider a direct sequencing system with a PN clock frequency of 40 kHz, requiring a multiplication factor of 800. For proper receiver operation the synchronization error must be kept below l/10 of the PN clock p period /2/. So the maximum phase shift an !I between the line voltages at transmitter receiver is given by YL& & = 40450 . The result of eq.(l) may initiate some doubts, whether the proposed concept is realizable in practice, but experiments which are discussed later in this paper have shown that it will function satisfactorily. System concepts This paragraph gives a short review of basic spread spectrum system concepts, which are applicable to data transmission over indoor power lines. The concepts will not be discussed in detail, but only some characteristic features, which concern the topics of this paper are pointed out. Phase hopping systems Phase hopping direct sequencing spread spectrum systems are easy to construct when using the power line voltage as global reference. Some expense however is necessary to fulfil the the longwave EMC requirements concerning broadcast band: A low pass filter must be provided at the transmitter output to achieve a sharp cut-off of the transmitted spectrum at - 150 kHz. Furthermore phase not optimally exploit the tran&ission channel. 455 8202 - hopping systems do bandwidth of the Frequency hopping systems Frequency hopping systems are more COmplex than phase hopping systems, because appropriate frequency synthesizers are needed. Frequency hopping systems have excellent EMC because the transmitted signal specproperties, trum is precisely defined with sharp cut-offs. efficiency is high, Furthermore the spectral because each spectral line has equal weight. Chirp systems Chirp systems with e.g. linear chirp can optimally exploit the transmission bandwidth and offer excellent EMC properties due to a welldefined spectrum with sharp cut-offs. Multiple calling and synchronization as access, selective we11 as despreading are more difficult to achieve as for phase hopping and frequency hopping systems. REtElm Fig. 2: Experiments An experimental phase hopping spread spectrum system was built and tested in onefamily houses. In Table 1 the salient features of the system are given. : 102.4 kHz = 2048.5OHz spread spectrum modulation : O/180’ pseudonoise phase shift keying (PN-PSK) PN-code : carrier frequency length PN clock frequency PN-code selectable according to PN-code length: (8) 400 Hz, ( 16) 800 Hz, (32) 1.6 kHz, (64) 3.2 kHz, (128) 6.4 kHz, (256) 12.8 kHz period : 20 ms : power line voltage (50 Hz) sync. reference sync. intervals Tab. 1: Salient features tal set-up System selectable: 8, 16, 32, 64, 128, 256 chips 20 ms of the experimen- of the experimental chronization is performed. If synchronism has already existed, the set-impulses have no effect. The PN-generators in transmitter and receiver are equal and constructed with programmable memories (PROM) and’ settabfe read only address counters, according to the principle described in 161. Up to this point the functional blocks of the transmitter and the receiver are equal; the following blocks will perform different tasks. In the transmitter the carrier frequency is fed to a biphase modulator performing the spread spectrum modulation corresponding to the PN-code. The spread transmission signal is now appropriately amplified and fed to the power line by means of a coupler. At the receiver the spread spectrum signal from the power line passes a coupler and a bandpass filter and is fed to a biphase modulator, which is supplied with a synchronized version of the PN-code contained in the received spread spectrum signal. At the output of the biphase modulator we have the despread signal which is now filtered by the interference suppression bandpass. At the output of this bandpass a power meter is provided for the measurements which are discussed now. Measurements Measurements were made in three different one-family houses with a PN-code length of 256 chips and a PN clock frequency of 12.8 kHz, giving a chip duration of about 78 us. The measured quantities were: The synchronization error without load at the receiver wall-plug and with a load of 2 kW description In Fig.2 a block diagramm of the experimental set-up is shown. The transmitter and the receiver both contain a zero-crossing switch to which the power line voltage is fed. The switches output proper rectangular pulses for the following phase-locked loop synthesizers which generate the carrier frequency (transmitter> and the PN clock frequencies. With each leading edge of the zero-crossing switch outputs set-impulses are generated in the transmitter and receiver, setting the PN-generators to the same well-defined state. So every 20ms initial syn- Block diagram set-up signal attenuation due to cable without load and with 2 kW load receiver wall-plug interference switched off. ments within power with length at the transmitter Synchronization error. During all measurethe synchronization time error stayed the limits of 2 2.5us. There was no signi- - 456 ficant change to be noticed, when a 2 kW load was switched on at the receiver wall-plug, even with cable lengths close to 1OOm between transmitter and receiver. So a negligible degradation due to synchronization errors can be expected when using the proposed synchronization concept, because the maximum error is only 3.2% of a chip. At a PN clock frequency of 50 kWz, which is the maximum value for the transmission bandwidth considered here, the maximum error would be 12.5%. This error would lead to a worst-case degradation of about 10% - see e.g. 121. Signal attenuation. Fig.3 shows a plot of the attenuation as a function of the cable length in one-family houses. The solid line holds for no load at the receiver wall-plug and the dotted line was measured with a 2 kW load at the receiver wall-plug. The slightly higher attenuation for the dotted line however is not due to increased synchronization error, but due to the decrease of power line impedance caused by the load. - Conclusion Applying spread spectrum techniques for communication purposes over indoor power lines can be effective against electromagnetic interference, which heavily stresses that transmission channel. So a new promising access to using power distribution networks as communications media is opened. This paper shows that using the power line voltage as a global reference for synchronization of spread spectrum transmitters and receivers is a feasible approach to solve EMC problems and significantly reduces system costs. Measurements with an experimental set-up have shown, that attenuation and interference problems can be overcome, and that the global reference concept using the power line voltage exhibits excellent synchronization performance. For further applications to data transmission, frequency hopping systems can be recommended, especially because of their excellent EMC properties, together with the proposed synchronization concept. References 0, 0 Fig. , 10 20 CABLELENjTH Ill T. Dvo%k, H. Ochsner, “Low Tension Power Line as a Fast Digital Data Transmission Channel”, Proc. of the 4th EMC Symposium, Zirich (1981) /2/ “Spread Spectrum Communications”, edited by CE. Cook, F.W. Ellersick, L.B. Milstein, D.L. Schilling, IEEE Press and John Wiley & Sons (1983) 131 P.W. Baier, M. Pandit, “Spread Spectrum Communication Systems”, in: Advances in Electronics and Electron Physics, VOL. 53, Academic Press (1980) 141 R.C. Dixon, “Spread-Spectrum John Wiley & Sons (1976) 151 J. Gabel, “Elektro-1nstallationsnetz Informationsnetztt, etz Bd.104, (1983) 30 me 3: Attenuation of a PN-PSK spread spectrum signal transmitted over indoor power lines Interference power. The interference power, measured at the output of the interference suppression bandpass with the transmitter switched off, is generally 39dB below the power measured with the transmitter on. A dimmer decreases the ratio to 38dB. /6/ Systems”, Heft wird 1 P.W. Baier, K. Dostert, M. Pandit, West German Patents: DE 3020463 C2 and DE 3020481 C2 ( 1982) - 457 83 - 03 THE EFFICIENCYOF TRE CELLULARSPREADSPECTRUMRADIOTELEPHORE L.E.Varekin All-UnionTelecommunioation Instituteby oorrespondenae Moscow,USSR The efficienoyof the use of Prequenoy spectrumin cellularmobileradiotelephonesystemsis discussed.It is shown that the efficiency of oellular spreadspectrumradiotelephone is bigher with pseudo-noisesignalsand pulse-time modulation rather than with frequenoy modulation. use of a matchedfilterend a post-deteotorintegratorwhich resultsin the increasedSSRT effioienoy.The effioienoy of SSRT is shown to be muoh higher then that of Fb%T. 2.Cellularradiotelephone effioienox When oellularprinoipleof oonstruction of rad&ot$.ephoneand digital mobile oommunioationsystemsis used 1. Introduotion [I] a big city territory is divided into a large number of oells (sones) Cellularprinoiple of eonstruotion in the form of hexagons.Base stationa of radiotelephone systemsmakes itpos(BS) are situated in the centreof sible to increasetheir effioienoyand each cell and provide oommunioation capaoity h,2]. There appeareda great with mobile subaoriberstations(MS). number of reviewsand surveysdisaus- Suppose, radiusof a serviue 5one is sing oellularradiotelephone (see,for Ro, radiusof a oell is R. The number example,[3.08-j 1. In this paper the of BSs is efficienoyof the use of speotrumin L =1,21(Ro/R)2. (1) cellularmobileradiotelephone systems With Ro=30 km, R='lkm the numberof under oonditionsof UBF multipathtransmissionis disouseed.The effioienoy BSs is L*lO89. Proteotive distanoe betweencells using the same frequency here means the number of aotiveusers (or ahannels) per a ? NR5 frequenoy channelsis D. A minimum numberof frequenoy ohannela is C = (D/R)'/S. band. The results of comparisonof oommuoellularfrequently modulationradiote- Assume,one BS ia establishing and lephone (FMRT) with that of oellular niaationwith 1HSs simultaneously the total speotrum width, used by the radiotelephone (SSRT) spreadspectrum BS, Is FBS. The number of aotiveusers with pseudo-noisesignals (PBS) are (ohannels)is given.Suoh a oomparisonis being oarried out for the first time. In SSRT M = Ll = 1,21(~~/~)~1, (2) telephoneinformation is transmitted band with the aid of pulse-time modulation the totel transmission(reoeption) (PTR)and PM. In SSRT optimumreoepF=CF BS (3) tion of PNS is obtainedthroughthe - and the efficiency '6= M/F = Ll/CFRS . 458 (4) - Since with FM CMCS uses the principle % singlechannelto a singleuser:' the tidth of the ape&rum FBS=lFo,where F, - the bandwidthof the frequency channel. The less the cell radiusR,the greater L and henoe the higher the efficiency of the oellularmobile communioationsystem (CMCS). As the minimum Hence, the efficiencyof FMRT is number of frequenoychannels c grows (9) + FM = L/CF, . (i.e. as the tolerable interference decreases)or as BBS grows the effioiNumerous experiments showed that ency deoreaaee. Since the dependence of q on R is,quadratioit is reaeona- UHF attenuationindex within large cities is approximately n =3,5 with Vs 9 ble to use small radius cells of the Let F,=25 kH5, Roe30 km, R=l [9, lOI* order of R~0,5..~1 km for CMCS with a km. Table 1 shows the valuesof the large numberof users, minimumnumber of frequencychannelsC When the principle "a signalchanand the efficiencyr FM for different nel to a s9gnal user" is used, mutual qualityof informationreceptionwhich interference in CMGS is determinedby is characterized by signal-to-interfethe signal-to-interference ratio for rence ratio q&. powerc9, IO, 111 Table 1. FMRT efficiency -1 , (5) = (D/a)9* =Ps/Px where PS, PI - signaland interferenoe power at the-BS (or MS)receiverinput* 13. - UEF attenuationindex, V=V(Q,n)radiationfunctionof the web network of interferingstations, Q=R,/D- the number of ahords. Aaoordingly, the minimumnumber of frequencychannelsis c = (s2V)*'n/3 . (6) Henoe, the effioiencryof a cellular radiotelephone is '6= 3Ll/(J'2V>2'n PBS . (7) It is to be noted that V dependson n and Q, i.e. on the R, and D values. It shouldbe noted for comparisonthat the efficiency of centrelized mobile communicationsystemswith FM is 40 activeusers/MUz.Therefore with q&=20 dJ3the efficienoyof the cellularsystem is 68 as high even with multipathing. The last crolumn in Table 1 shows the valuesof the total number of a@tive users (channels) in FORT, with the total frequenoybandwidthbeing 16 MHZ. 3. CellularFM radiotelenhone Under conditionsof UHF multipath trensmiseionFM has no advantagein interference immunity. Thereforesignal-to-interference ratio q& at the outputof PM receiveris approximately equal to signal-to-interference ratio 9* at its input [I], i.e. (8) 4. Cellularspreadspectrum radiotelephone Since PNSs allow to providecode division of signals overlapping in spectrum,it is not neoessaryto use the princriple"a singleohannelto a singleuse+ in SSRT when spreadspeotrum PNSs are used. Thereforewe asaume PRS speotrumwidth to be equal to - 459 BS spectrum width, i.e, Fc=FBS. In the BS frequency band interference occurs among simultaneous users. When PNSs are used separation of user& is possible only at the expenoe of separation of PNSs acoording to PNS structure, i.e. to establish oommunication wfth 1 users BS uses 1 different PNSs. This separation is also known as aode division (CD)p,12]. Interference is especially dangerous when there is no automatio COntrOl Of levels (ACL) of signals arriving from mobile units situated at different distances. Without ACL the dynamio range of signal levels can reaoh very high values whioh results in signifioantly deoreased efficiency of CD with PNS. Therefore we assume that ev8ry MS has an ACL system which is working in aocordence with oontrol signal from BS and which is normalizing signal power at the BS input making it equal to signal power from MS situated at the boader of th8 working cell, i.e. *s- aPt,/Rn , Where a- normalising coefficient, Per - transmitter power. Acoordfngly, interference power is -4 Pt *xl-Rn l-l )+1(3c) 01) where the first summand is stipulated by interference in the working cell and the seoond aummand - by interference in the adjaoent cells of the web network. Assuming l>>l signal-to-interference ratio at the input of the optimum SSRT receiver is found from (lo), (11) 92=Ps/Pnx21"[1+(3C)-42V -11. (12) With PNS it is to have C=4. For n=3,5, V=9 the second term in (12) is signifioently less than the first one. Therefore, approtimately g2 = 2/l . 033 Numeral 2 in the numerators of (12), 83 - 03 (13) is the consequence of the use of PNSs, the duration of whioh T is twioe less than the inter@& of discretization equalling l/2W, where Cam4kH% ia the upper frequenoy of the telephone information spectrum. b Rultipathing and PNS reception UHF spreading within large c&ties is multipathing due to RFs reflection from different objects in the oitiessrr Under conditions of UHP multipath transmission and when PNSs are used it is possible to separate the beams and then to integrate (store) them p3].In this case the optimum receiver (Fig?) consists of a matched filter (IQ?),an en-' velope deteotor (D), an integrator(I), a solver (S) and a demodulator (DM). 1 Fig,lo Optimum PNS receiver with post-detector integrator 1 Fig.2a shows multipath signal at the envelope deteotor output. A means a beam delay time interval (pulse pack duration). Beams are shown idealized and with equal amplitudes and durations h/S, where S - the number of possib18 beams. The number of real beams is r. For Fig.2a 6~20, rell, To separate beams it is neoeasarg that A/Shl/P,. If the beams are separated, their noncoherent store becomes possible&et integrator memory be A and its pulse characteristic is of the form presented in Fig.2b. Fig.2o shows a pulse at the integrator output. Its leading end trailing adges are the result of oomposition of a random number of beams falling into na frame" of duration A . Therefore, in the general case leading and trailing edges of a pulse at the integrator output are nonsymmetric.But on an average this pulse fn its form is close to triangle with a base equal to When information is transmitted 2A. - 460 with the aid of PTM an important factor is unshifted beginn3ng of trailing edge which in the absence of noise always begins at the moment t= A. Therefore, the solver must find the peak's maximum at the integrator output and then to determine the moment of the beginning of its trailing edge, i.e. the beginning of fall time. - Law but at other momentatwhen there is no beam)obeys the Rayleigh Law. Therefore, at the moment t= A (Fig.2~) voltage at the integrator output is the SUUI where the first summand in the 35.&t part is the sum of "signal plus nOiSe" voltages for r beams and the second sutnmand is the sum of the remaining S-r noiae components. The mean value of a random variable p is f =ml{y] = and its dispersion =rcc+(S-x)&/26, is G g2=rG2+(s-r)(4-rr)62/2. For those values of voltage at the integrator output contribution into which is made only by noise components we have the sum '=$Yjn* 3: Its mean value % and dispersion 0 I ?ig.2. Time 1 diagram in a circuit with post-detector integrator Let us find statistioal characteristics of a signal at the integrator output. To do this let us firstly determine characteristics of random process at the output of the matched filter.At the moment of arrival of a certain beam the voltage at the output of the matched filter contains signal oomponent d and noise component with dispersion@*.At the beam moment Signal-to-noise ratio is q~=~2/G2=2Ef'Ro=2PsT/?Jo , (14) where E=P,T - PNS energy, Ps and T ita power and duration, W. - noise power spectral density. Distribution of voltage at the output of the envelope detector obeys the Rayleigh-Rice =m-,(~}=&'2G, @%2 - s(4--3~) G*/*. In the process of making a decision the quantity 7 = p-q is formed. Its mean value with an accuracy to small values of higher order with &/@>I is approximately t 2 rd, and dispersion @Iz X SCY2 *Thus, signal-to-noise ratio at the integrator output is q2 ‘.@g = 22 q, ; , 05) i.e. when I increases signal-to-noise ratio at the integrator output inorea88s according to square low.With r>fi signal-to-noise ratio at the integrator output q2 is greater than the initial signal-to-noise ratio at the output of the matched filter q:, i.e. onlY ff ryfithe integrator is expendient. 6. The efficiency of cellular radiotelephone with PIVSand PTM Signal-to-interference ratio at the - 461 output of the matched filter 2 =2Bp2, c131 40 (16) where PIG base With (131, B * FBs/4W . (161, we @;et 2 40 = (17) Q&w 83 - (17) . (18) %sx qz r2/J?-8WaS with r Ss indepen;ent of FBS and 1. Having values of q, it is possible to find q sS. It is to be noted that the applioation of (19) is possible provided that qz zf, otherwise the quality of reception will be poor* (22) . From (22) it follows that the gxeater q: and r, the higher the quality of the reception of telephone messages. Substituting (18) into (223, caloulating FBS and substituting PBS into (4),we find the efficiency of the cellular radiotelephone with PNS and PTM: . Tss = Lr2/fi 8qSs m2AS According to(4) the SSRT efficiency is (19) 03 (23) Table 2 shows the efficiency values yss for different q: values and different numbers of real beams r, other parameters being LB1089 for R,=30 km, R=l km; A=5 mks, s=20; w=4 wz; c=4. It has to be noted that with the requirement FRS = F/C = 4MHz is fulfilled. A /S s l/Fss Table 2. SSRT efficiency The quality of transmission of telephone information with the e3d of PTM is characterized by signal-tonoise ratio at the-.autputof the-PTM demodulator. As it was pointed out PES duration is equal to half an interval of disoretization,i.e~ T=l/4W. The amplitude of the useful component at the output of the PTM demodulator is proportional to half PNS duration, i.e, Atmx=1/8W. Let CfJT stand for RMS van lue ,ofshifting of a trailing edge of a pUlSe at the integrator output at the moment t-A (Fig.203. Signal-to-noise ratio for voltage at the output of the PTM demodulator is 1/8W . RES value of 14 is Cl trailing edge 6,~ tiA/q' 9 I I I I 2 I 21 I 33 I 40 I 1 4 1 271 39 1 46 I 34000 I 544000 17000 1 272000 1 8 1 331 45 1 52 1 8500 136000 It can be seen from Table 1 and Table 2 that SSRT efficiency is significantly higher than that of FMRT. It is accounted for by, firstly, FRRT efficiency decreasing due to multipathing and, seoondly, by SSRT efficiency increasing with the increase of the number of beams. (20) References shifting (21) 11 1 Microwave where q2 - signal-to-noise ratio at the integrator output. With (151, (20),(21) we get signalto-noise ratio at the output of the PTM demodulator: PI Mobile Communications. Ed. by W.C.Jakes Jr., A Wiley-Interscience Publication. Tr. from Eg. Ed. by M,S.Jarlykov,M,V.Tcherniakhov. M., Sviaz, (1979). Varakin L.E. Rural radiocommunioation and code division of chennels. - Electroaviaz. No 10, (1973). - 462 / - Cellular. Speoial Report. - "Comcxl munioation News", No 2, (1984). [I4 [I5 communi.cation systems with pseudo-noise signals. - "Electromagnetic Compatibility 1983", 5th Simposium and TeclhnioalExhibition on Electromagnetio Compatibility, Zurioh, (March 8-10, 1983), Williamson J, Cellular radio: a market on the move. - "Telephony", No 7, (1984). Stoffels B. Cellular - Up and Running. - "Telephone engineer and managment", No 8, (1984). Nelson M. Cellular Mobile Radio Systems Bring New Service FlexLbflity for Users. - "Communioation News", No 8, (1984). 7 Take your telephone with you. "Financial Times",(August20,1984). 8 Cellular radio takes to the road.L-3 '*CommunioationSystems Worldwide", No 5, (1984). Varakin I.E. Interferenoe Lmmunity [I9 of digital cellular spread-speotrum mobile oommunioations systems' - "Electrosviaz", No 12, (1982). 10 Varekin L.E. Electromagnetic oompatibility of cellular mobile 6 Cl Cl c3 p. 537-542. IllVarakin L.E. II [I21 c131 043 Spread-spectrum digital cellular mobile communicatfon system. - "4 World Telecommunication Forum", p. 2; Geneva, (1983). Varakin L,E. Theory of signal systems. - M.: "Sovetskoje radio'; (1978). Turin G,L.Introduotion of SpreadSpectrum Antimultipath Techniques and Their Applioation to Urban Digital Radio. - "Proceedings of IEEE", No 3, (1980). Tikhonov V.I. Statistical radioteohnology. - lH.t"Sovetskoje radie", (19663. 8404 - 463 - Comparison of Spectrum Efficiency of CDMA and FDMA Mobile Radio Systems Heinz Ochsner Federal Institute of Technology Institute for CommunicationTechnology Zurich, Switzerland It is a general belief that mobile radio systems using spread-spectrumtechniqueswaste precious bandwidth because of their moderate spectrum efficiency. The goal of this paper I_is to demonstrate where Spread-Spectrum coae Division Multiple Access (SS-cDMA) Systems are as effective as their Frequency Division Multiple Access (FDMA) competitorsand under which circumstancesthe former are inferior. Most of the figures presented stem from various authors, in this paper, however, they are adapted to the CCIR definition of spectrum efficiencyand consider the spatial organisation of a radio telephone network, as well as network the traffic behaviour of the subscribers. INTRODUCTION In a multiple access communication system, the subscribers use signals that are elements of a signal space. If W is the bandwidth that all signals have to share, and T is the duration of one signalling element, e.g. one bit, then the number of dimensions of the signal space is D = 2-W-T . whole signalling interval T. Nevertheless, in a signal space many such signals exist and can be found. A suitable set of codes has the is additional property that each code orthogonal, or quasi-orthogonal to its own time-shiftedversion. In reality, the signals of any of these multiple access schemes occupy more than one dimension. Especially practical codes are not completely orthogonal. Therefore, each active user using a particular code interferes with every other, such that the number of active users is limited by the cross-correlation interferencebetween individual codes. Now, spectrum efficiency is a measure of how many active users a communicationsystem can accomodate compared with the number of available dimensions. This measure, however, must also take into account that the interference produced by a signal using the same dimensions as the desired one is negligible if the interferer's power is small enough. This fact is considered in frequency reuse schemes like cellular radio [lo]. THE EFFICIENCYMEASURE (1) The number of dimensions is the number of orthogonal, i.e. perfectly distinguishable signals that exist in the signal space. Thus, in an ideal multiple access system, the signals of each subscriber occupy exactly one single dimension. There are many ways to divide the signal space among the users. Two common schemes are Frequency Division Multiple Access (FDMA), where each signal occupies the whole signalling interval T, but just one slot of the available bandwidth W, and Time Division Multiple Access, where one signal occupies the whole bandwidth, but just a narrow slot of the signalling interval. A third technique is called Code Division Multiple Access (CDMA). It uses special signals, so-called codes, which occupy the total available bandwidth W, as well as the The efficiency measure used in this paper has been proposed in CCIR Report 662 [I]. Although it dates from 1978, it is used very rarely, even in recent publications. In this Report, spectrum efficiency is the ratio of the communication achieved to the amount of spectrum space used. The measure of spectrum utilization is defined to be the product of frequency bandwidth, the geometric (geographic) space (usually area), and the time denied to other potential users: Communication Efficiency = ____________________---- . (Time)(Randwidth)(Space) (2) Obviously, this definition is well suited for multiple access communication systems which divide the spectrum space either in frequency (FDMA), in time (TDMA), or in space (sDMA). Hence, for CDMA systems a general more - 464 definition called "Throughput" is used by some authors [4,6]: Number of active users Troughput = -____________-______-Number of dimensions . (3) The throughput, however, can easily be to the CCIR efficiency measure. - To guarantee such a blocking hour. busy probability Pg the mean number w of users the network can accomodate is about 70 $ of the number of channels. It is the.^mean , number of active users which is relevant for the spectrum efficiency. The exact relationship between the number of channels, the mean number of users, and the blocking probability is given by the Erlang B equation [q]: adapted P ? -.=_I_ ml (4) CONSIDERING "COMMUNICATION" The communication achieved may either be the mean information quantity (in bits) or the mean traffic quantity (in Callseconds), carried by the whole communication system. Combining it with the time (one dimension of the spectrum space), one gets either the mean information rate (in bits per second) or the mean traffic (in Erlangs) which is conveyed by the whole network. Any potential subscriber of the network expects to find a communication channel with certain properties, e.g - A minimum bandwidth of the channel, since he wants the communication to be achieved in a certain time, - A minimum quality of the channel, expects a reliable communication, - A minimum probability that he will find a usable channel at any time, i.e. a channel offering at least this minimum bandwidth, and minimum quality. since A CDMA network can be assumed to have virtually infinitely many channels, because, the interference, due to crosscorrelation maximum number of active users usually is much smaller than the number of dimensions. It thus As constitutes a M/M/W queueing system. mentioned above, the channel quality becomes the number of channels in use worse as The case in which the channel increases. below the required minimum quality drops quality is called congestion. As in previous instance, this should happen only with a small probability, say 2 $. The relationship between the number of active users Nmax which cause the congestion, the mean number of active users N, and the congestion probability P of a M/M/a Poisson &e queueing system is given by equation: he The minimum required bandwidth of one channel is usually determined by the service requirements. A digital telephone system using adaptive delta modulation (ADM) needs at least 16000 bits per second [2]. Channel quality in a FDMA radio network depends mainly on the characteristics (distortion) of the particular frequency band in use, and the signal-to-noise ratio at the receiver input. However, in a spread spectrum environment each active user produces crosstalk in the receiver of any other user, due to the imperfect crosscorrelation properties of the code sequences used in practice. Therefore, the quality of the particular channel is a function of the number of active users, while the spectrum efficiency depends on the mean quality. Thus, efficiency may be increased by decreasing the communication quality. To assure a minimum intelligibility of speech, the error probability of ADM at 16 kbps should be better than 0.005, a mean error probability of 0.001 is desirable [2]. If a potential subscriber of a communication network with a fixed number of channels m - a FDMA system is such a network - fails to find a usable channel, the network is said to be blocked. The network must be designed in such a way that this happens with a relatively low probability. A common figure is 2 $ during the Note, that the use of the Poisson equation implies independence of the calling the behaviour of the subscribers on the network state. In practice, however, most of the active users probably would hang up their phone immediately if the radio network seems congested. CONSIDERING "SPACE" IN A CELLULAR SYSTEM In a cellular mobile radio network [IO] the service area is divided into hexagonal cells. Rather than communicating with other mobiles directly, the mobile units communicate with the The frebase station of the attached cell. channels that are assigned to a quency particular cell may be reused in a distant cell This at a certain protective distance D. protective distance depends on the propagation characteristics within the service area. Fig. 1 shows the channel assignment of a cellular network using a twelve-frequency plan. For such a channel assignment a cell using for example the frequency bundle A prohibits the use of these channels at eleven other cells. Thus, the total space denied to other users turns out to be K, = 12 cells. K, is the frequency reuse factor. Unfortunately, the frequency reuse factor of - 465 8404 - Fig. 1. A twelvefrequency-set plan Twelve frequencies are labeled A through L a cellular CDMA network cannot be determined by One must calculate such a simple reasoning. the efficiencies for networks with and without Usually cellular assignment and compare them. all interfering codes can be assumed quasireuse The resulting frequency orthogonal. factor then becomes K, = 1.5 to 2.5, depending characteristics 13971. on the propagation Thus, on the average, the number of cells affected by the activity of one cell is just in the order of two ! Note, that a base station receives signals from mobiles from its own cell, as well as from its neighborhood. Assuming that the average number m of active mobiles circulating in one particular cell-is the same for all cells, the total number Ni of mobiles whose signals the base receives becomes propagation concerning specified must be characteri.stics, required communication bandwidth and quality, channel availability, as The necessary well as the user's profile. overall bandwidth and receiver sensitivity can then be calculated. PROPAGATION CHARACTERISTICS - In order to consider the propagation characteristics, an expression which relates the error probability, the number of active users, the overall bandwidth (in terms of the number of code symbols, or chips, per data bit), and eventually the thermal noise in the receiver is needed. A formula for DS-SS in a non-cellular environment can be found in [6]. (6) (7) For a specific link, all users are interferers. but one of the pi The reason why narrowband signals "contaminate" such a large area compared with broadband codes is due to the fact that two narrowband signals of the same frequency are highly correlated, whereas broadband codes in same frequency the band are practically uncorrelated. Hence, the protection distance is much smaller for broadband codes than for narrowband signals. N: Total number of active users, M: Number of chips per data bit, PM: Error probability, Energy per data bit, Thermal noise density. Equal signal strength of all interferers at the receiver inputs is assumed, as well as no fading and purely random code sequences. A value of Kr = 2 is taken for the frequency factor. reuse THE DESIGN OF A SPECTRUM EFFICIENT CDMA NETWORK The following example traces the various steps in the design of a spectrum efficient CDMA network using Direct-Sequence SpreadSpectrum (DS-SS) modulation. The signals used by this technique are binary codes consisting of M so-called chips per data bit. The number M is closely related to the processing gain of the DS-SS system [8]. In a certain preparatory phase of system design network parameters and requirements COMMUNICATION REQUIREMENTS - Adaptive delta modulation with a rate R = 16000 bps is used. The maximum tolerable error probability PE, = 0.005 shall be attained with a maximum congestion probability P = 0.02. The mean error probability q shall 6e better than 0.001. USERS PROFILE - One cell must accomodate Npet = 200 potential users, each offering AM = 0.06 Erlangs of traffic during the busy hour. - 466 This results in users per cell. an average of x = 12 active MAXIMUM NUMBER OF INTERFERING USERS - Using equation 6, the total average number q of interferers - including the desired signal The use of the Poisson equation becomes 24. yields a maximum of N,ax = 35 users such that the congestion probability is less than the required value. REQUIRED NUMBER OF CHIPS AND MINIMUM SIGNALTO-NOISE RATIO - Equation 7 can now be used to combine the mean number of interfering users K with the mean error probability rj-,-on the one side, and the maximum number of interfering users N,ax with the maximum tolerable error probability P,, denoting congestion on the Solving these two expressions other side. yields the result - of active users per cell not exceeding z, and the maximum tolerable error probability PEc is obtained. In this case it is found na=$--&--{~(h+)-l + ;} (10) Note that in many cases the second term is much smaller then the first. This leads to the efficiency available the conclusion that depends only slight?3 on the number of chips and M = 127 it becomes per bit. For ";E = IO 1 a = 3.89 Erl/MHz/Cell Table I also shows a system which makes full use of the available efficiency. M = 131 Chips per data bit, E!4N0 = 13.6 dB. EFFICIENCY OF A FDMA NETWORK EFFICIENCY - The total signal bandwidth can be defined in several ways [ll]. As a consequence the result may vary from less than to more than twenty times the chip rate. A reasonable and widely used definition on the DS-SS bandwidth is the null-to-null bandwidth being twice the chip rate: W = 2.M.R . r . Assuming the same offered mean traffic per cell as with the CDMA systems, i.e. m = 12, the number of necessary channels to guarantee a blocking probability of less than 2 $ turns out to be m = 19. Assuming further (optimistically) that the required channel spacing is equal to the bit rate R, the total spectral efficiency may be calculated to QZ------- m mRK (8) The resulting spectral efficiency is then defined as the average number of active users per cell divided by the bandwidth m n=2.M.R r (11) With a frequency reuse factor of K, = 12 the efficiency becomes n = 3.29 Erl/MHz/Cell. CONCLUSION (9) and becomes n = 2.86 Erl/MHz/Cell. In equation 9, the frequency reuse factor must not be considered because it has already been taken into account when evaluating the total number of interferers (eq. 6). An alternative approach is to prescribe a practical code length of M = 127 or M = 255, and a maximum tolerable error probability PEc. This procedure yields the mean error probability and the required signal-to-noise ratio. Table I shows the results. AVAILABLE EFFICIENCY - If enough transmitting power is available such that the thermal noise may be disregarded, an available maximum efficiency may be determined. Starting with a given number of chips per bit and the mean error probability (and implicating an infinite signal-to-noise ratio) the mean number Table II gives an overview of the results of various authors for different cases, including fading and error correcting coding, of CDMA and FDMA networks. The figures are adapted to the CCIR definition. The comparison shows that the efficiency of CDMA is much worse than of FDMA in a non-cellular environment. This is related to the well-known "10 to 20 $ exploitation of spectrum" one can often read about in the literature, channel coding reduces the efficiency of FDMA systems because an increased symbol rate is necessary, whereas the efficiency of CDMA is enhanced, the efficiencies of CDMA and FDMA are the same in cellular radio networks. about This last conclusion seems natural, because it cannot be understood why one modulation scheme should be much more efficient than another as long as the system makes use of all - 467 8404 I I 1 System Using 1 Av. Effiency I I Suboptimal System Optimal System / I I m= PPE specified Parameters 12 I m= 12 -3 = 1.0~10_3 = 5.0.10 j PEC = 5.0.10-3 m = 12 P I Resulting Parameter.5 I I I I I -* j R=16000bps,K dB = 2.5.10~~ G / iT = 16.31 / -3 P = 5.7.10 EC ( IJ= 3.89 I / q = 1.47 I c Chips/Bit I dB Chips/Bit Erl/MHz/Cell 4 -t =2.0 r CDMA (FH-1 3) CDS-CLS ) I [41 [cl 7.87 / ~~z.~~~ / / 6.88 3.29 Cellular With Fading :::: I I CDMA FDMA [31 9.90 10.9 ] 3. 13.10-2 i ... 1.37 + 3.89 I 1 advantages offered by a particular technique. One advantage of using spread spectrum systems is that the area "contaminated" by radio energy is much smaller for CDMA than for FDMA. Noncellular radio networks do not enjoy this benefit. It is, therefore , possible to conclude that cellular mobile radio using code division multiple access offers all advantages of spread spectrum systems, including a good spectral efficiency. [71 I 1 + Sel. Areas July 84. Table II. Spectral efficiencies in Erl/MHz/Cell. CCIR Report 662, "Definition of Spectrum Use and Efficiency," Documents of the XIVth Plenary Assembly, Kyoto; 1978. G.R. Cooper, R.W.Nettleton, "A SpreadSpectrum Technique for High-Capacity Mobile Communications," IEEE Trans. on Vehicular Technology, vol. VT-27, pp. 264-74; November 1978. for the on SAC-2, pp. 482-6; R.N. Lane, "Spectral and Economic Efficiencies of Land Mobile Radio Systems," IEEE Trans. on Comm., vol. COM-21, pp. 1177-87; November 1973. [61: G.L. Turin, "The Effects of Multipath and Fading on the Performance of Direct Sequence CDMA," IEEE J. on Sel. Areas Comm., SAC-2, pp. 597-603; July 84. r71: P.S. Henry, "Spectrum Efficiency of a Frequency-Hopped-DPSK Spread-Spectrum Mobile Radio System," IEEE Trans. on Vehicular Technology, vol. VT-28, pp. November 1979. 327-32; [81: R.C. Dixon, "Spread Spectrum Sys terns," John Wiley & Sons, Inc, New York; 1976. [91: L. Kleinrock, "Queueing Systems, Volume I: Theory," John Wiley & Sons, Inc, New York; 1975. [lo]: J. Oetting, “Cellular Mobile Radio - An Emerging Technology,” IEEE Comm. Magazine, vol. 21, No. 8, pp. to-5; November 1983. [II]: F. Amoroso, "The Bandwidth of Digital Signals," IEEE Comm. Magazine, vol. 18, No. 6, PP. 13-24; November 1980. c21: J.L. Flanagan, et.al., "Speech Coding," IEEE Trans. on Communications, Vol. COM-27, pp. 710-37; April 1974. Comm., [53: REFERENCES J.Y.N. Hui, "Throughput Analysis Code Division Multiple Access of Spread Spectrum Channel," IEEE J. I + 1.56 . . . 3.13 / i I [41: 1 / Eb/No + m ] M = 131 I I 131: % Design of different CDMA systems i [II: / 1 M = 255 I Eb'No = 14*1 / II= 2.95 I + Table I. -3 Eb/No = 8.9 Eb/NO = 13.6 M = 131 17= 2.86 General Specifications: = 5.0.10 1 M = 127 9.4-10 t -3 = 1.0~10 EC EC t - INTERFERENCE ANALYSIS 469 85 05 - OF A LAND MOBILE CELLULAR RADIO SYSTEM G. K. Chan Department of Communications Canada Ottawa, the A detailed analysis of and transmitter channel, adjacent receiver intermodulation interference in a cellular system is conducted. The inter-cell and considers analysis situations for interference intra-cell on both the uplink voice transmissions and downlink channels. Results of the show that inter-cell adjacent analysis channel and intermodulation interference may be ignored but not the intraWorst situations. cell interference case assumptions are used such as the allocation of consecutive channels to a location cell site in some situations, of desired mobile station being on the edge of the cell coverage area, etc. moduction Land Mobile Cellular Radio Systems have been proposed in a number of countries over the past few years. These systems are generally recognized spectrum-efficient for radio comas munications mainly because of their freq’uency reuse capability. Similar to other land mobile radio systems, cellular radio systems are affected by the three most common types of interference: co-channel, adjacent channel and intermodulation interference. In this paper, the effects of adjacent channel and intermodulation interference on a cellular system are discussed. The intra-cell and inter-cell interference situations for both types of interference are analysed. In the first situation, the base station at the centre of the cell is assigned n consecutive channels for communications with the mobile stations. In the second situation, the base stations in under the blocked-calls-cleared situation (i.e. Erlang B) with the appropriate number of channels, U, allocated to any one cell. Adjacent Channel Interference Analysis Intra-cell Adjacent Channel Interference Analysis Adjacent channel interference in the intra-cell situation arises from the possibility that the desired signal path may be longer than the interfering signal path. Adjacent channel interference would occur if: P adj 2 pd - Q f . ..(l) J the adjacent channel P,dj is where interfering signal power level, Fd is the desired signal power level, Q is the protection ratio and J is the amount of attenuation that the interfering adjacent channel signal would suffer at the carrier frequency of the desired signal. All parameters are in dB’s. On the downlink channels, same as Pd and adjacent channel ference would occur if: Q,J Pad’ is in t!er(2) . . . The value of J varies from one transmitter to another. If J is chosen to be equal to 26 dB [ll and Q to be 20 dB, (2) will not be satisfied even with both adjacent channel signals on at the same time. It can therefore be deduced that adjacent channel interference would not occur on the downlink channel. On the uplink channels, assuming that received power is inversely proportional to the 4th power of distance, the desired mobile is on the edge of the cell at a distance r from the base station and the interfering mobile is at a distance d smaller than r from the base station, ‘d - padj p 401og d - 401og r - For adjacent occur, channel P adj d or 2 s interference pd r - Q + to J .1O(Q-J)/40 IE total traffic in total number of interference area available - values different Table increases With for 1. Q = 20 dB and J = 26 dB, the P D may be calculated for va Pues of Vt and are shown in It can be seen that PAD as the cell radius increases. (3) . . . Given that the desired mobile is the probability, Pa, that transmitting, one of the (n-l) remaining channels is area within the interference active with radius d may be found and is given by: Pa = 470 channels Inter-cell Adjacent Channel Interference Analysis The worst case situation is for the mobile station to be located at the intersection of three cell sites, point M in Fig. 1. Since the namely, signal paths from D to M, 11 to M and I2 to M and vice versa are about the channe 1 interference same, adjacent would not occur on both the downlink as well as the uplink channel. Intermodulation 77d*Vt/(n-1) . . . where V is the number of mobiles per sq. km., t is the traffic in Erlangs generated by a mobile station and the mobiles are uniformly distributed in Note that Pa would the cell. not exceed unity under a grade of service of 0.02. Considering that the desired mobile may be operating in any one of the n consethe probability of cutive channels, PAD, is: adjacent channel interference, ‘AD = * 1 (n-*)(*-P,) pa + 2 I/n . . . (5) It should be pointed out that the value of d obtained is based on the assumption that only one adjacent channel is on. With both adjacent channels on, the value of d is slightly larger. Hence PAD should be slightly higher than the value predicted in (5). Analysis In this section, only two-signal third-order transmitter intermodulation and receiver intermodulation (TIM) (RIM) interference is considered. Three-signal third-order intermodulation is less likely to occur since all three signals have to be on simultaneously before it would cause a Higher-order problem. intermodulations are of less significance due to their lower power levels of interference. Intra-Cell TIM Interference -The Bownlink Channel Referring to Fig. 2, transmitter emits a signal at frequency fA TX4 which mixes in transmitter TX with its The tntermodusignal st frequency fglation product is a signal at frequency 2fg - f equal to frequency, fC, emitted 6 y transmitter TXC. The output power Pti of the TIM product at 2fB fA is then given by: Pti r Interference (4) = Pt - B - C + Ct . . . *AD n (km) ~~ 0.16 0.36 0.49 0.56 0.60 0.2 0.4 2 6 0.36 3 10 0.49 4 14 0.59 6 26 0.68 2 8 0.53 3 IS 0.61 4 24 0.66 TABLE PRoaABILItY or 1NTERFERENCE INTRA-CELL ON THE I FIGURE *nJ*.CENT “eLINix CHbNNEL CHANNEL 1 Inter-cellAdjacent Chance! Interference (6) - 471 fA TXA: A: Pt I The probability of TIM interfermay be calculated as follows: ence Assuming that the amount of traffic generated by the base station on the that same as downlink channel is generated by the mobile station on the then given that the channel, uplink the transmission is on, desired probability that both the interfering and victim transmitters are on is: 1 (_ Ce 85 05 - - P ti fA vt. = c 3 fi X r2/2(n-1) 3 JY . r2/2(n-21 Vt: Vt.3fi.r2 .ri . 2 i 1 2 (n-l)(n-2) . ..(9) Since there is an average of Tav IM products per channel in the given frequency band of n channels, where Tav = (n-2)/2 when n is even and T,, = (n1j2/2n when n is odd as derived in [41, one or more IM products in considering the to be on at the same time, interference = probability of TIM 'ti FIGURE 2 Ptim FORMATION OF TX IM PRODUCT the transmitter antenna where Gt is gain in excess of the circuit losses in C is the conversion loss in dB, dB, defined as the difference between the levels of the interfering signal power from an external source and the intermodulation product, both measured at the output of the transmitter and B is the coupling loss in dB from the output of TXA to the output of TXB. The value of C is estimated in [21 to be about 11 dB for a frequency separation of 30 to 500 kllz between fA and fB* According to [31, if the antenna gain is 12 dB, carrier frequency is 460 MHz and the distance of separation between the antennae is 0.5 metre, B ranges from 0.2 dB to 7.5 dB depending on the orientation of the two antennae. Hence, = B+C (Pt + Gt) -Pti = 11.2 dB . ..(7) if the separated, antennae are = 1 - (1 - Pti) Tav It should be noted that this expression since some is only an approximation frequencies are shared between a number of IM products, hence the actual probability of interference is higher than that predicted in (10). The values of Ptim for various values of r are graphically depicted in Fig. 3. o- I- horizontally TIM interference occurs when the interfering TIM product power level exceeds the desired signal power level less the protection ratio. Since the transmission characteristics from the co-located transmitters to the mobile station are the same, interference occur0 when B+C 5 Q . . . (8) dB to reception, be ignored. be 4 ” If we consider required for TIM interference a Q of good quality must not 20 CELL FIGWE 3 RADIUS PRCWBlLfTY THE taMLINK ,N I KM OF INIRA-CELL CWWEL VS TIM INIERFERENCE CELL RADIUS ON - 472 Intra-Cell TIM Interference -- The Uplink Channel In the intra-cell uplink channel the coupling loss B from one situation, another is transmitter to mobile Therefore, for all usually high. practical purposes, the probability of TIM interference on the uplink channel is not considered to be significant. Jnter-Cell TIM Interference the On the downlink channel, interfering and victim transmitters are separated by a distance almost equal to the coupling twice the cell radius, is high enough to dismiss any loss possibility of TIM interference. On the uplink channel, similar to uplink channel intra -cell the the interfering mobiles from situation, different cells have to stay close together before any TIM interference would occur. No TIM interference would be encountered here as well. Intra-Cell RIM Interference -- The Uplink Channel In the intra-cell uplink channel situation, the desired mobile station is assumed to be located on the edge of the cell. The two interfering mobile stations are referred to as the near transmitter (with frequency fN) and the far transmitter (with frequency fF) closer to the victim since fN is receiver frequency fV than fF in the IM product: 2fN - fP = fV. The interfering received level Pri is given in [5j power P ri 2PN a f PF - K signal by: - dF .dN2 . . . (14) KSim 5 RIM interference Kim is the where criterion and is equal to: 40 where CT is the 10 log Pf and 'fd" standard deviation of the power level of the received signal. This means that RIM interference occur if mobile with the would frequency fN transmits at distance d from the base station and a secon ! mobile with frequency fF transmits within distance dp from the base station. Assuming that the desired is transmitting on the edge Cell, the probability that product is on is, . Vt * ndP . d(dP) XnVt mobile of the a RIM / (n-l) * Kim I dF.(n-2) Figure 4 shows the relationship between dN and dF. Note that both dP and d would not exceed the cell radius r an 9 that dP = Kim / r2 when dN = r. For probability (7TVt)2 2 P rl Kim/r2 < dF of interTerence ( r:, the rs: . Kim E (n-l)(n-2) . . . (11) PN and PF are the power levels where of received signals from transmitters N and F respectively, K is the RIM conversion loss factor in dB and is given by [21 to be: K = 60 where df is the MHz between the er frequencies. occur if: Pd where power - pri Pd is the level. log ... df (12) frequency separation in near and far transmittRIM interference would _i . . . (13) Q desired signal received Using the transmitter output power model outlined in I41 which predicts the amount of transmitter output power required in order that the received signal power level is above a certain threshold power level Xtd (dR) for Pf percent of the time under Rayleigh fading and lognormal shadowing conditions, it can be derived as shown in /41 and 161 that for RIM interference to occur, r FIGURE 4 RfLATlONSHfP BETWEEN dN AND dF dF - For 0 < probability of dF ( ( 1T Vt)2 Pr2 2 ~~~~~ inter?erence 473 85 05 - the , is: . Kim2 . r2 (n-l)(n-2) Hence for 0 ( of interference Pri(S) dP = ( is, r, 0.1 the 0.04 probability 0.2 t Prl pr2 0.4 (nvt)2 = ’ (n-l)(n-2) r 2 Kim* - - K.im 2 r2 0.1 where S is the frequency separation in channel spacings between f and fP in Hence an IM product. Pri Y S) is the probability of RIM interference due to one IM product and df = S x fs where fs is the channel bandwidth. For a there are on given S, the average P(Sl/n IM products falling on the desired channel where P(S) = (2n - 4s) according to [41. The probability of Pi,(S), for a given S RIM interference, is therefore, Pim(S) = [ 1 - 1 - P(S)/n Pri(S) 1 . . . (16) and if Pim(S) Pim(S) i.S = Small P(S) . P,i(S)/n .a. (17) This probability takes into account the possibility that more than one RIM product may be active at one time. However, as in the TIM interference analysis, this is only an approximation since some frequencies are shared between a number of IM products and the actual probability of interference is higher than that predicted in (17). The interference, total probability is then, Prims of P rim = c Pim(S) (18) S=l where Smax is the maximum frequency separation in channel spacings between fN and fP in any P-signal third order IM product formed in the band of n frequencies. According to [41, S,,, = (n-1)/2 for odd n, and = (n-2112 for even n. and The values tabulated of in P . THk?e are obtained 2. In the u - ,.0004 0.0017 ).0018 0.0016 0.0035 0.0146 0.005, 0.0121 o.oo*s 0.00?9 O.OOSL o.o*?.o 0.0080 0.0344 O.OOS5 O.OISl 0.0071 0.0305 0.0100 0.0417 0.0009 o.oo,* 0.0040 0.0169 0.0017 Cl.0351 O.OLL9 0.0507 O.OO4L 0.0176 O.OLL5 0.0490 0.0180 0.0769 0.0079 0.0337 0.0159 0.0671 0.0122 0.0955 0.4 IZ calculation, it is assumed that f, = 0.03 MHz, Ktd = - 132 dBN, Q = 20 dR and Pf = 0.1 and 0.02 separately. It can be seen that the probability of RIM interference increases when the cell radi.us increases or when Q increases. But in’ general, RIM interference is insignificant in almost all of the situations. Intra-Cell RIM Interference -The Downlink Channel In this situation, different signals from the base station are received at a mobile station in which an IM product may be formed. However, since the interfering signals travel the same distance as the desired signal from the base station to the mobile receiver, Hence, ..a 0.04 0.2 RIM S max r-6 km . 1 l9. p Prim Yt .?.r,. per PN Pd - zri for all practical and d . So no occur. pF RIM >> = Q pd values of K, interference Xtd :oU’f% Inter-Cell RIM Interference Referring to Fig. 1 on both the uplink and downlink cha)nnels the worst situation would be for the i6terfering mobiles from I1 and I2 to be located at M, the intersection of the three cells. However, since the interfering mobiles are at approximately the same distance from base station D as the desired mobile no RIM station, interference would occur. - Cone lusion In adjacent this the paper, channel, transmitter receiver and intermodulation interference situations in a cellular system are analysed. The intra-cell and inter-cell cases are considered with the downlink and uplink channel operations discussed separateThe results of the analysis show ly. that inter-cell interference situations may be ignored but not the intra-cell interference situation. Even though the RIM and TIM interference analysis is performed for the low UHF band due to the lack of empirical data for equipment in the upper UHF band, the interference characteristics are expected to show similar behaviour at the 800 MHz band. consecutive Normally, channels are not assigned to the same cell. This is partly because of the difficulties and limitations in coupling the equipment and also because of the fact that interference normally be can reduced using frequency larger In practice, separations. filtering, isolation techniques and variable mobile power features have been used to minimize interference problems. The assumptions used in this analysis represent the worst case 474 - The effect of this on our analysis is that the uplink interference situation may be slightly worse than what the results have indicated as the desired and interfering signals not be may fully correlated. References [ll Department of Communications, Canada. “Radio Standards Specification” RSS-118, Issue 1, Oct. 1983. [21 McMahon, J.H. “Interference and Propagation Formulas and Tables Used in the Federal Communications Commission Spectrum Management Task Force Land Mobile Frequency Assignment Model”. IEEE Transactions on Vehicular Technology, Vol. VT-23, No. 4, pp 129-134, November 1974. [31 International Committee 524-1, pp I. [41 Chan, G.K. “Design and Analysis of a Land Mobile Cellular Radio System Under the Effects of Interference”. Ph.D. Dissertation, Carleton University, Ottawa, Canada, Department of Systems and Computer Engineer1984. ing, [51 International Committee pp 217-224, [61 Approach Chan, G., “A Practical to Determine Culling the Mechanism in Electromagnetic Interference Analysis Models”. Fifth Symposium and Exhibition on Electromagnetic Compatibility pp 503-507, March 1983. situations. It should also be noted that our analysis has considered only long term median power levels and the effects of fading and shadowing on the desired and interfering signals are not differentiated. In other words, the desired and interfering signals are assumed to be fully correlated and suffer the fading and shadowing losses in exactly the same fashion. This is not usually true for transmissions on the uplink channels since the mobile stations transmit signals from different locations in the cell. But for downlink transmissions from the base to a mobile station, this assumption is correct. Radio (CCIR) 118-131, Consultative Report No. Study Group Radio Consultative (CCIR) Report No. 522, Study Group I. - PLANNING OF TlMDJ’&ZIm PMlD K Department 86.06 475 - III FOB USE n IdOmLI:S=VIC8iS Fisher oi Trade and Industry London UK f. NTROIXJCTION In ths United Kingdom the demand for land mobile radio is growing at an annual rate of 5-l& and this increase has continued right through a period of recession and shows no immediate Signs of abating. In 198~ a study was made of the spectrum requirements and availability of the land mobile radio service to the end of the century. The report noted that 730 dual-frequency channels were then available for 17,000 private users having in excess of 20,000 base stations The report also 6nd more than m,OOO mobiles. calculated that by 1~85 conventional land mobile services would require a total of 1142 dualfrequency channels, and by 1990 a total of 1784 would be needed. Additional spectrum would also be required for other land mobile needs such as radiotelephones,cordless telephones and services in operational support of broadcasting. The report recommended, inter alia, that negotiations designed to clarify the amount of spectrum that could become available for land mobile radio should be pressed to a conclusion as quickly as possible. The Government responded to this recommendation by setting up an Independent Review of the Radio Spectrum (jo-96~ MHz), with this task as part of its work. The Review reported in two stages. In September 1~82 it reported on the future use of television Bands I and III and recommended that these bands be withdrawn from broadcasting use and re-used for a combination of land mobile services and services in operation61 support of broadcasting; with priority where It further necessary for land mobile cervices. recommended that the obsolete 405-line television system be closed down by the end of 1984. These recommendations were subsequently In July 1983 the accepted by the Government. Review presented its final report in which it noted that while. the situation in the land mobile bands is likely to be manageable until the late 198Os, the problems thereafter are likely to become acute unless significant use can be made of other bands and/or new technaIt noted also that there was no prospect logy. of any significant reservea of unused or underused spectrum being identified in the 30-960 URz range and xc-allocated, and that long timcscales w6re involved in bringing about major change6 in the pattern of use. Sharing with Broadcasting At the World Administrative Radio Conference in 1979 the International Frequency Allocation Table wa6 amended by footnotes to enable the land mobile service to operate in a number of J!bropcan countries in TV Bands I and III. Although the United Kingdom ha8 decided to USC Band III exclusively for mobile 8erviccs, neighbouring Administration6 are continuing with television broadcasting, at least for the tima The new land mobile service thus ha6 to being. ahare with the broadcasting service and this has led to a requirement for new sharing criteria. Television services typically operate with considerably larger radiated poware than mobile sarviccs (typically 40 dB more) and mobile services operate with considerably smaller usable field strength8 than television (typically 30 dB le66). This total difference of 70 dB between the planning criteria of the two services suggests initially that sharing will have very limited application. Bowavar, in practice the following features assist the mobile services: - the signal to interference protection ratio required by the mobile services is much less than that for tha talcvision service (typically 10 dB a8 opposed to !50 dB); - the sarvice areas in the mobile service are smaller; antenna height6 ara lower th6.n broadcasting antenna height8 and the height gain corre6pondingJ.y less; -’ the television service amploys a much wider bandwidth than the mobile service (typically d MHz against IL.5 kHz) se the full power of the television signal will not bc present in any one mobile channel; - the telavision pslarisation. Land Mobile servica may employ horizontal Protaction Figure 1 chows the limit of the power spectral density of a television signal when measured in a 7 kHz bandwidth (appropriate to a 12.5 kHz Thi6 maek is channelled mobile radio receivar). for the SECAMsystem used in France and is for normal picture 6cenc8. The frequency separation between the varioue carrier6 is different for PAL systems, but the aaek otherwise does net - 476 - I Fig 1: Power spectral density of a television signal measured in a 7 klizbandwidth greatly change. It can be seen that most of the power in th,esignal.is concentratedin the visia and sound carriers and that for a wide band of frequenciesbetween these carriers the power is at least 5C dB below the vision carrier power, and for much of the television channel it is at least 40 dB below. Figure 2 shows the contour of the acceptable interferenceto the mobile services for the parts of the television channel where the power is 30 dB below the carrier power. It can be seen that this contour olears the coastline of Great Britain with the exceptien sf part of Wales. A kJ dB contour also clears the coastline of Wales. It can be concluded that mobile service operation will be possible all over Great Britain as long as the television carrier frequenciesare avoided. The level ef acceptable interferencewas calculated as followa. The minimum median usable field strength was dervived from CCIR Report 358 : Emin = - 41 + d + 20 lsg f dB(uV/m) 0 0 HI 100 200 sxl”.C Table 1 shows the calculation for the case of a wanted FM signal and an interferingAH signal. TABJE 1 - Calculationof the maximum interfering fisld strength Base The maximum interfering field strength was calculated fraa: (9) where a is the signal to interference protect&en ratio hg is a height gain factsr to convert the field strength to that at a height of 10m above the ground b is the discriminationfactor of the receiving antenna taking account of the azimuth and polarization. Mobile Station Station VP Taking a value af d for a high quality signal (Rrade 4) aives an average minimum median field strength (Emin) fa; Band III of 22 dB(uV/m). dB(uV/m) ), Fig 2: Contours of acceptable interference -3C dB relative to visxon carraer power (1) d is a degradatien factor which takes into account fading due to multipath propagatien, and man-made noise. hg + b ,TALY BORmwX where f is the frequency in Mie E .=E -amax min Ii Y .LIONS JIMOFES BP VP BP 22 22 Emin dB (uV/m) 22 22 a dB 10 IU hg dB b b b dB 0 18 0 8 dB (uV/m) 6 24 16.5 24.5 %lax 10 10 - 4.5 - 4.5 It can be seen from Table 1 that appropriate values for the maximum interfering field strength (E J are 6 dB (uV/m> for vertically polarized (Vfkvtelevisiontransmissionsand 24 dB (uV/m> for horizontallypolarised (BP). Centours of these field strengths are shown in Figure1 for the cast ef 3% locations, yC% time by using CCIR Recommendation570. The Land Mobile plan The use made sf Band III in North Vest Europe is shewn an Figure 3. On the Continent there are the Western European systems, as used in Belgium and Holland, which have channels spaced at 7 MBz, and in Ireland a aystem with channels spaced at o MBz. In France - 477 86.06 - ..-. . SYSTEM - ,.,& ,F ,p ,M) ,q2 ,(,, ,u y FREQUENCY (MHZ 1 ,y ,y ,p ,p 199 200 zoz 204 206 201 2Jo 2!2 214, 2!6,2?9 .'2;o.2?2.2?‘.226.2?a.2?o. U. K. 405 lines IRELAND 625 lines IBaseTxlIlMobilefxlIIMobileTx2jI Fig 3: Use of Band III in North West Europe the obsolete 819-line system has been closed dovn but a new system has been introduced. This employs 6 MHz channel spacing, as does the Irish system, but has the vi&en carriers offset from the Irish carriers by 0.75 MHz. In fact in France there is aleo mobile operatien and the sharing is being achieved by restricting the mobile service to the main mctrepolitan areas, i.c Parish and Lyon, and to no more than four of the available six television channels. interference contour from television carriexm are shown on Figure 4 from which it can be seen that the potential interference is now very severe and mobile operation ipi impractical over meat of the country. No area ef the country is however affected by more than two e&s of carriere and there is a small corridor around The London which is affected by only one set - the French carriers. This is fortuitous as the London area has the highest demand for mobile radio. It is possible ta maximise the number of paired two-frequency channels in London by aligning the mobile channels with the French vision carriers aa shown on Figure 3. The plan gives 3 sub-bands for mobile use each of 7 + 7 MUa with a transmit/receive spacing of 0 MHZ and 1 MHz guardband between transmit and receive. This plan net only optimises the uee for the United Kingdom, but is also identical to the mobile plan uaed in France and therefore ha6 potential to become a European standard. Outside of London less spectrum is available due tn other carriers, but there is at least 5 + 5 MHz,er 400 paired channels available in each of the 3 sub-bands of hand III. *Wl”ltS FRANCE Fig 4: Centourls of acceptable from vision carriers SWKER /LAN@ \ interference - 478 Television Protection The minimum median usable field strength for which protection may be sought in planning televisionservices is given by CCIR Recommendation 417 as 55 dB (uV/n) Band III. The percentage of time for which protection should be sought is recommendedto lie between m A figure of 95% is often used if a and!@. sea path is involved between the interfering station and the television service area. The signal to interferencepratection ratio for television is given in CCIR Report 306 and Recommendation418. For the case of narrow band interferenceunder fading conditions, the protection ratio has a maximum value of 50 dB but is less in some parts of the television channel. Figure 5 shows the protection ratio curve for the L/gECAM system used in France. The pratection ratio curves for the PAL systems used elsewhere on the Continent and in Ireland are similar, but the frequencyspacings between the vari4us carriers are different. - where i is the number of transmitterson the site a. is the protection ratio associated with 1 the i-th transmitter bi is the receiving antenna discrimination factor e. is the field strength at the edge of the 1 television service area from the i-th transmitter. For the case of a television receive antenna 15 dB is an appropriatevalue allowable for the case of orthogonal polariaation. The usable field strength E, is then calculated by way of iteration from: PC = l-l- L ‘Eu - Esj) (4) j where p, is the coverage probability (50% locations) L is the probability integral for a normal distribution Eu is in dB (uV/m) E is the nuisance field associated with sj the j-th base station site in dB (uV/m), as given by e uV/m calculated from equation (3). This methed laads to a lengthy computationbut has the advantage ever a direct power summation of taking into account the statistical tin locations) nature of the pr4blem. Fig 5: Protection Ratia for L/BECAH television The curve given in Figure :,assumes a single source of interferencebut due to the large difference in bandwidth used by the mobile and broadcastingservices there could be 300 or 400 mobile service channels in use in a televisionchannel. A new considerationthus arises of the effect on the television signal of multiple interferingsources. A method called tho simplified multiplicationmethad (CCIR Report 945) has been used over the last 30 years for dealing with small numbers of interfering sourcea and it is proposed to adopt a method baaed canthis for Band III planning. The method first calculatesa "nuisance fiel.d" (e) from each base station site given by: e=Jw uV/m (3) In the case of mobile stations the calculation becomes complex as the interfering field strength 4. is not constant but various with the location 03 the mobiles. From measurements conducted in the United Kingdom it has been found that, on average the interfering field strength from a mobile station is at least ;rOdB below the correspondingfield strength from its associated base station. Calculationscan therefore be made for the base stations and reduced by 20 dB for the mobile stations. Potential Use by the mobile services To obtain greatest USC of Band III for the mobile services the nuisance fields given by equation (3) must be minimimed. This can be achieved by minimising the interfering field strengths, i.4 by restrictingantenna heights and radiated power using techniques such as directionaltransmittingantennas, and by using the smallest values of protection ratios. Bowever, from Figure 3 it can be seen that low protection ratios are confined to a limited band -width around the sound carrier, and this part of the spectrum must thus be reserved for base station sites which Produce large field strength i.e high sites and coastal sites. Other sites can use frequenciesnearer to the vision carrier When these nuisance fields are accumulated in equation (4) the difficultywill be in not exceeding the CCIR recommend4dtelevisionusable field strength of 55 dB (uV/m) which will then determine the ultimate use of the band. This - 479 will be a particular problem in the Boulogne arca of France which is both nearest to the United Kingdom coast and opposite the area of densest Negotiations with the mobile use in Britain. French Administration have, however, resulted in an agreement to plan to a value of 70 dB (uV/m) at Boulogne and 65 dB (uV/m) in other sensitive areas of the coast. To test the scale of possible use of Band III a computer model was constructedusing the details of existing VHF base stations from licensing records. This calculated equations (3) and (4). It was found that all transmitters within 3OOkm of a test point could contribute to the accumulated nuisance field but that with @U channels in operation the agreed usable field strengths on the foreign coasts would not be Therefore, the &Xl channels per subexceeded. band found previously to be available which are free of interference from television will all be usable without causing interferences to tclevision - the assumption being that the geographical distribution of the new Band III stations will be the same as the existing VHF stations. It will be necessary however to exercise care in assigning frequencies to mobile users in Band III and a new computer algorithm has been developed for automatic assignment. This calculates the nuisance fields to the relevant test points on foreign coasts from a proposed new base station site and adjusts the protection ratio until the nuisance field is about fl dB less than the usable field. In this way each base station wiIl contribute equal1.y to the interfering effect and it will be possible to operate around 10~0 trane_ mitters in each JO0 km sector. Equipment Standards in Band III There is a need to adopt new technology to make better use of the available mobile spectrum in order to cater for growth in the demand for mobile services until the end of the century. Against this, however, is the need to develop Band III equipment rapidly so that services can commence in 1.985. Accordingly a new performance specification, MBT 1323, has been issued for the initial implementation of Band III which is based on existing performance specifications but has some new features. Band III equipment is required to be FM only with 12.5 kHz channelling. The mobile equipment must be synthesisedand have a switching range of at least 5 MHz as users are likely to be assigned different channels in different parts of the country. The intermodulation performance demanded of transmitters and receivers have been improved and permitted levels of spurious emissions have been reduced to allow for the greater density of radio use. in the future. Finally, the USC of 1200 bit/s FFSK modulation has been made a requirement for signalling systems to allow for the developmentof sophisticatedcontrol systems using techniques such as trunking. 1 The opportunityhas also been taken to align the new specificationas closely as possible with CEPT Recommendations. 8606 - It is the intention to introduce new technology into Band III as soon as it becomes To this end at least one of the practical. three mobile sub-bands will be resrved for the A possible new technology is time being. amplitude compounded single sideband in which there has been much interest in recent years. This shows the promise of at least doubling the number of available channels within the constraints of the sharing criteria. In order to make best use of the band with the FM equipment it is the intention that the majority of the mobile services in Band III This gives better will employ trunking. spectrum efficiency,allowing more users on a system, or alternatively gives users a better This leads to the approach grade of service. of %ervice prsviders" who will set up community repeaters of between 5 and x) channels offering attractive communication packages to smaller users. A common signalling standard is under consideration for use by service providers. This is intended to give users the flexibilityto switch from one source of supply to another if they wish to do so, and to permit a user to access a service provider’s system when he travels service to an area provider. Other Mobile not covered by his own Services In addition to the @XJ paired channels in each sub-band which are suitable for conventional mobile radio, there is additional spectrum suitable for other uses. There is a requirement for wideband radio microphones for use in support of broadcasting and 0.7 MHz of each sub-band will be reserved for thtir use. The actual allocation will be the lower 0.7 MHz of each sub-band a6 this is the part of the subband for which the television channel has the highest protectisn ratio and thus is least usable by conventional mobile services. Radio microphones being low power devices operated close to the ground will create little disturbance te television reception. There is also a requirement for cordless telephones and consideration is being given to the use of leaky feeder systems operating on the television carrier frequencies. Conventional mobile operation is net possible on the carriers but leaky feeder systems are found to have a goad performance when operating in interference fields from distant transmitters. This development will allow very high densities of cerdlcss telephones to be employed and may make possible the “wireless PA0P where all telephone extensions in a large building arc cordless. - 480 CONCLUSIONS A channellingplan for the new land mobile service in Band III has been developed to make available the greatest number of paired frequency channels in London. This plan provides three mobile sub-bands,with a transmit/ receive spacing of 8 MI%. The plan has also been agreed with the French Administrationfor use in France and has the potential to become a European standard as televisionbroadcasting in Band III gradually declines in Europe. By using current CCIR planning criteria1 highquality land mobile systems can be introduced over the whole of Great Britain free from interferencefrom foreign television broadcasting. Foreign broadcastingcan also be - protected using current CCIR criteria to make available in Sand III at least as many paired mobile channels as currently exist in all the mobile bands. A new computer algorithm has been developed to automaticallyassign frequencieswhilst protecting broadcasting. There is then additional spectrum in Band III to meet other land mobile needs such as cordless telephones and services in operational support of broadcasting. ACKNOWLEDGEMENT This paper is published by kind permission OX the Director of Radio Technology, Department of Trade and Industry. - COMPARISON OF FIELD COMPUTER 481 STRENGHT PREDICTION 8707 - MEASUREMENTS IN LAND MOBILE AND SERVICE BBeriC Federal Radiocommunication Direction Beograd, Yugoslavia ABSTRACT: Relief digital model has been used for computer-aided field strenght predictions in mobile radiocommunications. ln order to find the objectivity of this method, detailed ,,on-site” (terrain) measurements have been performed using the special measurement vehicle of the Federal Radiocommunication Dik?CtiOn Of the SFR of Yugoslavia. The route on which the mentioned meawrements have occured was more than 1500 km long. For field Jtrenght recording the whole route have been devided to 302 path incriments of 1 km langht and the vehicle speed was 36 km/h and also 16 km/h. Field strenght measurements resultscompared to the results obtained by computer aided field strenght prediction using relief digital model are showned in this paper. 1. INTRODUCTION High-grade and fast determination of mediane electric field value is undoubtly the greatest difficulty in mobile radiocommunication planning. Field strenght data am the barnstones for determination of the receiving zone of base radio station as well as for determination of interfering signal zone and interfering zone of two base stations operating the same frequency. The phenomena in electromagnetic wave propagations, such as reflection, difraction, refraction and scatter, as well as all terrain profile elements are to be constantly changed due to vehicle motion and to the fact that the receiving antenna is low, being installed on the vehicle (often aprox. at 1.5 m above ground). Therefore, it is obvious that one of statistical method should be used for field strenght predictions. The application of any statistical methodes requires numerious arithmetical operations, as more as higher accuracy is wanted. Namlly, for incontestable statistical picture it is necessary to performs high-scale cakxdation, and this could be done only by the aid of a computer. Field strenght predictions by computer-aided techniques are possible using the digital relief model. One of the most important questions is the accuracy of computer-aided method using digital relief model. In order to compare the results of mentioned computer-aided method to real situation, a detailed and sophisticated measurements have been undertaken. These measurements could be considered also as confirmation of digital relief model accuracy for application for field strenght predictions in mobile radiocommunications. The method of computer-aided field strenght predictions is showned in this paper, following the comparison to results obtained by measurements. 2. SELECTION OF THE METHOD FOR COMPUTER-AIDED FIELD STRENGHT PREDICTION Several statistical methods for field strenght median value are known. To mentioned some: Langley-Rice method, CCIR methode, “cleamnce angle” methode, methode proposed by the Polish Administration, Bullington’s method end Okumura’s method. By comparison of measured and calculated values performed on the basis of mean error, mean squem error and standard deviation, carried out insofar examinations at the Faculty of Electrical Engineering at Belgrade University [4,5] it was found that the ,,clearance angle” is to be considered as the most appropriate methode. By the CCIR Recommendations [ 1 ] , the geometric configuration of the terrain over which the propagation is performed is to be determined by only two parameters: the effective transmitting antenna height (that is, by average terrain height in transmitter surrounding) end by the terrain irregularity. The average terrain height in the surrounding of the transmitter is to be considered as terrain height mean value at the distance from 3 km to 15 km from the transmitter in the direction of receiving point. The effective height of transmitting antenna isequall to difference of transmitting antenna ground to top height and average terrain height. The terrain irregularity is determined by the difference of terrain heighs exceeded in 16% or 99% cases respectively, considering the part of path between 10 km end 50 km from transmitter, taking the direction to receiving point. As stated above, a better prediction of field stmnght could be achived by introduction of another parameter into the path model, in order to define the quality of selected receiving point [ 6 ] . This quality could ba represented by clearance angle [21Figure 2.1. The clearance angle is determined as an angle between 26 6 9 % ‘Z 0 f 0 10 O A: VHF *lo 5: -26 -30-W -6. -4. -3’ -T -1. Crq5 Clearance anc$e Fig. 2.1 UHF - 482 the horizontal line from the receiving point towards the transmitter and the first streight line exceeding all the obstacles up to the distance of 16 km from the receiving point and in the direction of transmitter. In depandence of terrain’s clearance angle, one can find the correction factor to ba added (or subtracted) to the values readable at the CCIR diagrams. The method used in this work for the field strenght predictions is based on clearance angle methoda based on CCIR diagrams (having as results median values of field strenght [I ] 1, A h = 50 m and clearance angle for terrain irregularity factor correction factor (detarmened taking the terrains configuration in the surrounding of transmitter [ 2 ] ). Following this way, the simplicity of the method was retained and the personnal approuch in the estimation of correction is avoid. The CCIR curves and the correction due to “clearance angle” are converted to computer readable form and permanently stored at disc-package magnetic medium with semidirect computer approuch. It should ba emphasized that digital data describing relief and stored on one disc-package are used for the purposa of this work. Calculation of the effective transmitting antenna height and clearance angle in selected receiving points are to be performed by the aid of the existing communicating programme modules and computer readable relief model, immidiately in the moment of naed[6]. Fig. 3.1 - 3. DESCRIPTION OF TERRAIN CONFIGURATION WHERE THE MEASUREMENTS WERE PERFORMED The measurements described under section 4 where performed on the road of more than 1500 km lenght during the month of May 1664. The transmitter under testing has been located at aprox. 1200 m height above see level. There wera some mountains in the vicinity of the transmitter site having heights between 700 and 1000 m. There were also higher mountains (up to 1500 m heigh) in the wider surrounding of the transmitter site. Therefore, it is obvious that the terrain was mountainous, that is very unfavorable for the use of present non-sophisticated methods without use of digital relief model and computer of field strenght predictions. Measuring points (from Table 4.1) am shown by numbers in Figure 3.1. 4. FIELD STRENGHT MEASUREMENTS At the very begining, it should be considered to which extent computer-aided field strenght prediction using digital mleif model coresponds to real practical results. To find this, particular measurements have bean undertaken being not only the confirmation of “clearance angle” methode but also the confirmation of accuracy of digital relief model for field strenght prediction in mobile radio- communications. The measurements have been carried by special measuring vehicle owned by Federal radiocommunication direction of SFR of Yugoslavia equipped by non-directional calibrated rod antenna (having centre 2 m above ground), measuring receiver connected to the antenna and automatic signal level recorder. The measurements of median field strenght was taken in the vicinity of each measuring point along to 1 km of path. The 1 km long incriment has been choosen by two reasons: a) the differences between all points belonging to the path part of 1 km and the transmitter’s site are almost the same compering each other. Following this, the defined median value could be compared directly by field strenght values obtained as the results of calculating .methodas .._ .._ and b) the lenght of 1 km coresponds aprox. to the lenght passed by the vehicle during one usual conversation: Typical diagram of field strenght values (In dB releting to pV/m) vs. the lenght of the road passed is showned at Fig.4.1. 1 As stated above, the field strenght has been registrated along the 1 km segment of the path. The vehicle speed was 36 km/h (IO m/s) and on some segments 18 km/h. The recording tape speed Table 4.1. Measu- Ep rement(dB/u) point II :: 80 81 :3 “8: 86 13g 9:os 9,41 “3% 26:35 %58 14180 22,03 23,50 43,85 45,07 :‘s50” -2128 -3,96 z;,g -5:56 -2;; 2’12 ;: 4:61 7,58 18,81 14,66 -1,43 -0,91 -3,Ol -9,63 8,90 :z j3;95 32,88 24,19 X! 2s:15 x 38:38 24,85 35,79 31,66 25,61 26,21 27,08 ::,“3; 21180 28,54 27,20 %8 30:41 39,85 35,89 zx 21160 20.70 zo;90 18.12 x 39122 22h8 ;;;;; 5,32 :3;: 11:os 6,32 2,12 2,44 14,64 ::,;; -9:14 los,$ 16:67 :% 9:92 12;; $33 6,21 16,71 x 11:n ;t:: 17104 152 153 :“5: ;:s(: 23’29 26:94 40,84 40,53 39,78 :Lz 102 103 56’48: 11104 9,83 10,lO 10,45 -3,37 39,87 ;; :% $55 %f 38:47 39,32 39,29 4x: 40196 40,.55 Maasu- Ep rement(dB /u) Cd%0 paint !Z ;.! 9.5 96 x 9:51 4,651 $51 if%:: 37:64 17,94 41,15 41,76 42,20 42,85 &!I :% 8:12 8,78 1.11 8,86 x4” 39121 39,20 33x 39188 (d% % 108 109 110 111 112 113 114 115 ::: 118 119 :;‘: 122 : 123 124 12; 42,21 41,44 42,19 48,79 29,66 19,21 $17 :% 12,54 13,82 13,63 11,49 x0” 10:31 12,12 2’5: 5:Sl 7,46 I,23 f% 16:49 18,06 x’: 35:13 35,54 :% 17106 17,46 14,12 12,00 12,80 12,70 Table 4.1. E -Em ?dB, -0,OS -5,96 12,00 13,61 r”,,g 13,20 -5174 15,50 -I,11 24,70 -13,24 27,90 -12,33 $;,;g -11,ll 37,15 31,23 35,83 :“6: 161 15,91 12;; f% 232 5104 8,14 z: 5,51 5,38 235 236 231 “7% 6161 % 4:21 :z 166 167 49,61 32,12 21,56 242 :z; :‘,: “3% 35153 112 173 174 175 x 8145 9,88 9,04 17,SO 12,98 13,80 14,28 176 111 178 179 :tt: 9,14 9,08 9,19 36,67 42,90 :% 18:50 182 183 184 185 186 187 %dj 43:61 d% 42:21 41,36 43,40 39,84 38,33 28.43 30,06 E 4op49 ;3z E ;98$: 38154 21165 192 200 201 Z18 2:; 41121 41,87 % 8:97 ;z 206 207 208 209 210 211 212 42,SO 42J5 41,92 41,25 40,49 39,44 38,34 31,42 36,86 gs; 37,65 g,;y 6,24 6,97 -6,08 11,45 ;:: 215 36,31 35,64 33,92 :9’: 142 :1: 38,89 38,s 1 37,88 38,66 z:o” 42:42 44,41 48,62 4,12 3,Ol 11,21 6,14 216 211 218 219 145 146 147 148 34:39 34,87 21,45 33,44 :;,“2! 2111 9,08 1;‘;; % it;: 224 225 226 221 -4,s3 -3,92 -S,24 -5,09 3,87 I,69 0,17 3,37 3,94 12,93 10,43 :x:. 16:89 % 26129 193 139 :z: 36:28 29,16 33,21 26,88 228 229 238 239 % 36,20 lo,36 rement(dBP /u) 32,99 34,53 %% ;“,:;; 19,70 19,oo 22,86 23,92 21,71 25,20 Measu- E :z: 126 127 128 129 130 131 132 133 134 135 136 137 138 34116 35,06 35,ll 35,56 35,34 35,21 12,74 156 157 159 % 196 191 198 199 27198 -1% 30,72 -17:ll 29,Ol zl: x 24;92 10,74 8160 E -Em ?dB) a:21 % 15132 16,32 1954 20,13 19,67 44,37 29,89 24,38 29,53 %5: $0” ;;:“8; 10:70 9,44 8,63 “2:: 248 XIl $2 ;:o’ 251 ;:; 6182 6,91 10,04 254 25s 2.56 251 258 ‘9% 11:57 11.00 8,72 2,40 % 2,48 2,66 261 262 % ;:;; ;:4” 265 266 261 268 269 8:90 9,80 10.65 1135 16,90 %% 20,68 -6,69 z;: 275 39:30 ;;:g 40,14 x 17;43 216 211 :% % 41,84 41:69 $R 282 283 284 285 BASE 40:99 41,16 41,65 lo,26 9:11 14,24 3:: 9184 27:02 36,38 -0,07 -1;,;; %: 35:25 -1167 26,91 -16,75 16,20 -14,09 23,26 -14,18 18,92 -9,36 16,21 -6,34 16,3-i -8,14 17,so 19,00 1:‘:; 17,50 -2:y 18,OO -4,36 24,33 -lo,69 2288; ;:9” % E -Em ‘IdB) 12::: -6’36 -7:26 %! z: 245 !f ;x 11:11 12,91 1;‘;; (df%lu) 9,80 8,48 8,49 I,81 7,lO % 294 295 x 6:39 5,89 6,36 296 291 % 2:: 4181 lo,58 % 302 8,12 7,89 lo,65 13)s ::‘z -I,16 -6,03 13,04 13,lO I$;; -3:24 L-1:; x -2100 13:oo 12,90 -4,18 13,00 -lo,60 13,30 -lo,64 13,lO -lo,62 14,16 -4,36 ;;,g -12,lO -8,OS 20:38 ZIf 15:71 x: 28125 :2;t 13105 13,oo 13,91 14,06 13,59 16,41 -4,51 I:,:; -7108 :x -4;; -8:31 14:27 166’::: 16,lO -10114 ;‘y; “4:; 13:50 -4:78 -6,ll -5,s - 484 - was 0,2 cm/s. At this way, the measuring segment of 1 km is represented In any case, these results shall be the objects of further researchs in order to determine the most adecvate method for field on the recording tape lenght of 20 cm (by the vehicle speed of 36 km/h) and of 40 cm (by the vehicle speed of 18 km/h). Obtained measuring results have bean statisticlv treated on the following way: the part of recording tape coresponded to the path lenght of 1 km has been devided into 10 smaller sections for the vehicle speed of 36 km/h (and to the 20 small sections for the vehicle speed of 18 km/h). Therefore the lenght of the path passed of coresponding to the small section is 100 m (for 36 km/h) and 50 m (for 18 km/h). The variations of field strenght values shown on small SeCtiOn are mainly due to superposing of numerous waves propagating over various paths. The field stmnght, beinga statistical value, is pardoned very approximately according to the Rayleight distribution [ 5 1. In this case the median field strenght differs from the mean value for aprox. 5% and therefore it will not be the remarkable error if the mean value would be determened instead of median value. This is also so due to the accuracy of reading the recorded tape. Establishing now the set of median values from IO (or 20) small sections and finding cumulative function of partition, it could be found that this one is very similiar to Log-Norm distribution L 5 ] . After determining median value of this 10 (or 20) median values, the median field strenght value for the path passed away (1 km lenght) has been determined. The median value determined following the described procedure represent the value necessary for mobile radiocommunication planning. In fact, field strenght will be at 50% of points of 1 km segment higher and at 50% of points lower comparing to the median value. As such, thfs value is good enaugh to describe the possibility of mobile station to communicate with base station [ 5 ] . This value is marked in the Table 4.1. as Em (dB relating to 1 pV/m). Calculated field strenght values obtained by the method described in section 2 are also included in the same Table. Differences between calculated and measured values of field strenght ammarkedas A=E -E (dB). The mean vaue ‘f oT absolute error is determined from the Table as: strenght prediction in mobile radiocommunications. zIEpL- Emil =482 5. CONCLUSION Respective high-standard and adecvate frequency planning is the must for optimal frequency managements and frequency coordination procedure to be used between governmental administrations. By applying the digital relief model, today is possible to fulfill these requirements. To proof these by the practical way, detailed measurements along the 1500 km long part have been take in order to record field strenght at 302 parts each beeing 1 km long. If the value of IO dB shall be taken as accteptable error, it could be considered that these measurements have been produced applicable results by using the computer, digital relief model and the clearence angle methode and for the terrain configuration for which the ordinary methodes would be unpractical, long and not enough exact. According to the presented results, it is to be concluded that the future frequency planning is to be carried out using computers and exact digital releif model taking in mind the engeenering propagation models. REFERENCES (I) (2) !Sl (4) (5) dB * m= -n with n = 302 being the total number of measurements taken. Considering the error of IO dB in field strenght predictions as accteptable [ 61 , it is to be concluded, according to detailed measurements, that the obtained results am generally accteptable, particulars having in mind unfavorable terrain configuration described in Section 3. By comparison the measured field strenght values (Em) to computer calculated values (Ep) it is to be found: a) for calculated field strenght values higher of 35 dB the deviations are in most cases positive (optimistic results), b) for calculated field strenght values lower than 20 dB deviations am in most cases negative (pesimistic results). (6) (7) (8) (9) CCIR, Rec. 370-3 and Rec. 567-1, XIV Plenary Assembly, Kyoto, 1978, Vol. V CCIR, Rep. 239-4, XV Plenary Assembly, Geneva, 1982, Vol. v DPaunovit, IStojanoviC: One probabilistic approuch to definition of service zone and computer method for determination it (In Serbo-Croatian), ETAN Proceedings, Mostar, Yugoslavia, 1981. Z.StojanoviC, D.Paunovif, I.Stojanovit: Selection of most suitable method for field median value in land mobile radio systems, Proceeding of ETAN, Subotica, Yugoslavia, 1982. (In Serbo-Croatian) I.StojanoviC, D.Paunovid, ZStojanovif, A.MarinEiC, N.Simid, H.BeCa, ZPetroviC, MDukiC, M.StojkoviC: Application and verification of one method for field strenght prediction in land mobile radiocommunication, Journal ELEKTROTEHNIKA, No. 31 (1982) 1. pp 79-82 (In Serbo-Croatian) B.BeriE, D.StarEeviC: Mobile operational radio communication system planning by an aid of computer (In Serbo-Croatian), XVII Yugoslav symposium on Telecommunications, Ljubljana, Yugoslavia, 1983. B.BeriC: Determining the receiving and interfering zone in mobile radiocommunications by the aid of computer, VII International symposium on Electromagnetic compatibility, Wroclaw, Poland, June 1984. B.Beri6, MZec: Field strenght measurements (In SerboCroatian) internal paper, Federal Radiocommunication Direction of the SFR of Yugoslavia, Beograd, May 1984 Set of computer programmes of the Federal Radiocommunication Direction of the SFR of Yugoslavia, Beograd - COMPATIBILITY 485 OF TV AND .MOUNTED ON 8808 - UHF THE COMMUNICATIONS ANTENNAS SAME TOWER Golas Ashok Telecommunication Research Centre, P&T New Delhi, India summary This study was carried out to determine the feasibility of mounting TV transmitter antenna and communications antennas on the same tower. Detailed investigations were carried out to determine the variation of field strength along the tower face and to determine the effect of second harmonic of TV signal on the UHF communication signil. Measurement of field strength variation was carried out by hoisting an antenna along the towa face with a calibrated rope and pulley arrangement and field strength different readings were taken at heights and for different frequencies. This data was analysed to determine the nature of variation of field strength along the tower face. A model of TV signal spectrum was developed based on measurement of TV and spectra of different scenes This model was used to patterns, determine the interference reduction factor for computing the baseband noise in UHF FDM-FM systems. Introduction Telegraphs Indian Posts and Department has set up microwave radio relay systems which require large towers and buildings in various parts of the country. A dedicated rf channel for transmission of TV signal has also been provided on a number of routes. The TV coverage in India is being planned for expansion at a very fast Utilization of the towers pace. presently being used for mounting communication antennas in UHF and microwave bands for mounting low power (upto 100 watts) TV transmitter antenna also was one of the alternatives objective. Board used to achieve this The separation between the UHF antenna and TV antenna is expected to be around 10 to 30 metres. The frequencies for the low power TV transmitter have been chosen from As the Band III (174 to 230 MHz). frequencies for UHF FDM-FM radio relay lie in communication systems 367.0 to 461.5 MHz band, there is a possibility of harmful in,terferenceby the emission of spurious signals (at and around second harmonic) from the low power TV transmitter. This study was carried out to develop a suitable model for estimating interference to the UHF systems. Scope of Investiqation The proposed layout of TV transmitter antenna and UHF and microwave communication antennas is shown in 'figure1. The type of TV antenna which is proposed for use is either a stack of 3 or 4 crossed dipoles mounted on the top of the tower or stacks of 3 or 4 dipoles mounted on the four tower faces. Interference coupling paths are also shown in figure 1. For the case of FDM-FM systems, the signal power to noise power ratio (S/N) at baseband due to interference can be evaluated by using the following equation : S/N = C - I+B+NWdB ... (1) where c: paver of the wanted signal carrier in dBm; I: power of the interfering signal carrier in dBm; - 486 - I I’+- ‘\ E&em-m microwave antenna _ _ + indicates EM1 coupling Fig. 1: Interference coupling paths from TV transmitter B NW: : interference reduction factor in dB; weighting factor in dB (= 2.5 dB). noise value of c can be either measured or taken from the link engineering documents. Value of I can be computed by adding (i) the power of interfering or signal ( either at fundamental spurious in dBm), (ii) the gain of the TV transmitter antenna towards the communication system antenna (in dB ), and (iii) the gain of the communication system antenna towards the TV transmitter antenna (in dB), (iv) the path and then subtracting loss (in dB), Values for antenna gains and transmitter powers are taken from the technical data sheets. As the two coupling antennas are quite close to each other, it was considered necessary to verify the nature of 05 strength variation field experimentally. Remaining unknown variable for The formula estimating S/N is B. for evaluation of B is given in cl] . For evaluation of B, it is necessary to have data about the spectrum of the interfered and interfering signal. The interfered signal is a carrier frequency modulated by frequency division multiplexed signal and it is Fig. 2: SJhematic showing the arrangement used for measuring field strength possible to determine the spectrum of such a signal mathematically. However it is not possible to determine the Spectrum of TV signal mathematically. Therefore, it was necessary to develop a model of TV signal based on measurements of spectrum for various types of scenes and patterns, Field strength variation alonq the tower face The scheme used for measurement of field strength along the tower face is shown in figure 2. Conical antenna which covers the frequency range 200 MHz to 1.0 GHz was hoisted using a calibrated rope along the tower The separation between tower face. and face antenna kept was approximately one metre and this was ensured by means of three guy ropes which are also shown in this figure. The field strength was measured using a Field Strength Meter. Appropriate antenna correction factors and cable loss values were used to calculate the field strength values at different heights and at different frequencies. The above measurements were carried out at different sites and using both types of TV transmitter antenna. The data collected was analysed to determine the nature of variation of field strength. The least mean square error fit to this data was - Field strength Vertioal 380.0 at Maa8ursment MHz with 487 versun LNA at 8808 - separation Surat on from antanns .3-1zI-02(Souroa SO) Ragraesion 80 - 2 I equation C + M * where 2 = Field in t ‘; C I 70-_ Tha : 5 CI c 00.. Q * t lllo 1 10 D strength dB uv/m constant M = Slop8 : i log is I Separation in I 20 v8rtioal Fig. 3 : 30 direotion I : 40from 80TV ~ * * : I 00 70 erntsnna 60 in regression crnalyeis Qivea -23.02 M106.89 C” corralat~on coefffoisnt - -0.02 Z’- , ““““,~:~ffX 3.43 90 matresb) Field strength versus separation obtained using regression analysis. The variables were the field strength and logarithm of separation between the TV antenna and measurement antenna. A sample result is shown in figure 3. The analysis carried out indicated that the slope obtained for different sets of measurement data is around minus 20. This leads to the conclusion that the variation of field strength along the tower face would follow a variation proportional to l/d where d is the separation from the TV transmitter antenna. The path loss computation was therefore carried out using the formula for free space paths. Characterization of TV Siqnal spectrum Spectrun of a floral pattern picked up by a TV camera is shown the in figure 4. (Note : All ~~~~e9 in I. _.... . ^ i parenthesi$ _+%_ ..,.$ 1 _.,_... _ __ “_^__, __...-_. / ..__+-_ IF' % %dth / L-!EE!. .-... __- _I.,-- _,._- ’ ^. I ..._1 _pJ _r_.__-..+ Model of TV signal spectrum The waveforms for different scenes and patterns were observed on a spectrum analyzer which was kept in the 'maximum hold' condition in order to obtain maximum value of spectrum amplitude at different frequencies for each type of signal. Various stages in development of this model are described in following Paragraphs. Fig. 4 : Spectrum of floral pattern signal - determining the vestigial and distributed part of 'TV signal spectrum. - 488 - IF bandwidth = 4 kHz (_.I___ Fig. 5 : Spectrum with highest levels of video sidebands. spectrum analyzer displays drawn in this paper were recorded with 2 MHZ per division on the abscissa and 10 dB per division on the ordinate). The spectrum was recorded with different IF bandwidth settings on the spectrum analyzer and different traces are shown in this figure. It can be seen that levels of video carrier and sound carrier do not change with change in IF bandwidths. Thus these part of the vestigial form the The modulation sidebands spectrum. and the colour Of video carrier carrier vary linearly with changing bandwidths and these form the IF distributed or continuous part of TV signal spectrum. Modellinq of video part of Spectrum Spectrum of colour bar pattern signal generated by a video cassette player (WI?) is shown in figure 5. In this case, the spectrum had highest level of video sidebands as spectrum of other with compared scenes and patterns. As the VU? did not have vestigial sideband filter, symmetrical spectrum appears the the video carrier in this about figure although the spectrum will be assymetric due to the effect of vestigial filters. Based on above information and the model of an amplitude modulated signal in figure 3.14 of C21, model of video part of the TV signal spectrum it3 shown in figure 6. (Note: All the models have been drawn with the video ;7t;--;-:-:!: ----- .: r-vestigial carrier 3 T -I, ; 6-l r : -34 :A: . Frequency offset from video carrier in MHz Fig. 6 : Modelling the video part of TV spectrum carrier frequency normalized to 0 MHZ). The video bandwidth is 5 MHz and therefore B = 5 MHz. The video carrier pave? has been normalized to 0 dB. The continuous part of video spectrum in figure 5 is 20 dB lower than the video carrier when the IF bandwidth is 100 kHx. IF bandwidth of 4 kHz has been chosen for the model of the spectrum as this is convinient in the present case. Therefore, the continuous part of video spectrum for the range + B 2 has been shown at (-20-10 log-(188/4)) dB = - 34 dB. The shape of spectrum beyond 2 B VI2 is based on the information in [2 s . J Modellinq the colour carrier part of spectrum Spectrum of floral pattern picked up by TV camera is shown in figure 7 far the case of IF bandwidth of 10 kHz. The colour carrier part of spectrun consists of energy which is distributed. This can be modelled by the triangular envelope drawn over the colour sub carrier in figure 7. The level 8f this'carrieris 26 dB below the video carrier when the IF bandwidth is 10 kHz. For the present case, the model is being developed for an IF bandwidth of 4 kHz and, therefore, this part - 489 8808 - IF bandwidth T 0 Modelling the colour carrier part of the spectrum Freq. offset from video carrier in MHZ Fig. 9: Complete model of TV signal spectrum B 2 ( fd + fm ) TS = fm = max. modulation freguency fd = frequency deviation and given in figure 3.16 of L23 For this is drawn in figure 8. the sound carrier, fm = 15 k~z and f = 50 k~~zz and substitutinq values value of i$ the equation for B kHz = is obtained as 2*&5) B 158 kHz. The peak of sound carrier level is shown as -10 dB as the sound carrier is 10 dB below the video carrier. T Complete model of TV signal spectrum 1 I. X’ a g 27 . 2 . .I ‘T’ . c2 T In EL offset . . . ul u; - ‘A 1. St a” “I- In & . - .I - /;;, E-r :. 3 fram video carrier inMHz Freq. Fig. 8 : Modelling the sound carrier part of the TV spectrum will be at (26+ 10 log (104) = 30 dB the below the video carrier in complete model of TV signal spectrum. Modelling the sound carrier part of spectrum The sound carrier part of the modelled on spectrum has been the basis of FM modulation envelope Different models of spectrum of TV signal have various parts of been canbined and the complete model shown in spectrum is the for figure 9. The effect of vestigial filter has not been included in this However, in an actual TV model. filter is transmitter, a vestigial and the spectrum model is also used this filter modified by suitably interference evaluating the while effects fran the TV signal. Minimum attenuation Of vestigial sideband filter is as given below : Frequency offset from video carrier in MHz - 1.25 - 3.0 - 4.43 Attenuation in dB 20.0 20.0 30.0 Evaluation of B factor The calcula*ion for B factor - Interfering carrier I - : II harmonic of TV 'Interfercid carrier : 60 chl. FDM-FM I 490 20\ I Frequency offset in MHz Fig. 10 : B factor as a function of frequency offset can nw UBFs#tem be carried out. Data is as given below: for Baseband : 12 - 252 WHZ : 60 No. of channels rms test tone deviation : 100 kHz pre-emphasis : CCIR Rec. 275-3 Transmit and receive Xf filters : passband ripple : 0.1 dB over +l MHz 3dB bandwidth :+2MHz 40 dB bandwidth : rf:10 MBz Receive IF filter : passband ripple : 0.1 dB over _+.SMBs 3 dB bandwidth : 2 1.5 MHz 20 dB bandwidth : + 3.3 MHz Since the interference effects from the second harmonic of TV signal are to be considered, the model of TV signal spectrum has to be further modified. Based on the observations of second harmonic generated by the TV transmitter on spectrum analyzer, it was found that the model at second harmonic frequency can be obtained doubling the value of simply by frequency offsets on the abscissa of the model shahm in figure 9. The factor was value of B evaluated by using a computer program. The result is shown in figure 10. References (13 CCIR Report 388-4, "Methods for determining the interference in terrestrial radio-relay systems and systems in the fixed satellite service'&, International Telecommunication uhion, Geneva, 1982. (2) William G..Duff and Donald R.J. White, “A handbook series on Electrcmagnetic Interference and Canpatibility - vol. 5 'EM1 Prediction and Analysis Techniques", Don White Consultants Inc., Maryland, 1973. - 491 - 89 09 BZSIGN OF CGMPATIBLZ ..-I_-VEHICLES -1 JQUIPMENT FOR LAND MOBILE S Satyamurthy Combat Vehicles Research & Development Establishment Avadi, Madras 600054, India Modern Xlectronic d,,rrices are fast replacing conventional electri.r::5:l., electromechanicaland electrohydraulio devices in land mobile vehicles especially in military tracked vehicles in 'a random mannr. This is primarily to t!l:.l:::t the challenges of 100% operational reliability and varsatality of sysis not only tsms. In this arca, ,:%I1 just a system problem but it is also a great deal of different :3lIconfigurations right from chip level to*the highest level of deployment of vehicles contributed by different :;MIcoupling paths. As such design of compatible equipment for this application calls for high degree of lllMC design synthes,is,inexpensivehardening techniques and realistic oompatibility tests. Clear insight to L;MIsituation at various levels, optimum design goals and accurate testings are essential tools for successful design of compatible equipments. In this paper, an attempt has been made to discuss the design objectives of developing a totally compatible vehicle;:using case by case EMC analysis of systems, their modifications and,oompatibilitytests. Introduction_ Land mobile vehicles especially tracked and wheeled vehicles are densely packed with large.number of mode n elec$ onic devices, heavy duty efeotricaI machinery, sophisticated control systems, multipleswitching devices, etc. depending upon their application in the operational area. If we'aee the &m$ development of ground vehicles from various countries almost 408 of the cost of the vehicle is towards eleatronicsr In order to meet the challenges of modern electronics wdifferent rating of machines installed in a prototype vehicle is shown in Fi;r.4.Of all, curve (c) is noteworthy. 'Thisis a 5 K.VDC motor which drains 1200 Amps initinlly and poses a sev;3retransient problem to sensitive electronic devices connected in the same power supply. EMC design for such heavy maohine is complex for economical approach. EMI Problems of Communioation Electronics These are multifolded whether it is a single radioset installation in a support vehicle/truck or multiradioset installation of combat radio net in tactical vehicles. They,are both susceptible to emit and receive unwanted electromagneticenergy from different coupling paths. Of all, important arc (i) antenna to cable coupling (ii) antenna to fic;:ptors and (i) antenna to box couoling (LI) antenna to cablte coupling (iii) cablo to box coupling when they are emitters. To cite Lan intorusting case, thermal picture of a low level video monitor was masked half the scresn by a potential YN field from a VKF transmitter installed in the same vehicle. Similarly a sensitivo VW receiver located nr?Xtto the thermal scanner could not pick up messages from a transmitter located 1 km distance ins@te of several chsngas over to standby freguoncics in VHF. Mu.B~~~'l,~~er~s~~~e~.a~~-~ess~~-f&~~-l;~~~:cation performance of antenna are very popular in land-mobilevehicles. Eiadiation characteristicsof a typical thermal scanner is shown in Table I. Table -I.I _-__-_--... _",_.. I -.--*..*"*..__ Frequency Intorferonce Snec. in MHZ level in limit dbuv/m/iNZ -_. -1II~--~---~_---0.01 ;*z 0:15 0.20 0.25 0.30 0.60 0.70 0,RO 15.20 30.50 141 129 105 1:: 100 88 102 ;: 89 :5: 86 u2 1:: 82 31 139 78 6‘7 79,. 64 _.-I___,_-,1-^-W- _--__- ElectromagneticCompatibilityof Di@tal _-.-_SquiJments "._ -P_-Microprocessorsare revolutionalising almost in ev~.?ry field and more so with land-mobile vehicles. Engine parameters monitoring and control, transmission gear control, computing devices (for navigation and weapon aiming in the case of tactical vehiion panel,frequency cles), instrunonta-t - 494 hopping techniques and auto euning of antennas in communicationfield are some of the %ypical areas where Microprooessors have simplified the operations and size of equipment. In such devices clock, interface cardS/ back plane wiring and ?CW traces conduct and radiate EM1 significantly to a wide frequzncy range. IIighspeed circuits; g"tes and display devices emit wide band -3M1,while clock emits narrow band E):lI. They normally affect the whole VHF communicationband masking their rac?ption. ~limissions from digital equipment upto 490 MHZ (please see table II) and at levels more than IQ db above the NIL stsndards are not uncommon in vehiclt+s. Improper wiring R:shiulding of logic boards have ros,ultedin CW emissions at clock frequencies from gear changing equipments. Such emissions are threat to communicationsreceivers and also they act as potential jammers to navigation signals, di:splaysand sensors. Speed pick up daviccs using pulse techniques are rich in harmonics and hence potential threat to communication receivers. Table II ----II..,~~L--~ -_WP-_--.+-.*Frequency Interference Spec. ifi..#&Z. 1“ I$jl%fe.l_ i&p,?. limit dbuv/m/MXZ L._u*-__... PC,*0.15 0.8 xi 71 69 59 ::: 1::: E ig*: 50:o ii80 g*; 1oo:o 200.0 300.0 400.0 I; 92 * :; ;ft 50 2: 62 ,,:i ;"2 GO ;"o 2: 50 50 - EMC Aspects Devices --- of O&@l There are several optical techniques for detqtion and display of objects/targetsboth in day light and in darkness for security observations, .,weaponaiming, battle field survcillante and gun firings. Of alltechniques thermal imaging is becoming vary popular. Now we shall SIX?the b‘asic .XMI producing devices in this. They are (i) switched mode 20Wr supply (ii) DC-DC convertors (iii) Motor operated thermal scanners (iv) video converter (v) compressor operated cooling systoru (vi) Xlectronic processors (vii) video monitor. Conducted XT41 observed from thermal scanners and electronic ;?rocessorsare vary high, - right from 0.01 MHZ to G or 7 PJIIX at a level of 170 to S5 dbua/hliIZ s@nst the specification of 130 to 50 dbua/ HI%. On the other hand, video convertars and monitors are most susceptible to H.F. signals from VKI?transmitters located in the same vehicle as discussed earlier. Several horizontal lines will be seen in tho monitor totally blocking the image of the object being viewed. Rmm;lnt in vehicles. Shield terminals of ICs and - 89 09 logic circuits should bo properly connected to ground. Pibre optic cables are great boon for 13lCas they are immune to EMI, SW, lightning, cross talk and over voltage stresses in addition to their light weights. However system gcnarated noise (due to TTL signals) in the electronics needs careful handling of EMT generation and emission control. Use of smart integrated sensors such as optical and hall effect devices greatly reduce WI1 threat and simplifies encoding of signals directly to processors. Optical transmissions,signal i/p filtering, shielding, floating are useful techniques in connecting sensors (thcrmocouplt?s, prt2ssuretransducers, strain gages etc.) i:lho.stile, %MS. Transient is an inevitable phenomenon in inductive loads of vehicles and solidstntc d::vices.The excess onorgg of transient pulses are to be absorbi!din addition to protection against component damage from over voltages by suitable moans. Vsristors transzorbs, zeners, capacitorsand surge filters of correct ratings are must to tackle transients. i%lCDrsinn ..--.e”,, Gui.dPlj,nes_.I k-w-W.-L.“.& All the techniques discussed above ~1*e&+32H*-e~tions during _ design stages and neoessiates XP!C measurements when systems are developed.:In addition, bofor: system integration in the vehicle, degree of shieldin@ requir~?m,?nts are to bc+i-?sti.. mated to m+tt the standards of the systems and the vehicle. As a general 'SMCdesign guideline, ZXC plans with r-ga:rdto shielding, bonding and grounding schemes, filtering and harness routing are necessary in any vohicle installation.Since cables and connectors form a major portion in vehicle:integration they need careful look wi'thregard to nature of signal; level, no.of cnbles, separation from adjacent cable, branches, terminations, shielding etc. CommcrcialLyavailable special cablo systems such as lossg line filters, electroloss filter line cables, and optical fibres are viable solutions to critical 22KCproblems. Conductive heat shrinkable shields, feryite beads, etc. largely rcduco iZMI. Mesh type shields, foil type shields, flat ribbon cables, optimised screening cables and zipper cables are also extremely useful for .EMIreduction, While using connectors of different origin, right COnneCtOr having sufficient filterpins meeting the specification for 1011,!1MTshould be chosen. Invariably backshell of connectors should be terminated with cable shields for effective shielding. Over design and undar design (random selection of pins, less number of - 496 ground pins) of connectors should be levels avoided. Pin .topin co~l>li!l~$ of analogue and digital si&nals should be examined by tc?stingwith circuits workin& in lIOrU;lal conditions. i;iETestin? .-..-wIr_ --_ +v?lWtti Oil and . *. A--.m-._.M We shall dovidc 311 t::stinc; of vehiclr:sinto throc diffcr3nt stsg~s for simplicity. They arc (i) lo:{ 11?vr:l (ii) middle level and (iii) high lev;?ltestings. :Jh:?ro prototype Subsyst(+msare devolopcd, ::NCcompliancc shall be probed by sample menaurcments knovn.,as.. t..iSKL pr:3t.23tr for ths bb>n:::fit of subsvstem disign::rsto cater for wave shaping, filt<:ring, by passing, shieldin@ ,?tc.Once it is done, subsystem will be retested for mee%ing KE1; stdr; whi.cYiiis known a3 low i3.d tostin;l;. Still this - dznands. Succ:zssful3EICcoizpatibility of s;ystemsand vchiclt+scan be verifi;!dwith IA9 following t ::sts : (a) By injectin!;; emissions Lit critic21 Points in thy syst,:ms/v :hiclstSon major :syst.?;.lly in till? vehicll: nntl v::IIiC:l.~ to vehicle/externalworld's intcr:Lctions can bc regarded as lXC evaluation' which is tht: ultimate test for proving AX dasi@ns. The zvaluation phase needs intelligent plannin,gwith regard %o time, manpower, instrumsntation, t3st site, etc* Transient measurements and susceptibilitym::asurcmants assume special importance at this stage. Automatic test systons controlled by desktop computin:;; systems facilitate.&11 testin,; ‘by providing good repeatability,fast and accurate results, besides other advantages like red!lcinsmanpower, making large computationsand storage of data. The mission profile is excited by prop::rloading sequ:.nces as in field operation. JIorethe mission scleneriodisctatoa the naturo of loadins and schedules as per I_ 000 (5) Dy studyin:;the r :?ctionOE SJWti3r.D ~~~x~l:?llts (3spO0idl;[ syst,:ms) ,$n~uish bIi:ZLpOn to 1~ois.e each time to dis- periodical and random ?hznom(:!non of ':+:14I . Conclusion W"I Dofinins th:evehicle 313, careful syste?n,_la,yputs with .intalligqnte_npJ,,7 tostin~s necring practices, accurnt,:: with repeatability, r,ooaomicnl harXonins keeping pace with th::! growth of technology and avoiding ovl:r designs make compatible desi.;;ns aim\?la and ~x~r~)o:3c:ful in ion,? ti;rm vehlclc devolopm~nt programmos and in turn bulk production of vehicles espocinlly for military upi?lications. Rof ori3nccs I-.-P_ (1) Shin Yama lloto etl. ::l ATTENUATION VI RUBBER FERRITE H!XCII 4L’THICKNESS PRINCIPLE OF OPERATION This paper discusses the use of a corrugated metallic structure (C.S) as shown in Fig.l(b,c,d) instead of a plain metal plate backing to the rubber ferrite. The linear or slant 45' corrugated structure is fixed to metal sheet of oven's door as in Fig.l(e,f,g). Absorbing material namely rubber ferrite is fixed on top of the corrugation. This is called MRS with corrugation backing, The principle of operation is based essentially on the fact that the oven's mainframe and its door when it is slightly ajar can be considered as equivalent to a parallel plate transmission line and an axial component of the field is produced on the surface of corrugation, then the lossy material (rubber ferrite) laid on the corrugation sustains a power loss. The corrugation structure was covered on tno opposite sides with adhesive backed Aluminium foil which gave fairly good results. Hence by choosing appropriate values for corrugation depth, pitch and thickness and also suitable type and appropriate thickness of lossy rubber ferrite, a fairly large - 499 90 - Pl attenuation of electromagnetic waves leaking out between oven’s mainframe and door can be obtained. c.s r FIG. l( k------72 f). Enlarged view mm------A FIG.1 (b).View of corrugation structure FIG. l(g). FIG.l(c). Elevation of fig.1 (b). Slant 45O C.S on door panel THEORY OF OPERATION Morita and Suetake [77 have shown that in the case of a lossless corrugated waveguide the surface impeclance 2.1 and Z2 of non corrugated and corru- FM(d). Slant 45OC.S with Same as fig.1 (CL gated section respectively should be equal because the tangential field components from which the impedances are derived are continuous. Hence the following equation is obtained taking into account the boundasy conditions and each corrugated slot in a sense considered as the waveguide which has a sectional area (axd) and extends from y=O to y= -12 and d=p-t and ‘t’ is the thickness of teeth of the corrugation and each guide is excited in HIO wave by the field component E, p,t,d,lj ’ generated on the surface tion and short circuited lower (e). Corrugation on door panel H plane of the of at main corrugay=-l2 by guide. Yl is the transverse propagation constant and is determined from above relation. r, the lonqitudinal propagation constant is derived t = Cr+j/.J - . . .-- the from 5 (4) - r. =j /a,=2 n/&j, 500 - - - - ( 5) where '20 is the guide wavelength in waveguide from which the corrugated structure is removed. a is attenuation constant. /3 is Phase constant. Theory of the corruaated wavequide havinq a resistance card or MRS When a resistance film card or a MRS is laid on the surface of the corrugated structure it can be considered that the surface wave impedance of the COrrUCjdtiOn Z2’ and the impedance 'MRS (ZRFC for resistance film card) are in parallel, then the look-in surface impedance Z2 can be written as II73 d z2 =’ zbazM!?s . . _-.( 6) * z;+ $,RS where L * is the look-in impedance of 2 each small guide in the corrugated structure. Hence cl Z;-ZMRS yltanhq $=p. Z;+ZMRS - - - (7) - 12 should be taken as %$4 - CORRUGATION STRUCTURE In this paper 3 configurations of corrugated metallic structures that have been tried are described. A) The first called linear corrugated structure (L.C.S) is shown in Fig.l(b) The and its elevation in Fi .I(c). M and magnerubber ferrites H C(l:43 tically loaded ep2xide cillsd Eccosorb MF 124 LB] have been attempted as MRS. A part of this workras presented by one of the authors (S.R.H) @I . B) Since in microwave ovens a fan is used and its blades rotate in front of the waveguide aperture perhaps as a mode stirer it was thought that a slant 45' corrugation structure called slant C.S as shown in Fig.l(dBg) in combination with MRS (rubber ferrite) could give better attenuation of microwave energy leaking from ovens. Indeed it was so as can be seen from Tab1e.i. In both the L.C.S and slant C.S cases MRS thickness doubled from 2.4 mm to 4.8 mm and it was observed that increased attenuation was achieved in both the cases. C) Further the idea of using double corrugated structure was thought of a&shown in Fig.l(hBi). In this 'case also experimental results indicated that double corrugation is bett,e; ikan single corrugation structure. the case of single corrugation that where d = gap between two consecutive pitches = pitch of corrugation structure P 11 = width of air gap --ale 12 = depth of corrugation+MRS thickness zMRs = Surface impedance of MRS The corruqation effect The above discussion dealt with an ideal structure for calculation on the boundary impedance at the surface of the corrugation and corrugated surface had been macroscopically considered as a plane having some average impedance. In reality, however, it is corrugated, then the corrugation effect on the propagation constant should be taken into account. When the pitch is small compared with guide wavelength, it is considered that 'the corrugation strongly modifies the value of Gc rather than fl . Morita and Suetake [77 have shown that a' (the modifieda ) =(P/~)cx..@) and t/p are<< 1 and to when p/X make Q large the depth of corrugation FlG.l(h). Double side linear C.S with MRS on dog- panel and mainframe of MW oven ,slant FIG.l(i). FIG.l. Double side slant with MRS SCHEMATIC WITH MRS 45O C,S 45O C.S VIEW OF C.S ON A MW OVEN. - 501 90 - TABLE Pl 1 COMPARISON OF THEORETICAL AND EXPERIMENTAL RESULTS -MRS with slant 45O C.S dB dB dB ____________________~~~~~~~~~~~~~~~~~~~~~~~~~-~-~Calcu- ExperiCalcu- ExperiCalcu- Experilation ment lation ment lation ment MRS alone Sl. No. 1. Description MRS (MF 124)Thickness=2,4mn Magnetically loaded epoxide - 14 24.50 26.10 34.65 35.74 2. MRS (MF 124)Thickness=4.8mm 3. MRS(H5C(1:4) - MnZn) Thickness = 10 mm . - 10 - 40.50 17.00 42.20 - 24.00 38.32 - 4. MRS (M3-MgCuZn)Thickness=lO mm 11 _ 19.78 - 27.00 - 8.10 8.60 11.50 12,90 21.20 23.10 32.70 33.83 40.03 - 34.04 5. Corrugation slots filled with absorber (MF 124) 6, Corrugation with slots filled with absorber and MRS(t=2.4mn) 7. Corrugation with slots filled with absorber and MRS(t=4.8mn) _ _ - - slant 45' C.S was better in double corrugation also it gave improved results when compared to double L.C.S. An enlarged view of the single linear corru ation structure is shown in Fig.19f). Corruqated_structuE with absorbers Further the idea of fitting an i) absorber in the slots of the linear and slant 45' corrugated structure was attempted [103. The propagation constant in this case is where 1 MRS with L.C.S Since all the parameters Eo,,Uo , 2~ , can be calcu/ur ,121 are known 'c;e lated. The results of calculation and measurement are shown in Table.l.There is fair agreement between calculated and experimental results. The absorber used in this case was Eccosorb MF 124. However, the attenuation obtained is far less than the required level. ii) Then the idea of using MRS on top of the corrugated structure whose slots were filled absorber was attempted. The attenuation measured both in L.C.S and 45' C.S cases were 23.1 dB and 33.8 dB for MRS thickness of 2.4 mm respectively. They are correspondingly lower compared to the case when the c.s slots were not fitted with absorbers namely 26.1 dB and 35.74 dB respectively. iii) The same characteristic was observed when MRS thickness was doubled to 4.8 mm as can been at Tablelat Sl.Nos.2 and 7, It is felt that the corrugation effect is dominant in modifying the attenuation constant,CY when the slots are not filled with absorbers while it modifies both a and phase constant, /a in the case when slots are filled with absorbers due to the complex CT &fir and lossy characteristics of absorbers, Hence overall attenuation is marginally less in the case of (iii) when compared to (ii). CALCULATIONS The attenuation,a has been calculated numerically on PDP 11/70 Computer by solving equation (7) in conjuction with equations(3,4,5 8, 8). The calculations and plotting of attenuation, Vs thickness of ferrite for various gaps both for MRS alone and MRS with corrugated structure for M3 and H5C(1:4) samples using equations (1,2 & 8) were done on PDP 11/70 and HP 9825A desk top computers respectively* Attenuation of 22 dB and 16 dB have been calculated for a pitch of 0.7 cm and slot width of 0.4 cm for the L.C.S combined with MRS (in this case H5C(1:4))when gap between door and mainframe of oven is 0.5 mm and 1.0 mm respectively. Attenuation of 24.0 dB and 27.0 dB were calculated - 502 for slant 45' C.S for MRS H5C(1:4) and The dimensions of p, M3 respectively. d and 12' were kept same as L.C.S for the slant 45' C.S. Since a microwave cooker was not available with the authors it was decided to carry out the measurements in a S band WR 284 waveguide set up as shown in Fig.4. HP 4iwR AUTOMATED 6asQR MEASUREMENT is so in both the cases of L.C.S 450 C.S. Both from cost, aInd slant #reight and performance point of view it is not recommended. 5. rhe double side corrugation confiJuration with linear and slant MEASUREMENTS FIG.4 I :t SETUP 3-5'C.S was also attempted. Prelininary experimental results indicate that double structure gives better attenuation to leakage than single C.S. Here also among the two, slant 45O C.S. shows better attenuation characteristics than L.C.S in that attenuation is higher in the case of the former. But the theoretical model has to be derived taking into account that the wave travels between two MRS layers and hence the boundary conditions are different, But the gap through which energy leaks can be divided into two equal halves by considering an electric wall of e = 00 Fig.l(h&i). That is, it effectively reduces the gap to half its size and hence the previous analysis carried out holds good. Thus it shows a much higher attenuation can be achieved double The corrugated structure with MRS was positioned inside the waveguide. Though the results are given for 2450 MHz only in Table the measurements were carried out from 2200 MHz to 3000 MHz and trend of the results indicate almost similar performance except for marginal variations since p,d and 12* of the corrugation structure were not altered. Measurements were repeated to check for consistency of the results. CONCLUSIONS The following conclusions are drawn based on the above study: 1. The corrugated structure provides much higher attenuation when combined with MRS than in the case of MRS alone. a 2. Among the two types of corrugations studied the slant 45' C.S provides higher attenuation when compared to L.C.S as can be seen in Table. 3. The slots of C.S filled with absorbers give much less attenuation when compared to MRS used alone. Hence this is not recommended. 4. The combination of MRS and C.S slots fitted with absorber does not give as much as attenuation as the C.S without absorber in slots. slant 45’ using corrugation. 6. For still better results the orientation of upper and lower slant 45’ corrugation structure can be oriented in opposite directions as shown in Fig.l(l). The upper slant 450 C.S is fitted to the mainframe and lower slant 45’ C.S is fitted to door of the microwave oven the res- pectively. 7. If the attenuation reported here for L.C.S is reasonably sufficient for practical purposes then L.C.S itself can be adopted ease of manufacturing view. from the point of of p,t,d of corru8. Optimization gation may be attempted for further improvements in attenuation level and also for compactness. 9. Among the materials studied Eccosorb MF 124 seems to be having better attenuation proMF 175 and MF I.90 may oertv. be still better when corn ared R to MF 124 as seen from t e material HowevE r, compared may be characteristics. better materials H5C(1:4) available & M3 materials in Jaoan. 90 10. As seen at Table that absorbtion of microwave energy increases from 26.1 dB to 42.20 dB when MRS thickness was doubled from 2.4 to 4.0 mm in case of L.C.S. But the increase is only from 35.74 dB to 38.2 dB for doubling the MRS,thickness in case of slant 45O C,S which is only marginal. From this-two points emerge namely that the optimum thickness of MRS (in this case MF 124) has to be selected. It is somewhere between 2.4 and 4.8 mm. It is proposed to measure and plot the locus of surfaceZof MF 124 and select the correct thickness of MRS from this plot. Secondly the 45' angle may not be the optimum for the slant C,S and it appears that there is an optimum angle which lies between 0' and 45'. This has to be identified. It is proposed to investigate this aspect in detail. 11. The idea of combining MRS with corrugated metallic surface backing and a resistance film card was attempted to examine whether improved attenuation levels could be obtained. The resistance film card was kept on top of the MRS. Resistance film card and MRS thickness were 0.5mil and 2.4 mm thick respectively. The same corrugated structure was employed. Both in the case of linear and slant 45' corrugated structure the attenuation level decreased by 4.01 dB and 2.99 dB respectively when compared to MRS alone with C.S backing was used . See Table-2. This indicates that MRS with C,S is better when compared to combining MRS with C.S backing and a resistance film card. Also the thin film card may have to be replaced often due to wear and tear. Hence this is not recommended. TABLE -2 S.No. Nomenclature 1. Linear C.S with MRS (2.4 mm thickness) Attenuation level in dB 29.12 2. L.C.S and MRS combined with resistance film card 25.11 3. Slant 45' C.S with MRS (2.4 mm thickness) 32.02 4, Slant 45' C.S with MRS (2.4 mm thickness) + resistance film card 29.03 -- PI In this paper the idea of using a C.S along with MRS and absorbers in different configurations has been reported. The above study indicates that a high level of attenuation can be obtained in ovens to reduce the leakage of microwave energy and thus reduce its harmful effects. This idea can also be applied successfully in other electronic equipments to reduce leakages and interferences. The C.S can be built-in in the body of the oven, The MRS can be fixed on the C.S by means of a suitable adhesive. The MRS is not likely to affect the food being cooked in the oven and it is a cost effective solution. ACKNOWLEDGEMENTS The authors express thair thanks to Dr. E. Bhagiratha Rao and Mr. P.N.A.P. AA0 for their encouragements. Thanks to Mr. Rabindranath Saha for the discussions. Assistance from Mr. Devender in the preparation of sketches and running the program on PDP 11/70 computer is greatly appreciated, Help rendered by Mr. M. Sankaran and Mrs.Neeraja S. Gopi in the measurements is gratefully acknowledged. Thanks to Mrs. Udaya Lakshmi for her typing of this paper. REFERENCES Scott, J. : Is today's standard fox Microwave Radiation Safe for Humans? Microwaves, Vol. 10, l,, 9-14 January (1971). Mumford, M.M. : Some technical aspects of Microwave Radiation Hazards, Proc. IRE,(USA), 427-447, (1961). Watkins, C.F. : An Evaluation of ANSI C95.1-1982, Criteria for distance determinations, Table 1, 244-251, Interference Technology Engineers' MasterITEM, (1984). Watkins, C,F, : Determination of Safe Distances for Human Exposure to Radio Frequency Electromagnetic fields based on ANSI C 95.1-1982 Standard - A Graphical Method 151-156, 1984 National Symposium on Electromagnetic Compatibility, San Antonio, Texas, April 24-26 (1984). - 5 r3 American National Standards Institute, Inc., "Safety Levels with respect to Human Exposure to Radio Frequency Electromagnetic Fields, 300 KHz to 100 GHz", American National Standard, ANSI C 95-l1982, 7-24, September 1, (1982). c6] Ramasamy, S.R. : Attenuation of E.M. Waves from Microwave Ovens by Magnetic Resistive Sheet, J,I,E.T,E (India}, 305-306, July (1978). 171 Morita, K., Suetake, K. : A new Waveguide Attenuator element utilizing corrugated metallic surface combined with resistance card. Bulletin of Tokyo Institute-of Technology, (Tokyo), No. 40, 15-33 (1961). [8] Technical Bulletic 2-6/4-82 Eccosorb MF, Magnetically loaded epoxide M/s. Emerson & Cuming Europe N.V., Nijverheidstraat 24, Westerlo, (Belgium) 2431. 504 - [9] Ramasamy, S.R. : Attenuation of Leakage from Microwave Ovens, Indian Institute of Technology, Bombay, Silver Jubilee Seminar (India), 10-17, January (1983). bo) Harrinqton, R.F. : Time Harmonic Electromagnetic Fields, McGraw Hill Book Co., 170-171, 193(4-25), (1961). BIBLIOGRAPHY 1 MARTIN MINTS, GLENN HEIMER. : New Techniques for Microwave Radiation Hazard monitoring, IEEE Trans. EMC, Vo1.7, June (1965). 2 FREY. J. : Biophysical Hazards of Microwave Radiation, North East Electronics Research & Engineering Meeting, Cornell University, NRECM-72, P-136, Part I Record, Vo1.14, IEEE Catalog No.72 CH 0693-2 NEREM, - 505 - 91 P2 LO\J-FREQUENCYMAGNETIC SHIELDING EFFECTIVENESS OF STEEL-REIWJiORCED CONCRETE PLATFORMS W. Hadrian Technical University of Vienna Austria This lecture is concerned with the 1. Structure of the Model shicld.ingeffectiveness of platforms of reinforced concrete against lowfrequency magnetic fields (16 Z/3 Hz , 50 Hz). The fields are created by extended loons. This problem occurs when constructing station buildings over railway tracks with electrical traction. This type of construction is becoming prevalent as main train stations along with their We set up a model in order to test the measured results. Our task was to determine the distributionof the magnetic field over the platform. Therefore it was necessary to know the current distribution in the platform. The model structure is based on the following considerations. The field caused by the electric tracks are built in the centre of cities. Thus administrative as well as bank traction can be approached through the superposition of the fields of two concerns make use of these centrally straight conductors with currents in built offices. All. of these offi_ce opposite directions. This magnetic field constructions are established with causes so-called eddy currents by induc- clcctronic equipment, especially with tion in the conducting and bonded concrete steel. computer terminals. Therefore the influence'from currents caused by electrical traction should be considered Let us suppose that the dimension of the platform is very large in the direc- and dealt with. I;iithin the buildings tion of the current-carrying loop. Then electromagnetic pollution from magnetic fields of electrical traction can the eddy currents in the steel netting of the concrete steel will primarily greatly reduce the economic efficiency flow in current-conducting paths parallel of the building. Therefore we must estimate these fields when a buildingcompound is still in its planning stage. to the current-carrying loop. These Measuring of field distributions in a completed structure of a station building shows that the shielding e.f.fectiveness(SE) is very locally dependent, not larger than 7 dR and is even negative at the edges of the platform. This means an enhancement of the field compared to the unshielded Si.tUatj_on. current-carrying paths close at the end of the platform. Therefore we can justify replacing the steel netting of the concrete steel with conductors which lie parallel to the current-carrying loop. Fig.1 shows the structure of our model. The model measurements of the current distibution were carried out by about 10 kHe. At frequencies lower than 10 kHz the distribution is frequency-dependent. - 506 - Fig.1 : Structure of the model When we know the currents in the conductors and the coordinates of a point number N of conductors of a given P the magnetic field can be calculated by the following formulars: sin ai Bx = - 2 $'i*T (1) . change in the current-distribution. l B Y = + sin ai = PO z d l1 id cos I-j_’ ai 1 ) COS ri = increase of current causes the enhancement of the field and consequently a negative SE. 3. Measurements OLj_= d(3)r. 1 The task of the model eXaminati.OnS t r. 1 The current increase at the edge of the platform is guite notable. This x - x. Y-j_ Y - (2) r. platform width does not cause a great {Cx- xj_l2+ (Y - Y-j_)* ,C4) was to verify the.results .of field field measurements. The magnetic field was measured by means of a calibrated 100~. The situation is shown in Fig. 5. ,,,,,,,,,,,,,,, “W,,,,“,~,,” Fig.2 : Calculation of the magnetic field Our model was built on a scale of 1:lOO. ,,,,,,,,,,,,,,, ..,,,” ,,,, ,,,..7,.e7,cr,,m~,,~,,,-” ,,,,,, x,~~~~,~“,,,I~,~ ,,,, ,,,, l,ll,,,r Fig.5: Measuring set-up Fig.6 shows the calculated and measured Via a step by step increase of the number field by I = 100 heff and the SE in of conductors we studied the influence on the current-distribution. Fig.3 shows the laterial axis to the currentcarrying loop. the measured current-distribution for N = 19, the calculated induction Be(x) The fact that the SE is decreasing at the edges of the platform can readily of the current-carrying loop (I,), the calculated superpasition of all field seen in Fig.6 . At z = +40m the SE is zero, at z = -35m the SE is negative. components created by the loop and the This measured negative SE was the reason the currents Ii , and the shielding to start a model examination because effectiveness SE in dB. The results for IbT = there had been no explanation for this 37 are sllcswn in Fig.4 . The increasing: LJ phenomenon. We supposed the influence 91~2 N-3? c=99cm Pig. 3: Model. examination, N = 19 Fig. 4: Model examination, N = 37 of another fields. The negative SE is not particularly critical because in this area of the platform the field is weak compared with the field directly over the loop. 4. Conclusion The measured SE of the steel-reinforced concrete platform is locally dependent and even negative at the edges of the platform. To study these effects we built a model. The results of the model tests prove good with the measured data. 5. References 1 Fig. 5: Field measurement, field without platform a) fiel.dwith platform h) Buzzi, R.: Modelluntersuchungen der Schirmwirkung von Gittern gegen Magnetfelder. Diplomarbeit, ausgefiihrt am Institut fi.ir Elektrische RnLagen und Hochspannungstechnik der TU Wi_en,l98~ - 509 - 92 P3 A NEW APPROACH TO ELECTROMAGNETIC SHIELDING B.L.Michielsen Philips Research Laboratories, P.O. BOX 80.000 5600 JA Eindhoven The Netherlands A formulation of electromagnetic shielding theory is presented, which concentrates on the calculation of equivalent sources induced in electronic systems by incident fields. Inteqral representations are derived for the vector of source strengths applying to an N-port Thevenin representation of the electronic system. It is shown that the influence of a shield on this source vector can be expressed as an integral over the shield. A simple elementary configuration is analyzed explicitly within the formulation as an example. 1. Introduction and general method of analysis In electromagnetic shielding theory, one studies the effects of applying shields to reduce the unwanted interaction of an electronic system and an incident field. Publications in this area of research mostly treat the influence of shields on the structure and strength of the electromagnetic fields in the configuration [l-5]. However, the behaviour of the electronic circuits is described in terms of voltages and currents. Therefore, in order to determine the effects of fields and shields on an electronic circuit one has to relate the fields to the voltages and currents. This can best be done by means of induced equivalent sources representing the fields (see [ 61). To the authors knowledge no satisfying analysis of electromagnetic shielding exists in the literature, which provides these relations mathematically. Therefore, we set up an analysis concentrating on induced sources from the beginning. We shall first explain the general arguments leadinu to a well-defined problem. We intend to analyze the electromagnetic shielding of electronic circuits. As we want to use linear electromagnetic theory and have to deal with nonlinear electronic components, we adopt the following reasoning. Any electronic circuit can be thought to be composed of a linear N-port network with a set of terminations (which can be active and nonlinear), see Fig.1. The interaction of Fig.1 Decomposition of an electronic circuit. the electronic circuit with some incident field can be separated into an interaction with the terminations and an interaction with the linear N-port system, We shall assume that the former is negligible compared to the latter, because of the smallness of the nonlinear parts. Then, the problem is completely described by the source representation of the linear N-port system, [ Zl [ 11 -[VI where =[vl (1.1) (Thevenin) [VI = column vector of port voltages, [I] = column vector of port currents, [Z] = matrix of impedances, [V] = column vector of voltage source strengths. The fundamental unknown quantity of our problem is [VI, so we shall derive expressions for it; this is done in Sect.2. In Sect.3 we analyze the effect of a shield on [Vland in Sect.4 we study an elementary one-port system. The analysis is carried out in the frequency domain (complex time factor exp(jwt) ). 2. Derivation of an integral representation of the source strength vector The theory presented in this section is in fact a modification of the one - 510 presented by de Hoop [7]. We start with Lorentz' reciprocity relation between two electromagnetic states satisfying reciprocity conditions in a domain D with boundary &S(see Fig.2), / B uJg (EJ&EJ$T, (2.1) dA = 0, where rE_,I$TrR= electromagnetic field in two reciprocal states labeled with T and R. \ --- ,- a) b) Fig.2 Two electromagnetic states in the configuration: a) Transmitting state, b) Receiving state. The domain D between SO and S is assumed to contain the linear (and reciprocal) part of the electronic system, while SO surrounds all other parts. All component surfaces of SO are assumed to be small enough for the low frequency approximation to hold on them. Then the integral over SO reduces to an expression in port voltages and port currents (see Appendix then becomes, A). Expression (2.1) [v? 4 IS - 1 IT1 5 VT = fs). (gTx~R-&gT, dA (2.2) ([ It means transpose) For the T-state we take the state where all sources of electromagnetic fields are contained within So the 'Transmitting' state. In the R-state we take all sources outside S, the 'Receiving' state. Because of the passivity, linearity and reciprocity of the N-port in the T-state, the following impedance relation exists, [ VT1 = [ 21 1 IT1 with [ Zl’ - (2.6) where ET BT are fields in the m-th -m ’ -m T-state (ph sical dimensions Qrn-land m-1 resp.) Relation (2.5) is the N-port equation in the receiving state, so (2.6) provides an integral representation for the equivalent sources. We shall call integrals of the type of the right-hand side of (2.6), integrals of the Lorentz type. They have some properties which are summarized here: - Integrals of the Lorentz type Only yield a non-zero value when the (continuations ofthe) two fields have their sources on different sides of S. - Integrals of the Lorentz type are not affected by deformation of S through domains where the (continuations of the) two fields satisfy the reciprocity conditions. - Parts of S on which both fields satisfy the same,or the reciprocal, boundary condition, do not contribute to the value. A particular case of the latter is the radiation contition on spherical sectors at infinity. 3. The influence of a shield on the induced sources In deriving (2.6) we only assumed that the fields satisfied the reciprocity conditions in D. Therefore, the same integral representation remains valid when a shield is contained in D. However the influence of the shield is then implicit in the "weight functions" 'E$$and &$ , and we have no means of analyzing the contribution of the shield. Therefore, we follow a different route. The shield is assumed to be present in the R-state, but absent in the T-states (see Fig.2). Now the reciprocity condition is not satisfied in the shield region,Ds. For simplicity, the shield is assumed to consist of a simple conducting volume. Theh, instead of (2.6) we obtain, = [ Zl . (2.3) Substitution of (2.3) into (2.2) yields, where [ IT 51 Zl [ I9 4 ?I 1 =jsv’ (gTx~R-gRxgT,dA (2.4) T By choosing N different T-states {I&_, , HT [ITIn} n=l,...N, such that ILf, n=Sm n -n, I , we arrive at, ~Zldl - rvItl = rv3 with VR = (IT m,m)-l/sF cg;I~R-~~ldA, m or (2.5) JR = aE _R the volume current density in the receiving state. The field ER IHR - . is decomposed as . . rE_R,gRl = @n,$nl + @sc,Iyc;3 . 2) where ign,kynl= incident part, in absence of system plus shield, {Esc,$'c]= scattered part. Because the scattered part satisfies the - same constitutive, boundary and radiation conditions as the transmitted field, it does not contribute to the surface integral. The remaining part of the surface integral is the same as in a configuration without a shield and we write 511 92 - P3 advantage that evaluating the effect of a shield requires knowledge of the currents in the shield in the transmitting states, which mostly are fixed ones in contradistinction to the incident fields. 4. Calculation of an elementary example. [ V] = [ VU] + [ VSh] , (3.3) where 1 vu1 = j&J- [ VSh] = j in -sinx[ET])dA, (3.4) *JR dV. (3.5) ([ ET1 XH &ET1 Here [ Vu1 is the source s&ctor in the unshielded system and [ V ] is the contribution of the shield. It is the task of the shield designer to obtain as close as possible the equality [Vsh] = -[Vu] for the interferences. As can be seen immediately from (3.5), it is of no use to ap ly shields in regions where either [E !] vanishes or such that the resultiiig JR will be orthogonal to [gTI. Furthermzre we remark that (3.4) can be used in an experimental procedure to predict interference levels. Then the "weight functions" t@Tl and [ 91 must be measured and together wi‘eh some incident field, substituted into the integral. This method can be practical for one-port systems with a simple structure of the T-fields,and has actually been carried out by the author to predict interference levels of an NMR-apparatus at a hospital location from measurements at the production site. Fgom (3.5) it follows that knowledge of J is required for determining the influence of a shield. However, it is also possible to obtain the following relation (see Appendix C for a proof), [VI =[V systl where + [ vs4 , l EindV, l gin dV , ]V sYst]= /Dsyst]“1 = IDsh[ zTl [ VShl D syst = In this section we consider a small conducting loop, which serves as a first step towards th,eanalysis of more complicated networks (i.e. superpositions of loops). In Fig.3 the configuration is depicted, where a thin conducting spherical shell is added as a shield. Because the field transmitted by the loop is a magnetic dipole field,the spherical symmetry and the orthogonality of the vector spherical waves ensure that all field- and current distributions transversal with respect to the radiusvector have the same magnetic dipole structure. We have the following relations on a sphere, h*x L, = y*e’ (4.1) wherg -& {e ,h ]= magnetic dipolefield:+ regular at infinity (i.e. satisfying the radiation cond.), - regular at the origin, Y+= jnodkr[krhl(2)(kr)]/krhl(2)(kr), Y'= jnodkr]krjI(kr)] /krjl(kr), (2)= spherical Bessel and Hankel j’lthl function of the second kind, k = w(so LIO 4, no= (EO/UO)L'. (3.6) (3.7) (3.8) domain of the system. In cases where the system currents in the transmitting states are not significantly influenced by the shield, we can identify the first term of (3.6) as the source vector of the unshielded system. The contribution of the shield is then given by (3.8). It may seem that we have put more restrictions for (3.8) to hold than we did for (3.5). However, when using (3.5) the complete interaction between system and shield m st be calculated and is implicit in J# . The same effort is needed when usTng the complete system (3.6) to (3.8). Only in special cases, e.g. as discussed in the next section, it is useful to make decompositions like (3.3). Expressions (3.7) and (3.8) have the Fig.3 A simple loop as part of a circuit (fat contour), in a spherical shell. In order to deterqC\ine the shielding we must calculate J in the shell. From the arguments above it follows that the next representation holds, y’ = yT -GT (4.2) where BT= electrical field of magnetic dipole with unit current, YT= as yet unknown amplitude. The factor YT is determined by a boundary condition at the spherical shell (see Appendix B), - 512 y* = do /[l + da/(Y+-Y-)1 (4.3) where (5= conductivity of the shell, d = thickness of the shell. Now we are in a position to calculate the shielding factor defined by, s = (VU + VSh)/ VU. (4.4) Using the representations (3.4) and (3.8) we obtain, VU = (Y--Y+) \,&T*e_in dA, (4.5) VSh= Y* ,,&T*gLn dA, wher n = electrical field of the dipole 2 component of the incident field. With (4.5) and (4.6) it follows, s = 1 + Y*/(Y- - Y+) = l/( 1 + da/(Y+ - Y-)). (4.7) In this particular case the shielding factor is independent of the structure of the incident field. This is because the (shielded) system only responds to the dipole component of the incident field, which is fixed by the field vectors in the dipole's centre. The shielding factor in this example is identical to the spherical wave transmission factor (cf. [1,2 I). As a second example, we shall superficially discuss a conducting loop close to a parallel conducting plate (see Fig. 4). Again we want to calculate the shielding factor of the conducting plate. We assume that the distance between the loop and the plate, d, is so small that the current induced in the plate by the active loop is concentrated under the loop. Now, the strength of this current is simply determined by the ratio of the mutual and selfinductances of two loops with radius R at a distance d apart. I Fig.4 A conducting loop parallel to a conducting plate. This ratio is given approximately as 19 I, C=l- l/( 4&n 8R/d - 7). (4.8) It is now a straightforward calculation to obtain the shielding factor for incident plane waves, S=l- C exp(-jk*d) , -- (4.9) - where k_ = wave vector of the incident plane wave, d= dv (see Fig.4). Using the inequality k-d -- gl, S=l - C(1 + jk-d). -- (4.10) In this case the shielding factor clearly depends on the incident field. 5. Conclusion A formulation of eleC+XOIIIagnetiC shielding theory has been presented which concentrates on the calculation of equivalent sources in an N-port representation of an electronic circuit. The influence of a shield on these sources is expressed as an integral over the shield. It is emphasized that within the formulation of shielding theory presented here, one can directly calculate the quantities one really wants to know. It is often unnecessary to do complete diffraction calculations in shield configurations, because it is proven here that the currents induced in the shield in a transmitting state already determine the shielding properties. In many cases these currents can be calculated more easily or can even be guessed. It is also important to note that in this way one is able to do the calculations on the fixed part of the problem (i.e. the system to be shielded) instead of the variable part (i.e. the incident fields). Two elementary examples were analyzed within the formalism. Acknowledgement Some stimulating discussions with Prof. A.T. de Hoop, on the various representations used in this paper, are greatfully acknowledged. References 1 11 Franceschetti,G.,"Fundamentals of steady-state and transient electromagnetic fields in shielded enclo: sures",IEEE EMC-21,1979,p.335 [ 21 Harrison,C.W. and C.H.Papas,"On the attenuation of transient fields by imperfectly conducting spherical shell8 'I ,IEEE AP-13,1965,p.960 [ 31 Senior,T.B.A., "Electromagnetic field penetration into a cylindrical cavity ",IEEE EMC-18,1976,p.71 [ 41 Mgndez,H.A.,"Shielding theory of enclosures with apertures",IEEE EMC-20, 1978,p.296 [ 51 McDonald,N.A.," Electric and magnetic coupling through small apertures in shield walls of any thickness",IEEE MTT-20,1972,p.689 - 513 l'Topological concepts for internal EMP-interaction",IEEE EMC-20,1978,p.60 "The N-port receiving [ 71 de Hoop,A.T., antenna and its equivalent electrical network",Philips Res. Repts.z,1975, p.302* and G. de Jong,"Power [ 81 de Hoop,A.T. reciprocity in antenna theory",Proc. IEE 121,1974,p.1051 theory [ 91 StraEn,J.A.,Electromagnetic McGraw-Hill,NY & London,1941,p.264 [ 61 Tesche,F.M., Appendix j 2. 8 Appendix the integral, V -- where I 04.2) -V X Hb - = -Jb , where Jb only differs from 0 Tn the crosssections of the wires, D (A.31 P' Substitution (A.l) gives of (A.2) and (A.3) into (using Stokes' theorem), I; = ,,x*J" dA 1). v 2 = unit XE (B.1) -1 = -10 J-s = ad(vxE)xu,(B.2) “51 = -vector d = thickness Because of charse conservation we have =0, together with gauge invariance e potentials one oole can be eliminated-to yield N-l ports. By interchanging a and b it follows immediately js$* (Ea Hb-Eb Ha)dA = 1 VaIb - vbIa P P'P P P' (A.51 to the shell, of the shell, of the shell. Because of the arguments of Sect.4 we have only one transversal field distribution. The"field inside the shell consists of the primary dipole field and a reflected part. The field outside the shell consists of a transmitted dipole field. Therefore,we write for (B.l) and (B.2) using (4.1), g A%T - ArgT - - - T= 0 ' Y+AteT -YAe- r-T - y+zT= -&&jTr where = reflection sion factor From these equations ately that, = ,D +agb/Ib dA = /D$adA/]Dl P P P D = cross-section of p-th P conductor, IDpI= area of D . P normal 0 = conductivity Ar,At where in a spherical shell B. Current due to a magnetic dipole source inside at the shell ap{E_.,H_ll I 2= fields poaching from side 1,2 , The surface 9 is assumed to be penetrated by N conducting wires, constituting the poles of the system. In the low-frequency approximation it holds on all of SO, Ea = -pa - Xg2- -2-v XH (A.1) (EaxHb)dA. P3 Here we consider the configuration of Fig.3. The currents induced in the spherical shell follow from the boundary condition at the shell. For thin conhave the following ducting shells,we conditions on the tangential fields (cf .[ 11 1, description A. Low-frequency of electromagnetic fields Here we consider 92 - it and resp. (B.3) (B.4)' transmis- follows inunedi- At = 1 /[ 1 + do/(Y+-Y-)I , 03.5) JT -s = da/[ 1 + da/(Y+-Y-)] gT. (~.6) and Appendix C. Derivation of an alternative expression for the contribution of the shield. In this appe.ndix we shall derive an expression for the induced source vector, where apart from the incident field, only currents appear which are - 514 induced in the shield in a transmitting state. Because the shield is present in the receiving state too, we have to do some effort to eliminate the scattered field in the receiving state. We start with the following equality, - Because the transmitted fields and the scattered field in the R-state satisfy the same constitutive, boundary and radiation conditions outside S (cf. Fig. 2) we have, ~s~~([~T]~~sc-~sc~[~Tl) dA = 0. (C-5) gR-E_RX[ET])dA. (C.l) The field in the receiving state is again decomposed as in (3.2). Then, a decomposition of the source vector follows, [ V] = [ vloadl + [ V’ ] , Using a D=SU -SO, where -So means:So with the opposite orientation (see Fig.2), we obtain -jaDx- [ I”]= = (C.2) ([ET1 “~sc-~Sc~~~Tl )dA, I, ([ET]•~c-~sc*[J_Tl)dV. (C-6) _ where . . [ vloadl = I$* ([zTl~$~--$~x[ fif] )dA, (C.3) From the Maxwell equations it follows that JSC = 0 (Eln + Es'). (C.7) [ V’l Substitution of (C.7) into (C.6) yields, The first term in(C.2) given by (C.3) is interpreted as the voltage source vector induced in the terminations of the system. This term is assumed to be negligible compared to the second term representing the system and the shield. [ V’l=jD [zTl *gin dv. (C.8) The integration in (C.8) is over all conductors contained in D. - 515 93 - P4 FIELD l’JONUl!WORHfTY RBDWCTIOl4 IRSIDE A SPHERICAL HAOXETICSHISLD V.A.ltorosov,lV.V.Rodionova, Institute of Radioengineering & Blecrtronice USSR Academy of Saienoer roscow,UsSR It Is ehown that the field nonuniformity of an external interfering eouree5 is reduoed ineide magnetic ehielda 611,s.) in magnetoetstio regime. This property of m.8. ie treated a0 a relatively high attenuation of the higher epatlal modee of the interfering field,compered to the fundamental mode, which is a uniform field for our problem. Introduction. Some modern experiment& in magnetioe are poseible only in a high magnatio...Yacylum bath_-;ln.fiqld. ~eral.aad. its spatial derlvatives.An example of euoh experiment5 ie the lnveetigation of magnetlo field5 of bfologi&sl orlgin in magnetocardiography and magnetoencepholography,oarrie# out with extremely low aoirs threrhgld =gnetometers and gradientometers fl] . The purpose of thla paper ia to show that the pawiv6 apherlcal m.e. in mm#neto5tatio regime reduoer not only the field level of the external aouroe inside the 5hield,but alao the field nonuniformity (or relative spatial derivatives) within the shielded enclosure. Thirr effeot llsr examined here for two types of m.e.ta)multi-lamellar magnetoatatic rpherical enaloewre of 1 ferromagnotio ehells separated by (H-f) nenmagnetlo gapt:b)an anisotroplo spherical shell with high anisotrepy of pemeabillty at radial and tangential direction5 to the ahell,Such anisotropic ehell ie treated ao an anymptotio (a+-) model of multi-lamellar magnetostatic shield with fixed Inner and enter radii.Both types of arhieldn attenuate substantially etronger higher modes of the external nonuniform field than its fundamental mode,ao that the homogeneity of the lnrlde field la increamed, MultlZamellar umerlcal inamdtoetatlo emeider. Suppoae that the PPsgnetio potential nonuniform magnetic field ia given as 5um of 5patial modest of an external r , ,!I -spherical coordinaterr,PnlcosB)the Leeendre polynomials, *) Equation (1) holde e.g. in oaee of a magnetic dipole when the dipole moment vector ir in the O-5 direction and the dipole ia itself at a distanoe “aa from the ori n* shield ia mede of $ e multllamellar a ferromagnetic materiel with a hl@ pammbility $~a$, The .~Wattio~. of. ,thda problem with proper boundary oonditions show5 that the oomponenta of the field veotor H Inride the 8hield are equal to where index iii denotes ihe &Bide region of the ehield, j= !,2,3; oLfiJ,&r,O,Y))i hL2,3=(f,rjr&B). &@-the llrhielding coefficient of the 2Llamellar shield for the n-th mode. w$-the reaction o) the B-lam&w ahield for the n-th mode:the index “‘111” in these equations irr for magretlo shella,i,e, dM ,dH are the thiokneeetetl of mametio or reapeotively ~Yrhe final reetalt6’az-e also eorreet when U” Is expandid In terms of Pi@@ (the a55eelated Legendre polynomials). This fite,e.g, the oarno of the *herIaontal magnetic dipole*,f.e. the d&pole moment veotor ia in the x-0-y plane. - 516 nonmagnetic shells. &-the outer ra(N-‘-l and the thickness of a single shell decreases correspondingly so dius of the shield consisting of (N-l) that the Inner and outer radii of the inner magnetia shellsR~5Rg+d"*(~-z~[dn+d~), enclosure are fixed (T=const,(&+dn)-,O, &-the innar space radius of the dH/dH=const).In this case the permeashield.The first term in (2) corresbility of the magnetic material may ponds to the uniform,tho second onebe taken as a tensor of the kind: to the nonuniform part of the field. With N=l,2 equations (3) and (4) tran+ sform to the classical Kaden's results [2],generalized to a nonuniform field iErz%e of the shield structure. Here WC introduce the equivalent parameters of the shield with high for the n-th mode (geometry parameters and permeability) to estimate how fast the second term in (2) decreases with inoreasing n for such a shield.This significantly simplifies the analysis when compared to the treatment of the general equations (3),(4).In accordance to &,.>>Iit is possible if in (3),(4) one takes (5) $&- =jR,R:;,d:dMF (6) An unilamellar shield (Nzl). For relatively thin unilamellar shield of thickness d with strong ma netostatic shielding effect (S~ti=,~>I, nt Ra=Ro-+d . It fdilows from (7) that the coefficients of the nonuniform component tend to zero as&/Ut>>$b,this tendency being stronger with increasing mode number (supposing that (8) is correct).The desired increase of field homogeneity of the field within the enclosure is thus aohieved. An example. The nonuniformity of the field inside the magnetic shield made of "isotropic" (M$Ln=&) and "anisotropic" (Jut=&-/2, &*I) material was calculated for the case of an interfering dipole'at a distance a=Q2dfrom $hc origin,when d/Ra%10-2,Ko=q=10 . It is shown that the nonuniformity of the interfering field4 HPJ/fiIO) lies below 5% inside the whole enclosure of anisotropic shield.At the se.mctime for the isotropic shield a 12% nonuniformity was found for r=O,l.%O . Conclusion. It is shown that a passive sphcrical magnetic shield permits a considerable reduction of the relative nonuniformity of an interfering magnetic field inside the shield. References. l.WI$liamson S.J. et al.Biomagnetism n "Superconductor Applications: SQUIDS and MachineP,ed.by Schwartz B.B.,Foner S.,Plenum J?ress,N.Y.,1977/ 2.Kaden H.Wirbelstrome und Schirmung in dcr Nachrichtentcchnic (in German),2nd ed.Berlin:Springer (1959). 3.Cohen D.Rev.de Phys.Appl.,5,53-58, (fev.1970). - 517 94P5 - F,MP ENCLOSURE PENETRATION AND CABLE COUPLING Habibur Rahman Department of Aerospace Engineering Parks College of St. Louis University Cahokia, IL 62606 Jose Perini Department of Electrical and Computer Engineering Syracuse University, Syracuse, NY 13210 Abstract A very efficient formulation for the problem of a bundle of straight wiers in a rectangular cavity is presented. Any number of wires in the bundle can be excited and/or terminated by arbitrary loads and voltages at each end. The excitation is assumed to be in the form of a unit voltage source at the point of entry of the wire in the cavity. The currents in all bundle wires and the input impedances of all excited wires are calculated. Introduction When aircraft, ships or shielded enclosures are subjected to intense electromagnetic radiation, such as that of an EMP pulse, it is well known that very large currents, of the order of thousands of Amperes, can appear on cables or cable bundles that interconnect equipment inside such enclosures. It has also been observed that if one cable in a bundle is excited it will excite all others propagating energy everywhere the bundle goes. If these cables carry power or data for computer equipment, even small currents can cause irreparable damage. It is therefore important to study the mechanism by which the enclosure is penetrated and how the cable coupling occurs so that ways may be devised to eliminate or attenuate the induced currents. Usually the penetration occurs because small portions of power cables are left unshielded or because they are connected to a sensor that cannot be shielded such as a nav.,gation light or antenna. The prediction of this coupling and the calculation of the voltage at the entry point of the enclosure can be done by using the many available computer programs for wire coupling. In this paper we will be concerned primarily with the prediction of the current in a cable inside an enclosure, and the coupling to other cables in the same bundle, once the voltage at the entry point is known. Wire bundles usually run through several compartments of the enclosure as shown in Fig. la, where the compartments are assumed to be rectangular cavities. This problem can be analyzed if the simpler canonical problem of Fig. lb is solved first. Here a voltage source V is applied at one end of a straight wire W which is terminated by a load ZL at the opposite cavity wall. Once the input impedance seeing by the generator V and the currents on the wire as well as in ZL are known,the original problem can be analyzed as shown in Fig. lc. Given the load ZL3 the input impedance ZL2 be calculated which in turn allows the calculation of ZLl. Once ZLl is known and Vl is set to l+jO volts the current on wire WI and the voltage V2 across ZLl can be calculated. With V2 we can proceed in a similar fashion and calculate the current in all wires. Figure 1 shows the case of a single straight wire in the cavity. If other wires exist, as in a bundle, the several wires can be specified and a similar procedure used to calculate the currents in all wires. The canonical problem for this case is shown in Fig. 2 for the case of two wires. Note that the problem is solved for a voltage of lCj0 volts. Once the input voltage Vi is known, by solving the external wire coupling problem, the wire currents can be calculated by simply multiplying them by Vi. Problem Formulation To analyze this problem a rectangular cavity geometry was used and the wire currents were expanded in a truncated Fourier series. Since the wires are parallel to one of the cavity dimensions, as shown in Fig. 1, the Fourier series expansion is orthogonal to the cavity modal expansion simplifying the problem. Next the cavity modal coefficients are expressed in terms of the currents Fourier Coefficients and therefore eliminated from the computation. The problem in principle requires three nested infinite summations. One is eliminated by the orthogonality with the Fourier series expansion and a second one was carried out in closed form. This reduces the computation time considerably. Usually it is necessary to use several thousand modes for the solution to converge. However this does not cause any problem since only a very small amount of computer time, of the order of a fraction of a second for a single wire, is - required. The surface impedance of the wire can be specified by the user so that loads can be placed arbitrarily on the wire. This also allows the treatment of the problem where the wire does not run the whole length of one of the cavity dimensions. In this case it is only necessary to specify a very large surface impedance for the wire from the point where it terminates to the wall where the wire would normally run to. The advantage of this approach is that the current is still expressed by a Fourier series that runs the whole length of one of the cavity dimensions, preserving the orthogonality properties. The formulation for two wires will be used here in order to simplify the algebra. The extension to any number of wires is straightforward [2]. The geometry of the problem is illustrated in Fig. 3. The wires are located at r(i) = (x,yi,si), i=1,2. Assume that the radius of the wires, ri, are much smaller than the wavelength so that circumferential currents can be ignored. Time harmonic with angular frequency w and the factor .jW are understood throughout. The magnetic vector potential A has to satisfy the Helmoltz equation V2& -t k2A = - J (1) where k is the wave number of the medium of the cavity interior, and _J is the source or impressed current. y Btcos T v=o x i=1,2 (2) Thus, from Eq. (2) we have x S(y-bi>S(s-ci) and permittivity respectively. The x-component of -E is readily written from Eq. (5) as ' (* + k2)Ax E, = -z---JWEo ax2 (3) where 5 is the unit vector in the direction of x and 8 is the impulse function. Note that J is x-directed and the wires are thin; so it'is expected that a x-directed A is sufficient for representing the fields inside the cavity. Thus-Eq. (1) reduces to (V2+k2)Ex= f &- d(y--bi)c+.ci) i=l 0 v=o tk2 - (?)2}Bi cos 3 B;cosTx&(y-bi)d(z-ci)(4) 5 i=l v=o In terms of the magnetic vector potential we can express the electric field E as [ll E=- V(V*A) jwvo A + -LjWEo (5) where u. and EO are the free space permeability x V (7) The general expression for the x-polarized electric field $ inside the cavity is Ed= mIo nI,prl *mnp cos Ex a sin=ysin=s b c (8) where A's are the unknown mode coefficients. Obviously, the field represented by Eq. (8) satisfies the boundary conditions on the walls of the cavity. However, it remains to satisfy the boundary condition on the surface of the wire. If we substitute the value of Ex from Eq. (8) into Eq. (7) and performs integrations on both sides by forming suitable inner products, we have mnn = (y)2} 2 1 Bisin 7 bc(k2-E2 )WEo i=l mnP bisin y ci (9) where KLp = (?)2+ (7)2+ ($)2 (10) In order to evaluate the coefficients Bi's, we consider the boundary condition on the surface of the wire which is given by iixE = &(x)1(x) -tot (11) where ii is the unit normal vector on the surface of the wire, Etot is the total electric field and g(x) is the surface wire impedance. In view of the assumptions made earlier, Eq. (11) reduces to E X ’ - &'(x)Ii(x) = - ETan i=1,2 (12) in where Etan is the tangential component of the impressed electric field. Let us define a testing function of the following form wuz+cos y ;. i=l (V2+k2)Ax=- (6) Eliminating A, from Eq. (4) by Eq. (6) we have j4tK2 - where the B l's are the Fourier coefficients yet to be determined. The representation of Eq. (2) converges to Ii(x) on the closed interval 0 -< x -< a. B;cos f - A Let us assume that the current on the wires inside the cavity are x-directed and that they undergo variations with respect to x. In view of this, the unknown currents on the wires are represented by Fourier cosine series in the interval 0 < x < a as Ii(x) = 518 + 6(y-yi) {CE;(Y-Yi-r)+6(y-yi+r)16(z-zi) [b(z-zi-r) + S(z-zi+r>l I (13) Substituting Ii(x) and E from Eqs. (2) and (8) into Eq. (12), multi$lying both sides of the resulting equation scalarly by W and finally performing integrations with% the limits of cavity dimensions one obtains a set of equations that can be put in matrix form as B1 = B2 V1 (14) V2 References or .. first problem is straightforward and the third should not presentanytheoreticaldifficulties. ZB = V [ll Harrington, R. F., Time Harmonic Electromagnetic Fields, McGraw-Hill Company, New York. [2] Rahman, H., "Analysis of Wires in a Rectangular Cavity," Ph.D. Dissertation, Electrical & Computer Engineering Dept. Syracuse University, August 1984, also SU TR-84-8, November 1984. (15) The matrices ZLJ represent the self and mutual coupling between wires. B", i-1,2 are the vectors of the coefficients of the truncated Fourier expansions of the currents and the vi, i=1,2 are the wire excitation vectors. Note that by selecting the testing functions of Eq. (13) the boundary conditions are being satisfied at four generatices of the wire at (yi?r,si) and (yi,si+r). The expressions of Eq. (14) are very lengthy but easy to obtain [2]. Equation (15) is now solved by inverting the matrix Z B = Z-l" El Y ” z (16) L b) giving all the Fourier coefficients of the currents on the wires. 2, "2 2L2 "3 "1 2Ll C) Fig. l-a) Original problem. b) Canonical problem. c) Solution of original problem using the Canonical solution. v ZO:! “Cl\ ___ ___ WY 1117i “, “2 II /, 1' Fig. 2. Canonical problem for wire bundles. Conclusion and Recommendations The computer program predicts the current on straight wires or bundle or wires, in rectangular cavities, when one or more wire is excited and the others are terminated by specified loads. It confirms the practical observation that very high currents can be induced in wire bundles or parallel wires even when the wire separation is large. The analysis can also predict the currents in waveguide multiple probes without the usual assumption that only the fundamental mode is present. The obvious extensions of this work is to treat the case of cavities of other shapes, the case where the wire makes one or more bends before reaching the cavity wall and the case of surrounding the wire bundle with lossy dielectric material to attenuate the currents. The second problem is being analyzed presently and is the more difficult of the three. The L "1 "2 Numerical Results A few numerical calculations are shown in Figs. 4 to 9. All cavities are rectangular and have x,y,z dimensions of 3,4,5 m respectively. Figures 4 and 5 show two wires of diameter 0.001 m at locations (x, 1.5, 2.0) and x, 2.5, 3). In both figures wire 1 is excited at x-3 m by V=l volt and wire 2 is either connected to the cavity wall or left open. Note that when open the parasite wire current is of the same magnitude of the excited even though they are about 1.5 m apart. Figures 6 and 7 show three parallel wires half a meter apart. When wire 1 is excited, due to symmetry wires 2 and 3 have the same current. When wire 3 is fed, as in Fig. 7, the symmetry is broken and wires 1 and 2 have different currents. In these two examples the wires not fed were shorted to the cavity walls. Figures 8 and 9 show a similar situation but for a bundle of five wires. The wires not fed are shorted to the walls. The induced currents are about five to six times smaller. This is still significant taking into consideration that the bundle diameter is one meter. "1 Fig. 3. Geometry for the two wire problems. . - 520 =: .oo . (x10-'] LENGTH IM .oo LENGTH METERS IN METERS Figs. 4 and 5 - Current magnitude for two parallel wires in a rectangular cavity (3,4,5 m>. F=60 MHZ, 1000 cavity modes, 10 Fourier coefficients, for different parasite wire termination. :: (x10-‘) .oo LENGTH IN METERS (X10-1) LENGTH IN METERS Figs. 6 and 7 - Current magnitude for three parallel wires. Same conditons of Figs. 4 and 5. Different wires fed. Parasite wires shorted. LENGTH IN METERS Figs. 8 and 9 - Current magnitude in a five wire bundle. Same conditions of Pigs. 4 and 5. Different wires fed. Parisite wires shorted. - 521 95 - P6 MEASUREMENTS OF TRANSFER PARAMETERS OF SHIELDED CABLES AT FREQUENCIES ABOVE 100 MHz B. Demoulin, P. Duvinage et P. Degauque Lille University, Electronics Dept. 59655 Villeneuve d'Ascq Ckdex, France We propose a method to measure the transfer parameters of coaxial cables at high frequencies. The usual test setup made by concentric lines is keeped but the cable under test is coveredwith an additional shield except along a small length. This part, exposed to the disturbing electromagnetic field, behaves as a shield discontinuity. Matching the disturbing line is not necessary since measurements are made in time-domain. Results in frequency domain are determined through FFT procedures and are given until 1000 MHz. Limits of the usual benches Measurements at low frequency The simplest bench which could be used is represented in Figure 1. COWhI $"?&g fine I 1 cd) I ,,,,------‘a#J----- Introduction _---The usual method for the measurement of transfer impedance and admittance of coaxial cables uses two propagation lines : a disturbing line and the cable under test. In most cases, the disturbing line is made by a hollow tube which acts as an outer concentric conductor, the inner conductor being the cable shield itself. This disturbing line is driven at one end and the transfer parameters are deduced from the voltages appearing at each end of the coaxial cable. At low frequency, thus when the cable length is much smaller than the wavelength, the transfer parameters are easily determined from the measurements. However, if the frequency increases, the propagation along the line cannot be neglected. We shall see on an example that corrections based on a mathematical analysis does not allow to get results with a high degree of confidence for frequencies above about 100 MHz. The only solution is to decrease the cable length but putting a cable few centimeters long in a usual bench is not realistic. We propose a new method of measurement badiscontinuised on the behavior of shield ties. The coaxial cable has, as usually, a length of about Im. An additional homogeneous shield (copper adhesive tape) is added but it remains a small length of the braid which can be exposed to the disturbing field. This cable is then put in a hollow concentric tube such that the disturbing line is the same as in the usual benches. To avoid reflection effects due to mismatches, the line is excited by pulses of current with a fast rising time. We shall describe this method and give results in the 100 MHz-1000 MHz frequency range. Fig.1: Experimental set-up Both the disturbing line (subscript d) and the coaxial line (subscript o) are matched. At low frequency, the disturbing voltages at both ends of the coaxial cable are related to the transfer impedance Zt and to the transfer admittance Y t by the following approximate formulas : Vo(o) = -~(Zt+ZcoZcdYt)Id(0)L (I) Vo(L) = ~(Zt-ZcoZcdYt)Id(0)L If the tranfer admittance is negligible as for shields with a high optical coverage ratio, the disturbing voltages become equal and proportional to the transfer impedance. Otherwise, both V (0) and V (L) are measured and Z and Yt are'deduced fgom equations (1) and ($1. Limits for high frequency If the reaction of the coaxial cable on the disturbing line is neglected, the voltages VAL) and V,(o) are given in the general case by Vo(o) = ~tZt+Z~oZcdYt)Fot~o,~d,~~~dt~~Lt3~ Vo(L) = %Z -rZ Z Y )F (y y L)Id(o)L (4) 2 t co cd t L o' d' F. and FL are functions of the propagation contants of the coaxial line y and of the disturbing line yd. They chara:terize the propagation of the waves into the structure. - 522 - As an example, the curves in Figure 2 represent their variations as a function of frequency. The velocity of the waves into the two lines is v. = 2.10Sm/s and vd = 3.1OSm/s. oddihnal ,Shield / I1 t 4 &Id of Me I aable 1 ~___GL.._?!~~~ -fm 2 I ______‘__,, Oscif f0sCope Fig.3: Test setup for high frequencies Fig.2: Influence of the frequency : variations of F. and FL If the transfer admittance is neglected, the low frequency approximation (F =F ~1) is valid with and error which does no ? eiceed 10 % if f < 10 MHz if V,(o) is measured or f < 100 MHz if Vo(L) is measured At higher frequencies, the calculated values of F, and FL must be introduced for the determination of Zt. However this supposes that the propagation constants in the two lines are well known and that these lines remain matched whatever the frequency. If these conditions can be verified for the coaxial cable, difficulties occur for the disturbing line since it requires complicated mechanical parts at the ends of the bench. Even if all these conditions could be satisfed, we note from Figure 2 that F and F L have small values at high frequencieg and are equal to zero for few frequencies 111. This leads to a low precision in this range of frequencies. From our experience, one can expect to reach 100 MHz but not higher values 121. A new method is thus needed. Measurement at high frequency Principle As we have previously outlined, the solution to increase the frequency limit is to decrease the length of the cable. In order to avoid to test a cable few centimeters long (and thus to measure the end effects), we propose to start from a cable Im long but to add on the braid an additional shield, made by a copper adhesive tape for example, except on a length 6. The configuration is represented in Figure 3. The cable under test behaves now as a wellshielded cable (perfect shield at high frequency) except along a small discontinuity. To avoid to match the disturbing line, measurements are made in the time domain with pulses of current such that its rising time is much smaller than the time of propagation along the line. The additional shield being very effective, the disturbing voltages appearing at both ends of the coaxial cable are due to the penetration through the braid of length 6. This length must be chosen so that the "low frequency" approximation previously described, can be applied. This leads to a length of few centimeters (3 cm to 5 cm). Signals processing To simplify the explanation, let us assume that the transfer impedance behaves as an inductance L . In this case, the effect of the braid appearE during the transition time of the disturbing current. The shape of the various signals are given in Figure 4 if the discontinuity is situated in the middle of the cable. We note that the effect of the shield discontinuity occurs at a time 0 /2 + Qd/2 after the beginning of the pulse of'current where So and 8d are the durations of propagation along the entire disturbing and coaxial lines. The effect of the discontinuity on the voltages V (0) and V (L) corresponds to the derivative xf the curgent. Then they are reflections of the disturbing current at the end of the line which give rise to oscillations of the measured voltages. TO process these signals, we first make a shift in the time domain such that all the signals appear at the same time. (correction of the time of propagation). Then the signals are "cleaned" to eliminate all the reflection effects after the significant response. Before applying Fast Fourier Transform procedures, there is a convolution of the signals (almost Id(O t)) by the Nicholson Transform Method 141. - 523 95 - Low @ical Cowyage P6 ratio I -- “~----_----_-- Fig.6: Variation of the phase angle of Zt To obtain these results in the 10 kHz 1000 MHz frequency range, three experiments have been made successively with various rising times of the disturbing pulse of current: rd = lus, IOOns, 6OOps., the window for the Fourier Transform being respectively 4Ous, 4~s and Sons. We see from Figure 5 that above 1 MHz, the amplitude of the transfer impedance remains proportional to frequency. This behavior agrees with a theoretical model base on the diffraction by small apertures 131, 151. If we now consider the case of a coaxial cable with a high optical coverage ratio (90X), we show in Figure 7 and Figure 8 the variation of Zt between 10 MHz and 1000 MHz. As a comparison the previous results have also been noted. We see the limit of validity of the results at about 1 GHz especially on the variation of the phase angle. Fig.4: Shape of the signals I and V Experimental results The variation of the amplitude and phase angle of the transfer impedance of a coaxial cable with a low optical coverage ratio is given in Figure 5 and Figure 6. (Coverage ration A = 60 X> 0 0’ 0- OfLf cbvercye mhb Fig.5: Variation of the amplitude of the transfer impedance Fig.7: Amplitude of the tranfer impedance for two cables (low and high coverage ratio) - 524 - An other limit is the sensitivity of this method of measurement. Indeed for cables with a high shielding effectiveness, the signals have to be amplified in a wide frequency band and with a good signal to noise ratio. Conclusion Measurement of the transfer impedance of coaxial cables above 100 MHz is not easy if the propagation along the lines cannot be neglected. We have shown that the usual setup made by the cable under test and a concentric tube can be used if only a small part of the cable is exposed to the disturbing field. Furthermore the approach in time domain allows to avoid to match the disturbing line. _/t --------References III t Martin, A.R.: The shielding effectiveness of long cables : maximum leakage. 5th Symposium on E.M.C. (Zurich), Proc. 379-384, (1983). 121Demoulin, Fig.8: Phase angle of the transfer impedance for two cables Limits of the method As it has been previously outlined, the length of the discontinuity must be small enough so that the electromagnetic field is nearly uniform along it. On an other hand, this discontinuity must characterize the braid and thus it must have a value of at least few centimeters. A compromise between these two assuptions leads to a discontinuity of 3cm to 5cm long. The rising time of the pulse of current is chased to about 400 ps. In this case, the upper limit of frequency is between 1 and 2 GHz. However the behavior of the cable for frequencies smaller than 1000 MHz is sufficent for nearly all the applications. B.: Etude de la penetration des ondes Clectromagnetiques a travers des blindages homogenes ou des tresses h structures coaxiales. These de Doctorat d'Etat, Lille, (1981). I31 Vance, E.F.: Shielding effectiveness of braided-wire shields. I.E.E.E. Trans. on E.M.C., vol. 17, no 2, 71-77, (1975). 141 Duvinage, P.: Etude et caractdrisation electromagnetique des discontinuites de blindage. Application B la mesure des parametres de transfert de csbles coaxiaux aux frequences &levees. These Doctorat 3e cycle, Lille, (1984). 151 Demoulin, B., Degauque, P. et Cauterman,M.: Shielding effectiveness of braids with high optical coverage, 4th Symposium on E,M.C. (Zurich), Proc. 491-495, (1981). - 525 MAGNETIC OF TWISTED TO MULTICORE TWISTED AND AG, Power Erlangen, The following means can reduce the interference runs : be used to along cable - Connecting cal paths to - Using screened - Separation different cores conducting the points are dealt with 1. Voltage induced in a twisted twowire line by the field of an untwisted two-wire line 2. Magnetic multicore stray cable field of a twisted, 3. Current induced in a rectangular loop by the field of a twisted, multicore cable 4. Voltage wire line multicore THEIR TWO-WIRE COUPLING LINES Gonschorek and Automation Republic of Group Germany Introduction The simplest way of estimating the worst case interference of a constant magnetic field on a twisted two-wire line is to consider an effected area formed by a half length of twist. As shown in Fig. 1 the vector of the area influenced by the magnetic field changes its direction every half twist length. This is indicated in the figure by a plus and a minus sign. S = length RO = core of twist radius cables This paper deals with the effects of twisting of power and signal lines to reduce the magnetic stray fields and the magnetic coupling. The following individually: AND symmetri- of cables conducting signal and power levels - Twisting of same signal CABLES Federal Knowledge of the active and passive interference behaviour of the devices and components is necessary in order to be able to analyze the EMC of a system. Quantitative values regarding the interference radiation of the cables interconnecting the devices as well as the interference coupling into these cables is necessary, signal FIELDS Engineering The electromagnetic compatibility (EMC) between power and electronic systems is becoming increasingly difficult due to the increasing power ratings on one hand and the reduced signal levels and increasing packaging density on the other. the STRAY NON-TWISTED K.H. Siemens 96w - induced in a twisted twoby the field of a twisted, cable Fig. 1: Schematic representation of a twisted two-wire line Assuming any given length of a line, the maximum area which can be subjected to interference is, as stated above, defined by half the twist length. A rough approximation of this area is given by A = ROSS and the voltage “i maximum induced is (1) open-circuit = w.B*RO.S. The effects of large changes in the magnetic stray field, such occur in the fields of twisted cables, can at present be only inadequately estimated. To improve this unsatisfactory situation this paper deals in detail with twisted lines. Simple formula are derived for analyzing the interference effects. (21 Voltage induced in a twisted twowire line by the field of a nontwisted two-wire line The arrangement in Fig. 2. studied is 526 The resulting a function of parallel path shown .04 0.00 I Fig. 2: IT‘ X T-- 3: ---!+ 1 r-d 1 * i r+d >> d, this equation as follows: spacings r be simplified = po. &.G d2 0 ‘F M = 37.4 (3) The coupled r2 the for value of spacings .lOo%. voltage 2’d.RO.S (5) i-2 * it was assumed that Furthermore, the magnetic flux density in the influenced area remains constant and is equal to the value along the centre axis. ‘i = uo . f.1 For I = 1 A, d = 0.5 cm, r = 10 cm, RO = 3 mm, S = 20 cm and f = 10 kHz , we obtain the value of the induced voltage = 7.5 uv . “i In order to assess the accuracy of the approximation (51, the arrangement shown was calculated for mutual inductivity by means of a computer program. The program has already been described in detail in /1,2/. 2. = pH . voltage w.1.M is given = 2.35 uV . by exact solution in (5) is F = 10 dB . The approximation (5) can be used for all practical cases as described above. Extending equation (5) by a correction factor k=0.5 derived from Fig. 3 reduces the error to F = 4 dB . Magnetic , the coupled r >> d is .20 x [ml = 10 kHz, d = .5 cm, RO = 3 mm, S = 20 cm The error between the and the approximation Assuming that for estimation purposes the area which can be influenced (based on the twist data) is given by A=RO*S .is It can be seen that for x = 10 cm and 20 cm the mutual inductance and thus, the induced voltage is zero and for x = 5 cm and 15 cm the mutual inductance has a maximum, equal to In this case the error between the exact solution (3) and the approx. is proportional to the solution (4 ratio (d/r) 4 : F= .i2 Mutual inductance for the arrangement shown in Fig. x = Length of the parallel path. ui BZ .OS I q 1 A, f r = 10 cm, The magnetic flux density of the non-twisted cable for field points in the xy-plane is given by the equation For can Fig. Arrangement of a non-twisted pair of wires causing interference in a twisted pair Bz = po+* mutual inductance as the length of the is shown in Fig. 3. stray field of multi-core cable a twisted, The magnetic flux density of a conductor of finite length in space is given by the equation with the transformation ratios (7) a’ E _ = /q-To/ (8) /(;+-~o,x(?l-Fo~/ _______ (9) /F1-iro/ y= (??-Fo)* ‘F1-Fo’ (10) ‘=x3 See also Fig. 4 and the description in /l/. detailed - 96 527 - ~7 with BXI (Px=atan BXR (15) (16) BY1 AY2 = BYR2tBY12 , cPy=atan BYR (17) (18) (19) (20) AX2=BXR2+BX12, AZ2 =BzR2tBz12, Cp =atan Z s 6 X4+Az4t2.K2*AZ2*cos(~K-2~z- ;) AX4tAy4t2. AX2.AY2.COS ( 260x-2qy) 4: Fig. Geometry equation - used (6) for AX2.sin(2$t deriving (PX=atan The magnetic stray field of One core of a twisted cable is obtained approximately by representing the core by a polygon and applying equation (6) to each side of this polygon. The representation of one turn of a core by 12 straight lines has proved to be sufficiently accurate, Considering these bases, the treatment of multicore cables becomes only a programming problem. A current I. in a wire or in a core having amprttude Ii and phase angle can be represented by the equation Ii This flux = Ii*(cos current density vi+ Ii Tf j - generates C Bi=BXi.~x’BYi.b sin Each summand in equation the current 1, according (11) as its %efficient. (11) (Pi). a magnetic +BZ..~ -1 Y i, ‘Pi T)tAY2*sin (21) (22) (2%+5) (23) AX2.cos (2’px’ ; )tAY2. cos (2c”y’ ; ) If we now use these equations to study the magnetic stray field characteristics of a twisted two-wire line, we obtain rather suprising results. Fig. 5 show the magnetic flux density of a twisted two-wire line for I = 1 A. The line data are: length of twist RO = 2.76 mm. S = 90 cm, core radius The radial spacing r as measured from the centre of the cable is drawn On the abzissa. Parameter x for the individual curves is the axial distance from the beginning of the cable. The lines of the field points thus run perpendicular to the axis of the cable. (12) Z’ (12) has to equation cable not twisted, x = 5 m If a system contains a number (N) of wires or cores carrying currents of different amplitudes and phase angles, the total magnetic flux density is given by the vectorial, phase-correct addition of the individual components. Fig. i? -ges= +(BYi’cos vi+j*BYi.sin = (BXR+j+BXI)$ qi)a + (BYR+j*BYI)p Y at one from 3 = $ (AX2+AY2+AZ2+L2) values the The curve is harmonic for x = O! the field strength decreases approximately in inverse proportion to the square of the spacing. field the 1 / of Magnet ic flux density a 10 m long twisted cable, carrying a current of 1 A, on field point lines perpendicular to the cable axe, Parameter x is the axial distance from one cable end It can be seen that for various Of X widely different curves of stray field are obtained. Y + (BZR+j-BZI) .< The maximum amplitude point can be calculated formula 5: (14) For x = 80 cm, there is a sharp drop to a local minimum value close to the - cable. Thereafter, the field to a local maximum and then ly approaches zero. 528 - increases continuous- A better picture of the magnetic flux density is obtained by a three dimensional representation (Fig. 6). In Fig. 6, the display range in the axial direction is from x = -1 m to x 5 m and in radial direction from r = .I to r = 6.1 m. q Fig. 7: Magnetic flux density NYM 4 x 4 cable Display range: x = -1 m to 5 m 1 m to 6.1 m r=. for of the twisted two-wire line. This curve runs above all the curves of the twisted line. Fig. 6: Magnetic flux density for a twisted two-wire line; Display range: x = -1mtox=5m 1 m to r = 6.1 m r=. As can be seen minimums in the axial direction occur at regular intervals corresponding to the length of twist. It is also interesting to note that in close proximity to the line weak minimums occur in the axial direction at distances corresponding to half the twist length. Fig. 7 shows the magnetic stray field of a twisted, four-core cable used for electrical installation in buildings, type NYM 4 x 4 (length of twist S = 90 cm, core radius RO = 2.76 mm). A symmetrical three-phase current of I = 1 A was impressed on the cable. There was no current flowing through the remaining core. The display range is the same as in Fig. 6. It can be seen that there is hardly any difference between Fig. 6 and 7. The behaviour of cally the same. both From these figures one that the maximum fields the ends of the lines. lines is Thus formulas (3) and (4) can be used for a worst case estimation of the magnetic stray field of a twisted cable, replacing d by RO. Current 1 00 Here example induced multicore in a rectangular cable again, we shall from /2/. consider It is required to calculate the current in the sheath of a twisted cable NYKY 3 x 185195 which is laid with a spacing b above a metallic cable tray and whose sheath is connected regularly at spacings x*to the cable tray. See also Fig. 8. l:YKY 3x 185195 \I/l kH), we obtain a maximum cable sheath current M - max L M max = 55.5 Y loop (;;td2) ~7 above equation r must be reby b/2 and d by RO. We then the following value (24) From Fig. 8 in /I/ we obtain for xw .6 m, being half the twist length, = 54 nH as the maximum value M oFathe effective mutual inductance. “i 96 - M max=+R).$$-. Mi, M2, M3 are between the inaffected loop. The induced voltage in an open is thus given by the equation 529 --___________ 9: M - 52 _- x0=1. 5m Arrangement of two-wire lines - -__-___ X * two twisted The influencing (culprit) cable a twist length Sl = 1 m, a core RI = 1 cm and a length of 5 m. has radius The influenced (victim) cable has a twist length S2 = 0.5 m = S1/2, and a core radius R2 = 1 cm. The influenced cable begins at = 1.5 m and runs parallel with a spacing of 10 CIIJto the influencing cable up to x = x . X0 Fig IO shows the mutual inductance of this arrangement as a function of the length of the influenced cable. The mutual inductance is zero x* - xo = S2. It can be seen a local minimum occurs at x* - xo = S2/2; local maxiums approx. 0.46 nH occur to the and right of this point. At a current of I = 1 A and f in the influencing cable, the for that of left = 1 kHz maximum 530 - induced voltage cable is Ui = in Mmax 0. - the I be used accuracy. influenced 2.89 PV. = M in spite of the large in- I .15 .20 661 [PHI 646 431 215 T 0.00 .20 .I0 0.00 .30 .40 Xf - 1.5 Fig. 10: Mutual inductance arrangement shown .50 [ml for the in Fig. 9 = p, .f *I .=LFQ.S. r2 (30) The shorter of the twist lengths of the two cables is to be used S; here S = S2. We thus obtain “i = 12.5 for pV. The error is relatively large (factor 4 = 12 dB). One should not however overlook that we have here assumed ideal conditions which exist very rarely. The second example handles the coupling between an electrical cable NYM 4 x 4 carrying a three-phase current of two-wire line I = 1 A and a twisted laid close to it. Referring to Fig. 9, line 1 is the NYM 4 x 4 cable with a twist length Sl = 90 cm and a core radius two-wire RI = 2.76 mm. The influenced, line 2 with S2 = 20 cm and A2 q 1 mm is laid at a centre to centre distance of r = IO mm from the influencing starting at x0 q 0.5 m. cable, ‘i From equation of ui = 6.9 we obtain pV .5 [ml Mutual inductance for the arrangement of a twisted NYM 4 x 4 cable and a twisted two-wire line remarks Knowledge of these peripheral ditions add weight to the use approximate formula. conof References /I/ Gonschorek, K.H.: Numerische Berechnung der durch Steilstromimpulse induzierten Spannungen und Strijme; Siemens Forschungsund Entwickl .-Ber., Bd. 11, 1982, ppa. 235-240 /2/ Gonschorek, K.H.: Application of Computers for the Determination of Magnetic and Electromagnetic Coupling; 5. Symposium on EMC, Zurich, 1983, Sci. Contrib. 70 M3 131 Bridges, J.E.: Study of LowFrequency Fields for Coaxial and Twisted-Pair Cables; 10. TRI-Service Conference on EMC, 1964, ppa. 475-495 a value . The error in this example is 8 dB, For a conservative assessment of the magnetic interference between twisted cables equation (30) should x* - I .25 Exact assessment of the magnetic stray fields and of the magnetic coupling of twisted lines is only possible by the use of computer programs. Four arrangements are investigated extensively. The value of computer calculations is however not entirely undisputed as the actual physical condition is not the ideal one which is assumed when carrying out theoretical studies /3/. When twisting the conductors together the coupling is reduced by the compensation effect of opposing field or area components. The result is almost always a small difference between very large figures so that any manufacturing inaccuracies lead very quickly to large deviations between the theoretical and the actual values. The magnetic field of a coaxial cable is a good example. I_Iv . (30) I .10 An effective method of reducing the magnetic coupling between power supply and signal cables or between signal cables themselves is to twist the cores belonging together. Fig. 11 shows the curve of the mutual inductance for the above arrangement. From this, the maximum amplitude of the coupled voltage in line 2 for f = 50 Hz can be calculated as = 2.7 11: Concluding If a non-twisted two-core line is taken for the influencing line, the coupled open-circuit voltage can be calculated according to equation (5): ui Fig. 8 .05 - 531 97a1 - PROJECT OF A RAILWAY ELECTRIFICATION FROM THE EMC POINT OF VIEW M. Di Stefano FS (Italian Railways) Roma, Italy G.L. Solbiati SIRTI S.p.A. Milano, Italy Introduction Basic criteria for the project The expansion of railway traffic in Sardinia, one of the largest Italian islands led, to the "resolution of electrifying in the 25 kV, 50 HZ ac single phase traction system the most important FS lines of the island. The Italian rules 'concerning the problem of interference [I] define as induced the telecom lines which are, at least for a part of their route, at less than 3 km from an inducing line. These lines, having a total len gth of about 400 km, connect CagliariT the main town of the above district, with the greatest commercial ports,fun damental for the import-export move ments with the continent. In Sardinia the cablesforp; blic service which fall in this case, relating to the traction lines to be electrified, have a total ley! gth of about 1,053 km (fig. 1):more in detail the cables are: The same towns are also linked by a system of main roads, along which telecom cables for public service are laid: the cables are both for long dis tance service, with coaxial pairs an;i for short distance service with symmetrical pairs. symmetrical pair cables : 178 km coaxial pair cables 875 km 0.7/2.9 mm : 91 km : 848'kmgpair 'l.2/4.4mm :439 km :2772 km-pair 2.6/9.5 mm :345 km :2252 km-pair Owing to the orographic configuration of the island, the road system covers a considerable length in proximity to the railway network, and therefore, the problem of the electro magnetic compatibility ( EPIC) betwee;; telecom cables and traction system takes up a great importance. The same problem alsoarises for signalling and telecom railway cables which are laid inside therailwaytrack area. This paper shows the project approach followed and the results of the project under the EMC fea tures. A very short description of the calculation method for evaluating voltages and currents in a multiconductor system, widely used in this project, is also given in Appendix. A preliminary analysisofthe problem showed that a large percentage of these cables wouldhavebeen affected by interference, with re markable effects, if no protective measures had been taken. Railway signalling and telecorncables, laid along thetrack.and thus more affected by electromagnetic coupling, are expected to be: telecom : 46 symm. pairs : 400 km signalling: 1200 km From a general pointofview, the protective measures can concern both the induced cables andthetrac tion line. On the basis of economical considerations it was considered more suitable to look over the trac tion system only in order to insure the EMC with all theexisting cables for public service, and thentolook - 532 - W ne the type of cables, as to the shiel ding characteristics. railwayline t0 be electrified To this purpose it is possible to evaluate the compatibility using a single pair railway cable-inducing section where the latter is a reference section with a configuration proper to abridge the most severe conditions existing in the 13 actual sections. telecomoables The admissible limits for the different cable types are shown in table 1. COAX CABLES FOR PUBLIC SERVICE SYMM. 4cABms FOR PUBLIC SERVIGE SIGNALLINGAND TELECOM RAILWAY CABLES *I -- 60 1200 -_ 60 430* - - 60 430 Tab. 1 Limitsused in the project w 1 Schematicroute of railwaylines and telecompubliccablesin Sardinia ble 1 which pling lines The figures marked with x in ta are given by the Italian Rules only refer toelectromagneticcou between telecom lines and power under fault conditions [I] . Other limits were ,,defined by the telecom system owners on the basis of the CCITT Directives [2] . the railway cables only in order to assure the EMC with the traction system so shaped. o%$ Looking at the power SUPPlY scheme of the traction line, the 13 sections between a feeder station and the adjacent midpoint track set tion cabin (MPSC) are regarded as inducing lines: their lengths vary be; ween 19 and 38 km (fig, 6). To take into account the speci fit characteristics of each inducea cable for public service ( screening, length, distance from the railway, ...) in connection with the specific characteristics of the inducing sections (length, train locations, absor bed currents, .,.I 52 inducing line= induced cable pairs have to be considered. Different considerations be made for the railway cables. can Goal of the project is to defi Preliminary project This stage of project was car ried out by FS (Italian Railways)in or der to make the most proper choices foi! the project, taking into account the EMC with telecom cables for public ser vice. For this stage of project the following assumptions were considered: - the traction line is powered by 6 feeder stations each feeding radially line sections as far as the nearest MPSC(s); - traction currents, for each inducing section, are picked out in the most severe 5 minutes, with reference to the maximum expected traffic (fig.2). Among the possible choices, the - 533 - 97a1 MPSC FS telecom cable FS traction line 1,' ' 1, + 1 1 1 &_I 11 13 12 li lh MPSC Mid-Point Seotion Cabin train location (at time t') Ii FS Feeder station li train location (at time t) Fig. 2 1 l'lj Faction currents for two inducing section, both inducing the same telecom cable (asan eX!%Wle). - classic, with or without the wire; - with autotransformers; - with booster transformers. return contact olassio following ways to carry out the 25 kV, 50 Hz, ac traction system were COnSidg red (fig. 3, 4) Ill<<&:; rails ground z analysis The first step of the involved the optimization of signifi cant parameters (conductor section, ma terials, geometric arrangement, ...)of the traction system for each mentioned choice. with autotransformers 1 1 feeder 25 kv contact line -_______--_______- Then, following the calculation method described in [3] , the results of coupling between the 13 inducingsec tions, for every possible choice of eand each telecom cable lectrification, for public service, were computed in the form of emf induced on the cable sheath [4] . --------wire (retur rails mnmmmnnmmnmmmnmmmmmmground with booster transformers = The analysis of the results to the following considerations: led it is possible to use classic elec trification in some sections only; in some sections electrificationboth boos with autotransformers and with ster transformers can assure the EMF with the cables for public service: in some sections electrificationwith booster transformers is necessary to Fig. 3 contact line return wire The three possible ways tooarry out the 25 kv, 50 Hz, a.o. traotion system l feeder telecomcable l return emf E wire wire I railwayoable Fig. 4 Geometrical arrangementof the traction line conductorsand of the railwaycable electrification system: guarantee the compatibility. In fig. 5, from [4] the values of the emf induced on thecablesheath are plotted, for thecableaffected by the highest induction. Considerations notdependingon the EMC problem led to the conviction that the solution with autotransfor mers would not have been quite satisfactory and therefore itwas stated to consider the choice between classic system and system with booster transformers as the project trend for elec trification. Definitive project The EMC definitive project was developed by SIRTI SpA as a member of the TEAM Consortium to whom FS entruE ted the project and the implementation of the Sardinian line electrification. The first step of the definiti ve project consisted inevaluatingthz measures to be taken on the traction system in order to assure theEMCwith the public telecom network.This phase was developed according to the preli- classic withoutreturnwire l2- classicwith returnwire 3 - with autotransformers 4- with boostertransformers Fig. 2 The publictelecom cableaffectedby the highestinduction:emf inducedon the cablesheath minary project, by considering the al_ ternative between the classic system and the booster transformer one. To this purpose we basedonthe following elements: - FS and MPSC positions; - position of traction system condug tors and characteristics; - cables position and their screens characteristics; - admissible values for 'the induced voltages. Use was made of the calculation method adopted forevaluatingvol. tages and currents in a multiconductor system and shown in Appendix, as wellasofthe calculationmethod shown in [5] and [6] for evaluating the vo_l tages induced on a telecom cable. The calculationbas been per formed for each one of the 52 indu - 97a1 - 535 cing line-induced cable pairs in 02 der to determine the sections for which: to account requirements different from the EMC ones. Fig. 6 shows theresultsob - no particular measures are requi- tained. red; - just the use of a return conduc tor in parallel with rails is required; - the use of boostertransformersis required. The second step of the definitive project was the evaluation of the screencharacteristics and other protective measures to be applied to telecom and Signalling cables for railway service. This phase was developed ontheba sis of the traction line configuration coming from the first step. equipped For the sections with booster transformerstheirmost convenient positioning has been determined, while theb.t. characteris tics have been chosen eventakingig AS the cables position is constante along the line (fig. 4) cable the the emf induced on sheath depends on thetractioncur rent distribution and on the indu ted length only. For these reasonswedeemed it useful to perform the EMC calculations on one reference pair railway cable-traction line. The reference cable has been supposed to have one end in the FS: the re ference inducing section is thz longest FS-MPSC sectionassociated with a traction currentdiagramob tained as a wrapping of the ac tUa1 current diagrams onthe 13FSMPSC sections. For evaluating the emf per unit length induced on the cable sheaths, use was made of the calculation method adopted for ob taining voltages and currents in a multiconductor system, shown in Appendix: the emf induced on the cable conductors was evaluated V. sing the calculation method shown in [5] and [6]. This calculation procedure has been applied to the reference inducing section considering the three alternatives coming fromthe first step of the definitive project. MPSC 0 FS 2C$ classic,withoutreturnwire p.ppw 4% olassio,with returnwire wti 35$ with boostertransformers (4,5 km waced) 5$ with boostertransformers (2,5km spaced) Fig. 6 Eleotrifioation type,as the result of the definitiveproject On the basis of the screening characteristics ofsomecable types, suggested by the experiences, this step of the definitive project allowed us to know per ca ble type and per kind of traction system the maximum electrical len gth of the conductors accordingt_'E; the admissible limits for the induced emfs (tab. 1). Where required the electri cal continuity of the conductors is cut off by using isolating transformers. Fig. 7 shows a flow-chart summarizing the calculation proce dure. - 536 - Conclusions bhe sing how to carry out system. The evaluation of the electrothe magnetic compatibility between traction system and telecomsystemsis in the proof remarkable importance ject of an ac 25 kVi 50 Hz traction apsystem: when the electrification plies to a railway network placed whe devere telecom networks are quite loped, as in Sardinia, thisevaluaticin chofor in fundamentally important traction approach we have The project followed usescalculationmethods (Q pendix,[5],[6]) which for their gene rality and flexibility prove quite proper and lead to detailed choices for the precision of the results. preliminary project 1 public cables position I _ possible measures traction side _ Et railway net. configuration alternatives for screens Y dyiGGzJ I railway cables position F/ rr -1 I I actual inducing section I -- "(IF;\ 1 1 1 1 ! ! I i i voltage induced on the public cables oonduotors L I t eleotrification type for each section i ] ! ! I STEP 1 Fig. 7 inducing section 1 multi cona, * Plow-chart of definitive projeot I induced e.m,f. on the railway cable sheath I I reference pair: inducing section -induced cable I /----a I e.m.f. induced on the railway cables conductors I lengths of the * electrical continuity L tion of the iso@ ting transformers STEP 2 I 97m - 537 in the k-th Appendix [J] Computing voltages and currents multiconductor system. in a The calculation method is a 2 solution of nified approach to the conductor system in which voltages t U, and currents, I, depend on genera tors inserted in the system, G , on the coupling from external sources , C, and on the system self and mutual parameters, 2, Y; (U,I) = The whole system is formed by return ciras many conductor-earth cuits as the conductor are: the sysinto fl sections W tem is subdivided (they can have different lengths and the different parameters 2 and Y) : the the V sections ' are ends of I'pointsII. inserted in Both generators the system and coupling fromexternal sources are described through voltage sources inserted in the sections and current sources inserted in the points. The solution, currents in the sections, voltages in the points, is given by the following formulas (n is the number of points): [II, = @Jl,-blk+,+ bl,j k=l,...,n-I k=n cul k , . . ..n-1 [H]~+ [uIki,k=l where: k=l k=2 ,...,n k-l k=2 ,...,n [A],= k=l CM1, [M];[D\*[A$QHI,_, k=2,... ,n The input data of the calcula tion are: the impedance matrix of bJlk tern in the k-th section CYI, the admittance matrix in the k-th point PI, the voltage vector the current vector k in the k-th point. the of the sources such a solution has no limits for the number of conductors and the number of sections: butof course the computer program has such limits. viously f (G,C,Z,Y) section The calculation method can ok be applied even if noEMC pro blem occur, to study the propagation of voltages and currents in a given system (telecom , power , traction , etc.). When EMC problems are invol ved, both inducing and induced systems can be studied in one step putting them together in order to form a "big system" which is not induced by external sources. But this is not always the best way and it is not always possible due to the limits of the compu ting system. When some alternatives to the induced system must be studied for a given condition of the inducing one, it is better to split up the studyin two Steps separately applied in series to the inducing and the induced systems: the condition for this is that the effects from theinducedsys tern on the inducing one are nq.ligi= ble. When the induced systemis far enough from the inducing one , the first step ends with the calculation of the inducing current in every set tion, as the sum of the currents i;i the inducing conductors (from this, the emf induced on the induced system can then be evaluated). When the induced systemis clo se to the inducing one, but is quit; small compared to it, the first step ends with the evaluation of the emf induced on the induced system. In both case the firststepon ly studies the inducing system, not induced by external sources: the second step is the study of the indu ted system, induced by the previous one. SyS- of the sys- of the sources The description of this calcu lation method will be insertedinthz new edition of the CCITT Directives, which is expected to be published by ITU in 1988 c7]. 538 - References PI Norme CEI 103.6 - Norme concernen linee dy ti la protezione delle telecomunicazione dagli effetti elettromagnetica dell' induzione linee elettriche provocata dalle vicine in case di guasto. Cd concerning the CCITT-Directives protectionoftelecommunication li effects harmful nes against the from electricity line - ITU-1963 (1974-1978) c31 G. Guidi Buffarini - Impostazione generale de1 calcolo elettrico delle linee di contatto della tra zione a corrente alternata - IngZ gneria Ferroviaria - n. 8 - 5 gosto 1983 Dl S. La Rosa, M. Di Stefano - Elet trificazione della linea dorsale sarda. Scelta delle caratteristithe realizzative de1 sistema 25 kV, 50 Hz per limitare le fern lon gitudinali indotte nei cavi di tz lecomunicazione - Ingegneria Fer= roviaria - n. 12 - Dicembre 1983 c51 R. Pomponi, G.L. Solbiati - On the caused calculation of the effects by the electric power lines on the telecommunication cables - CSELT Rapporti Tecnici - n. 4 - Dicembre 1974 ISI effects Italy - Evaluation of the caused by the electromagnetic coupling between power lines and tele communication lines - CCITT Contrz bution Corn. V n. 42 - Period 19771980 L-71 GRD - Voltages currents in a and Numerical multiconductor system . solution - CCITT Contribution Corn. Vn. 29 - Period 1981-1984 _. -- - 539 98 - 02 COUPLING AND FILTERING POSSIBILITIESOF TRANSIENTS DURING EMC TESTS Harald Kunkel, Martin Lutz, Otto Frey EMILE HAEFELY & CO. LTD., BASEL, SWITZERLAND 1. Object Manufacturers and users of elCtriCa1 engineering products encounter more and more the problems presented by an electromagnetically polluted environment; in other words, with electromagnetic compatibility (EMC) problems - all the more as it is becoming widely appreciated that the quality of a product is extensively determined by its electromagnetic compatibility characteristics. The required standard quality can only be obtained by EMC testing during 'the design stage as well as during the final inspection of a product. In order to obtain meaningful results from an EMC test, the tester must possess indepth knowledge of the prevalent interference sources, the coupling links and the test setup. Some basic facts were already explained in references /l/&/2/. A very important aspect of this EMC test is the systematic and reproducible coupling of simulated interference into the test equipment under actual working conditions. For that purpose, various coupling elements are required. At the same time, suitable decoupling elements are needed to protect the untested system sections from the effects of interference. /6/ Various standard committees /3/ have prepared and continue to define test specifications and proposals applicable to certain applications and products. This approach has resulted in a number of different test set-ups, the interference impulses of which must contain the same spectral ranges. In this paper, an attempt is made to ,cover the entire frequency range from '10 kHz to 300 MHz with a minimum number of filtering and coupling elements. /7/ 2. The coupling and filter elements for the EMC test Fig. 1 Schematic diagram EMC test set-up of an During the EMC test, the coupling and filter elements must fulfill the following requirements: The coupling element is responsible for coupling the interference test parameter onto the equipment under test with as little attenuation of the required spectral content as possible. On the other hand, there should be only a minimum amount of damping of the useful or operating signal by the coupling element and the test generator. The filter must ensure that the interference variables are directed towards the test specimen and that no second system, such as the supply source, the ,AC power supply system, etc. is contaminated. The filter should also offer the highest possible impedance to the test interference variables to assure that the interference generator is loaded only by the test specimen. 2.1. Possible coupling elements a) Conducted coupling - electrical coupling by means of actual components - parallel coupling - 540 b) field coupling Coupling elements - indirect coupling by means of a parallel strip line type antenna or coupling clamp r-bwl A’8301 8 1 I Possible coupling elements Fig. 2 Coupling elements Varistors Gas discharge tubes Bipolar diodes Capacitors . . Fig. 3 E-Field coupling by means of a strip line type antenna Fig. 4 inductive-coupling in series inductive coupling, differential mode Table 1: Types of coupling and measures to reduce , coupling represented on the basis of a bipolar system. Direct coup11ng Common Mode system I Radiated coupling Radiated coupling S/S WE Common Mode system Differential II System Capacitive, Coupling Mode inductive Differential Mode RC power supply unit Common + Differential Mod I system II General transformer Interferencesource H/E field Coupling - 20 log ug E/H From H/E field calculate VEMr, then as for direct coupling Interference H/F field Coupling - SOUICC 20 log “0 fix Differential node EM2 Cowling 1. 2. . . Install choke in,.atth 1inss optoco”plar~ I “crease circuit inrredonce 3. raduction Reduce the h/l loop Insert ferrite cores t.0 absorbxxm”wn mode Screen the h x 1 loop Red.!. co&n.~ area Cou~llng r.duction 1. Reduce ‘pacing 2. Twist cable, 3. SCrewI parsu.1 conductors I Reduce cowlinq Couvlins s area . Select conductor, I. Select height Over earth plate (stray capacitance 8. Spaca between conductor, 1. t ncrcase 3. wscin,7 2. Split COINIIO” mOde EnI reduction connect primary screen to winding Differential node EMI CDnnsct acreen t0 e*rfl mode EN2 Optc.CO”pler I cOuDlinq CaPacitance 90 Fig. 5 r inductive field coupling itCtUAtin9 And monitoring device Measuringinstrument 02 I 8 IF- Power supply lines 0, rest specimen uI Clamp-on current transformer \ power signal generator \ 0 MAtChin Interference network base coil vindiws 2.2. Possible filter elements 2.3. problems which can be encountered After the description of various coupling elements and their configurations this section will deal with the implementation of filter circuits used in EMC testing. The use of inductive components, such as coupling and filtering elements for mv2 test purposes may present some problems. In the case of the SuperPOSition of test impulses - also called test surges - on the power or data circuits, peak amplitudes of up to 12 kV in the frequency range of 10 kBz and lower, are specified. Based on the typical test set-up shown in Fig. 1, the filter must comply with the following requirements: 1) The attenuation of the AC power supply lines from system I to system II for the high frequency interference must exceed 40 dB. 2) The attenuation of the power supply lines from II to I for the power supply frequency operating voltage should be as small as possible. (less than 1 dB). The maximum longitudinal voltage drop with the nominal current should be less than 10% of the rated voltage. 3) The dielectric strength of the filter should exceed the maximum occurring impulse voltage. Depending on test operations, an impulse voltage of up to 12 kV can occur. 4) The interference impulses coupled into the supply lines must not be influenced by the filter circuit. This means that the impedance of the decoupling device in range I must be significantly higher than the source impedance of the interference generator. 5) The series chokes used as filter elements must not exhibit any saturation in the specific frequency range; as otherwise, the inipedance of the filter will be changed and the filter attenuation will be adversely affected. With these test specifications it is difficult to meet the requirements outlined in items l), 2) and 5) of the above paragraph. Items 1) and 2) define the behaviour of the filter in a certain frequency 'range as well as the behaviour under powered conditions. The calculation resp. design of such filters resp. Coupling elements is well known and no special problems are met. The caraCteriStiCS of such components, however, as a rule are given for AC voltage operation. If used in impuls'eoperation in contrast to AC operation much higher voltages and also relatively high DC components occur. These two parameters can be expressed as voltage/time surface. If the voltage/time surface is too big, i.e. the impulse voltage too high, resp. the impulse time to halfvalue too long, the inductive components are driven into saturation. Such neither the specified maximum amplitude nor the specified time to halfvalue can be superposed, respectively filtered. The consequence is, that the specified test impulses cannot be' coupled unchanged on the equipment under test. Fig. 6 Fig. 7 Fig. 6 Fig. 7 Erroneous design of the filtering element in act. with Fig. 1 Correct design of the filtering element EUT Filter Mains Fig. 8 principle diagram superposition of current impulses 3. Typical EMC test system COnfigUration /4/ Example: Test system CR0 pictures above show impulses of 6,8,10 c 12 kV peak. Burst generator type PB 4 Coupling/filter element FP 16/3-l Coupling Clamp IP 4 1 to 300 MHz In Fig. 7 the specified impulse 1.2/50 us is correctly superposed on the POwered circuit, whereas in Fig. 6 with the erroneously designed filter the series chokes go into saturation above 6 kV so that the test impulses are distorted. If coupling and filtering elements available on the market are used, the rated operation data with respect to rated voltage, rated current and rated frequency must be very carefully checked. If the rated data are given with regard to AC operation it cannot be taken for granted that these data are also applicable for impulse operation. The lower limit frequency can be assumed to be around 20 kHz, i.e. impulses with times to half value > 50 us can hardly be superposed unchanged on the powered circuit. The superposition of current impulses by means of a capacitive coupling element also is possible up to some limits only. In the case of capacitive coupling the coupling impedance is part of the test generator's discharge circuit and such has the effect of an additional damping circuit. Frequency range: Fig. 8 The photograph shows a test set-up where the data transmission between a table top computer (lower right) and a printer (upper left) are tested. The data transfer is via series interface RS 232. The coupling of the interference pulses on the data line is made with a capacitive coupling clamp. Ph The higher the discharge energy, i.e. the higher the impulse current, the higher the coupling capacitance must be chosen. Here the justifiable limits are reached relatively soon. Another alternative is to connect the test generator without a coupling element directly to the powered circuit as the generator does not have a parallel connection to the output terminals. The generator's spark gap such acts as coupling and decoupling element. Example of a test set-up MEAll N computer . 1 RS 232 l- Fig. 10 Schematic test set-up Pfinter 3.1. Description of interference parameters Fig. 11 time domain Fig. 14 coupling behaviour of capacitive coupling clamp 3.2.2. Coupling filtering unit Fig. 12 frequency range Fig. 14 Amplitude Density spectre 3.2. Description of coupling and filtering elements 3.2.1. Capacitive coupling clame The interference pulses are coupled capacitively on a power line (max. 4 phases). The filter protects the remainder of the circuit. It is such safeguarded that only the powerline feeding the equipment under test is surged and peripheral equipment which need not be tested, but belongs to the test setup, are not interfered with Over a length of 1 m the interference transients are coupled into the tested data lines. Fig. 13 coupling clamp Z = 50 n Fig. 15 Block Diagram - 544 References: Fig. 16 /l/ Interference generated by switching operations and its simulation; Rodewald, Kunkel, Lutz IEEE/EMC Symp. Tokyo, 1984 /2/ Priifung elektronischer Systeme und GerBte auf EMV; K.Feser/ M. Lutz Industrie-Elektrik + Elektronik, 25 Jahrg. 1980 /3/ The Origins, the Effects and the Simulation of Transients, as well as their international standardization / 0. Frey Electra 82/ Boston /4/ IEC 65/WG 4 draft proposal /5/ IEEE 587.1 /6/ Guide on surge testing in low voltage AC power circuits / F. Martzloff, P. Richman /7/ Emil Haefely & Co. Ltd., Base1 brochure E 111.20 Filter behaviour Filter attenuation Coupling - A NEW PULSE WIDTH MODULATION 545 CONTROL 99 - FOR LINE COMMUTATED 03 CONVERTERS MINIMIZING THE MAINS HARMONICS CONTENT _________________________-_-_--_____________________--__- ___________-_________ _____________________-____ FRANZ C. ZACH Institut fiir Allg. Elektrotechnik und Elektronik - Power Electronics, University of Technology, Vienna, Austria phase conAbstract - Application of conventional trolled power electronic circuits causes reduced power factor and increased harmonic content in the electric mains. Therefore, a new method has been investigated here in order to eliminate to a large extent these effects mentioned. The optimization objective has been to minimize the rms harmonics current content in the mains while achieving a power factor of one. It should be pointed out that the problems treated here are in the lower frequency range as most important for EMC problems concerned with influencing the electric mains. The method used has been found to have a duality with PWM (pulse width modulated) inverters: the voltage patterns for PWM inverters are governed by the same switching patterns and control laws as the current patterns for the (improved) phase controlled circuits. The improvement requires switching devices having a turn-off capability. While this formerly did require thyristors commutation circuits, with force today this feature is easily implemented by using power transistors or GTOs (gate turn-off devices). The control laws for minimizing the rms harmonics current contents in the mains, the circuits and the results are shown in the paper. Introduction The application of power electronic circuits connected to the mains (such as line commutated rectifiers with phase control) is connected with introducing current harmonics and reduced power factor in the mains. These effects lead to strict regulations, especially in European countries. Corresponding codes are, e.g., EN 50.006, as one major guideline to be observed. Furthermore, various electric utility companies have introduced their own standards, limiting the use of line commutated circuits without special permits usually to very low levels (typically to a few kW, in some cases even below 1 kW), These reasons make it almost mandatory to look for methods to improve the power factor and the harmonic content associated with power electronic circuits connected to the electric mains. A very promising method is the so-called pulse-time-control [l, 31. This method basically adds force commutation to line commutation, as will be shown later in detail. Since the advent of power transistors this so-called force commutation does not require any additional devices as was the case formerly when thyristors had to be used. We also just can say that the switching devices are assumed to be of the gate turn-off type (i.e. transistors, GTOs etc). Whereas the exact elimination of lower order harmonics as shown in [ 1, 31 leads to an increase of the remaining harmonics, the method used here is based on the minimization of the overall current harmonic contents. (Here - as in [1, 3] - the power factor achieved is unity.) The method of [I, 31 is limited to a restricted load voltage region (where the restriction is the more severe the more harmonics are to be eliminated). Other methods have been proposed in the literature in the last few years. They are based, e.g., on the subharmonic oscillation method [8]. This method does not allow a rigorous optimization because the control is generated by intersections of a sine wave and a triangular wave without the basic possibility of freely adjusting the switching instants. A more sophisticated method, basically related to the method given in [4, 71 and also applied here, has been proposed in [9, IO]. The merit of these papers is the discussion of filter sizes (both on the supply and on the load side). However, the control angles are fixed values, basically not allowing to control the load voltage mean value U, and the line current fundamental $1 in order to’ control the energy transfer just by c angmg the control angles. With the method of [9, ib] application of tw”o rectifiers operating out of phase would be required which is recommended for higher power applications only. The new method proposed here does not have such limitations because the control angles can be adjusted according to the Ud and iR1 required. Phase Control and Concepts of Power Factor Improvement and Mains Harmonic Content Reduction Figure 1 shows one of the best known linecommutated circuits, the three phase fully controlled bridge 131. Also shown is a free wheeling diode DF. This diode reduces the phase shift of the line current with respect to the line voltage. Figure 2 shows the resulting line current for one phase and the load voltage. It can be seen ,h,s;es~;;n~tween . from Fig. 2 that a certain and the fundamental of iR ‘R. Another method for reducing the phase shift is given by splitting up the bridge in Fig. 1 into two bridges connected in series. (Then each bridge will - Fig. 546 Three phase fully controlled bridge using diode. thyristors. DF .,... free wheeling and resistance. I-, R . . . . . load inductance 1. uI’ - Fig. Transistorized 3. trolled rectifier time control. three circuit phase bridge consuitable for pulse a wt b wt * wt Fig. 2. Electric variables for the circuit (a) line to neutral voltages (b) load voltage (c) line current Jo-t- -c of Fig. 1. deliver half of the maximum ud as compared to the one bridge of Fig. 1.) Furthermore, if for one bridge a Y connected transformer is used and a A for the other, a reduction of line current harmonics will result. Both methods have been described many times in literature wherefore more details shall be omitted here. The whole area of influencing the mains by power electronic circuits including the means for reduction of such influences is covered in great detail in [II]. A new method adding turn-off capability (by the switching devices) to the conventional line commutated circuits has been introduced in [5] and extended in [I]. This method has gained much mot-e importance since power transistors and GTOs for relatively high currents have become available making additional devices for force commutation for thyristor applications not necessary any more. The basic circuit is shown in Fig. 3 where the thyristors of Fig. 1 are replaced by transistors; Fig. 4 shows the electric variables obtainable with this circui’t. As can be seen from Fig. 4, three control angles are introduced. The pulse patterns “,!i fzr$edOls&h that symmetry about the peaks of voltage u is achieved. This leads to zero phase shift betwe $ n the current fundamental and the voltage in the mains. Fig. 4. Electric variables for the circuit of Fig. 3. Once ud (the mean of ud) is chosen, the three angles “i have to be determined such that this ud is obtained. It is seen that two degrees of freedom remain for optimization of the line current harmonics content. This optimization in a basic manner has been performed in [l] such that t e lowest order harmonics (i.e., the 5th and 7’?l ) of the line current are eliminated completely. This method, however, shows a limited range of controllability: for complete elimination of lower order line current harmonics the load voltage has to remain below a certain level; this level is the lower the more current harmonics have to be eliminated [6]. A breakthrough in this respect has come from optimization of PWM inverters. There, the goal is to minimize the output voltage harmonics contents in order to minimize the load losses or the speed and torque ripples of an AC motor [4]. (It is interesting to note that the first approach in this area has been also to eliminate lower order harmonics [2], also yielding only limited control range.) The most important feature in this approach is the fact that the control law optimization has to be performed for currents for purely ohmic load for the PWM inverter in order to be applicable to the cqntrolled rectifier. Although here an output voltage optimization is equivalent, this has to be mentioned because usually the inverter control optimization has to take the particular load into - making current Control Laws for PWM Inverters account, considerations 547 - inevitable. Figure 5 shows a transistorized (PWM) inverter circuit with constant DC VOltage input (supply). The usual application of the configuration of Fig. 5 is in variable speed AC motor drives. There, the output voltage shall approximate a Sine wave as close as possible. The frequency and rms Vahe Of this Voltage shall be adjustable. This leads to output voltage patterns as shown in Fig. 6. In order to minimize the the general harmonic expression a1 = F a 5- -. . contents Fig. 5. Transistorized (I) (1 - 2cos “1 + 2cos “* - 2cos a,), 4ud (1 - 2cos 39 a3 =v 4ud harmonic is used: Application to Controlled circuit. Rectifiers (2) + 2cos 302 - 2cos SC+ (3) (1 - 2cos 5a1 + 2cos 502 - 2cos 5+ 5n 4ud an = x (1 - 2cos no1 + 2~0s no2 - 2~0s no13), with all even order harmonics inverter (4) Based upon the previous section one now can say that a duality exists between optimization of the line current harmonics of the controlled rectifier (Figs. 3, 4) and the output voltages of the PWM inverter with DC voltage input (Figs. 5, 6). Equations (I) - (4) are equally valid for both cases, only “d has to be replaced by Id for the controlled rectifier. The inverter output fundamental peak Value alo is given by Eq. (1): to be zero. The optimization approach now to be followed for the PWM inverters would be to calculate the inverter output current based upon the voltage expressions given above and based upon the particular load. As mentioned, the load to be used here would be purely ohmic because then the line current patterns of Fig. 4 have the same characteristic shape as the patterns of Fig. 6. a 1 + 2cos a2 - 2cos a,). (5) The line to line voltage can be easily determined: due to the phase shift of 120’ between ulo and u20, we receive Q12 ‘\/“;“10 . (6) For the following considerations the duality of the two circuits considered is of advantage: the line current peak PR is easily determined by starting with pulse patterns as shown for u,~. If we assume that a. pulse pattern equal to that for u12 in Fig. 6 is introduced for transistors Tl, T4 in Fig. 3 we receive Fig. 7a. This pulse pattern IS realized by appropriate turn-on and turn-off of the transistors. Then we assume the same pulse patterns for the other transistor pairs T3, T6 and T5, T2. load is tal P state 5’ output The energy transfer from the mains to the characterized by the line current fundamen. (The phase shift in the mains is zero as before.) We have (similarly to the inverter voltage) PRl =flal (7) because the pulse pattern of iR of the controlled rectifier is the same as the line to line voltage of the PWM inverter. Here, a, Fig. 6. of (a) (b) Output voltage Fig. 5. for C( - 0 for a11 f 0 patterns for the circuit 4Icl(1 =y - 2cos al + 2~0s a2 - 2~0s a,). (8) Since i is based on two pulse patterns having a 120’ Pease shift with respect to each other, we receive as for line to line voltages in three phase systems: q iR3n and 0 for all n=l, 2, 3, . . . . . (9) - ’ for all k odd, but k 1 3n Rk = fiak (with n = 1,2,3, (I 0) . .. . .). One most interesting and fundamental feature can be observed: the currents in the upper transistors T T3, T aL;;zrs behave such that they yielding the required complemLAt eat 2 smooth and constant load current Id (Figs, 7a - c). This is equally true for the lower transistors T2, T4, T6. In some cases (e.g., as in Fig. 8) also intervals are observed where the currents do not flow in the mains line; they have to flow through the free wheeling diode in order to maintain constant load current as required in controlled rectifier circuits with sufficient load inductance [ 51. The most important fact, however, is that the control law and the resulting transistor current pulse pattern do not require the same Id, to flow in two different or even in all three transistors (either in the upper or in the lower half bridge) at the same time. This is the limiting fact of the control laws gained from eliminating only certain harmonics: there for various regions of iR (especially for values more or less close to iR ax) Id would be required, e.g., to flow through q and Tj simultaneously. This would mean a sudden lump of the load current from I to 21d and back to Id which of course is not feadsible. The calculation the switching leading angles a to the optimization of has to minimize 1,2,3 PH = C al 548 and the minimum PI, determined. Then the value of al is changed in steps and the optimization iS performed for all these steps of a . This finally 9 - II. leads to the solutions shown in 6 1gs. These solutions all constitute pronounced optima. In order to illustrate this point, a three dimensional representation is shown in Figs. 12 and I3 [7]. One can say that altogether five different solutions can be clearly differentiated in Figs. 12 and 13, not to mention other local optima which are only valid for a very limited range of aI. Figures 9 to II show only the three most important region solutions, i.e., they are valid for the entire 0 G al G 41d/ti (I 5) and they lead to better values for PI, than the other local optima. Figure 12 corresponds to Figs. 9 and 10, Fig. 13 corresponds to Fig. II. Of the three solutions shown in Figs. 9 - 11, tht c 66 ;I I, (11) k=5,7,11,13,17,19 ,..... i with ak taken from Eqs. (1) - (4), where ud is replaced by Id. The sum does not contain harmonics with the order 3 and multiples thereof. This is due to the generating principle: as explained before, the current pulse patterns in the transistor pairs are chosen such that they are the same as for the line to line voltages of PWM inverters. These in turn are generated by two voltages (u and ~20) having a phase shift of 120’ with re@ect to each other. Because u12 = uIo all harmonics k=3.n; n=1,2,3, . . . . . cancel. u20, 4 ‘7 Fig. 7. Current waveforms for the rectifier with the new control scheme. iR, is, iT, . . . . . waveforms ang,es as for tro, are u?,z,;; Fig. 6a (the same con- Fig. 8. As Fig. 7, but than used in Fig. 6. It is obvious that the optimization can be done only on the digital computer. It does not seem to be necessary here to discuss the different optimization routines existing for this purpose. It has to be pointed out, however, that several local optima exist. This makes methods as the steepest descent method very unreliable because it usually only leads to local optima. This would require to operate with a whole range of starting points and then selecting the overall optimum. It has turned out that it is less time consuming altogether to apply an entire search over the whole region of realizable ai. For realizability, 0 < aI< “2< “3< -rr/2 (12) has to be observed. For optimization purposes, it geous at first to select al which Eq. (1) ) to is advantaleads (with I - aI coso 3- --- ?‘ cos “1 + cos “2. (13) Then only al, a2 have to varied on the computer; for each set ( cx I, o 2) Eq. (5) has to be calculated for other control angles - 549 solutions given in Figs. 9 and 10 show about the region 0 6 same results for PH for the entire al rel Q I. Either one of these solutions can be recommended. The solution shown in Fig. 11 leads to spectra somewhat higher PH. The line current corresponding to the solution of Fig. 9 for a1 re, 2 0.5 are shown in Fig. 14 in comparison to the conventional phase control. __ 99 the first alternative is to let ulo have a3 a positive value where the fundamental of LII~ goes through zero (as assumed in Fig. 6a) or vice versa (similar to Fig. 6b) for the second alternative. The details are rather involved; therefore, the reader is referred to [4, 71. While this paper has instantaneous treated pulse trol) angles, patterns certainly with three a also switching (conhigher number The difference between Figs. 12 and 13 lies in the “modulation sign”. This means that the same fundamental can be generated by two pulse patterns being dual with respect to each other: 1 a “2 Fig. Basic principle of a three dimensional 12. representation of PH versus a , a 2 fqr al = 0.5 ( c1 id given by Eq. (I) ). ihe regions realiza 2 le solutions ~1 are discussed l’“2 [4, 71. of in r 0 Fig. 30 9. Control anglesai 80 90 “i for line current harmonic content minimization. al rel = al / (414 /71) (14) al re II Pk Fig. “1 a: 13. As Fig. 12, but showing tions [4, 71. additional solu- a “2 & ak rt :I w 0 Fig. 10. 30 As Fig. 80 60 a second 9, but showing Ii _____ I I0 - conventional control new optimal phase control 9O cli solution. at 1 30 Fig. 11. As Figs. solution. 9 and 60 10, but 90 cli showing a third 5 7 11 a k Fig. 14. Comparison of mains line current spectra for the same a, r , (being l/2 of its maximum). One has Eb note that control of power in conventional phase control is performed by phase shift of the current blocks in the mains, therefore leaving the absolute values of the line current harmonics unchanged. - 550 is applicable. Because there is no basic difference and because there would not be any three dimensional representation possible, such cases are omitted here. How the procedure would be can be seen from PWM inverter drive optimizations as performed in [4]. The practical realization can be easily performed as shown in [l], where also the line filter dimensioning is discussed. The line filter Will be much less involved with this new type of COntrOL - [31 ZACH, tronics). II41 ZACH, F.C. and H. ERTL: Efficiency Optimal Control for AC Drives with PWM Inverters. IEEE Industry Applications Society 1983 Annual Meeting. Mexico City, Oct. 3 - 7, 1983. (Also to appear in Trans. IAS.) [51 ZACH, F.C.: Optimization of the Harmonics Contents and of the Power Factor of Power Electronic Circuits by Pulse Time Control. ETZ 94 (9), pp. 535 - 538,1973. [61 of Influences on ZACH, F.C.: Minimization the Mains for Line Commutated Converters the F#se-TimeTControl. Proceedings of by International Power Electronics and Motorcon Conference. Geneva, Switzerland, September 13 - 15, 1983. [7Li ZACH, F.C. and F. THIEL: Pulse Width Modulated (PWM) Inverters for Efficiency Optimal Control of AC Drives - Switching Angles and Ef iciency/Loss Profiles. Proceedings of the 3t-6 IFAC Symposium on Control in Power Electronics and Electrical Drives. Lausanne, Switzerland, September 12 14, 1983. [81 KATAOKA, T., K. MIZU.MACHI and S. MIYAIRI: A Pulsewidth Controlled AC-to-DC Converter to Improve Power Factor and Waveform of AC Line Current. IEEE Transactions on Industry Applications, vol. IA-15, no. 6, pp. 670 - 675, November/December 1979. [91 ZIOGAS, P.D., Y.-G. KANG and V.R. STEFANOVIC: PWM Control Techniques for Rectifier Filter Minimization. Conference Record 1984 Power Electronics Specialists Conference. Gaithersburg / Washington, June 18 21, 1984. Conclusions It has been shown that the control of the transistors of a three-phase controlled rectifier bridge is best performed as for PWM inverters. This means that the pulse patterns of the currents in the controlled rectifier are the same as the pulse patterns of PWM inverter output voltages. This further means that the basic principles and relationships of PWM inverter output voltage control also can be applied here. One essential feature is that the overall harmonic content of the line current is minimized. This leads to realizable solutions for the entire region of line current fundamentals between 0 and its maximum. (This is opposed to elimination of only a few selected lower order harmonics (while increasing the remaining harmonics) which would lead to unrealizable solutions in many cases.) The resulting control law is especially easily realizable when the new high current power transistors or GTOs are used in controlled rectifier bridges. The method should gain importance in all the countries where the application of conventional controlled rectifiers so far has been highly restricted by regulations. The new method avoids power factor reduction in the mains and reduces the harmonics content. The reduction can be further improved if more than three control angles are applied. With this method it is certainly possible to meet the standards and regulations determining the upper limits of line current harmonics generated by controlled rectifiers connected to the mains. References [l] [2] ZACH, F.C. and R. DEMATTIO: Pulse-Time Modulated Converters for Minimum Harmonics Contents and Ide 1 Power Factor in Supplying Networks. 5t% EMC Symposium. Zilrich 1983. TURNBULL, F.G.: Selected Harmonic Reduction in Static DC-AC Inverters. IEEE Trans. Comm. El., vol. 83 (73), pp. 374 - 378, 1964. (Power ElecF.: Leistungselektronik Vienna / New York: Springer. 1979. LlOl ZIOGAS, P.D., Y.-G. KANG and V.R. STEFANOVIC: Optimum System Design of a ThreePhase PWM Rectifier-Inverter Type Frequency Changer. Conference Record 1984 Annual Meeting IEEE Industry Applications Society. Chicago, Sept. 30 - Oct. 4, 1984. 011 BUECHNER, P.: Stromrichter-Netzrilckwirkungen und ihre Beherrschung. Leipzig: VEB Deutscher Verlag fiir Grundstoffindustrie. 1982. - 551 100 - Q4 COMPUTER-AIDED ANALYSIS OF THE RF1 VOLTAGE GENERATION BY SMALL COMMUTATORMOTORS J. Sack, H. Schmeer Hochschule der Bundeswehr Miinchen Elektrotechnik Fachbereich Elektronik Institut 4 Federal Republic of Germany Neubiberg, 100 ABSTRACT dBoJ/) The conducted emission of radio frequency interference (RFI) voltage by small commutator motors is investigated, with emphasis on permanent-magnet DC The current commutation in the motors. armature coils is identified as the main source of RF1 generation. Therefore, commutation theory is briefly reviewed with regard to the features of small It is shown that two types of motors. commutation-dependent voltage peaks can An RF-domain equivabe distinguished. lent circuit of the motor is presented which models the influence of the distributed inductance and capacitance of the motor winding on the emitted RF1 spectrum. By combining the results of commutation analysis and the RF-domain model of the motor, a computer program for simulating the conducted RF1 voltage emission of permanent-magnet DC motors is obtained. Some computed RF1 spectra in the frequency range 0.1-100 MHz are compared with corresponding measured results. Input data requirements and limits of applicability of the computed simulation are discussed. 1. 1.1 Generation INTRODUCTION mechanisms, problems commutator motors emit a broadband RF1 spectrum (Fig.11, due to voltage peaks which are produced by two mechanisms: the current commutation in those armature coils that are short-circuited by the carbon brushes /l/, /2/; and abrupt variations of the contact resistante, in extreme cases short-time contact breaks due to the relative motions between the brushes and the collector /2/. Both types of voltage peaks can be distinguished by means of their spectral properties, Those caused by commutation occur in a nearly regular periodic pattern, which means that the associated spectrum is of a coherent type (surroundings of spectral zeros excluded). Variations of the contact resistance and 15 OF: OS Fig.1: motor firing 1 IO MHz 100 Conducted RF1 voltage of a DC (1) as compared to a dimmer at angles 100 (2) and 900 (3) contact breaks through the relative motions between the brushes and the collector, however, occur in a quasi-random manner. They produce a spectrum of 2 stochastic type. Two peak voltage (U) measurements at a given frequency using an EM1 receiver with variable bandwidth (B) will yield: 51/62 = 0l/B2 (coherent spectrum) &/ii2 = 4%; (stochastic spectrum) (1) (2) For the permanent-magnet DC motors, which we examined, we found that Equ.1 is approximately fulfilled within a wide range of operating conditions (Fig.2). We concluded that commutation is the most important cause of RF1 voltage generation, at least within the shaded area of Fig.2. Some results of time and/or frequency-domain measurements published by other authors indicate that this is also true for DC motors from different Fiq.2: age is caused mutation shaded (n:speed, rent; RF1 voltmainly by comin the area I:curDC motor) - 552 manufacturers /l/, /3/ as well as for universal motors /4/. The limit between the two areas in the I/n-diagram of Fig.2 is marked by an empirical equation of the type Fig.4: System of b commutating coils 1.n = K. K depends on the construction the motor, some of which will cussed in SS2 and 3. data Of be dis- We will treat the following problems which have not been solved so far: - In which way is the emitted RF1 spectrum influenced by the distributed inductance and capacitance of the winding? - How can the RF1 voltage be calculated, purposes? e.g. for prediction 1.2 RF-domain system model commutation current self-inductance of the coil LS of the coil RS : ohmic resistance contact resistances rlrr2: trailing edge of the brush) (‘1: contact currents il,i2: i2=is2-is1 i =I/2tisl, (i: armature current) transformer voltage due to mutual ut1 : induction from coils No.Z...b of rotation due el : induced voltage is1 i to Inter- Motor connection LISN RF-domain calculations of the conducted differential mode interference voltage presented in S4 use a two-port network model shown in Fig.3, where the motor is represented by an RF-voltage source !o and a complex, frequency-dependent internal impedance 21. The principles of calculating U, an7 Z are explained in SS2 and 3. FomparaF‘re to RF1 voltage measurements, the RF-impedance of the power supply is simulated by a line impedance stabilization network (LISN) which has an impedance Z2. The influence of the interconnectiEn motor/ power supply is included in the calculations, but will not be discussed here. The interference voltage c2 across z2, referred to uo, is 1 u2 -L: L!O &11+!!12/22+!!21* Zl+A22’ WZ2 (5) LSdiSl/dttutl+RSiSl+rlil-r2i2+el=O (4) ’ of where Al . ..A22 are the coefficients the campi ex Chain matrix of the interconnection motor/power supply. 2. COMMUTATION ANALYSIS Detailed analyses of the current commutation in small motors are presented5 in /4/, /5/, and /6/. The purpose of S2.1 is to give a risumC of the results for DC motors. Supplements necessary for universal motors are discussed in S2.2. (parasitic) interpole field The contact resistances rl and r2 depend on time because of the varying contact areas Al(t) and AZ(t). Due to contact fritting /i’/, they are also influenced by the current densities gl and g2 (Def.: g=i/A) . Since gl and g2 vary in the course of a commutation period, Equ.5 is a nonlinear differential equation without a closed-form solution. It can be solved by numerical integration. The starting value of i is I/2 at t=O if b-l. For b>l a set o s1 starting values iSb must be calculated by an of iSl... iterative method /9/. The commutation in coil No.1 ends at t=tR when isl=-I/2. Some important influences on commutation are shown in Fig.5, with special regard to small motors: 1.1 The reactance voltage across Ls has a retarding effect on commutation If LS increases, the differ(Fig.Sb). ence at t4tH between the actual value of iS1 and its desired final value of -I/2 increases, too. This is equivalent to a larger contact current il and contact voltage rlil. 2.1 For the absence of interpoles, el can be varied only by shifting the brushes out of the neutral zone. Growing el in Fig.Sa is achieved by a backward brush shift, compared to the rotation direction of the motor. The brushes in small motors are usually shifted backward, which partially compensates the reactance voltage across LG. 3.) At t4tK the current density gl under the trailing edge of the brush takes high values, which in addition change rapidly with time /8/. The effect 0) b) C) 2.1 DC motors A system of b commutating coils under one brush is represented by the equivalent circuit of Fig.4, leading to a system of b coupled differential equat ions. For coil No.1 we obtain: Fig.5: Influences on commutation d) - of contact fritting and thus, the effect becomes then of gl on rl (see above) This is called the dynamic negligible. contact behaviour. rl can then be described as a mere function Of time: rd: (6) (t+tE) r1 = rdA,/Al(t) dynamic Contact reSiStarm? full contact area A~: At t+tE the reactance voltage across ~~ and the contact voltage drop across in the commutating coil. rl are dominant using Equ.6: Equ.5 can then be simplified LSdiSl/dt+rdilAo/Al(t) k 0 (tdtE) (5a) From Equ.5a one can where il=I/Z+iSl. deduce the influence of rd together with which is shown in Fig.5c at LS at t4tEr undercommutation ( isl> -I/2 1 and in Fig. 5d at overcommutat ion ( islC-I/2 1 by introducing a fictive “time constant” LS/rd. (Overcommutation can be achieved by making el sufficiently large, see Fig.5a.) Thus, increasing rd at a given LS reduces both the contact current 11 and the rate of change disl/dt at t-+tE. 2.2 Universal motors Additional parameters appear in universal motors: 1.) The armature current is no longer constant, but varies with the frequency i=$!*I*sin( 2rf. t) . As of the AC supply: a consequence, the commutation analysis mUSt be Set up at various times within a half-wave of the sine function. 2.) The inductor pole flux varies Its first harmonic has also with time. the same frequency and phase angle as i. The flux variation causes an additional transformer voltage in the armature coil which influences commutation. 3. 3.1 RF1 VOLTAGE CAUSED BY COMMUTATION Generation of voltage peaks As a consequence of commutation, two types of voltage peaks appear in the armature. To demonstrate this fact, we use the equivalent circuit of a symmetrical simplex armature winding of a DC motor including the brush contacts shown Each of the brushes short-cirin Fig.6. cuits two armature coils. At t4tE this corresponds to a brush-to-segment width _ Fig.6: Symmetrical simplex armature winding and brush contacts (low frequency) 100 553 - Q4 The armature branches ratio of ll /9/. For ~$1, the single commutating coil under each brush is magnetically decoupled from the branches. A third contribution to RF1 voltage appears in motors which have an inductor winding, like universal motors. This is due to the flux coupling between the commutating coils and the inductor windining. Again a rapid diS /dt at t+tE duces a voltage peak. 4 he principle is the same as explained above for Ui, except that the value of 13 is irrelevant. - 554 - 3.2 RF-domain model of the motor So far we have described the motor by equivalent circuits valid at low frequencies. For RF-domain calculations we have to take into account the parasitic capacitances which are distributed along the basically inductive motor winding. These may appear between adjacent pieces of wire, as well as between the wires and the motor casing. Their effect can be seen from the measured impedance of a DC motor armature winding plotted versus the frequency in Fig.7. The resonances at f>2 MHz are due to the interaction between the inductance and capacitance of the winding. They are damped by ohmic and magnetic losses. traduced as complex quantities Ur and Ui at suitable locations in the eqriivalenf circuit of Fig.8. Finally the complete network can, by complex algebra, be reduced to the representation as a voltage source I_Jo plus internal impedance 21 according to Fig.3 (left hand side).Thus, by combining the results of a time-domain commutation analysis with the RF-domain models of Figs.3 and 8 we can calculate the RF1 voltage generated by a motor. 4. COMPUTED AND MEASURED RF1 VOLTAGE FROM PERMANENT-MAGNET DC MOTORS 4.1 Results Fig.7: Impedance Of a DC motor armature winding Simulating these effects by introducing suitable lumped elements in the equivalent circuits of Figs.4 and 6 would imply the necessity of solving a high order system of differential equations in the time-domain. Instead, we decided to choose a method involving less computing time and offering more flexibility: First, the commutation-dependent voltage peaks u and u. are evaluated according to §§s and 3?1. The number of first order differential equations to be solved simultaneously is equal to the number of coils commutating in parallel under each brush, for example two at lea42 (Fig.6). Second, the motor winding is modelled as a system of lumped elements simulating the distributed capacitances and inductances plus the ohmic and magnetic losses, An example for a simplex armature winding is given in Fig.8. The computed voltages u and Ui are now transferred into the rrequency domain and in- The following examples of conducted RF1 voltage spectra refer to a permanent-magnet vehicle accessory drive motor. All input data for the computations were taken from a real motor. The discontinuity in the spectra at 30 MHz is due to the measurement method (german standard VDE 0879, part 3; quasi-peak voltage at receiver bandwidth 9 kHz below and 120 kHz above 30 MHz), which is also simulated by the computer program /9/. Computed and measured results are therefore directly comparable. First an operating point of 24 A at 3200 RPM is considered. The brush-tosegment width ratio is chosen S>l in order to make both ur and Ui appear according to S3.1. From the computed contributions of Er and vi to the RF1 voltage U2 across Z2 it can be seen (Fig.91 that_Ui is mosy effective below 5 MHz. At hi?jherfrequencies, v, delivers the predominant contribution to U2. This can be explained with the help of Fig.8: At high frequencies, rJi is short-circuited by the distributed capacitances of the winding. Contrary to this, yr becomes most effective if the internal impedance Zl is small, i.e. at f=lO...lOO MHz TFig.7). By the way, a comparison of Figs.7 and 9 shows that the minima of the spectrum coincide with maxima of the impedance. It can be concluded that the characteristic curved shape of the 25 Fig.8: RF-domain equivalent circuit of a symmetrical simplex armature winding 0.1 Fig.9: 1 IO MHz 100 Computed contributions of gr (1) and ui (2) to g2 (31 - 555 - 100 1”11”“‘12 spectrum, which is similar for any other motor, is just an image of the impedance curve. By choosing %l one can eliminate gi, as indicated in 53.1. A comparison between s=O.9 and B=1.3 is shown in Fig.10 (computed) and Fig.11 (measured), at the same operating point as before. The RFI Voltage at 0=0.9 is smaller in the frequency range fhc3MHz, because Ui is eliminated. At fL30 MHz a slight a& vantage of H=1.3 appears. This is explained by the fact that the portion of the total RF1 energy contained in ci 1S dissipated within the motor winding at high frequencies. The brush shift (see 52.1) can be optimized for minimal RF1 voltage generation. However, the brush shift also influences the operating characteristic of the motor. A reasonable condition for an RF1 optimization is constant voltage combined with constant mechanical power. At 12 V and 155 W, the influence of the brush shift on the RF1 voltage at 0.5 MHz and 50 MHz is shown in Fig.12 (computed) and Fig.13 (measured). Negative abscissa values correspond to a backward shift. Critical parameters with regard to RF1 voltage generation are mainly the self-inductance Ls (Fig.14) and the dynamic contact resistance rd (Fig.l5), each within a preferred frequency range, Q4 v / 155 w sot-““““““~ -400 -100 -20° -300 Fig.12: Computed influence on U2 of the brush shift 9oc SO! -400 -300 -100 -200 Fig.13: Measured influence on U2 of the brush shift dB(pV) 15 24 A / 3200 RPM 0.1 1 IO MHZ 100 Fig.10: Computed RF1 voltage 22 at H=1.3 (1) and a=O.9 (2) 0,1 1 10 MHz 100 Fig.11: Measured RFI voltage LJ2 at 0=1.3 (1) and B=0.9 (2) IO 15 20 /JH - 25 Fig.14: Computed influence on g2 of the self-inductance L.g I&#,,1 20 1 ,,,,I,,,, 25 30 mR 35 Fig.15: Computed influence on g2 of the dynamic contact resistance rd - 556 - 4.2 Input data requirements, applicability of the simulation A full set of input data, as was used in the computations for Figs.9, 10, 12, 14 and 15, consists of the parameters needed for the Commutation analysis: - geometric dimensions of the armature, collector, and brushes, - inductor field and armature reaction field data, - current-voltage characteristic of the brush contact, - ohmic resistance, self- and mutual inductances of the armature coils, all of which are determined either according to the usual construction rules or by field computations. Further the - RF-parameters of the winding according to the equivalent circuit of Fig.8 are required, which can hardly be calculated from the construction data of the motor. They must be found by estimation, or empirically by choosing a set of initial values and adjusting them for good agreement between measured and calculated impedance of the winding. However, the simulation can be used for parametric studies even if not all input parameters are accurately known. This would mean that the tendency of a variation of the RF1 voltage through variation of an input parameter is correctly predicted, although the computed absolute values differ from the real ones. An example was given in Figs.12 and 13, where the optimum brush shift of about -25' is correctly predicted by simulation, while the computed absolute RF1 voltages are partly incorrect. The simulation is based on the assumption that commutation is the clearly dominant cause of RF1 voltage. Therefore it cannot be used to predict the effect of a bad quality of resulting in contact Moreover, it should the sliding contacts breaks (see $1.1). not be used at oper- DC motor shows that - in addition to a contact voltage peak ur under the trailing edge of the carbon brush, possibly with commutator sparking, there is also an induced voltage peak Ui in the branches due to the flux coupling between these and the commutating coils. Both ur and Ui appear at the end of a commutation process; - at high frequencies (fLl0 MHz) only Ur is effective, while u. is short circuited by the distributea capacitances. One can eliminate Ui in a symmetrical winding using a brush-to-segment width ratio BSl, yielding a reduced RF1 voltage in the lower AM range (fi3 MHz). However, at FM frequencies there is an advantage of B>l, since a part of the RF1 energy is dissipated within the motor winding. ACKNOWLEDGEMENT This work was supported by the Robert Bosch REFERENCES /l/ S. /2/ Properties of High-Frequency Conducted Noise from AUtOmOtiVe Electrical Accessories IEEE Trans. Electromag. Compatib., Vol.EMC-25, No.1, Feb.1983, p.2-7 R.M. Labastille /3/ /4/ ating points in the unshaded area of Fig.2. /5/ 5. CONCLUSIONS The conducted RF1 voltage emission of small commutator motors can be evaluated by a computed simulation. The method presented consists of - a time domain evaluation of the commutation process, - an RF-domain modelling of the winding including the distributed capacitances, - and combining both in the frequency domain. A comparison of computed and measured RF1 spectra of a permanent-magnet DC motor demonstrates a satisfactory accuracy of the simulation and its ability for RFI voltage prediction. The computed RF1 voltage generation in the armature of a permanent-magnet GmbH, Stuttgart. /6/ /7/ Yamamoto, 0. Ozeki Die Funk-Entstorung von Gertiten mit und Kleinstmotoren KleinETG-Fachberichte l/1975, p.114-121 D.P. Motter Commutation of D-C Machines and its Effects on Radio Influence Voltage Generation AIEE Trans.,Vo1.48(1949),p.491-496 D. Roye, M. Poloujadoff Contribution to the Study of Commutation of Small Uncompensated Universal Motors IEEE Trans. Pow. App. & Sys., Vol. PAS-97,No.l,Jan./Feb.l978,p.242-248 A.W. Mohr Der funkenfreie Drehzahl- und Belastungsbereich bei Universalmotoren ETZ-A,Bd.81,H.23,Nov.l96O,p.812-816 G. Bolz Die Stromwendung der GleichstromKleinstmaschinen ETZ,Bd.60,H.32,Aug.l939,p.949-953 R. Holm Electric Contacts, Fourth Edition ;pr;;;;f;Verlag, Reprint 1979 /8/ Die Strom-Spannungs-Charakteristik des kommutierenden Kohlebiirstengleitkontakts in der Endphase des Kommutierungsvorgangs Elektrie 28 (19741, H.4, p.206-209 /9/ J. Sack Storspannungsemission kleiner Gleichstrom-Kommutatormotoren im Bereich PhD thesis, der to HBrfunkfrequenzen be published - 557 101Q5 - Control and Reduction of Spurious Emissions from Small DC to DC Power Converters J.M. Firth Herzberg Institute of Astrophysics National Research Council of Canada Ottawa, Canada KlA OR6 INTRODUCTION The spacecraft for the International Solar Polar Mission (ISPM; now renamed UlYsses) will carry nine groups of experiments designed for studies of the sun. Each experiment is separately powered from the spacecraft 28~ bus by its own DC to DC converter, which serves the unique power requirements of the experiment and isolates the experiment from the spacecraft power systems. This paper outlines the problems encountered in meeting the stringent EMC specifications and describes a neutralising technique used to meet them. This technique is completely general and may be applied to any converter. d. conveniently measured by lifting the unit from chassis and measuring the current in a connecting strap. All converters synchronised to a master clock. d Fig. 1. Primary CM, structure current limits. Fig. 2. Grounding protocols. EMC on the ULYSSES Spacecraft: The need to control EMI Two experiments have a particular need for good EM1 control on this spacecraft. The radio science experiment measures antenna signals of less than a microvolt from a few Hz to over 1 MHz. The magnetometer measures fields of less than 100 pTesla from 0 Hz to a few tens of Hz. These experiments need the following conditions: low currents on all cable shields, ba: low currents in the spacecraft structure, c. defined frequencies for the possible interfering signals, d. low magnetic induced fields from a few kHz down to DC. Differential mode emissions, carried on by a pair of wires to and from the load or source, are easily handled by well-known filtering techniques. Spacecraft EMC Specifications: Controlling EM1 a. b. C. Isolation of both primary and secondary sides from structure, plus primary to secondary isolation of better than 1 Mohm and 1 nF. Differential conducted currents and voltages (time domain) on primary lines not to exceed 20 mA and 280 mV pk to pk in a specified bus impedance. Common mode (CM) and structure currents less than 500 uA from 1 kHz to 10 MHz. See Fig. 1. The structure current is Radiated emissions will occur from transformers and chokes; these are readily controlled at frequencies above a few tens of kHz by a case which is thick compared with the eddy current skin depth. Defined frequencies (phase-locked converters) for interfering signals prevent the possibility of intermodulation products in the power system and allow sensitive experiments to devise comb filters for troublesome interference. - 558 The most troublesome problems in this context are common mode emissions, the result of electrical unbalance in the switching clrcuits. Such currents may flow around fairly large loops, including the structure. These loops may radiate or couple into other Circuits via a common impedance. LOAD Designing for Low Common Mode and Structure Current Mechanical and electrical symmetry in design is a fundamental way to addressthis problem. This achieves a balance in the coupling to structure (case), thus cancelling the transients caused by the switching process. The inclusion of shields around the switching circuits returned to local ground will help to keep circulating currents out of the structure. Other electrical design considerations include use of toroidal transformers for low leakage inductance, low radiated magnetic fields, and low phase shift between primary and secondary sides. Bifilar winding creates balance in magnetic coupling and interwinding capacitance. Active devices should be selected for matched characteristics and the sync signal supplied from balanced lines. Finally, the mechanical design should stress symmetry of layout to create a balance in stray capacitances to case. Mechanical stability of components and wiring is vital to achieve a lasting balance. Fig. 4. Structure current induced by differential voltage. The solution is the use of common mode transformers (Tp, Ts, 1, Fig. 6) outside all the differential filtering applied. This forces net balance of currents in the lines. The result closely parallels the way a coaxial cable can keep RF currents off the structure, clearly described in ref. 3. Common mode transformers are fairlv standard practice on the primary side in splacecraft applications. Used on both sides, together with capacitors of adequate size to case, Cs, Cp, they greatly reduce primary side common mode currents by keeping these local ised within the unit. Use of a high permeabil ity core toroid is recommended (ref. 1). The use of feedthrough filters (a low pass L-C section in a single package) is the frequent resort of a designer plagued by unwanted emissions. This may in fact make the structure current worse! Fig. 5. Fig. 3. Filters on primary and secondary lines. The source and load are inherently unbalanced with respect to structure; the noise voltage appearing across Cf will drive current to the load which may return via the system ground return and structure; this may be worse from a system point of view than the higher levels of differential ripple before filtering. In Fig. 4, note how the internal noise voltage, En, has created an external structure current, In, as the result of the Cf to case. Only a large, and maybe unattainable, value of Lf will control this. Common mode transformers and capacitors to structure. Note that the OC current balance is implicit in the common mode connection and there is no problem of saturation, even when using a very high permeability core. Applying this technique to the secondary side as well produces a dramatic reduction in the structure current (see Fig. 7). However, the use of CM transformers alone was insufficient to reach levels required by the ISPM specifications (see Fig. 7). Consider that on the primary side at least the switching of 28v is equivalent to 4Ov RMS which, at 60 kHz, when driving unbalanced capacity to case of only 10 pf, will exceed the 100 microamp spec. Practical limits on Lcm and Cp and Cs may make the filter insufficiently effective at the first few harmonics. (In the case solved, Cp=Cs=IO nF was needed to bring the L-C series resonance below the firt harmonic frequency. This exceeded the stated spacecraft isolation criteria, and a waiver had to be requested.) The key to the solution of the overall problem was the use of a neutralising capacitor from transformer to - IO-IQ5 559 chassis or ground return on the other side of the transformer (see Fig. 6, capacitors a, b or c). Here, the requirement for low leakage inductance should be noted. Minimum phase shifts between primary and secondary sides enables one trimmer capacitor to cancel over 20 db of current driven into the structure via various stray capacitances. Position a, b or c is arrived at during testing. Thus we achieve nearly ideal electrical balance (see Fig. 6). Fig. 8. Balancing of sync signal. CONCLUSION I Fig. 6. =j= =P cs : I Connection of neutral sing capacitors to case. DC to DC power converters operating at high frequencies need not necessarily be the source of troublesome emissions. Use of symmetry in both mechanical and electrical design, together with the use of primary and secondary side CM transformers, is a powerful tool in reducing CM emissions. Further reductions of more than 20 db can be made by neutralisation techniques applied to either or both primary and secondary sides. REFERENCES 1. 2. Mullenheim, T., "RF1 suppression chokes with soft magnetic cores", Proc. Symp. on EMC, E.T.H., Zurich, 1983. Ott, H.W., "Ground, a path for current flow", IEEE Int. Symp. on EMC, 1979. ACKNOWLEDGEMENTS Fig. 7. Measured structure currents: x: Ts, Tp, Cp and Cs installed o: a, b or c optimised 9: d optimised. Fig. 7 shows the before and after measurements on the structure current. Note the dramatic reduction of first harmonic, 40 db of improvement. At the higher harmonics, effects of phase shift make the simple capacitive trimmer ineffective; some further improvement may be had by adding variable R-C series elements, d, to case and adjusting until the best reduction is judged to have been achieved. These can then be replaced by fixed components of the closest value. The synchronisation signal itself may be the cause of CM currents, if not supplied on a balanced pair with a shield connected to primary ground. Again, a common mode transformer may be necessary to achieve balance in the differential current. Note that should the sync signal be supplied from a floating but ungrounded winding, grounding one side will result in large circulating currents due to the line to shield capacitance. Virtual balance can be created by matching the input impedance of the circuit, Zi, with its equivalent on the neutral side, Zb (see Fig. 8). The COSPIN particle experiment power converter was designed by SPAR Aerospace Ltd.; solutions to the EMC problems were arrived at in collaboration with the National Research Council of Canada. I thank my colleagues, Drs. M. Bercovitch and J.D. Anglin, for helpful discussions, and Mr. W. Blore for suggestions on this paper. - 561 102 - Q6 HIGH CURRENT FAST PULSE MEASUREMENT WITH A ROGOWSKI COIL B. Brandli, J. Bertuchoz, R. Steck Swiss Armament Technology and Procurement Group, NC-Laboratory Spiez, Switzerland Summary A review of the two most important methods for the measurement of current pulses is given, and their specific advantages and disadvantages are discussed. For most applications the Rogowski current probe is the optimum solution. The theory necessary for the understanding and the main features of Rogowski probes are resumed. Formulae for the design of a Rogowski coil are presented and a practical example is calculated. After a short discussion of the calibration systems used, some measurement results are presented. Further results will be given at the Symposium. 1. Introduction In the field of EMP-Simulation the measurement of induced currents is of great importance. Current sensors in accordance with EMP-requirements (e.g. peak values up to 10 kA and bandwidths from 10 kHz to 130 MHz) are commercially available. For some applications, however, it may be neccessary to have a probe adapted to very specific requirements. In this paper the experience gained with self-made Rogowski current probes is presented. 2. Methodes for wide band current measurements The principles used for the measurement of pulsed currents may be summarized as follows: - measurement of the voltage drop caused by the current flow through a shunt or a current viewing resistor [l], [2] - measurement of the current in an inductively (or field) coupled circuit of the maqnetic field associated with the current {e.g. by Hall effect or magneto-optic Faraday effect) . and it is difficult to design and use (e-g - feed back of the shunt impedance on the current to be measured - influence of the current distribution - interference of electromagnetic fields - skin effect, thermal behaviour) Field coupled probes require special calibration facilities and have a low-frequency cutoff. They have the following important advantages: easy installation; avoidance of ground loops and hence, in general, no coupling problems; comparatively straight forward to design. Field coupled current probes may be loosely classified as fluxmeters, current transformers or Rogowski coils, depending on their winding geometry. A fluxmeter is a single or multiple turn loop. For current measurements it generally must be calibrated for each measuring configuration. The current transformer couples the magnetic flux from the primary current to be measured into the secondary sensing windings. The Rogowski coil is a simple type of current transformer and will be described in detail in the next section. 3. The Rogowski current probe The Rogowski coil geometry consists of a shielded helical solenoid bent to form a torus. This principle is illustrated in Fig. 1. [21* Only the first two techniques are in practice being used for the measurement of high-current pulses. The field coupled probes are the favoured current measuring devices in the field of EMP-Simulation. The resistive shunt is DC coupled and although very easy to calibrate has the disadvantages that . it is directly connected into the current path (hence the possible creation of ground loops) Fig. 1: I+$iple I i Pr of the Rogowski-coil current - 562 The advantage of the use of this method is that the flux linkage between the probe and the current is independent of the distribution of the current to be measured - i.e. the probe may be calibrated independently of the measuring configuration. The flux linkage in a coil is described with the following equation CA is the path of the windings of the coil where: C A is the cross-sectional area n is the number of turns per unit length of the coil iit is the local magnetic field dl is the area element of a turn dl is an element of length along C If (1) the cross-sectional area of the coil, A, is everywhere the same; (2) the number of turns per unit length of the oil, n, is constant; (3) the magnetic field 8 is homogeneous within the area A; (4) the individual areas of the turns are oriented perpendicularly to C, then equation (1) becomes: &di (2) / C If further (5) C is a closed curve; and (6) C links itself no flux, equation (2) becomes Q = nAuipr (3) o = nA where: ipr P where ~7 = magnetic flux linking the turns ic = current in the coil = coil inductance = R,+R,= impedance of the coil circuit Rm = impedance of the measuring resistor Rc = skin effect resistance ot the coil L R (1) (1 n(iadi)dl 42 = - The field coupled current probe may be designed to be used in a differential or an integral mode which depends on the value of Tpr, the pulse width of the primary current, compared to the time constant of the sensing probe (= L/R). Differential mode If Tpr >> L/R then: L di, -edt R < ic di, 1 & ..--._*-_ dt L dt Rc ic CD CII-=L ipr N (6) where N = total number of turns of the coil. dg Rm dt Its are: t advantages over the different iating device PI, PI frequency-independent output voltage Fig. 2: Equivalent circuit of a Rogowski current probe when the coil is symmetrically excited its risetime is as short as the transit time through a single turn The circuit is described by the following equation: due to the low impedance of the currentviewing resistor the device is less sensitive to noise 1 R -'x d@ L di, =+i R'dt ' (4) less cable attenuation (output is frequency independent) - 563 - 102 Q6 Table I: Symbols 4. Design 4.1 Construction This is illustrated in Fig. 3 d coil wire diameter a mean minor coil radius A a*%: effective cross section of 1 turn r mean major coil radius b %r: total mean major circumference N total number of turns n number of turns per unit length #+E: pitch of windings P 'pr Fig. 3: Construction of a Rogowski-type current measuring system The two essential components are the probe coil and the current viewing resistor. The coil is usually"circular and has to fulfill as well as possible the conditions stated in section 3. That is Meaning ymbol] S. primary current threading the coil iC coil current R RM+Rc = total coil resistance RM measuring resistance ("current viewing resistor") RC skin effect resistance of the coil f - the turns should be equally spaced along the toroidal axis fundamental frequency of the current pulse to be measured CD - the effective cross-sectional area of each turn should be the same _/BdA: magnetic flux linking the windings of the coil B - the plane of the turns should be accurately vertical to the coil axis to avoid coupling with the axial flux. pipr/b: magnetic field intensity (cylindrical symmetry) Bsat saturation field intensity of the ferromagnetic core Urn Rmic: voltage on Rm Zt LIm/ipr:probe sensitivity (=transfer impedance) Tpr pulse width of primary current The coil has generally an air core or a solide dielectric, however for measurements below a few 10 kHz, a ferromagnetic core is needed. The coil is mounted in a conducting housin to serve as a shield against electric fie -78 and to act also as a short circuit of the the major turn to eliminate axial magnetic fields. An air gap is necessary to allow the radial flux to penetrate to the coil. The current viewing resistor is, in our experience, the most critical component for the design. It should be mounted very close to the active windings and have the following characteristics - Non-inductive - Frequency independent within the range to be preserved (skin effect) - small temperature coefficient (Joule heating) Damping resistors may be introduced to eliminate spurious oscillations due to capacitive coupling between the windings of the coil and the housing. TC T L/R: coil decay time V povr: magnetic permeability E EoE)-: P conductor resistivity C velocity of light in vacuum dielectric constant The following equations are used to calculate the main properties of self-integrating Rogowski coils. Most of these equations are explained in References [3] and[4]. Condition for a Rogowski coil to be selfintegrating: if L/R > Tpr then ic = g/L (7) Magnetic flux linked with the coil lziprNA (8) 27Fr Inductance of toroidal coil with circular core section described by Grover [5]: $= 4.2 Formulae for design parameters The symbols used in the equations below are explained in table I, all units being in accordance with the SI-System. coil transit time L = L'-aL where L = 1 2 and AL = ImaN* (9) b paN(G+H) G is the winding space correction given by Rosa [6] (10) (11) - 564 and with G=$-ln$ (12) H =k$o gk(lnW)k (13) go g1 g2 gs = = = = arising from the effect of the primary current distribution. Hence Tc < 5 ns. To realize this short transit time with the large coil diameter requires that it has only a few windings. If in addition to that the coil is split in two halves which are connected in parallel, the coil inductance is reduced by a factor of 4. This results, however. in an increased low cutoff frequenCY9 and-a higher current in the coil and subsequently in the current viewing resistor. 0,00070 0,17730 0,03220 0,00197 Skin effect resistance of slab with width nd, length [p2+(8Bi2 resistivity (flfpg) Rc (nf~g)12N[p2+(2Ra)2]12 (14) = R,: With the assumptions rd ZT = 6 IEJ; ic = 3 kA; Probe output voltage: R, = ZTipr/ic = 30 ~IXJ Probe sensitivity on transfer impedance i! : 5 ipr @Rm =_____= L +NARm 7=- Decay time constant: Core: Assuming a non-ferromagnetic core and R = Rc+Rm = 40 mu, the inductance can be calculated for the low cutoff frequency (16) 271rL L (17) R Low-3 dB cut-off-frequency: fl = h L = R/(2xfl) = 1.3 IIH (15) 7c which can not be achieved with the transit time constraints stated above. Transit time of the coil, neglecting capacitive coupling between the turns: Tc = F [erpr(p2+(2xa)2)]12 For this reason, a ferromagnetic core has to be used. Its characteristics are listed in table III. (19) Table Maximum of product iprt before core saturation: I N2Asat ipr(t)dt = _ Rc+ Requirements The design of a Rogowski coil current probe has to match the signals to be measured; the case of a probe to measure a signal with the parameters given in table II is now considered. Parameter peak current current rise time low 3 dB cutoff frequency transfer impedance minimum inner diameter (with housing) Value 15 kA < 20 ns 5 kHz < iJ lm62 7 90 mm 24x24 mm2 > 10 = 1 1T 50 pm Coil: The constraints arising from the already defined core size and from the maximum transit time leave only a few parameters to be varied within a small range. Three different coils have been constructed with the basic specifications listed in table IV. > 140 mm Table b) Value The core foil has a high magnetic permeability and is wound up to the core size of 24 mm. This construction results in a high conductivity in the core direction and a high resistance for circular currents in the plane of the cross section. Table II: Design parameters for a Rogowskl coil t Core characteristics mean radius of circular core r square shaped cross section relative magnetic below 100 kHz permeability br 1 MHz above suturation flux intensity B,,t Thickness of core foil (20) 4n 4.3 Practical design example a) III: Characteristic %at t0 15 kA the corresponding value of the current viewing resistor becomes (15) ZT = ipr = IV: Basic core dimensions Calculation of parameters T,: The transit time should be 4 to 5 times shorter than the rise time of the signal to be measured. This avoids problems due to spurious oscillations Symbol Meanings , I d I coil wire diameter Value 1 - 565 major mean coil radius 90 mm equivalent mean radius of the circular turns resulting in the same winding crossection like the coil on the square shaped core 14 mm 14 Total number of turns, consisting of two half-coils with 7 turns each Coil 1: With ferromagnetic core as described above Coil 2: With plastic (dielectric) core Coil 3: With ferromagnetic core again, but with doubled cross-section using two parallel wires as shown in figure 4. 102 - Contrarily to the coil, the construction of the measuring resistor revealed to be critical. Since there weren't any commercially available resistors with the required characteristics (no inductance, no skin effect, value of 30 a~), it had to be designed by ourselves. The finally chosen solution consists of a coaxial thinfilm resistor on a ceramic substrate with the shape of a disk. Fig. 5 illustrates the difference between two transfer impedance measurements at the same coil with (a) a frequency dependent (skin effect) resistor and (b) with a nearly frequency independent resistor. The different levels at 100 kHz are explained by the different values of the measuring resistors: R, q 10 rr~l(a) respectively 100 m62 (b). Frequency _.---- wi) , F( F, (‘&) - the energy spectrum of 4 he dis I!orted and nondistorted prooess respectively. If the harmonic interferences, asymmetry or voltage fluotuatioas are independent of time the EM01 coinoide with the generally accepted ooeffieients which oharaoteriae these inter f erences . The following would be obtained forrthe harmoni interferenoes where i.e. the oo%?fioient which characterizes the rem&, value of harmonic content. Three-phase voltage system asymmetry could be c&aracterized by the known ooefficient of asymmetry which whereUl,U - pornas. values of the dire& and Ll verse sequenoe voltages. If the voltage fluctuations are amplitude modulated acoording to the harmonio law then f =cl-zE 6U UL i ’ if it is square-wave ooltage(ourrent) mod;Iio$. 6u P- 9-W $Uf "' where magnitude of voltage ohauge. Thus, when voltage fluctuation, EMCIL * . It is more’ difficult to obtain simple formula for EMCIduring the action of the eleatromagnetic interferences changing in an accidental way. Rowever, comparatively simple expressions are managed to be obtained in some cases, For instance, if the harmonic interference amplitudes are accidental and independent on time the probable process of the harmonic interference voltage ohange might be presented as follows : where - random mlues. oonditions are observed (in a wide sense) the random process of the harmonic interferenoe voltage ohauge manifests itself as a proaess of the descrete spectrum, where values represent a distribution h”fh,a When the stationary where the r.mis. value of the first harmonic. It is necessary to point out that the proposed quantitive evaluations of EM01 oorrespond to the physical nature of the random processes of changing three-phase voltage system in electrical networks. In some oases these evaluations are simple expressions being wed in practioe for a long time. The above desoribed approaoh for evaluation of EMCI Gould be easily used in praetioe of mxintenance and designing as it proves to be a uniform methodological basis of measurement and caloulation of EMCI. Efficiency of the electrical equipment operation as well as maintenance of the required technological ohasacteristics as far as eleotrical equipment is ooncerned are determined to a reat extent b the ower supply sys&em zualits, Tge qdity of E;T”,;,E;Pply a understood to be two neoted conoepte with speeifio peculiaritiest reliability of .power supply and quality of elecltrical energy. The latter is a set of properties of electrioal energy stipulated by the generation, distribution and consumption of energy prooesses and which make the eleotrioal equipment operation with ~h&de~rmined technical oharaoterisIt is evident that the quality of ihe eleotrier energy is closely conneclted with the eleotromagnetio compatibility of the equipment during the action of interferences propagated in the network. Quality requirements of the eleotriaal energy are determined in the acting USSR standard GOST13109-67 “Norms of quality of eleotrioal energy for appliancsee oonneofed to public networks*‘. GOST 13109-67 determines norms of quality of electrical energy on the terminals of the equipment supplied by the public three-phaae and one-phase 5OHs stationary networks and by the d.o. networks. Requirements of the standard are extended to normal and after-break-down duties of the network operation. Seven &dices are regulated in the standard. They determine quality of the electrical energy in the a.c. three-phase networks. There are indioes among them that &araoterise the voltage curve sinusoidelity distortion, asymmetry and unbalance of the three-phase system voltages as well as voltage fluctuationa. Mathematical form of these indices - electromagnetio interfeoharacterizes rences asaumlng that they do not change in time. The limiting values of harmonic distortion and-asymmetry indiaes are determined in GOST13 109-67 with the allowanoe for susaeptibility of the most widely used energy receiver induction motor to the respeative interferences. Therefore the harmonic diatortion index value up to 5% or asymmetry index velue up to 2% are allowable for a long time on the terminals of any electrical aonsumer. The limiting values of the unbalance index are determined in every paxtitular case on the basis of the allawable voltage deviations on the terminals of one-phase eleatriaal applfawes since the unbalance of voltages mainly manifests itself in the additional voltage losses in the network and it means that in additional voltage deviations on the terminals of the electrical consumers. According to GOST13109-67 voltage fluctuations are evaluated by the ranby the frege of voltage change , i.e. quency of the voltage a&nge by the number of voltage changes per unit of time and by the time interval between the consecutive changes of voltage of the voltage Experimeital stu periodic changes infds uence on visuality and its fitness (the latter has been charaaterised by the clear visuallty stabillt ) permitted to set bounJ lowable voltage ahanges lower boundary oorresponds to d? . The to the those voltage changes when the reduation of visuality begins to take place The upper boundary aorresponds to the voltage changes which cause the lessening of illumination by 10% allowed for light installations. For instance, if the voltage ahan e fre uenoy equals 20 per second (1OHzf the Pover and the upper limits will equal to 0.32 and 0.45 % respectively. Besides the energy quality indioes already ooneidered three more indioes are determined in GOST 13109-67: the voltage deviation, the frequency deviation and the frequency fluctuation. These indices might be related to the eleotromagnetic situation oharaoteristios. More over, the voltage and frequency deviations define, to the most degree, technical and economic factors of the electrical network and e uipment operation, quality and quantjty of the manufactured products. Variation of the energy quality indices on the terminals of the electrical equipment depends on the great number of reasons most of which are of accidental nature. This clraumstance makes approach the energy quality indices evaluation from the probability point of view. It is accepted in GOST13109-67 that energy quality index values must be within the aIl.owable limits with 569 103 - 07 the integral probabllity of 0.95 per the determined interval of time. Besides, to prevent after-effeats conneoted with coming of the energy quality indiaea beyond the allowable litits which integral probability do not exoeed 0.05 GOST 13109-67 requires to restrict their values and duration of their effects by the limits approved by the Ministry of Power Energy. This regulation permits to introduoe the respective norms of the energy quality in aase of more strict requirements of oonsumers into branch and departmental standards under condition of their agreement with the requirements of the state standard. The probability approach to the uality evaluation accepted by ~~~~~3~09-67 permits, to our view, to aome more expediently to solving the problem of the necessary expenditures for keeping uality indices within the require 8 ljmits. To use the probability approach in the energy quality evaluation in designing and espeoiall maintenance practice GOST13109-6f established the durations of measurements whioh are differentiated for various energy quality indices and types of electrical appliances. The durations of measurements are determined on the basis of the uite lar e experimental material abou 8 the sta f istloal aharaoteristica of the random process variation of this or that energy quality index, of the required aoouraoy in index evaluation and expenditures for making meaSwemt3nt8 l BesuIts of the Research Researoh into the interference levels has been aarried out in urban, rural and industrial networks. Index of harmonio distortion in the low voltage networks for urban needs has not erceeded 5% and in rural networks aoming sometimes up to 7% due to great impedances. Greater ,distortions, up to 10-158 have been observed in the points of the network near the traation substations. Unbalance of voltato 7% has prevailed in the ft:‘vt Etage r&al networks Measurements were oar&d out with the help of standard apparatus duxing a long per&od (not less than 7 days). Influenae meotriaal of Interferenaes on Eauiument Oaeration Electromagnetio interferenoes are conducted In the network to the eleatrioal equipment eusceptable to their aation. As a result the supply voltage is being distorted. Action of the distorted voltage, might produce an instantaneous effect or manifeat itself gradually. In the first oase the eleotxical equipment as a rule stops to function or has a ielse operation, for example, relaying, automatio devices, - computers. The damage here comprises not only the oost of the broken equipment but also the losses due to the disturbances of the technological process, losees of the computer infolc mation, etc. Iu the seoond case after-effects of the electrical equipment operation due to interferences take place gradually. Operation of the electrical machines in such conditions is a typioal example of it. As a result eoonomio and technical factors of the electrical machine operation deteriorate. Por example, heating of windings increases, lorsses in machines also increase, servioe of life decreases, the motor rotation speed and the power consumption are &anged. The change of the given operating oapacity of the motor might cause a damage to the technologiaal process, faulty produotion and shortage of it. Therefore two components of the damage might be considered: electromagnetio, cancerning teohnioal and economioal parameters of the e uipment including the reduction of sts service life, and the technological, concerning the shortage of production and we&age. Inve~tflgation show that the technologiaal oompoaent of the damage is domineering and in some cases it comes to 85096 This circumstance should be taken Into aooount in limiting of interferences in the network, i.e. the problem of the allowable limits is to a considerable degree an economical one. 570 - Conclusion For the recent years a wide spreading of electric equipment which creates long-term electromaguetic dieturbances through electrio systema for supply of that equipment, inoreased number of eases when the electromagnetic disturbances influence badly on sensitive eleotrio equipment. In the USSR in order to guarantee an effeotive and reliable operation of the sensitive eleotrio equipment there is a standard limitated electromagnetio disturbance levels at a network point where the sensitive electric equipment has been connected or may be connected. The application of standard requirements in the uourse of design and performanoe prevents expenses due to BMCI decrease, References -1 ~~tantinov B.A., Zheehelenko Nikiforova V.N. and others: TLe*kyatem of indices and stendartization of the energy quality. ELECTRICHESTVO, No 9, 11-19 (19782 Levin B.R. Theoretical fundamentals of the statistioal radioengineering. SQVIETSKOIE RADIO(1977) 3 GOST 23875-79 oQuality of electrical energy, Terms and definitions” 4 Marusova T,P*, Jagovkin G.N. A question of normalizing volta e fluotuations in elec ric etworf8, SVETOTECHNIKA, No 2 119777 - 571 104Rl - RADIO FREQUENCY SPECTRUM MANAGEMENT K. Deutsche Bonn, Federal Bundespost Republic Introduction The usable radio frequency spectrum is a limited natural resource. Its use is open to everybody and it can be reas well as polluted. The reused, source is not adequately used, when the objective can be easily achieved in other ways or the parameters of the use are not correctly applied to the task. To use the resource efficiently, it is necessary to question the need to manage its use by imto use it, plying efficient sharing criteria and to coordinate the assignments nationally and internationally. Developed countries have difficulties to achieve these goals in view of the ever increasing demands in the field of radiocommunications. They however, developed tools and have, procedures that permit acceptable aolutions. The problem of developing countries is lack of infrastructure: To develop or improve their economical and cultural infrastructure, an efficient telecommunication infrastructure is essential. To cope with this demanding requirement, governments have to look at priorities to achieve nar tional planning objectives. Radio networks can rapidly improve domestic’ telecommunication at moderate costs. In 1979, the WARC recognized the need to improve spectrum utilization and considered the application of computer methods in frequency management essential. The conference also expressed the need for guidance in this It therefore in Recommendation field. 31 invited the CCIR “to prepare a handbook describing the various aspects involved in applying computeraided techniques to radio frequency management discussing the approaches which have been made, providing guidelines for various levels of practical application and making recommendations for those aspects involving international cooperation”. In Resolution No. 7 and No. 37 the need to develop national Radio frequency management is highlighted and seminars are en- 1. Olms of Germany visaged to assist developing counA first seminar was held in tries. Geneva in October 1983. CCIR Study Group 1 in 1978 sat up Interim Working Party l/2 which has prepared a “Handbook on Spectrum Management and Computer aided Techniques” which has been published. This presentation is largely baaed on the handbook. 2. International Regulations The radio frequency spectrum as an international resource, has been divided into frequency bands which may be used by one or more radio services under specified conditions. This division of the spectrum is the basis for the Table of Frequency Allocations set out in Article 8 of the Radio Regulations and constitutes an agreement between the members of the ITU for the sharing of the spectrum among the various radio services operated in different regions of the world. This table is a fundamental instrument of the national and international frequency management. For frequency allocation purposes, the world has been divided into three geographical regions, the boundaries of which are defined in Nos. 392-399 of the Radio Regualtiona. Besides the geographical division, different categories have been eatabliahed for radio services. These categories are defined in the Radio Regulations as follows: primary services permitted services, secondary services. The radio frequency spectrum is an asset which is common to all administrations. The spectrum must be shared -among administrations -among radio services -among stations. On the other hand, each administration is autonomous. It thus becomes apparent that the beat way to serve the interest of every administration is to obtain an international agreement on rules and procedures for the manage- - ment of the spectrum. The ITU Radio Regulations constitutes a base for this work. The main objective is to avoid harmful interference between stations of different administrations. For this purpose coordination procedures have been agreed upon to advise administrations how to exchange information and how to take all necessary steps to ensure that harmful interference will not occur. Coordination procedures can be divided into three principal parts: administrative provisions, exchange of information, and the technical calculations. (Typical international frequency management procedures can be found in flowcharts constructed by the IFRB). The relevant technical data to be exchanged between administrations and between the IFRB and the administrations are given in the Radio Regulations (e.g., Appendices 1, 3, 4, etc.), in Final Acts from ITU Conferences, or in CCIR Publications, depending on which services are concerned. Calculations of coordination distance and coordination area are sometimes rather easy to carry out manually. In other cases, the calculations become more complex and timeconsuming. Examples of computer programs for coordination calculations are given in the handbook mentioned. The geostationary satellite orbit is also a limited resource that must be shared by all administrations. In the case of coordination between apace services, it is also necessary to consider the efficient use of the geostationary satellite orbit. 3. National Spectrum Manaqement The requirements of individual radio stations or services can be met only by sharing time, space, or frequency. Moreover, bilateral or multilateral agreements will increasingly be required. This means that general assignment procedures in the form of simple network plans for the use of frequencies cannot always be employed in the near, and especially in the far future. In the future, it will become necessary to make electromagnetic compatibility (EMC) calculations before frequencies are assigned. The bulk of technical and administrative work involved will not only require detailed knowledge of the equipment used but also of the physical characteristics of propagation over the whole frequency spectrum. The solution of the aforementioned problems neceaaitates the introduction of new spectrum management methods. Spectrum management is the combination of administrative and technical procedures necessary to ensure the efficient operation of radiocommunication services without causing harmful interference. When an administration ratifies 572 - the International Telecommunication Convention or accedes to it, it should enact domestic legislation to make the provisions of that Convention and of the Radio Regulations annexed to it applicable to its Administration. Generally speaking, use of the frequency spectrum can only be efficient if it is properly organized: by imposing restrictions on certain technical characteristics that must not be less strict than those specified in the international agreements; by providing for the future well in advance through national planning of frequency usages. The first condition necessitates the issuing of national regulations which should not be confined solely to the international provisions and may comprise everything which the legislature considers necessary to: enable each user to carry on a service under specified conditions; make sure that the administration’s international obligations are fulfilled. The second condition implies the existence of an administrative body to ensure coordination between the different users. This body must have the technical resources and administrative means to check whether the domestic legislation is being applied, in other words, to manage frequency spectrum usage in the national context. The National Spectrum Management Authority will thus have several roles, which may be summed up as follows: which consists of a) Standardization, effecting coordination between various users with a view to defining technical standards to be imposed on users and, if necessary, on equipment manufacturers, in implementation of domestic legislation; which consists of effecb) planning, ting coordination between various users with a view to defining the use to be made of the frequency bands as listed in the Table of Frequency Allocations, and in planning future uses so that they may be included in this Table by an administrative conference competent to revise it; proper, consisting of c) management granting authority for frequency usages in accordance with international regulations and domestic legislations, and ensuring that those frequencies are actually used in conformity with the terms of the authorization. This role also includes compliance with the obligations embodied in the Convention and the Radio Regulations with respect to other Administrations (for instance, the for coordination handling of received requests from - other Administrations, etc.). 3.1 Domestic Leqislation. Since domestic legislation is considered to include the pertinent provisions of the Radio Regulations, an assignment in conformity with it is necessarily in harmony with those Regulations. A document containing the provisions of domestic legislation greatly facilitates the task of each national soectrum manaoement authority. Licensing is the ’ 3.2 Licenses. orocess of conferrinq the leqal authority to operate a-radio station under specific and stipulated condi- tions. No. 725 of the Radio Regulations stipulates that no transmitting station may be established without a license issued by the Government of the Administration to which the station In some Administrations the belongs. right to use a radio receiving installicensing. lation is also subjected to Administrations may charge users of the spectrum a fee for their licenses. The fee may reflect the degree to which the spectrum is used, as well as the economic benefit derived. 3.3 Inspection of installations. In connection with its responsibilities for issuing of the Government, licenses, on to frequency behalf users, including radio amateurs, the national spectrum management authority must be able to confirm that stations comply with the relevant provisions of the Radio Regulations and domestic law, and with the terms of the license. For that purpose it must have the staff and’ equipment necessary for conducting inspections of stations and checking their operation on the spot; as far as checking the quality of emissions from a distance is concerned, it is understood that the authority is able to do this with the monitoring facilities at its disposal and also to see that the nature of the traffic exchanges by amateurs remains within the limitations laid down in the Regulations. These various checks apply not only to non-mobile stations but also to mobile ststions, that is to say, to ship stations, aircraft stations and land mobile stations. In the case of mobile stations, the Radio Regulations stipulate that, when they are in the territory under the jurisdiction of an Administration other than that in which the license was issued, the stations may be inspected for the purpose of examining the license, discovering any irregularities in the equipment or its operation and reporting them to the competent authorities of the licensing administration. In the case of ship and aircraft stations, inspection also covers examination of the operator’s certificates. 104 5 73 - RI It is also the function of the national spectrum management authority to issue ship station and aircraft station operator’s certificates specified in the Radio Regulations. 3.4 Monitoring. The national spectrum management authority should monitor the emissions of radio stations to check their technical characteristics and to ensure that they are op- erated in conformity with the stand- ards and various conditions on the basis of which their licenses were issued. A monitoring station can help a great deal in solving problems of harmful interference and in finding suitable frequencies not subject to such interference. The monitoring service should be designed to meet domestic since radio emissions needs. However, are no respecters of frontiers, the stations of the monitoring service of an Administration should be prepared to cooperate with other Administrations as well as with the IFRB and the international monitoring system. 3.5 National file of frequency Once a frequency has been usaqes. assigned to a transmitting or a reall the technical and ceiving station, operating data indicating the spectrum space occupied by this assignment should be entered in a government master file. Such a file not only serves as a reference for subsequently usable assignments, but also provides the basic material for taking effective measures required to adapt national planning to the real requirements of the various users. The greatest care should be taken in compiling the national file and keeping it up-to-date; it must have room for a sufficient number of assignments and for all the information needed for the clear and complete description of each of them. It is very useful to employ modern computer processing and recording techniques suited to the size of the file and its usage. primarily 4. Analyses Models Efficient spectrum management can be accomplished only through a joint technical and administrative effort requiring the application of rapid record keeping, quantitative analysis and experiential judgement. In many cases the results of an analysis process serve as improved advisory information to the human decision-maker that are used to expand the understanding of the problem. It is only from the application of rapid spectrum management techniques that timely and complete solutions to the problem can be obtained, The nature of some of these tasks requires data files, and much of the problem approach is oriented with - Efficient use of large this in mind. volumes of data logically requires high-speed digital computer capability, and its application for both data and analysis is considered. In practice, a hybrid approach using a combination of automated and manual techniques may be used for the solution to many problems. A variety of techniques are available as aids in the evaluation of These include: problems. Frequency-distance separation criteria Guard Band Design Parameter Sensitivity Analysis Antenna Dynamic Analysis Frequency assignment procedures Each of the functions performed during this approach establishes requirements for a variety of analysis techniques. The requirements for a particular item of data base information or a basic analysis tool may be established by serveral analysis functions, although each function generally imposes different demands on accuracy or quality of the data or models. A short discussion of these requirements for models and data is given in the CCIR-handbook. An initial list of System Models useful in performing basic EMC analysis and spectrum management functions include: File Selection and Review Technical Cull Received Power Prediction System Performance Prediction Frequency Assignment Generation of Frequency-Distance Separation Criteria The need to evaluate the basic power prediction equation require component (basic) models that are capable of rapid computation and that produce consistently conservative or “safe” estimate of each variable. Basic Model requirements include prediction of: Propagation loss Antenna Patterns Transmitter Emission Spectrum Receiver Performance Each of these model categories is discussed briefly in the following paragraphs. More complete descriptions for various models are available in the CCIR Reports, many of which are discussed in some detail in the handbook. In general, no environment input data will be required beyond that needed by the component models. 4.1 Propagation loss. Propagation loss models are available in a wide range of complexity, accuracy, and input requirements. Input requirements for these models vary considerably. In most cases, the less input required by a model the less accurate its outa characteristic common to most put, model classes. Minimum input models are used for cull purposes in most 574 - cases although for those physical circumstances that fall within the model limitations the results can be quite accurate. 4.2 Antenna patterns. Models are required to portray antenna gain characteristics both discretely, as a function of angle, and statistically. Analysis requirements can be expected to involve all possible angles in three dimensions and frequencies below and above those for which the antennas were designed. Model development to date has been based primarily on emand significant conpirical data, straints have been imposed due to the limited availability of measurements. 4.3 Transmitter Emission Spectrum. Models are required that will portray electromagnetic emissions emanating from a transmitter. Emissions include wanted and unwanted energy (primarily that delivered to an antenna) including in-band and out-of-band spuriIn most cases, models ous signals. will be a primary function of power output and modulation characteristics. While such complete descriptions are desirable, they are often unavailable. In many practical cases, knowledge of emission characteristics within the few channels adjacent to the tuned frequency and at dominant harmonics is adequate. Problem analysis requirements can logically be expected to include all modulation forms, both state of the art and advanced development. 4.4 Receiver Performance. Models are required that produce descriptions of the performance degradation characteristics of electronic circuitry (primarily receivers) that is potentially capable of responding to the presence of electromagnetic energy. This should include the fundamental response and selectivity of a receiver plus any additional responses that may occur regardless of frequency. Image and other spurious responses are included. Both linear and nonlinear effects are of concern. Requirements may also exist for output displays to summarize or highlight a large volume of output data. The most common displays are the histogram type. In addition, displays of functional relationships synthesized within the program are useful as a user selected option. In the detailed analysis case, a receiver input wanted-to-unwanted signal ratio is most frequently useful. The use of this expression establishes model requirements of emission spectra, receiver performance, receiver sensitivity, desired signal level, and off-frequency rejection. Off-frequency Rejection (OFR) may be derived from analysis of emission spectra and receiver susceptibility data but in some cases, it may be more convenient (and accurate) to develop OFR data directly for given combina- - tions of transmitters and Several computer programs toward this objective are in Report 654. 5. receivers. directed described, Data base considerations Planning for automation and computer applications includes provisions for the extensive use of automated Availability of data in data bases. computer readable form greatly enhances the flexibility, speed, and economy with which automated processing is accomplished. However, many of the analysis techniques can be effectively used without any automated data in which case users supply rebases, quired input each time new analyses are conducted. In practice, the need for automation should be judged on the basis of individual requirements. Prospective users should carefully evaluate benefits and costs of each data base file. The most common data base structure involves one or more files, each containing a number of records. Each record contains data located in a number of fields. Perhaps the most fundamental data and one that already exists in file, some manual form in any organization charged with the responsibility of administering frequencies, is a list of frequency assignments. In the simple case, this frequency assignment file will contain one record for each and each record will conassignment, tain a number of fields. One field might contain frequency, another the transmitter power authorized, a third the name of the operator, etc. Data useful in spectrum management and interference analysis can be grouped according to its degree of generality. Such grouping offers several advantages. It reduces the amount of data collection and storage required and improved quality control efforts by ordering like data togeththus highlighting anomalies and er, discrepancies. An anomaly (unusual data) and a discrepancy (an error or missing data) are not the same; awareness of both is important for problem analysis. For example, the costs associated with maintaining the accuracy and currency of frequency data in afile may not be justified for the mobile units associated with a base station. A useful distinction in planning for data bases is the difference between unique and generally applicable information. Unique information applies to one single frequency assignment at a specific place. The latitude and longitude of a transmitter operating under a specific assignment is probably the best example of unique On the other hand, general data data. applies to a number of situations or 104 575 Rl items of equipment. For example, the power output of the simple transmitter used at a specific location is not unique to that operator at that place. It is general data that describes the power output of all transmitters of the same design (manufacturer, model number, etc). Data groupings useful for spectrum management applications may be as follows : 5.1 Operational usaqe or frequency Files in this category dependent data. contain data unique to a specific sitLocation, frequency, antenna uation. and pulsewidth/PRF are typical height, of such data. In general, one might expect large numbers of relatively short records in such a file. A key item in each record is information (a cross index) that unambiguously relates to appropriate information on The make and model of a other files. simple transmitter (one which has only one power output) is one example of such an index. 5.2 Equipment characteristics data. These files contain data common to all equipment of a given nomenclature or of a given class. Antenna gain, power output, receiver sensitivity, and IF bandwidth are typically in this group. Parameters other than nomenclature may be used as the classifying criterion, modulation type for example. Such a file would normallly contain substantially fewer records than in the equipment dependent case, but each record may be quite extensive. 5.3.Path’or couplet dependent data. Files of this kind contain information unique to transmitter-receiver combination . Path length, path profile, mutual gain, off-frequency rejection, susceptibility criteria, path loss, etc., fall in this category. Note that this type of data may apply to a single discrete path/couplet situation, or it may represent a general class of situations. 6. Conclusions Efficient spectrum management requires an infrastructure that can satisfy national requirements and international obligations. Joint efforts require record keeping, quantitative analysis and experimental judgement. Data files and analysis techniques vary from simple to very complex systems. Voluminious data and complex analysis lead to change from manual to automated processes. The changeover will create new types of problems and may be costly in the initial phase. Such a changeover therefore requires time and careful analysis and planning. Computer systems are now available to perform these functions at reasonable costs, thus permitting improved handling of assignments, staff savings and better information availability. - 577 - 105 R2 THE BIG SQUEEZE A SELECTIVE LOOK AT ORB-85/88 H. J. Weiss Communications Satellite Corporation Washington, DC, USA* AN HISTORICAL NOTE - At the 1979 World Administrative Radio Conference (WARC-79), a strong bloc of mostly developing nations forced the adoption of a Resolution [l] which calls for the planning of certain space radio services and frequencies which involve the use of the geostationary satellite orbit. A two-part World Administrative Radio Conference will establish in 1985 which services and frequency bands should be planned (ORB-85) and will institute such planning in 1988 (ORB-88). The heavy and growing use of the geostationary orbit/spectrum for some services and in some frequency bands has engendered concern with premature orbit "saturation". Planning is to ensure the availability of viable orbit locations for tomorrow's would-be users of the resource: to guarantee, in practice, equitable access to it by all nations. In the predominant view of the proponents of planning this requires the long-term reservation of specific orbit locations and frequencies for the exclusive use and disposition by individual countries. Industrialized countries reject this "a-priori" approach to planning as too restrictive and hold that equitable access can be guaranteed by other forms of planning which accommodate new requirements'as they emerge under observance of agreed obligations and appropriate codes of conduct. There are precedents for a-priori planning of a qeostationary space service: the two broadcasting-satellite service planning conferences of 1977 (for ITU Regions 1 and 3) and of 1983 (for Region 2). However, these have limited relevance for ORB-85/88 since they dealt with a service for which no previous assignments existed and for which it was possible to agree on highly uniform technical characteristics. Neither applies to the service of most concern to ORB-85J88: the fixed-satellite service using frequencies below about 15 GHz, which will be the subject of this discussion. TECHNICAL BACKGROUND - The electromagnetic isolation needed to protect two geostationary networks (,anetwork being a satellite and its associated earth stations) against unacceptable interference from each other when they use the same frequencies requires substantial physical separations between either their earth stations or their satellites or between both. *The views expressed herein do not necessarily represent those of the Communications Satellite Corporation. Figure 1 illustrates that an assignment (the orbit location of a network, its service area on the earth surface which contains its earth stations, and certain other technical characteristics) can visually be represented by an elongated cone, its apex defined by the orbit location and its "aperture" determined by the exocentric 121 angular width of the service area. Figure 1. Assignment Geometry. 6 = Apparent IntersatelliteSpacing @ = Exocentric Service Area Separation Figure 2 shows that coexistence of two such assignments is possible only when their service area separation and the spacing between their satellites meet minimum separation conditions. These minimum separation conditions depend entirely on the radiation characteristics of satellite and/or earth station antennas, mainly in directions outside the assignment cone ("sidelobe gains"), the transmitted power levels in a network, and the ability of a network to tolerate another network's transmitted powers which it generally receives through its and the other network's antenna sidelobes as unwanted ("interfering") energy. The sidelobe gains of earth station and satellite antennas decrease fairly rapidly in directions having an increasing angular separation relative to the direction of maximum gain (monotonic sidelobe 'gain "decay") . Figure 2 also indicates that collocation of satellites serving sufficiently far separated service areas is possible, but this option is limited by geography and service area size and available electromagnetic isolation. In general, an increasing network population will - 578 “use up" an increasing number of necessary "separation arcs“ out of the orbit's 360'; ulti- Relationship between Triter-satellite Spacing 0 and Service Area Maximum SatelliteAntenna Discrimination33 dB. Earth Station Antenna Diameter 10 m. Satellite Antenna Beamwidth 2.5'. 00 20 e+ 40 6O mately so many that additional networks could only be accommodated if the width of these separation arcs could be reduced. COORDINATION - Current practice already seeks to minimize the size of required separation arcs to facilitate the introduction of new networks. It imposes constraints on and encourages compliance with minimum standards for the three major determinants of separation arc size: earth station and satellite antenna radiation char'acteristics, t.ransmitterpowers, and sensitivity to interference (the latter through "interference criteria") . Most effective, however, in the reduction of required separation arc size is the process called "coordination". To coordinate two potentially interfering networks with each other, their transmission parameters, frequency plans and frequently also other system characteristics are adjusted with respect to one another in such a way that the service requirements of both networks can be satisfied from the orbit locations chosen for their satellites. Occasionally, during coordination, rather than putting more effort into the process, it is found preferable to obtain additional electromagnetic isolation through the relocation of one or the other network's satellite. Generally, however, coordination is successful with the chosen satellite locations and could, with some additional effort, even tolerate smaller intersatellite separation. This latter fact and substantial pressure of demand has prompted one administration to impose upon networks under its jurisdiction a fairly small but - as shown by analysis - coordinable spacing between adjacent satellites and to mandate that networks be coordinated with one another for that spacing. One can hardly take issue with the objectives of the 1979 planning Resolution - they reflect the purpose and spirit of the International Telecommunication Union. However, one of the ways in which those who found it - necessary to reaffirm the Union's basic tenets through this Resolution seek to implement it by the reservation of orbit assignments for all nations - may create a fundamental problem which has not yet received widespread attention. The accommodation of new networks under the current regulatory process, as well as it has worked, is nevertheless apt to require increasing skill and effort on the part of administrations. Therefore, quite apart from the concern with resource depletion, rigid planninq is also seen to obviate the need for coordination when an administration chooses to implement its assignment - coordination requiring both skill and effort - and that such implementation should not be economically burdensome - implementation cost representinq another facet of effort. The problem is that, in certain frequency bands and orbit segments, the existing regulatory process may already have achieved an assignment density which is greater than probably achievable with a plan that would not require coordination between assignments. The first intimation of that was given expression by one Region 2 administration at the 1984 Conference Preparatory Meeting (CPM) for ORB-85. This administration claimed that, by its calculations, it would have only a 13% chance of finding an orbit location that would not require coordination with existing networks. This claim may well be optimistic: under the current regulatory provisions coordination between any two networks is required when emissions from either network reach the other as unwanted energy and set up, at a receiving earth station of the other network, an equivalent noise temperature increase in excess of 4%. This is a fairly stringent "threshold“, and there is a good chance that the actual probability of a new network not requiring coordination with existing ones is less than 13% when recalculated with detailed transmission parameters of the new network. The problem can be quantified by example. THE U.S. PRECEDENT - Current practice requires that a spacing of 3' be achieved between the satellites of U.S. FSS networks in the 6/4 GHz band. This presumes orthogonal polarization between networks using alternate satellites and thus is equivalent to a 6" spacing requirement between co-polarized satellites. An antenna reference pattern corresponding to a sidelobe gain of 29-25 log 6 dB is stipulated where 8 is the off-boresight angle (in degrees), e.g., the anqle to a neighbor satellite, and a polarization isolation of 10 dB is stipulated where it is applicable. Assuming that all networks under consideration have equal technical characteristics, including about equal net available mean power density-to-noise density ratios and equal satellite antenna gains and coverages, the available "isolation" between two neighbor networks (i.e., the net wanted-to-unwanted carrier power ratio between "identical" 131 transmissions in two neighbor networks) is a function of the earth station antenna diameter. With the assumption of a 1 dB "topocentric" advantage 141, Figure 3 shows the available - 579 isolation (in dB) as a function of earth station antenna diameter for the "homogeneous" example, for earth station antenna diameters between 3 and 20 meters and a 6" intersatellite spacing without polarization discrimination. Figure 3 also shows the available isolation under the assumption that earth station antennas conform to the current sidelobe reference gain equation 32-25 log EldB. 29-25 Y Earth Station Antenna Sidelobe Discrimination at 6' off Boresight. C-Band. Figure 3. "' 2olI 3 5 10 20 ; E. Sta. Antenna Diam.(m) Available isolation needs to be compared with isolation requirements. Figure 4 shows these between several combinations of transmissions having the same assigned frequencies. Function acity of Required Isolation as of Voice Channel CapInterfered-with FDM/FM dBI c-3 35 - c-4 30 _ 10 20 n + 50 R2 with artificial energy dispersal of 1 MHz (curve C2),2MHz (curve C3) and 4 MHz (curve C4) peak-to-peak deviation at frame rate. 1solation requirements between digital 4-phase PSK transmissions would, by current interference criteria, lie about 3 dB below the curves A and B, and those to protect such transmissions to acceptable interference against FM/TV transmissions about 3 dB below the curves C, when the assumption is made that there is approximate equivalence between the FRM/FM channel capacity n of the abscissa and a digital carrier bit rate of 0.26 no** kb/sec. When the interfered-with carrier is SCPC, the isolation requirements against FM/TV transmissions are approximately: log 0 dB a 105 - 100 200 500 Transmissions assumed to be interfered with are FDM/FM multichannel telephony having large rms modulation indices (1.25 < m 2 2.65); their voice channel capacities are given on the abscissa. The isolation required to cause no more than 600 pWOp of interference in an interfered-with transmission by an interfering transmission is given on the ordinate, in dB. The interfering transmissions are identical carriers (curve A), a 792 channel FDM/FM carrier (curve B) and an FM/TV carrier without artificial energy dispersal (curve Cl) r and . Peak-to-Peak Energy Dispersal QPSK SCPC 64 kb/sec CFM SCPC 27 kHz 1 MHz 2 MHz 4 MBZ 47 dB 45.2 dB 43.4 dB 41.5 dB 39.0 dB 36.6 dB Departures from homogeneity in the transmission parameters tends to increase isolation requirements. The use of low-index FDM/FM transmissions will generally increase the isolation requirements because such transmissions require higher powers and may themselves be more sensitive to interference. For example, an interfering 1800 channel FDM/FM carrier, to meet its own performance requirements, would produce required isolation values about 11 dB greater than those of curve Cl of Figure 4 for the interfered-with carriers represented by the abscissa. Other inhomogeneities (i.e., differences between systems) may be in the noise temperatures and the antenna gains. They can be additive. Therefore, the general tendency to consider values of available isolation in the 30-33 dB range as adequate to protect networks from each other without coordination is not supported in practice. The current provisions of the Radio Regulations take this fact into consideration with the 4% equivalent noise temperature increase mentioned earlier. Figure 5 shows curves of equivalent noise temperature increase as a function of available isolation with the interfering signal being FM/TV with 3 energy dispersal assumptions. Thus, with 1 MHz of peak-to-peak frame rate energy dispersal, an isolation of about 44 dB is calculated. From Figure 4 this isolation value is about 4 dB greater than required to protect a 12 channel FDM/FM transmission to CUrrent interferenCe criteria. While this isolation also suffices to protect a CFM SCPC transmission, it is insufficient to protect a QPSK SCPC transmission (it is generally understood that the 4% equivalent noise temperature increase is an inadequate threshold for QPSK SCPC). Referring to Figure 3, a 44 dB isolation would only be available with earth stations of about 17 meters diameter at 6' intersatellite spacing and without polarization - 580 discrimination, but in fact there IS in use a mix of earth station antenna sizes the majority of which lie in the 4.5 to 10 meter diameter range. at 30 spacing which is the actually pres'_ cribed intersatellite spacing, a 10 dB polar= Appendix 29 sation isolation is stipulated. of the'Eadi0 Regulations provides, subject to agreement by the interfered-with Party, for a 6 dB polarization isolation. Thus, while the expected isolation at 3O is 10 aB better than the values shown in Figure 3, the fact that only 6 dB are allowed for purposes of calculating the equivalent noise temperature increase makes the AT/T threshold for 3O spacing bY 4 aB more protective than at 6' spacing. - the express purpose of establishing whether a need for coordination exists. Nevertheless, even the fairly stringent 4% criterion may be insufficient to "protect" certain transmissions against interference from certain others (FM/TV to QPSK SCPC); - orbit use density is greater than would allow a significant percentage of potential carrier combinations to meet current interference criteria without coordination. The U.S. example is most relevant because it reflects real networks with a representative mix of earth station antenna diameters and Service offerings and is even based on more stringent earth station antenna radiation characteristics than are used in other networks; - high-density transmissions (FM/TV and low-index FDM/FM) are the major problems, but they are also likely to be in continued demand for future networks, as are low-to-very low capacity transmissions which tend to be those most sensitive to interference from high-density transmissions; - net isolation requirements between real and, for the most part, inhomogeneous networks are likely to lie, with current interference criteria, in the range 30-40 dB with larger values (due to low-index FDM/FM transmissions) not unlikely to be required. Such high isolation values raise increasingly severe problems under a trend to decrease earth station antenna size and/or antenna sidelobe standards. 30 40 35 Available Isolation (dB) 45 what this means is that, under cantinued aSSLlmptiOn of homogeneity between netwx.dcs, Calculation of equivalent noise temperature increases for 5 and 10 metex diameter earth station antennas would Yield the following values, at 3O and 6O spacing, respectively (interfering carrier is FM/TV): , E.S. Ant. Diam.(m) FM/TV P-P Spreading (MHz) AT/T @ 3O - % AT/T @ 6O - % 10 1 2 4 14 7 3.5 10 5 2.5 5 1 2 4 56 28 14 40 20 10 These AT/T percentages would be twice as large with the assumption of the current earth station antenna radiation reference formula 32-25 log 8 dB and could also be significantly larger with non-uniform (inhomogeneous) networks, notably with the use of low-index FDM/FM transmissions as are, in fact, used by some U.S. networks. The preceding discussion reaffirmed that - orbit use density is greater than would satisfy the 4% AT/T threshold criterion. No. surprise here: that criterion was developed for IMPACT ON PLANNING - Qualitatively, the preceding adds little to what is well known: that the current coordination process is a rational, highly effective mechanism for the efficient utilization of the geostationary orbit/spectrum, at least in the fixed-satellite service. Quantitatively, however, the material presented may drastically affect the perception of reservative planning. In reservative planning, it would be essential that national assignments be made in awareness of their potential interaction. One could deal summarily with this by describing existing as well as future (not yet well defined) networks in terms of their actual or projected "Appendix 4" 151 characteristics and give each an assignment which meets the 4% AT/T criterion with respect to every other. However, we know that, with current standards and interference criteria, even existing networks do not meet the 4% criterion with respect to each other at the intersatellite spacings for which they have been successfully coordinated. Therefore, even if current coordination arrangements were allowed to remain in force, the requirement to meet the 4% AT/T criterion among all assignments to countries not now having one, and between these and existing assignments, would necessitate the removal of a substantial number of existing networks. Quite apart from the thus resulting very low orbit .utilization efficiency, this approach could be expected to meet with resistance. - 581 There are two alternative approaches which are more pragmatic. CONSTRAINTS TO ZIMIT INTERACTION - The first alternative would seek to diminish the basic potential for interaction between assiqnments by prior constraints. These could relate to design characteristics such as a lower limit on earth station antenna diameter and more stringent earth station and satellite antenna sidelobe radiation characteristics. The U.S. example is illustrative of this, and the adoption of tighter standards or constraints is an option, notwithstanding the current trend aqainst it. Then there are two potential operational constraints that would be effective. The first of these is the acceptance of more interference than currently recommended by the CCIR. Of particular interest is the selective acceptance of greater interference in low-capacity transmissions which mainly determine the required isolation between networks. Generally it should be possible to compensate for increased interference to such transmissions by assigning to them somewhat higher powers in spectral regions occupied by high-density components of interfering tsansmissions without thereby placing an undue demand on the total transmitter power shared with other transmissions (e.g., in a transponder). - 105 higher assignment density. An example would be to adopt AT/T criterion in the range between 40 and 200% for the establishment of assignments, and to require that, at the time an assignment is to be implemented, it be coor'dinated with existing networks and other affected assignments. This approach would be facilitated by the adoption of constraints of the kind discussed in the previous section, in particular by agreement on frequency seqreqation between high-density and low-capacity transmissions. Figure 6 illustrates the difference between fully protective planning and planning with insufficient protection. Two frequency plans are considered: one represented by the interference "threat" it poses (A) for another which is represented by its sensitivity to interference (B). Full protection planning must A The second operational constraint would be aimed at systematically segregating the frequencies occupied by (interfering) kigh-idensity transmission components and (interfered-with) low capacity transmissions. Such a constraint has been proposed in the form of various "spectrum segmentation" arrangements of a fairly restrictive nature. However, there is one arrangement that might be both effective and generally acceptable: the identification of a suf-ficient number of specific narrow frequency regions across an allocated band, High density transmission components would be restricted to these narrow frequency regions by general aqreement (e.g., 2 MHz frequency slots every 20 MHz across the band). Prior knowledge of where high-density transmission components could be expected would allow the planners of low-capacity transmissions to make appropriate frequency assignments. This approach is fairly equitable since it is about equally constraining on users of high-density and low-capacity transmissions, not denying either any particular major segment of the spectrum. It would be very effective, producing on the order of 10 dB lower isolation requirements. Moreover, it would siqnificantly facilitate the selective acceptance of higher interference as discussed above. POST-PLANNING COORDINATION - A fundamentally different approach would be to acknowledge the benefits of coordination as an effective tool to produce high orbit utilization efficiency and to provide for it in planning. Under this approach, reservative assignments would be made by electromagnetic separation criteria which are intentionally insufficient to guarantee acceptable interference for all assignments under all possible technical and operational conditions, in order to achieve a R2 B Eii Figure 6. Matching of Interfering (A) and Interferedwith (B) TransmissionTypes to Minimize Isolation Requirements. III assume co-frequency assignments of the most interference producing and the most interference sensitive transmission (I). Planning under a post-planning coordination arrangement could ignore the worst situation and seek one which would allow many but not all carrier arrangements to be protected (II a or b). In general, coordination would then not only allow the avoidance of situation I but could even lead to situation III which represents the tightest possible "packing" of assignments. Empirically, the difference between situation I and situation III lies in the range lo-20 dB from frequency assignment considerations alone and may be even greater when other elements contributing to inhomogeneity are taken into account during coordination. A 20 dE reduction of required isolation produces at least a 6-fold orbit utilization efficiency increase; 10 dB represent a 2.5-fold increase. PRACTICAL CONSIDERATIONS - The potential need to coordinate reserved assignments at the time of implementation has a number of practical consequences: _ ..__~_._ _----- - 582 - first, it would be necessary to develop two electromagnetic separation criteria: one by which assignments are made (assignment separation criterion), and another one by which it is determined, when coordinating an assignment for implementation, which other assignments are affected (coordination criterion) i _ it must be decided how coordination is to be undertaken with assignments for which no detailed implementation characteristics have as yet been developed. One option is to ignore them since they would continue, for the time being, to be protected by the assignment separation criterion; - the nature and magnitude of the assignment separation criterion and the coordination criterion must be determined. While both could be AT/T criteria, there are at least two other ways in which these criteria can be formulated. The adoption of a prior frequency segregation arrangement would undoubtedly allow the assignment separation criterion to be fairly low (in terms of required isolation), thus producing a fairly high orbit utilization efficiency. However, since it still has to be capable of protecting assignments having undefined implementation characteristics, detailed coordination would still be capable of improvinq orbit utilization. Since coordination always produces the best orbit utilization at little or no implementation cost and since orbit capacity is a major concern, it should be in the interest of all to retain coordination provisions even under reservative planning. It would be absurd to replace a highly efficient orbit utilization method by one that is less efficient when the prime motivation is to guarantee access to the orbit by all nations. ESTABLISHMENT OF CRITERIA - It is necessary to establish assignment separation and/or coordination criteria, whether they are to be applied under reservative planning or under any other method to gain access for a network. As has been pointed out, continued reliance on the AT/T concept would be one option. The use of AT/T criteria would tend to allow complete freedom in the choice of technical and operational network characteristics, subject only to such constraints as may be generally adopted. However, two other methods have been described by which assignment separation requirements could be established. One relies on calculating, for any two real or hypothetical networks, values of "available isolation" which depend only on major design characteristics of each network (antenna gains, gain discrimination, noise temperatures and up link-to-total link noise ratio). Available isolation, compared with required isolation a) for all transmission combinations, b) for a reasonable majority Of tranSmi.SSiOn combinations, would yield a "coordination spacing" and an "assignment spacing" between networks' satellites. This could also be used under a purely regulatory regime (i.e., no reservative planning): the assignment spacing would be the spacing which networks would, if necessary, have to accept when a new network is introduced; the coordination spacing would identify all affected networks. - The other assignment alternative would follow the U.S. precedent: intersatellite spacings which, if necessary, would have to be accepted by all networks, regardless of their characteristics, are pre-established by a formula which qives suitab,le recognition of available satellite antenna discrimination when there is geographical separation between networks' service areas. Determination of affected networks, when a new one is introduced, would be by AT/T or isolation coordination criteria. CONCLUDING REMARKS - It has been shown that, - -in order to realize an adequate number of or- bital accesses to satisfy demand under a reservative planning regime such as may be considered at ORB-85, there is a distinct probability that a) certain constraints may need to be adopted, notably one relating to an a-priori segregation of high-density and low-capacity transmissions; b) there may be a need for coordination of an assignment, prior to its implementation, with existing networks. This casts doubt on the benefits of reservative planning vis-a-vis the alternative of accommodating networks as demand for them arises. To the burden of planning would have to be added the continued burden of coordination: at worst, and of grave concern, reservative planning would diminish orbit utilization efficiency relative to more flexible orbit utilization approaches. It is noted that the 1984 CPM for ORB-85, charged with developing technical bases for the work of ORB-85, did neither recognize this nor even addressed the matter in general terms. To prevent ORB-85 from making decisions for ORB-88 which are in conflict with technical reality, it must address the questions raised above. As a minimum, it should request the CCIR to undertake additional studies in the intersessional period, but would then have to leave enough leeway in its decisions to allow the findinqs of the CCIR to be accommodated at ORB-88. [1lResolution No. 3 (Final Acts of WARC 1979). r21 As seen from an "outside" location; here from a location on the geostationary satellite orbit. r31 Identical in their modulation characteristics and noise budgets, not necessarily their required powers. r41 The apparent (topocentric) angular spacing of two satellites as seen from a point on the earth is generally greater than their geocentric angular spacing. [51 Only information necessary to allow the calculation of apparent noise temperature increase to characterize the potential interaction between networks. - 583 106 - R3 DEFORMABLE LATTICES FOR EFFICIENT FREQUENCY MANAGEMENT Andrzej H. WCJNAR Warsaw Academy of Technology Warsaw, Poland summary Congestion of radio spectrum calls for frequency reuse in large broadcast and mobile-radio systems. Theoretical network planning relies - since decades - upon lattice models with spatial and spectral regularity. Here, more realistic and efficient models without geometric regularity are introduced. Classical lattice theory is supplementedwith engineering criteria of coverage and compatibility.In eq. /6/, admissible displacementsof stations in the lattice are determined. Deformability of actual lattice increases with spectrum occupancy, represented by the number n of channel sets. Small mobile-radionetworks /n<12/ exhibit reduced deformability because of adjacent-channelinterference. Extended concept of deformable lattices disregards topological 9esselation" rules, as given by the set of "rhombic numbers". Various natural numbers n> nmin can be matched to actual topography and data of stations, provided their spacings remain within the bounds in /6/. Thus, spectrum occupancy may be reduced. 1. Introduction In large broadcast and mobile-radio systems, frequency reuse is now imposed by spectrum congestion. Relevant planning of compatible networks relies customarily - since decades - upon lattice models. So far, of [1,2], regular ar- rays of cells are exclusivelyused; consequently,very restrictive assumptions from classical lattice theory [3,4] are not removable. This author submitted some time ago [!!I] the idea of extending the lattice theory with due attention to principles of radio engineering.Step by step, cf [6], theorems and application rules have emerged. By now, advanced network planning with elements of technological design can be founded upon deformable lattices, a novel tool in spectrum management. 2. Regular and irregular lattices Conventional lattice models of radio networks are bound to be completely regular in both spatial and spectral domain [1,2,61.Then, the primary array of fixed /broadcast resp. base/ transmitting stations is characterized by the shortest distance between adjacent stations /modulus M in Fig. I/. In the sublattice of cochannel stations the least spacing is called coordination distance, D. There are altogether n such subsets, and a cluster of contiguous cells with different frequency assignments consists of n cells. Semialgorithmicrules of channel assignment in regular lattices are explained in [1,2]. The basic, structural equation of a regular lattice reads [6] M.. D/\r; /I/ - 584 Complete regularity is achieved with n = a2 + ab + b2 /2/ Here, a&l; b>,O are natural numbers. Note that the set /2/ is considerably larger than the set of "rhombic numbers" in n,2]. The values of M and D depend upon radio-engineeringconstraints. Fig. 1: Sketch of a regular lattice. The cluster consist5 of n=7 hexagonal cells; natural numbers denote frequency channels /reused/. In any actual radio network , the topography of fixed stations is irregular. In most cases, the array can be modelled by deforming only the geome&y of a regular lattice without affecting its spectral struoture. Evidently, see Fig. 2, such quasiregular lattice is characterizedby two finite sets of Mi and Di distances. Unexpectedly, compatible deformations can still be simply described if lattice theory is supplementedby engineering criteria. Let us now summarize this approach, initiated by this author in [5j, evolved in [6] and broaden&d here. - 3. Radio ensineerina constraints Along with topological "tesselatioff rules reflected in /I/ and /2/, radio networks have to satisfy coverape and comoatibilitr criteria. Coverage of a regular hexagonal cell by a transmitter in its centre requires the service radius d, to exceed dimension c /Fig.l/. Fig. 2: Sketch of a deformed lattice with irregular geometry. Mi, i I 1,2,... denote spacings between adjacent stations. Spacings between adjacent coohannel stations are denoted by D1' i - 1,2,... The coverage rsquirement leads to an uoner bound for the modulus T IQOXPf%l /3/ With conventional /static/ approach, ds depends upon the median value of the received field strength. In advanced probabilistic analysis, see e.g.[7], ds is the interference-limitedservice range. With frequency reuse, the cochannel compatibility constraint can be presented generally as Dmln p fdi/*dsq /4/ 106 - 585 - where di stands for the interference radius/range. Interpretationof /4/ depends on the operations in the system considered. For instance, see [6], in simplex mobile-radionetworks di is determined between base stationsand the dsterm in /4/ is omitted. In general, di and ds can be considered as independent variables and Dmin as a dependent variable, all with known values. Eq. /4/ combined with /I/ imposes a lower bound for Mi in compatible lattices /5/ Mmin = DminNQ Thus, the Mi set in any lattice is bounded by coverage and compatibilitycriteria n-"2*[di /+ds/lbMinmin can be employed in deformed lattices, see later; Considering only the cochannel compatibility is not sufficient. In mob% le-radio networks with small nC12, adjacent channels in adjacent cells are unavoidable. Then, as shown by this author in [63, another more restrictive lower bound for M exists Mmin %d s /a/ Evidently, small clusters can be less deformable because of adjacent-channel interference.Our recent research on small clusters with 2 buffer rings shows that /8b dominates over /5/ only if /9/ This case is also shown in Fig. 3. By means of /6/, every actual structure can be checked and - in case of need - readjusted, e.g; by changing the elevations of transmitting antennas. Manoeuvring with the n parameter can be optimized; thus avoiding exoessive spectrum usage. Od, 4, Extended concept of deformable lattices Fig. 3: Admissible variation of the M /n/ values in the hatohed area; Regular lattioe is represented by one point 'M/no/i In compatible deformed lattice Mmin /r+,/. In this context it is noted that there is in any case a bias in the type of location qualifying for measurement in the assessment of broadcast Television services are planned coverage. for viewers at home, and although radio services must also be available for listeners in cars, the majority of the radio audience Therefore interest is too is at home. concentrated upon populated areas, where the size of buildings will play a major part in deciding the overall distribution of the field. Many observations have shown that the strict concept of a Gaussian distribution applies only to very small movements of the receiving aerial, say within a distance of ZOh, An important conclusion which emerges from the examination of present measurement techniques is that whilst a well-designed programme of measurements can reveal much information about a service area and the probability of providing reception to stated proportions of the audience, inadequate methods can easily produce serious and misleading errors. The Prediction of Field Strength There are two basic methods of Firstly there are predicting field strength. the propagation curves already mentioned, which are based upon the statistical analysis of many thousands of past field strength measurements. These are simple to understand and easy to use, but it has already been mentioned they are not suitable for predicting Another serious detailed local variations. problem is the inaccuracy which must be expected when using any statistical curve they cannot take account of the unique In an features of each propagation path. examination by the BBC several thousand measurements were compared with predictions obtained using the Recommendation 370 curves, The and the distribution of errors analysed. standard deviation of this was just over 13dB. The majority of this error can be attributed to difficulties in assessing spatial Past results reveal that fading variation. ranges can be fairly accurately predicted, although here again there is a need to improve - 589 the form of statistical analysis used. A second and more precise means of prediction takes account of the terrain With this details of each propagation path. information, attenuation along the path can be calculated, and field strengths estimated with The terrain details can be good accuracy. acquired by producing a profile for each path using suitable topographic maps, which may also provide information about buildings and trees at critical points along the profiles. Alternatively a national terrain data bank can be built up and profiles extracted from this. The accuracy of the prediction is closely related to the density of the data, and this will be illustrated during the presentation of Several organizations have this paper. developed field strength prediction programs for use on a compquter, and the BBC system is a Over the past 20 years or typical example. so this program has been progressively developed, especially for use in connection with UHF work, and comparison with measurement produces a distribution defined by a standard This can be deviation of about 5.5dB. compared with the result of 13dB obtained However, using the propagation curves. development work on this program is far from complete. It is less accurate at lower frequencies, and as mentioned above its accuracy in any case is limited by the data supplied, which in turn dictates the design of Prediction for low receiving the method. aerial heights is inadequate, although it is hoped that work in connection with the measurement of the influence of local clutter can be used to provide further improvement in this area. The Result of Improved Techniques Three obvious benefits emerge from the use of more precise methods of field strength assessment. The first is very tangible greater accuracy allows the transmitter requirements to be precisely stated, resulting in financial savings. Secondly, better coverage is achieved. Thirdly, there is increased economy in the use of the RF spectrum. The latter can be illustrated by considering the case of UHF broadcast planning in the U.K. Planned to provide four programme coverage, the network is based upon 50 high-power stations - which were planned 20 years ago using the basic CCIR propagation curves. Together these covered about 90% of the population, and a further 800 or SO low-power relay stations have been needed to complete the national coverage. Detailed prediction has been of considerable importance in the planning of the relay network, both to ensure the required coverage was achieved, and to avoid the very real and constant risk of interference to existing services. In any case the scale of the work demanded computer-aided analysis - a full appraisal of the interference situation on each UHF channel involves more than 100 transmitting stations and the prediction of several thousand propagation paths. Thus a complex computer system has been built up around the field strength prediction program 107 R4 - not only to provide information about service area fiel.d strengths and interference levels, but also to present the planning engineer with detailed analyses allowing him to concentrate It has been estimated that upon decisions. without the improved prediction technique the final population coverage of the U.K. television service might have been 3% (1.5M people) below the present level, and there would also have been a substantial increase in the number of people suffering co-channel An example of the impact upon interference. the planning of a part of the UHF relay station network will be shown during the presentation of this paper. Future Work The studies so far completed by the BBC have demonstrated that a field strength prediction technique has been developed which can compare favourably with the accuracy achieved by measurement in the determination and planning of broadcast service areas. Immediate objectives are to improve this work so that it can be used at lower frequencies, specifically those in the range 88 to 108MHz - where currently much planning is taking place in the expansion of the VHF/FM However, it is suspected there may networks. be a limit which will obstruct the attainment at the lower frequencies of the prediction It seems likely accuracy achieved at UHF. that the ground reflected component is increasingly significant at frequencies below about 150MHz, and the present technique may A new and not be able to deal with this. more complex approach might be needed, but inevitably this would require more detailed data, and this is a questionable investment. It must be remembered that about 80% of the receivers likely to be used in these broadcast service areas will be portable and car radios, and field strength values will need to be translated in terms of probability of coverage Similar to the different types of audience. considerations apply to mobile radio systems, which are currently undergoing rapid growth, and improvements are being considered within the limits of the present prediction program which will allow more meaningful statements to be made concerning the reception quality of these services. This is an area of particular interest to both broadcasters and mobile radio users, because the need to share the spectrum efficiently demands the best planning techniques. It is hoped therefore that the program will meet the needs of both interests. The advantages of the detailed path loss predietion methods are clear, but it is acknowledged that at this time they cannot be internationally adopted. Few countries in Europe have immediate access to a terrain data bank, and the manual preparation of profiles can be a time-consuming and expensive procers. Thus it is reasonable to expect that the CCIR Recommendation 370 curves will remain in widespread use for many years to come. Certainly they will be the basis for frequency planning and international spectrum management. But it is also clear that steps are now possible to improve the application of - these curves for particular circumstances, given more informationabout spatial and temporal conditions. In essence, the whole prediction process should be a logical sequence, starting with the basic propagation curves, and yielding greater detailed accuracy as more informationabout particular paths and conditions becomes available. The confirmatorypart played by measurement,and the type required to confirm the predictions must also be defined. The fundamentalcurves and certain correction factors could be applied manually to provide simple, quick assessments,greater precision would require investment in more sophisticatedfacilitiescomputer systems, data banks etc. The presentationof this paper will run through the phases of the proposal, and will illustrate the impact that this would have upon existing CCIR texts. An important objective of such work would be standardization,particularlyin the development of prediction programs, to facilitate internationalco-operationin the future. In this way is is hoped that the positive advantages of more precise prediction,demonstrated in the U.K. and elsewhere, could become widely available with consequent benefits to spectrum utilisation. 590 - Acknowledgement The author gratefully acknowledges the permission of the Director of Engineeringof the BBC to publish this paper. References 1. "A Comparison of SANDELL, R.S. Standards Used to Plan UHF and VHF Networks". Royal Television Society Journal Vol. 12 No. 7 Autumn 1969. 2. Recommendation370 Vol. V. Propagationin Non-Ionized Media InternationalTelecommunicationsUnion, Geneva. 3. LEE, R.W., CAUSEBROOK J.H., SANDELL, R.S. "An Investigationinto the Prediction of Field Strengths". BBC Research Department Report 1970/26. 4. Computer Prediction of Field Strength A Manual on Methods Developed by the BBC for the LF, MF, VHF and UHF Bands: BBC Research Department. - 591 108 - R5 OPTIMUMFREQUENCY ASSIGNMENTSTRATEGIES FOR RADIO CELLULARSYSTEMS G.A. De Couvreur Department Ottawa, of and M.C. Delfour Communications Ontario, SUMMARY It has been established that it is possible to achieve ideal spectrum utilization for radio cellular systems with a maximum frequency separation strategy and with an intermodulation-free assignment strategy. In the complex intermodulation-free case, the proposed method, which is based on necessary and sufficient conditions, reduces the calculations to the level of one cell and provides the means to select the intermodulation-free strategy with the highest achievable The method can easily frequency separation. be implemented on a computer with a simple A simple example is used efficient program. throughout to illustrate the concepts involved. CANADA strategies are presented in this paper. CONDITIONFOR IDEAL SPECTRUMUTILIZATION Let us consider a cellular radio system containing M cells, where N frequencies have to be assigned to each cell, without frequency re-use, from a set of contiguous frequencies. Since a total of MN different quencies is required, it is clear that spectrum utilization will be achieved, if the necessary particular constraint, contiguous frequencies does not need to tain more than MN frequencies. c = (fa,fl,...fMN1), Af = fn+l fO 2i - j # k MAXIMUM FREQUENCY SEPARATIONSTRATEGY The strategy will starting with Let us assume maximum frequency be established in the derivation of that the smallest separation two steps, an upperbound. frequency - 592 - separation AF between the two closest frequencies in any ce.11 could be larger than the number of cells M, and could be equal to M+l for Then, in the cell containing the instance. frequency would be frequency k, the closest k+M+l, and the M intermediate frequencies would have to be assigned to the remaining M-l cells, on a one per cell basis in order to satisfy the hypothesis AF = M+l; this is Therefore: obviously impossible. The frequencies in the second be, with a shift of three: AF 5 M The frequency can then be represented (5) This upperbound can all successive frequencies in constant frequency separation frequency distribution is as 0, 1, 0 + M, l+M, Mll, (M-l) be achieved if a cell are at a equal to M. The follows: . . . . 0 + (N-l)M . . ..l (N-l)M M, . . ., (M-l) i. (N-l)M 0 1 cell M-l INTERMODULATION-FREE ASSIGNMENTSTRATEGY The derivation of a strategy that achieves ideal spectrum utilization without intermodulation relationship within any cell is a complex problem that requires extensive mathematical treatment beyond the scope of this paper. Therefore, the presentation of the method will focus on the concepts rather than on the theory. The fundamental concept was to reduce the problem of distributing MN contiguous frequencies among M cells to the choice of an appropriate set of N frequencies for a first cell, from which the frequencies to be assigned to the other cells would be obtained by successive shifts, so that equally spaced frequencies in intermodulation relationship would be systematically in different cells. A simple example of 4 cells with 3 frequencies each will be used to describe the concepts. In this case, the 12 frequencies are first distributed among three subsets containing 4 equally spaced frequencies so that each cell would be assigned one frequency from each subset. These subsets are: S] = 11, 4, Fz = 19, 13 = 1, 81 Fk = ((sn distribution by: + kN)Mod MN , 0 ( method k < Ml (6) 111, (7) Where s are the frequencies for the ?irst cell: selected OLn Cn , 2 (11) - 2sn 0 , in the above example, For instance, Fo = {1,3,11) does not satisfy Condition (11) since 11+3-2.1 = 12 and Cz = 3 > 0 = Cl. However, the sequence F. = {0,2,7) meets all conditions. Finally, it has also been established that the minimum frequency separation between the two closest frequenci.es in any cell can be calculated directly from the set of N elements in the first cell. Theorem 3 The minimum frequency separation between the two closest frequencies in any cell is given by: (AFo min, MN - AFo max) I , (12) where: AF = Min 0,min AF 0 ,max = Max R5 CONCLUSIONS the set AF = Min 108 - (10) , # + MN if +s 593 (lSn,2 - ‘n,,]) (13) (lsn,2 - ‘n,ll) (14) With these expressions, it is possible to calculate AF for any set ~~ satisfying the conditions of Theorem 2, and to select the set Fo that guarantees the largest value of AF. The problem of optimum frequency assignment strategies for cellular radio systems has been investigated, and methods to achieve ideal spectsum utilization with theoretically maximum frequency separation or without intermodulation relationship within any cell have been presented. With these methods, it is very easy to obtain complete frequency utilization strategy is schemes : the maximum separation essentially trivial, and the method for the intermodulation-free strategy can easily be implemented on a computer with a rather simple program which is very efficient since all calculations are made on a single cell. Furthermore, it is also easy to write a program that selects the intermodulation-free strategy with the maximum achievable value of the minimum frequency separation, which can then be compared to the theoretical maximum. For instance, it has been found that 336 frequencies can be distributed among 21 cells, with 16 frequencies per cell with a separation of 17. In comparison, a frequency separation of 21 can be achieved with the maximum frequency separation strategy, at the cost of intermodulation relationships. This gives a good idea of the trade-offs involved. - 595 109R6 - A SECONDGENERATION MOBILE SPECTRUMMONITORINGSYSTEM P. Vaccani Department Ottawa, of Communications Ontario, SUMMARY The Canadian Department of Communications has been using for several years mobile spectrum monitoring vehicles to gather occupancy data and has recently been engaged in the design of a second generation mobile spectrum monitoring system. This second generation system builds on previous experience and utilizes a microcomputer to automatically control the frequencies to be monitored in osder to obtain The major differspectrum occupancy data. ences from the previous generation is the availability of processed and unprocessed data concurrently, the capability of automatically controlling and recording data of additional equipments that would support othes spectrum monitoring functions and the capability to generate all reports with the on-board computer. INTRODUCTION Currently the Department has a number of mobile spectrum monitoring vehicles which are used operationally to gather occupancy data automatically for the land-mobile bands at twenty monitoring sites across Canada. The monitoring format for obtaining occupancy data was jointly developed by the Department’s Communications Research Centre (CRC) and Spectrum Management Systems (SMS) and described in the literature (1,2,3). These systems were obtained in the mid 1970’s from SRI International and modified substantially in the ensuing years and are now approaching the end of their life cycle. Recognizing the need for replacing these systems the Department proceeded on the design and implementation of a second generation Mobile Spectrum Monitoring System; hereafter, referred to as MSMS. The MSMSwas to be designed using the departmental research and operational experience with the first systems; yet, taking advantage of the advances in technology in particular in the microcomputer area. This paper deals with the characteristics and functions of this second generation Mobile Spectrum Monitoring System. CANADA DESIGN OBJECTIVES During the feasibility phase of the design of a second generation MSMSa number These of design objectives were fixed. ob j ectives were : The microcomputer on-board the MSMShad 1. to be utilized to the maximum extent possible. channel occupancy and Channel histograms, other reports had to be produced at the end of the day and at the end of a full monitoring The microcomputer had to have the run. capability to be completely self-supporting without the need of an off-site computer. 2. For economic reasons, MSMShad to be capable of providing other monitoring functions besides occupancy monitoring. The MSMS had to be designed as a multi-purpose vehicle supporting the enforcement function as well as the occupancy monitoring. As other requirements would be identified in the future, additional equipment had to be easily incorporated under computer control. 3. The MSMShad to provide processed and unprocessed data concurrently for the same monitoring run. This would allow further research and analysis of the unprocessed data by programs using the on-board microcomputer or an off-site computer. 4. The MSMShad to be easy to use with particular attention given to the man machine interface and automated sufficiently that only one operator would be required. A diagnostic capability for fault identification and for carrying out preventive maintenance had to be included. 5. The system had to be installed in a removable enclosure so that the investment in the system would be protected from the higher wear rate of the vehicle and the removable enclosure could be moved from one vehicle to another. ' - 596 - Figure 1: MSMS System Architecture Length = 3.66 m Width = 2.29 m Height = 1.9 m .Access to Cab ccess Cover Figure 2: Exterior and Interior View of MSMS SYSTEM ARCHITECTURE Figure 1 shows the MSMS system architecture. The major components of the system are: 1. Computer and associatedperipherals 2. Monitoring receiver subsystem 3. Antennas and mast 4. Vehicle which includes the above components, air conditioners,heaters and power generator The first three components are described in more detail below and Figure 2 gives a pictorial view of the exterior and interior of the MSMS. Computer and Associated Peripherals The MSMS central processing unit is a DEC PDP 11/23 Plus with S12K bytes of memory and 64 bit parallel and IEEE 488 interface. The peripherals included are 2 x 20 Mega byte Winchester disks, a 40 Mega byte streaming cartridge tape drive, a CRT console and keyboard, a graphics display and a printer. The cartridge tape drive is used for bulk storage, transfer of occupancy data and for disk back-up. - 597 The MSMS programs run under the control of the DEC RSX-11M operating system, a real time multi-taskingenvironment. The MSMS programs are mainly written in FORTRAN and in ASSEMBLY language where speed of execution or device control is required. Receiver Subsystem The monitoring receiver subsystem includes a custom designed receiver and standard manufacturer supplied synthesizer, signal generator and modulation analyzer. The specificationsfor the custom designed receiver are summarizedbelow. Antennas and Mast To optimize the achievable coverage, the monitoring antennas are mounted on a pneumaticallyextendablemast. Eight sections telescope up from a stowed height of 2.286 m into an extended height of about 12 m. Three antenna elements are arranged on the mast. DATE: DETAILED FUNCTIONAL DESCRIPTION In order to obtain occupancy data and control other equipment, the MSMS has to perform the following major functions: 1. Acquisition of occupancy and enforcement data 2. Data display 3. Operator interface and control 4. Diagnostic capability Acquisitionof Occupancy and EnforcementData Frequency Coverage: [Band I 138-174 MHZ] [Band II 406-470 MHz] [Band III 806-870 MHZ] Noise Figure: 8 dB maximum across the frequency coverage Minimum Detectable Signal [in 10 KHz IF (SW)]: Band II -118 dbm Band I -121 dbm Band III -115 dbm Third Order Intercept Pt.: Band II +16 dbm Band I +16 dbm Band III +3 dbm IF Bandwidth: 3 KHz/l0 KHz Scan Rate: 2 ms per channel AVERAGE SPECTRUM OCCUPANCY ________________..___..-_.-_--___--~ EDMONTON 532645 SITE: 109 ~6. - The major function of the MSMS is to obtain occupancy data. This is achieved by listing the frequenciesto be monitored in scan tables, each specifying up to 1,000 channels or frequencies. Up to 10 scan tables can be selected for use in a single monitoring run. The selected scan tables are used in rotation with 600 passes, constituting a raster, being made through the current scan table before proceeding with the next. Amplitude samples on successivechannels are taken at a rate of 500 per second. A separate histogram is formed on each raster for each channel scanned and is analyzed at the end of the daily run to yield the threshold occupancy and the window occupancy. The other major function of the MSMS is to have the capacity to add other monitoring equipment under computer control. Initially,a modulation analyzer is being added to measure the frequency,modulation and the signal level for enforcementpurposes. REPORT 1133918 27-JUL-83 CHN (MHZ) P 25 50 I I 75 100 I 406.050 ******** TH 52% BS 24% 4”6.1c,” l **************t******************* ************ TH 88% BS 33% 406.250 *********** TH 82% ES 26% 406,350 l *****t******************************* *************** TH 92% BS 38% ,**************************t***t* 406.450 l TH 82% BS 24% 406.550 ******* TH 62% ES 22% 406.650 ************t*******t*t******** *****t******** TH 68% BS 34% 406.750 ******** l ****t*********t***** t******************************** ******* t******t********t******** t****t*************** **t*t****t****t****** 406.850 l ******* *****t*********t***** 406.950 l ******* .*t****************** TH 52% SS 24% TH 52% BS 24% TH 52% BS 24% TH 8s 52% 24% 407.050 ******** 407.150 l *‘***** TH 52% BS 24% l *t**t**t~*********** ******* TH 52% BS 24% r******,********+***** 407.250 l 407.350 tt******t***tt***t**t *****a** TH 52% BS 24% 407.450 ttt**t*************** ******** TH 52% BS 24% Figure 3: Average Occupancy Report 598 Data Display The MSMScan generate two types of displays ; those produced during monitoring and those produced after a monitoring session. The reports produced during the monitoring session are the Spectrum Activity display and the System Status display. The Spectrum Activity display is a real time display whose main function is to give the operator an indication of the activity and the validity of the data collected. The signal amplitude is plotted as a function of frequency on the graphics display unit. The System Status display indicates the mode of operation and the commands available in that mode on the console screen along with the date, time and disk space left. MSMShas three modes of operation: the command entry mode, the monitoring mode and the file preparation mode. In the monitoring mode the scan tables being used for the monitoring session, the scan table currently in use and the raster number are displayed. The reports monitoring session is on the printer or the be transferred to an produced when the complete are displayed graphics display or can off-site computer. A Daily Channel Amplitude Histogram report displays the channel amplitude histogram for a particular channel after all raster amplitude histograms have been combined. An Average Occupancy Report (as shown in Figure 3) displays average threshold occupancy and window occupancy for each channel scanned during the monitoring session. The Peak Occupancy Report is similar except the peak times are shown. displays basis. Operator The Mean Occupancy the mean occupancies Interface by Time report on an hourly and Control The operator interface function allows the operator to monitor and control the operation of the system via the console The operator can specify the freterminal. quencies to be monitored and the operational parameters to be used either from a tape cartridge produced from the SMS database or prepared on-line by the operator using the text editor. He can initiate and terminate monitoring runs, invoke various displays and print-outs and run the analysis programs. If he requires assistance the complete operational manual is available to him on-line on the computer and he can request assistance on any command or mode via the help facility. Diagnostic Capability A diagnostic capability is provided to verify that a particular unit within the Diagnostic programs system is functioning. are provided to test the computer system and The DEC 11/23 comthe receiver subsystem. - puter system includes diagnostic programs for exercising and verifying the correct operaDiagnostic protion of the computer system. grams are provided to attribute faulty operation of the system to the receiver, the synthesizer or the modulation analyzer or verify that all three are functioning correctly. CONCLUSION All of the design objectives have been included in the implementation of the testing was successfully MSMS. Acceptance completed during August 1984 and demonstrated that the system met the design specifications. The MSMSis currently undergoing field trials. REFERENCES [l] BURKE, M.J. and COYNE, T.N.R. Monitoring IEE Conference Land-Mobile Radio Usage. on Radio Spectrum Conservation Techniques, London, England, July 7-9, 1980. [2] AHMED, S.N., DE COUVREUR,G.A., McCAUGHERN,R.W. and RACINE, T. A Spectrum Management System for IEE Conference on Radio Spectrum Conservation Techniques, London, England, July 7-9, 1980. [31 DE COUVREUR,G.A., Canada. DROUIN, M., McCAUGHERN,R.W. and AHMED,S.N. Acquisition and Utilization of Channel Occupancy Data in the Shared Frequency Assignment Process. EMC Zurich Symposium, Zurich, Switzerland, March 1981. - 599 110 - Sl SELECTIVE INTERFERENCE IN HOME ENTERTAINMENTELECTRONIC DEVICES Henryk Cichob, Hubert Trzaska EMC Working Group of the International Amateur Radio Union, Region 1 Katowice, Poland Changesin the naturalelectromagneticenvironmentcausedby an ever increasingnumber of sourcesof radiationappliedin t&eCOmmUIIiCatj.on industry,medicineand household equipmentalter in turn the conditions under which home entertainment electronicdeviceswork. The sources are locatednear and nearerto inhabited areas / quite often they are locatedwithin these areas /# In this paper the authorscontinue their investigations,resented in the previouspapers,of t%e susceptibilityof home entertainment electronicdevices.In the paper selectiveinterferencecausedin radio- and TV receiversby external electromagnetic / EM / fieldshas been taken into account0However,the interference under studyappearsonly becauseof the receiversensitivity at frequenciesdependentupon the harmonicfrequenciesof the local oscillator. In the work typicalmethodsof sensitivitymeasurements have been applied,, The measurements have been made in the frequencyrange up to about 400 MH5. Inuring the measurements the levelhas been estimated of the interferingsignalat which remarkabledistortionsin sound and/ or in image were observedand the maximalsensitivityof the deviceunder test / WT / has been determined at severalharmonicsof the local oscillatorfrequency. The work has ermittedus to formulatesome conePusionswhich may be of interestto receivingequipment manufacturersr These conclusions, similarlyas thosepresentedin the previouspapers,show the necessity of applyingsome solutionstypicalof professional eauipmentin home entertainmentelectronicdevices.These are, 8,g.z - spectralpurity of local oscillator - linearRF amplifierand mixer - well tuned input circuits. Introduction The sourcesof electromagnetic radiationlocatedin inhabitedareas and in apartmenthousesare e*g.t transmitters of variousradio services / includingpolice,taxi,emergency and amateurradio service/, medical equipment/ RI!and microwavediathermy, lancetron/, small workshopequipment / inductiveheater,dielectric welder / and householdelectronic equipment/ microwaveoven,radiosand TV receivers/. These sourcesmay remarkablychangeEM environmentand disturbthe work of home entertainment electronicdevices,e.g.: another radio- or/andTV receiver,tape recorder, gramophone. Till now the susceptibility of home entertainment electronicdeviaes to undesiredsignalshas been taken into accountneitherby manufacturers nor by scientists.It shouldbe mentioned,however,that this problemhas satisfactorily been solvedin modern radiocommunication equipment.It is necessaryhere only to understand necessityand to adopt well known protectivemeans and use them in home entertainmentdevicesas well. In the previousworks Cl,&33 the authorshave shown that the above formulatedthesisis true with respect to widebandinterference. In the prssent work the authorshave tried to find not only reasonsof high susceptibilityto selectiveinterference but also simpleand inexpensive means making it possibleto improveimmunity of radio and.TVreceiversto selective interference. Some selectiveinterfe-, rence has been observedin otherdevices than radios / e.g.: phonograph, tape recorder,heart pacemaker,hearing aid /; these effects,however,are ratherincidentaland they appearat parasiticresonantfrequencies. Such frequenciesdependupon parasitic inductancesand capacitances and they are usuallyabove severalhundredb!H5. AB a rule these selectiveeffect8are qtitewell eliminatedby the means - 600 - applied outside chamber measurements -----_- to liquidate wideband effects. Selective interference which occurs mostly as signals passing through the pass band of IF stages of a receiver and which is a result of linear combination of frequencies radiated by a source and frequencies generated in the receiver itself, may seem to be easier to eliminate in comparing with wideband effects. It is, however, the first approximation conclusion. In fact it can be quite troublesome and its elimination difficult since the susceptibility of a device c6.nbe in the same level as its sensitivity to the signal being received. At the beginning of the work the authors had wide program of the selective interference measurements. During some introductory measurements an exceptionally high sensitivity of the investigated receiver was haphazardly found at frequencies corresponding to harmonics of the local oscillator. It was shocking and it made us limit the work only to intermodulation effects related to the loaal oscillator harmonies. Contrary to the case of wideband interference where susceptibilities measured in similar types of devices, produced by different manufacturers, were also similar - in the ease considered devices of the same type made by the same manufacturer differ considsrably. This allows us to conclude, that from the considered oint of view it ia not the design but Bts realization which i6 important, i.e. I tuning, control during production. Analyses and measurements were made from the point of view of the Amateur Radio Service / ARS /, and they should be representative for, or at least transposable to other services. The work presents result6 of measurements in the frequency range up to 400 MRa. In this range the majority of services of some importance from the considered point of wiev work. _L- !--- -- -- -_- -- __---_- -- -- I, -7 __-__A inside chamber measurements Fig.1 Test set for susceptibility measurements of receivers with ferrite rod antenna. f I SPL I - DUT - MAINS f- Fig.2 Test set for susceptibility measurements to RF voltages induced in IN and OUT wiring of DUT. vity of a DTJT/ at S/N = 0 / to interfering signal was observed, A complete description, including frequencies and field intensities, of the local BC and TV stations is given in Cll. The measuring methods applied are similar or exactly the same as those applied previously Cl'1 . Suscaptibility to EM field was measured in TEM cell and in standard EM field. Conventional standard transmitting antenna was applied to susceptibility measurements of radio receivers with built-in ferrite rod antennas / Fig. 1,/. The susceptibility to RI?voltage6 induced at antenna input and in other DUT wiring was measured as in Fig.2. All the radio- and TV receivers available for the authors were measured and no correlation between the type of device and its manufacturer was found. Neither was found a correlation between the repaired and neverrepaired devices / repairs were simply connected with exploatation of a device /. As all the measured devices were made available to the authors with no permission to introduce any changes in their circuitry, the presented results of measurements were obtained with the use of one device and verified with the use of another. The first one in a way represents all the measured devices as its sensitivity at local oscillator harmonic frequencies is intermediate between the worst / - IO + - 30 dB / and the best ones / 30 and more dB /. As the changes in the sensitivity due to the applied means were similar in both the reconstructed receivers one can suppose that the final conclusions are true. The measurements were made in the presence of a signal from loual BC or TV station and without it. In the latter case the maximal sensiti- All the measurements presented in the work were made in the Institute of Telecommunication and Accoustics of the Technical University of PlrOClaW. &asurin& methods - 601 In the general case an input voltage of the first stage of a recerver / usually mixer / is given by: = & Ai.cos 2riifit (11 i=O where : 9. - frequency of i-th frin@r its amplitude A= i A selective interference will appear if an arbitrary linear combination Of an arbitrary number of fringers / including their harmonic frequencies / is equal to intermediate frequency of the receiver: IF = nofo + &Ifl * .... Sl Results of measurements Basic theory u(t) 110 - (2) where : no, al ,.. = 0, +,1, f 2 I.. If IF is a result of linear combination of m input fringers and their amplitudes are quite low, then the amplitude of IF An is C5J : AIR = Ao/no/,Al/n,/..Am/nm/ (3) For the work of a superheterodyne receiver at least are necessary two frin ersg it is the first order intermodu!? ation. If we have another fringer it gives theoretical possibility of generation of infinitely large number of fringers interfering at IF. As can be seen from formula (3) the products of higher harmonics have the rapidly decreasing amplitude which should limit the number of fringers of practical significance / resulting in remarkable interference /. The above considerations allow us to calculate ,,apriori" frequencies at which selective interference can appear. This needs, however, some comments. As was shown in Figs.'land 2 measurements were made with the use of a generator as the source of a single frequency fringer. Irrespective of the purity of the generator spectrum it is in most cases impossible to foresee, basing on the above presented way, any product caused by the nearby transmitter and making interference in the RUT. In any transmitter, of quite modern construction, an output frequency is usually a result of combining two or more frequencies and their multiplication or division. All these signals, more or less attennuated, are represented in the output signal of the transmitter. At small distances from the transmitter each radiated miliwatt may cause interference / see [4J where interference made by a TV receiver oscillator is presented /. The results presented hereafter are related to measuremente of a poPUlar class, transistorieed, battery or mains-operated receiver. It is 0oUi.P ped with a built-in, ferrite rod antenna for.reception at long and medium waves and needs extsrnal s.ntenWS for short waves and OIRT FM band. The receiver is representative for all other receivers, measured during the work, as its sensitivity at the local oscillator harmonic frequencies is approximately intermediate between the worst and the best sensitivity measured in this work. The measurements were made by applying the "step by step,'method as the use of a sweep generator made it practically impossible to select and to identify particular fringers in the "jungle of fringers". The applied generator, of quite a good quality, had also some spurious fringers. This made the authors use the less advanced method of measurements, i.e.: accurate generator frequency measurement / with an external frequency counter / and the measurements with a calculator in hand for current identification of fringers. As the results of measurements of the maximal sensitivity / S/'lV k 0 / of the receiver at harmonic frequencies of the local oscillator are the . ._ _. most repeatable ones under any condit;ions they are presented in this paper. The results are shown as the relation of receiver Sensitivity im its working band to its sensitivity at undesired bands. This way of presentation seems to be picturesque and makes it possible to see directly from the figures t&e decrease of sensitivity at undesired bands due to the applied means and procedures. In Fig.3 the results of relative sensitivity measurements are showm. Fig.3a shows results of measurements in the long wave /LW/ range, Fig*3b in the medium wave /MW/ range and Fig.3~ in the short wave /SW/ range of a new, just bought and never-used-before receiver. It can be seen from these figures that the receiver sensitivity at undesired frequencies is sometimes almost equal to the sensitivity in the receiving range / curve 1, Big.3c /. All the curves show extraordinary high sensitivity of the receiver at local oscillator harmonics. In Fig.3 / broken line / the generaliz;edrun of the receiver sensitivity is shown as a function of the local oecillator harmonics order and frequency. Such a generalieatioa was used to compare the performances of the investigated receivers and to analySe the efectiveness of proposed and - [dBl 602 - ’ -10 -30 Fig.3a -40 wave range low!i -50 -60 SW 0 [dBl -10 ;,'a -40 z%Z" wave \ range 10 -50 40 -60, 20 IO 2015 IO 30 5o [MHz] loo MW Q) IdBl ' -10 !? k -20 Fig.3 Resultsof msasurements of a newly boughtreceiversensitivityat harmonicfrequenciesof the lo~l~scillator Indicationst 1. f, * IF, . o - IF, 3. 2f, + IF, 4. 3f, - II?, 5. 3f, + IF, 6, 4f; - DP 7. 42, + IF, 8+ 52, - ICIP, 9. 5f, + IF, 100 62, - n? 11. 6f, * II?. In the frequencyaxis amateurbands on 160, 80, 40, 20, 15 and 10 meters are denoted. appliedprotectivemeans+ From the resultspresentedit followsthat it is practicallyimpossible the use the 160 and 80 meter bands while receptionat long waves, neither160, 80 and 40 meter bands while receptionat mediumwaves any shortwave amateurbands / &%- ding 11 m citizenband and 7.5 m radiotelephonebaud / while reception at shortwaves. To improveattenuationof the receiversensitivityat its oscillator harmonicsthe first of all the reoei- - 603 -30 IdSI -40 -50 -60 -70 -80 110 Sl - at the oscillator harmonies. It is shown in Fig.4 / curves with **/. These changes in the circuitry of the receiver were introduced after its tuning and the curves show effects resulting from the two above presented procedures. To eliminate disadvantages of the use of a nonlinear mixer as well as troubles connected with optimisation of the stage playing two roles in the receiver these roles were separated. The original stage was adopted to work as the local oscillator and a balanced mixer of the SRA-1 type was added. As an ^ .. example or tne use such a design the results of measurements in the long wave range are shown in Fig.5. All the above mentioned changes in -20 id61 -30 -40 -50 -60 neither that changes applied are optimal. Capacitive couplings play an important role when attenuations at the level of above 40 dB are considered. Thus, provisional reconstructions can show only aualitative results. The results confirm the validity of the idea presented. Its realization, however, must be applied at the desigting atage where each factor may be considered and optimized. Fig.4 Results of measurements of the kiB1 same receiver as in Fig.3 -50 after tuning / curves with ' / and decrease of local oscillator -60 x amplitude / curves with *' /. ver was tuned once again and the re-70 sults are shown in Fig.4. The tunning has not changed the out-band sensiti-80 vity. However, the in-band sensitivity increased and as a result the sensitivity out the band decreased. -90 3 In the original design of the receiver the first stage is an auto-generating mixer. It has been found that its oscillations amplitude is too high. As a result of this and of nonlinear characteristics of the transistor, the oscilator voltage was distorted. Due to a change of the workin& point of the transistor and changes in the oscillator circuits the voltage was reduced by 2 + 6 dB. This has remarkably reduced the sensitivity of receiver ,j,I, ,I r (4...)"' <-9OdB \ 3"' 1 3 5 [MHz] '0 Fig.3 Results of measurements of the same receiver as in Fig.3 with separated local oscillator and balanced mixer of the SRA-1 typed Irrespective of the accuracy with which the optimal conditions of the input stage of the receiver were chosen it may be said that after these reconstructions the receiver was not susceptible to the nearby workin amateur radio station using FT.1%1 Z - 604 transceiver and TH3MK3 antenna. Both the Presented results of m@aSUrements and this qualitative liedi permit US to suppose that suggested way may be considered another step toward the design of a new generation of radio receivers destined for aide use and able to work in electromagnetic environment at the level permitted by national or international regulations for inhabited areas. Conclusions International Amateur Radio Union as well as national unions in majority of the developed countries try to indicate the necessity of thinking nowadays on the future situation in the field of compatibility of electric and electronic devices* The engagement in this subject is supported by experience of the Union concerning the coexistence of home entertainment electronic devices and located-near-them amateur radio stations as well as the related problems and conflicts. To eliminate further, more serious, conflicts between users of home entertainment electronic devices and previously mentioned services utilizing EM field generating eauipment it is necessary to think about this problem at present. Various effective protective means are knowlaand widely applied in proffesional eauipment, The only problem is to simplify them, adopt and apply in the!devices concerned, The authors have shown that this is fully possible in the case of wideband interference. In this work the problem was considered of a receiver sensitivity at frequencies corresponding to the local oscillator harmonics. The observed results are in most cases caused bY " technical insouciance *Iof manufacturers, namely : - receivers are carelessly tuned, - local oscillators are designed not for optimal work of the mixer but for easier oscillations. - Hence, the following factors should be taken into account from the considered point of view : - spectral purity of the local oscillator, - linearity of the RI?amplifier and mixer, - well designed and well tuned input circuits, - tuned RF amplifier recommended. All the factors are well discussed in the literature and, except for RF amplifier, do not affect the cost of a device / the only added here part balanced mixer - is available for 50 cents ! /. The above considerations allows us to state that it is possible to improve receiver performance in a simple and inexpensive way in considered case as well. The authors would like to express their gratitude to braves who decided to permit us to measure receivers of their own* Their favor have made this work possible. References Cichon H., TrBaska H.: Selected susceptibilit problems of the aeneral use e3: ectronic devices. &oc. 3rd EM2 Symp., Rotterdam 1979, pp* '+23-128. CichoraH., Trzaska H.t Susceptibility problems of home entertainment electronic devices. Proc. 4th EMC SYmP., Zurich I981, pp. 295299. Cichon H., . Trzaska H.: Immunity _ . .._ improvement of nome entertainment electronic devices. Proc. 7th EBdC Wroclaw 19&t, pp.'lO15-1024. [4jl?:%%ik M ., Slniezko0.: Television receiver local oscillator harmotic radiation: statistical data, Proc. 7th EMC Symp., F$roELPg4* PP. 731-7404 ., Walaszek M. Analysis 151of moaliloear phenomena occurimg fa the presence of narrow band additive interference. hCoc. 7th EMC Symp,, Wroclaw lVS4, pp. 879-830. Ills2 - 605 - CO-CHANNEL INTERFERENCE IN AN ON-BOARD PROCESSING SATELLITE Ikuo OKA, Kazuhito ISHIDA, and Ichiro END0 Department of Communication Systems, University of Electra-Communications Tokyo, JAPAN ABSTRACT This paper presents co-channel interference on-board techniques in an cancelling processing satellite system using orthogonal The techniques based on mean polarizations. criterion and error square (MSE) convolutional coding / soft decision Viterbi decoding (SDVD) are introduced on board. The bit error rate (BER) is derived for the mixed and an of a desired signal detection co-channel interference when the undesired techniques are applied to binary phase shift keying (BPSK), and is compared with that of interference. for the compensation no Performance results, obtained theoretically, show that the techniques of both the MSE processing and the SDVD realize an excellent improvementsof the BER, and the improvements become significant in a large interference environment. improve the system performance independently, because the performance of SDVD depends on the probability distribution function (P.D.F.) of an amplitude of the MSE processor output. Therefore, the compound effects of both techniques on the BER seem to be produced. A model for the on-board processing satellite system employing BPSK is presented in section 2. We derive an explicit expression of the P.D.F. of the amplitude of the MSE processor output in a form of the Hermite polynomials in section 3. The BER of SDVD is derived with the P.D.F. in section 4. Section 5 provides the numerical results of the BER and investigates the interference cancelling effects due to the MSE processing and the SDVD. 2 SYSTEM MODEL 1 INTRODUCTION 'CONVOLWIONAL ENCODER CHANNEL 1. A recent problem of the rapid increase of communication demand that faces in satellite services leads to the effective use of frequency resource. One of the solutions is the frequency reuse by means of a dualpolarization system. However, channel distortions due to atomospheric propagation anomalies such as rainfall cause a interference at receiving interchannel terminal. In this paper, co-channel interference techniques are introduced in an cancelling processing satellite which ’ on-board expected as a hopeful1 satellite scheme. ;: the on-board processing satellite, various including a techniques baseband signal become available on-board the processing practical satellite. As effective and techniques for reducing the effect of cochannel interference, we consider a compound strategy of mean-square error (MSE) signal processing and soft decision Viterbi decoding (SDVD) in binary phase shift keying (BPSK) systems [l]. The MSE signal processing has been shown to be a useful scheme to reduce the interference by Nichols [Z]. On the other hand, the SDVD yields the significant bit error probability (BER) improvements [3], in future and is expected to be used satellite systems. These two techniques never CONVOLUTIONAL ENCODER CHANNEL 2. EARTH STATION -- - "1 LUP-LINK PSK MODULATOR 4/ r---_------. RECEIVER ____________~ - SATELLITE _____/ Fig.1. On-board processing satellite system. Fig.1 shows the model for the on-board processing satellite system. All symbols in Fig.1 are denoted in a complex envelope form. In what follows, we investigate the system in Channel 1 and 2 a lowpass equivalent form. the same frequency band, and are utilize separated into two independent information orthogonal using bearing signals polarizations. At the transmitter, the input are convolutionally data (Xl, h2(=0,1)) encoded. The coded data are sent to the BPSK 606 modulator. In the up-link, the transmitted aPSK signals Si=exp( ihln), Se=exp(&R ) are corrupted by the channel anomalies of both TII, T22 in the direct path and Tvz,T~I for the co-channel interference, where T~I ,TI~ coefficients are the channel ,T~I ,T22 corresponding to the attenuations and / or phase shifts. Thus, on-board the satellite, received signal rl and r-2are expressed as complex the asterisk denotes the where conjugate, and the relations of ISm]2=1 and (m=1,2) are used. Determination E[ n?+,]=2p2 of the channel coefficients ( Tii ) leads to the choice of thzoptimum v$.ghts ('&J;& ]. are obtained and S2 The MSE outputs Sl immediately. Now we let Xi =T22 T21= = lexp( 01 b exp($) (6) T:!=bexpk@) the where nd and n2 are respectively, mutually independent gaussian noises with variance 2p2 . The on-board MSE processor makes the mean square error between the transmitted and received signals~minimum,,and and S2 . yields the MSE processor output S1 are and $2 The estimated values 31 converted into digital forms using Q-l soft decision thresholds. Then, the obtained Qlevel soft decision data is sent to the transmitted decoder to get the Viterbi information. 3 , where b corresponds to an amplitude of the interference, and phase offset (h between channel 1 and channel 2 is distributed uniformly on [ 0, 2X 1. Eq.(6) implies that channel 1 and channel 2 are symmetric with the same signal-to-noise power ratio (SNR) and the same signal-to-interference power ratio (SIR). Since the same approach is valid in the performance analysis for both channels, we go on the analysis of channel 1. For St =exp(jO)(symbol ~1 =O), we can derive the MSE output-$, from eqs.(3),(4), and(5). (7) PROBABILITY DISTRIBUTION FUNCTION OF MSE PROCESSOR OUTPUT where The MSE criterion is based on minimizing the error e(W): Sef = (7-j i+= where W is the vector of weights to be determined, Sm is the transmitted BPSK signal, and Frn is the estimated signal reformed by the _MSE prEcessor. The MSE processor outputs Si and S2 in Fig.1 can be expressed by linear combinations of the received signals r-1 and r2 as (3) and z&=tiln +2322r2, (Ilef= tlt2&-ti,+bZ(t!+2~2-I 1 (l+b2+2u412 -4bZ 4@b (l+bz+ 2u-212-4bz expC1N ’ (8) f (9) 2)nl+b(b2+2~z-11exp(-~9)n2 , (‘+*~;;b’+ 2@)2,,&b2 (10) We note that the MSE processor output 3 consists of the signal component sef , the interference component &f , and the noise component 0l.af the probability distribution fuActr&dr;geD.F.) of the MSE output $ The probability distribution function P.D.F. F(k) is expressed as where (z(rii ) is the weights to minimize the MSE in eq.(2). With the aid of orthogonal principle [4], the following equations for the optimum weights ( '&J-Y ) is obtained. where e+(t)= &@q)(-w2)dI.f, , (12) the absolute value of ief k is thl? is the variance of fief th: signal normalized by amplitude amplitude, and E;p(. ) denotesa;i;eexpgtati;g with respect to the phase expand eq.(ll) with the relation of - Ills2 60.7 _-. represents the Hermite where polynom~!!~s'o$x!~derL-l. Then, averaging the t we resultant equation with respect to @ The bit error rate (BER) is derived for the SDVD using the MSE processor output in this section. Fig.2 shows the concept of Q-level are the soft decision, where go, *. . . .,gQ decision thresholdswhich are spaced equally and to be optimized by minimizing the BER, and where In(n=1,2;* l,Q) represent symbol metrics. The soft decision probabilitycan be calculated by hbi p(n)=$ p(r) = F-ck,&- J (14) dr F(len~, n=1,2;..,~, (18) where the effective SNR p and the effective SIR r are defined by gn.=k*Sq 2 E%f I p= len- {(d-Q+l)+p(~+-l))2 (151 = ck{(ch-(3+1)2 + p(d+(3--1,2} iw-P+l)+P(d+g-1)2 = where CX= - i 2 &* : inverse of SNR (17) and b2: inverse of SIR p = (19) (n=2,~~~+J-1 ) and where kn is the normalized threshold is the space between the neighboring n thresholdsexcept for go and gg . We show the derivation of the BER of the SDVD using soft decision probability p(n). For the convolutionalcode with rate qu/nu, the BER Pb is tightly upper bounded by [5] (20) where d is the minimum free distance of the code, Cq is the total number of erroneous bits included in all incorrect paths whose distance from correct path equals to q, and the coefficient Cq is obtained by the generating function T(D,N) of the code. dT where D correspondsto the free distance , and N corresponds to the error bit. The first event error probabilityPq can be derived from the soft decision probability p(n) as follows, We let M be the metric of correct path, and M* be the metric whose distance from correct path equals to q . Then, Pq is given by BIT ERROR RATE OF SOFT DECISION VITERBI DECODING density) ? ived 1.C”‘ threshold VP&H 81 decision : ” soft decision probability :P(“) soft , dT( D,N) . (Prob. kn-1=A, t (16) 4cq 4 where F(*) is the P.D.F. of the MSE processor output, and EgiO” Q P(Q) Q-1 2 80-00 -es_ ____ , P(Q-1) ---- ---- PW ,P(l) symbol 0 IQ Iq_, .--- ---- I, l;l symbol 1 11 I* ___- ---- &, Ig metric: Fig.2. Concept of Q-level soft decision. where me and rnh* are the symbol metrics of the correct and the incorrect paths and & runs over q symbols wherein these two'paths differ. Fig.3 shows the probability density function pm(m) of mQ*-me in eq.(22). In Fig.3, the area of each impulse is the soft decision probability p(n) (n=1,2," * * ,Q). - 608 Since mQ '%-mLare statistically independent the probability density among all L function p (m) o; $1 (mb'k-mL) is obtained by the q-folds convolution of pm(m) [4]. Then we get where 5 is the positive number which is sufficiently small relative to the symbol metric IL. Substitution p(n) into eqs.(20) and (23) yields the BER of the SDVD. - The derivation of the BER of SDVD with the MSE processing is summarized below. (1) For the given channel SNR l/d and SIR Up, we derive the P.D.F. F(e) using the effective SNR P , and the effective SIR J- ...a * eqs.(14),(15), and (16) (2) The soft decision probability p(n) is given from the P.D.F. F(k). ""eq.(18) (3) The first event error probability Pq is derived by the probability density function of metrics with p(n).s**eq.(23) (4) The BER is obtained by Pq and the parameters of the code. **** eqs.(20) and (21) 5 NUMERICAL RESULTS PnI(1n) Fig.4 shows the BER of SDVD for Q=2 and 4 numerical In using the MSE processing. P(l) calculations of SDVD, the convolutional code A with rate l/2 and constraint length 3 is examined. The SDVD demodulation encounters however, the optimum threshold problem, P(2) in eq.(19) is optimized threshold space 0 my minimizing the BER. Since utilization of ’ . . rate l/2 code requires the bandwidth spread, P(a-1) . ’ . P(Q) nre reduce the SNR by 3dB in the SDVD case. T T ,rnz-f?e. For comparisons, we depict the BER without compensation for interference in Fig.4. It 2) (1 Q-11) (I l-1 $ (12-IQ-I)----o---*(1 Q-1-I is apparent from Fig.4 that excellent improvements are achieved by the techniques Fig.3. Probability density function of metric. of MSE processing and SDVD. To be specific, the BER improvements are remarkable in case of the low SIR. For instance, at BER=lO-5 and SIR=5dB, the techniques can reduce the required SNR by 2.8dB for Q=2 and 5.9dB for Q=4 in comparison with the BER without the compensation. Therefore, in this region, the MSE and SDVD techniques yield the SNR withoutsignalprocessing margin of 2.8dB for Q=2, and 5.9dB for Q=4. ----withSDVDforQ2 usingI%E The effects of MSE processing and SDVD are -.-.-.-.tith SDVD for Q=4 usingMSE examined independently in Fig.5 for MSE' and Fig.6 for SDVD. We also show the BER without the compensation in Fig.5 and Fig.6. First, from Fig.5, we note that the MSE processor has the SNR margin of 2.ldB for SIR=5dB, and 1.3dB for SIR=lOdB, and 0.6dB for SIR=lSdB, at BER-10-5. According to the increase of SIR, the improvements produced by the MSE processing become smaller. This fact agrees with our intuitive comprehension. Since the MSE criterion is based on co-channel interference cancellation, the MSE processing is effective at the low SIR. Next, we consider the effect of SDVD on the BER. In Fig.6, it is found that the SDVD for Q=4 is extremely powerful compared with that for At BER=lO-5 Q=2. and SIR=5dB, the SDVD for Q=4 has the SNR margin of 5.ldB. On the other hand, the margin for Q=2 is at most 1.7dB. The value Q more than 8 is expected to yield the remarkable improvements, and to be investigated intensively. We summerize the SNR margin at BER=lO-5 and SIR=SdB in Table 1. From Table 1 we found that for Q=2 _Lthe MSE processing and the SDVD are 24 68 10 12 14 16 effectively combined, but with more powerful SNR(dB) SDVD(QEL), the SDVD dominates the BER Fig.4. BER of SDVD for Q=2 and 4 using MSE performance. processing. A - 609 - 10° without signalprccessin __________withm without SDVD -------_-- with SDVD Q=2 -.-.+_ with SDVD 4~4 10-l 10 -2 -3 10 aJ c1 10-4 $! 2 k 2 10-5 c, ;;: 10-6 lo-' lo-8 0 246 8 10 12 14 SNR(dB) Fig.5. BER of MSE processing 16 14 16 12 SNR(dB) Comparison of BER of SDVD for Q=2 and 4. 2 Fig.6. 4 6 8 10 Table 1. SNR margin at BER=10R5 and SIR=5dB. REFERENCES I SDVD 1.7dB 5.ldB 6 CONCLUSION This paper has introduced the co-channel interference cancelling techniques in the onprocessing satellite communication board system utilizing orthogonal polarizations. In order to reduce the co-channel interference, MSE processor and the SDVD are the equipped on board the satellite. The BER both techniques has been derived using analytically for BPSK. The numerical results of BER have shown that the compound technique yields the excellent BER improvements. To be specific, the improvements become significant in the case of low SIR. [I] K.Ishida, K.Funayama, I.Oka, and I.Endo, Study on Co-Channel "A Interference Cancelling in On-Board Processing Satellites", IECE of JAPAN, National Convention Record, No.708, Oct.1984 (in Japanese). H.E.Nichols, A.A.Giordano, and [21 J.G.Proakis, "MLD and mse Algorithms for Adaptive Detection of Digital Signals in the Presence of Interchannel Interference", IEEE Trans. Inform. Theory, vol.IT-23, No.5, Sep.1977, ~~563-574. [3] Y.Yasuda, Y.Hirata, and A.Ogawa, "Bit Error Rate Performance of Soft Decision Viterbi Decoding", Trans. of IECE of JAPAN, vol.E64, No.11, Nov.1981, pp'OO-707. A.Papoulis, "Probability, Random [41 Variables, and Stochastic Processes, 2nd Edition", Tokyo:McGraw-Hill Kogakusha, 1984. [5] A.J. Viterbi, "Convolutional Codes and Their Performance in Communication Systems", IEEE Trans.Comm.Technol., vol.COM-19,No.5, Oct.1971,pp751-772 - 611 RELATION 112 - BETWEEN APD/CRD OF AUTOMOBILE IGNITION AND RESULTANT TV PICTURE DEGRADATION s3 NOISE * Shigeru YAMAZAKI,*cukihide NOGUCHI and*Hiroshi KURONUMA * Science & Technical Research Laboratories, Japa;*Bsoadcasting Corporation Nippon Electric Company l-lo-11 Kinuta, Setagaya-ku, Tokyo 157 Japan 4-14-2 Shiba, Minato-ku, Tokyo 108 Japan In this article, first, a newly developed Summary noise receiver is introduced. It has a wide IF bandwidth for measuring the APD and CRD of the newly developed noise receiver is A impulsive noise received through nearly the introduced, which has a wide IF bandwidth to same transmission bandwidth as TV signals. measure the APD and CRD of impulsive noise an APD/CRD measuring instrument, Secondly, received through nearly the same transmission which can called a noise level analyzer, bandwidth as TV signals. An APD/CRD measuring measure the APD/CRD of the impulsive noise with instrument, called a noise level analyzer, a frequency spectrum of this width is also which can measure the APD/CRD of the impulsive introduced. noise with a frequency spectrum of this width Thirdly, the results of outdoor measureis also introduced. ments of automobile ignition noise using this Outdoor measurements are made of automobile together with the equipment is described, ignition noise using this equipment, together results of the subjective evaluation of TV with subjective evaluation tests of TV pictures pictures impaired by corresponding ignition impaired by corresponding ignition noise. The correlation of the measured noise. The a very subjective annoyance level shows parameters of the percentage time and the good correlation with the percentage time and with the annoyance average crossing rate the average crossing rate of the 'envelopecaused in TV reception is also described. detected impulsive automobile noise, which are both derived from APD/CRD, in both cases the Wide-bandwidth noise receiver bandwidth of the noise is 3.5 MHz and 120 kHz. On the other hand, the quasi-peak voltage is In order to investigate the statistical not a proper parameter as far as the annoyance properties of impulsive noise with nearly the caused by automobile ignition noise in TV same bandwidth as a TV transmission channel, a reception is concerned. new noise receiver was developed which can ilr~c3SLk’C iis2 rx~;st: envelope ara>liLude of a,, arbitrary VHF and UHF TV broadcasting channel. Interference from impulsive noise affecting Fig.1 is a block diagram of the wide-bandwidth noise receiver. The receiver has a 3.5 MHz TV reception and causing the degradation of picture quality has become a serious problem bandwidth of 6 dB down and a dynamic range of with the increase of automobiles, household 50 dB. The lowest possible input voltage is 30 appliances, various kinds of electric facilidB$l with 50 ohm termination. The noise is ties, and other sources. Quasi-peak (Q-peak) detected after being compressed by a voltage measured by a CISPR measuring receiver logarithmic amplifier employed at a 57 MHz IF developed for use in amplitude modulated sound stage. The receiving frequency is variable by broadcasting has been said to be not always a *1.5 MHz with a 80 kHz step from the video proper parameter by which to gauge the annoycarrier frequency of any VHF and UHF TV ance in TV pictures. On account of this, there channel. The measurement error of the level has been a need for a new look at more suitable indicator is within f2 dB. Three different parameters from the statistical point of view, holding times of 0.05, 0.3, 3.0 seconds can be such as the percentage time of noise envelope used for a peak voltage measurement. amplitude or the average crossing rate obtained from the amplitude probability distribution Noise level analyzer (APD) or the crossing rate distribution (CRD). Only a few articles, however, have been The newly developed noise level analyzer published with regard to the relationship for the measurement of the APD and CRD of between the annoyance in TV picture reception impulsive noise has 24 high speed voltage and such parameters. comparators to discriminate the noise level RF input RF attenuator+ (lOdBx5) Logarith- Noise mic ampl.' detector Synthesized oscillator Fig. 1. Block diagram of wide-bandwidth noise receiver. - 612 - Gate 32 Bits binary #' counter for APD- Control signal 32 Bits binary counter for CRD- a P q CRT Printer Start/Stop IF Clock rate Ext. trig. Fig. 2. Block diagram of noise level analyzer. Switch settings are in the following status; Sl: IF signal measurement, S2: counting duration of measurement.(only for 24th counter) from 24 different reference levels set up beforehand by micro-processor. Fig.2 is a block diagram of the equipment. The frequency range of the input signal is from DC to 6 MHz. The reference level of each comparator can be arbitrarily chosen from among 256 levels at The clock rate for the intervals. equal measurement of the APD can be selected from 30 MHz down to 1 kHz. Consequently, the maximum duration is from about 143 to measuring 4,295,OOO seconds (about 1,193 hours) according to the selected clock rate, as a 32 bit binary co%Jnter is employed fcllawinp each comparator. 24 other 32 bit binary counters are employed Repetitive for the measurement of the CRD. measurements can be performed up to 128 times with less than 10 ms sleep time between each measuring block, required for data transfer to a micro-computer. It is possible to measure the APD/CRD directly from an IF signal whose frequency For APD spectrum does not exceed 6 MHz. measurement, the number of IF carrier cycles which exceed a specified level is counted ; while for CRD measurement, a train of IF carrier pu,l,ses corresponding to one impulsive noise are converted to a single pulse using a re-triggerable one shot multivibrator following each comparator. The multivibrator yields a pulse with a slightly longer duration than one period of the IF carrier. The equipment is fully controlled by microcomputer and is designed to measure the APD/CRD of not only impulsive noise but also kinds of base band signals. Measurements of automobile ignition noise Outdoor measurements were made of the automobile ignition noise caused by various kinds of motor vehicles running through the main street in front of our laboratories. Parameters measured were as follows; a) the APD and CRD of the noise envelope with a 3.5 MHz bandwidth, b) the APD and CRD of the noise envelope with a 120 kHz bandwidth, c) the APD and CRD of the Q-peak voltage. In the case of a) the wide-bandwidth noise receiver was used. While in the case of b) and c) a standard CISPR receiver was used. It is impossible to measure a), b) and c) simultaneously for the same noise using a single noise level analyzer. Measurements were therefore divided into two cases : simultaneous measurement of a) and c), and that of b) and c). In these cases, measured data related to c), i.e., the waveform data of Q-peak voltages, were recorded in a digital data recorder. After a sequence of measurements, the APD/CRD of the Q-peak voltages were measured by the noise level analyzer on reading out the data from the recorder. Fig.3 shows a noise measuring system. The noise was picked up on channel 2 (at a frequency of about 100 MHz ) on which no TV Antenna I, Ban pass filter (TV ch.2) I Power splitter Attenuator + CISPR receiver Widebandwidth receiver Power combiner Video sip@& TV signal modulator (TV ch.2) - I Digital --. > data recorder 12OkHz, envelope ----_ r---(offline) Quasi-peak > various Demodulator recorder Attenuator Fig. 3. Noise measuring system. - 613 112 - 10” % time ordinate (a) Average is exceeded APD 10l 102 10” crossing (b) rate 10’ s3 lEIS 10” (cps) CRD Fig. 4. Examples of measured APD/CRD for automobile ignition noise. programs are transmitted in the Tokyo area. Notch filters for adjacent TV carriers were introduced in a band pass filter block in order to prevent any intermodulation between the carriers and the incoming noise. The noise was distributed in two ways : one was by feeding it into the measuring equipment, and the other was by introducing it to a power combiner to produce interference in the TV signal. The level of the desired TV wave was set to 80 dBuV at the input terminal of the power combiner. The duration of one measurement was 30 seconds. Clock rates were chosen as 30, 5 and 1 MHz depending on the bandwidths of 3.5 MHz, 120 kHz and Q-peak respectively. The level interval of comparators in the noise level analyzer was properly chosen as 2 or 3 dB according to the dynamic range of the incoming noise. 64, 33 and 97 samples were obtained for the measurement of a), b) and c) respectively. Examples of measured APD/CRD are shown in Fig.4. APD is plotted on Weible graph paper. Subjective tests Subjective tests were made of the pictures impaired by the ignition noise, which were recorded on video tape at the same time as the The original measurement of the APD/CRD. picture is a still, colored one, as shown in Fig.5. The viewing conditions and the evaluation scale are listed in Table 1. Viewers were more than 20 expert TV engineers. Mean opinion scores CMOS) were calculated as follows; Table 1. MOS= Z? i=l i*qi (1) where, qi is the population of votes for each of the comments i, (i=l to 5), obtained from the comment data for each subjective test. Test results The noise amplitude was measured as an absolute value, for example in dBuV. It is, however, quite convenient to express the noise amplitude relative to the desired TV wave amplitude because the annoyance depends on their relative amplitude. Hence, we define the following quantity D/Uj as a measure of the noise. D/Uj = Dw - Uthj (dB) where Dw is the amplitude of a desired TV wave in dBuV, and Uthj is the discriminating envelope amplitude of the noise in dBuV, which corresponds to the reference voltage of a j'th comparator in the noise level analyzer. Relations between the MOS and D/Uj in the APD Fig.6 and Fig.7 show the relationship between D/Uj and the mean opinion score CMOS) at four parameters of percentage time, obtained from APD for the noise with different bandwidths. Solid line in the figure are logistic trend. lines expressed in the form of Eq.(3). The logistic function is often applied instead of a linear regression, because it has the advantage of being able to express the property that the MOS never does exceed the value of 1.0 and 5.0. (a) Viewing conditions . . . 370 x 275 mm Picture size Highlight brightness . .. 150 fL ... 50 : 1 Contrast ratio Ambient illumination .. . 75 lx at the cathoderay-tube face ... 6 times the height of Viewing distance the picture screen (b) Five-grade evaluation scale Comment descriotion Comment Imperceptible 5 ... Perceptible but not annoying 4 . .. Somewhat annoying 3 ... Severely annoying 2 ... Unusable 1 .. . (2) Fig. 5. Test picture (NTSC,color). - 614 - 5.0 g B %: time * ; 10-1 0 ; 1o-2 4.0 t + ; 1o-3 E s ; 3.0 , n PJ/ 0 ?I a 2.0 ‘8 1 0 ’ ’ ’ ’ 10 20 ’ ’ 30 * D/Uj (dB) D/Uj (dB) Fig. 7. Relations between D/Uj and MOS for 120 kHz bandwidth at specified percentage time. Fig. 6. Relations between D/Uj and MOS for 3.5 MHz bandwidth at specified percentage time. Fig. 8. Relations between Q-peak voltage and MOS at specified percentage time. % time ; (a) .. 10 % (:b) . . 5 % (c) .. 1 % 11G 10 20 30 40 50 60 10 20 30 40 50 60 1 10 20 30 40 50 60 D/Uj (dB) crossing rate * ; 100 (cps) : 4.0 : 4.0 s s $3.0 .ti E? 2.0 9 2 -20 -10 0 10 20 30 40 50 D/Uj (dB) Fig. 9. Relations between D/Uj and MOS for 3.5 MHz bandwidth at specified average crossing rate. Y = 1 / (1 + exp ( A-X 1) A,B:Bregression constan::) There is a satisfactorily good correlation between D/Uj and MOS, especially at the percentage time of lc2 for a 3.5 MHz bandwidth, and 163 for a 120 kHz bandwidth. Fig.8 shows the relationship between D/Uj and MOS for Q-peak voltage at the percentage time of 10, 5 and 1, which are obtained from the measurement of a) and c) in the previous section. In this case the correlation is somewhat poorer than in the case of Fig.6 or Fig.7. Nearly the same poor relationship was obtained from the measurement of b) and c). Relations between MOS and D/Uj in CRD Fig.9 shows the relationshin between D/U.i and MOS obtained from the CRD for a 3.5 MHz bandwidth. There is a good correlation between 1.0 I. 0 8 10 a 1 20 * t 30 D/Uj * I. 40 t 50 *I 60 I i a (dB) Fig. 10. Relations between D/Uj and MOS for 120 kHz bandwidth at specified average crossing rate. them at an average crossing rate taken from about 10 to several hundreds cps, although it is difficult to decide an average crossing rate at which the correlation becomes maximum. For a 120 kHz bandwidth, in Fig.10, the correlation is also high when taken from a few cps to several tens of the average crossing rate. It can be confirmed that the ratio of an average crossing rate for the bandwidth of 3.5 MHz t0 that for 120 kHz, at which the correlation between D/Uj and MOS seems highest, coincides approximately with the bandwidth ratio of 3.5 MHz to 120 kHz, i.e. about 30. In the case of Q-peak voltage, the average crossing rate has an extremely poor correlation with MOS. For one reason, a 30 second duration of measurement seems inadequate to produce statistically meaningful data, but the average crossing rate of Q-peak voltage may be said to be an improper parameter as far as the annoyance of the automobile ignition noise in TV -7 -6 -5 -4 -3 -2 -1 10 10 10 10 10 10 10 % time (a) D/Uj= 41 dB (b) D/Uj= (d) 29 dB D/Vi= 11 dB Fig. 11. Relations between percentage time and MOS for 3.5 MHz bandwidth at specified D/Uj. % time (a) D/Uj= 56 dB (b) D/Uj= 50 dB (c> (d) D/Uj= 29 dB D/Uj= 41 dB Fig. 12. Relations between percentage time and MOS for 120 kHz bandwidth at specified D/Uj. Fig. 13. Relations between percentage time and MOS for Q-peak voltage at specified D/Uj. (a) D/Uj = 44 dB (b) D/Uj = 36 dB (c) D/Uj = 28 dB reception is concerned. MOS at a given D/Uj versus percentage time and average crossing rate Relations between MOS and percentage time at a given D/Uj can be derived from APD graphs drawn for each noise measurement, shown for example in Fig.4, and from corresponding MOS data. This is also possible for the relation between MOS and the average crossing rate. Fig.11 shows the:relation between MOS and percentage time at D/Uj ratios of 41, 29, 20 and 11 dB corresponding to a) through d) of the figure. The noise bandwidth is 3.5 MHz. Graphs indicate that there is a very good correlation between MOS and percentage time when D/Uj equals 29 or 20 dB. A similar relation holds for noise with a bandwidth of 120 kHz when D/Uj equals 50 dB, as shown in Fig.12. On the other hand, the correlation is poor for Q-peak voltage at any D/Uj, as is clear from Fig.13. Fig.14 and Fig.15 show the relation between MOS and the average crossing rate for 3.5 MHz and 120 kHz bandwidths respectively. Correlations similarly high, as in the case of MOS vs. percentage time, are obtained when D/Uj 29 or 20 dB for 3.5 MHz bandwidth and equals when 50 or 41 dB for 120 kHz bandwidth. Roughly speaking, in both relations for percentage time and average crossing rate, there is about a 20~ 30 dB difference between the cases of the bandwidths of 3.5 MHz and 120 kHz, with respect to D/Uj, at which the correlation seems highest. If we do not take into account the overlapping of impulse responses owing to the reduction of the bandwidth, the peak amplitude of the impulse increases in proportion to the banda 29.5 dB Considering this fact, width. decrease of the peak amplitude can be expected for the impulsive noise at a bandwidth of 120 The 20~30 dB difference in D/Uj is k&. 3 cl re’rifd l$lEW18 llQle'lB ldldl~ 1 Average crossing rate (cps) (a) D/Uj" 41 dB (b) D/Uj= 29 dB (d) D/Uj= 11 dB CC> D/Uj= 20 dB Fig. 14. Relations between average crossing rate and MOS for 3.5 MHz bandwidth a< specified D/Uj. Average crossing rate (cps) (b) D/Uj= 50 dB (a) D/Uj= 56 dB (C) D/Uj= 41 dB (d) D/Uj= 29 dB Fig. 15. Relations between average crossing rate and MOS for 120 kHz bandwidth at specified D/Uj. considered to originate from the peak amplitude reduction of the impulsive noise as mentioned above. Discussion In both cases of percentage time and the average crossing rate, the correlation between MOS and D/Uj is considerably higher for impulsive noise with a bandwidth of 3.5 MHz and 120 kHz than for Q-peak voltage measured by a standard CISPR receiver. The correlation for the bandwidth of 3.5 MHz seems slightly higher than that for 120 kHz, but no significant difference was obtained between them. This leads to the conclusion that the fabrication of a noise receiver with a wider bandwidth than necessarily standard 120 kHz is not the required for the measurement of the parameters APD or CRD for automobile impulsive noise, as far as the annoyance caused to the television reception is concerned. From the relation between MCS and percentage time or the average crossing rate, the noise caused by impulsive can annoyance quantitatively evaluated and probably be measured using a quite simple noise level analyzer, in which it is sufficient to employ only a single comparator and counter pair. For example, a noise level analyzer, which has a single comparator to discriminate the envelope amplitude of impulsive noise at 50 or 40 dB in D/Uj, and has a single counter to measure the average crossing rate, can evaluate the quality of the picture impaired by ignition noise, using the result shown in Fig.15. Conclusion A newly developed noise receiver has been introduced, which has a wide IF bandwidth to measure the APD and CRD of the impulsive noise received with nearly the same transmission bandwidth as TV signals. A noise level analyzer which can measure tne APD/CRD of the impulsive noise having a frequency spectrum of this width has also been introduced. Outdoor measurements with regard to the automobile ignition noise were made using this equipment. Subjective evaluation tests of TV ignition pictures impaired by corresponding noise were also made. The subjective annoyance has a fairly good correlation with percentage time and the average crossing rate of envelopedetected impulsive automobile noise, in cases of both 3.5 MHz and 120 kHz bandwidths. As for the parameters of quasi-peak voltage, the correlation was found to be poor. This suggests that the APD/CRD of quasi-peak are not proper parameters, as far as the annoyance of autois mobile ignition noise in TV reception concerned. References (1) Parsons, J. D., and Sheikh, A. U. H., "The characterization of impulsive noise and consifor a noise-measuring receiver", derations Radio & Electron. Eng. Vo1.49, No.9, 1979. (2) Matheson, R. J., "Measurements of Automotive Ignition Noise Using A DM-4 APD Meter", IEEE Int. Sympos. on EMC, 1981. - 617 113 s4 - ELECTROMAGNETIC RADIATION CAI!SE.n BY SILVER PALLADIlJlVl ALLOY CONTACT StlITCHING Keiichi Uchimura", Teizo Aida*, find Tasuku Takmi** * Kumamoto University Kumamoto, ** Tohoku Japan University Sendai, This paper deals with the electromagnetic radiation (EMR) caused by Ag-Pd alloy contact switching. In the case of a normal arc, the EMR level depends upon the alloy composition of contact materials, while, in case of a showering arc, it does not depend upon it. A parallel capacitance connecting between the contacts is effective for suppressing the EMR. Introduction An electromagnetic relay trends toward small-size and low-power, as an integrated circuit (IC) is being developed. There are many cases where the relay is installed closely to the IC iii an electronic equipment or a control system. A discharge due to the relay operation may cause the electromagnetic interference to the various IC equipments, especially the IC is irradiated by the electromagnetic radiation (EMR) from an arc which appears in breaking contacts of relay. Thus it is very important to clarify the characteristics and the generation of EMR which causes by the breaking electric contacts. Meanwhile, in a low current region of O-l-3A, various alloy contacts such as Ag-Pd, Ag-Ni, etc., are widely used in the relays rather than the pure metal contacts. This paper deals with the EMR caused by silver palladium alloy contact switching. The Ag-Pd alloy contacts have such good properties as low contact erosion and very few brown-powder. Ag-Pd alloy contacts are also not expensive compared with pure metal contacts such as Pd and Au. From these reasons, new manufactures of wire spring relay now practically employs the Ag-Pd alloy instead of pure Pd [l]. In this study, we operated Ag, Pd, and Ag-Pd alloy contacts at the source voltage of d.c.25V and 5OV and in air, and then measured various properties of the EMR caused by a normal arc and a showering arc. As a result, it was found that the EMR caused by normal arc depends upon the alloy composition of contact materials, while the EMR Japan caused by showering arc does not depend upon it. Furthermore, from the generation mechanism of showering arc, it is described that a parallel capacitance connecting between the contact is effective to inhibit the EMR. Experimental Method The experimental setup is shown in Fig.1. The circuit contains two contact pairs Kl and K2 connected in series. The break-make order between Kl and K2 is controlled so that Kl arcs only on break and X2 arcs only on make. The contact used here is approximately 3mm long and has a diameter of 5mm. The surface shape of fixed side contact ( cathode) and that of moving side contact (anode) are plane and spherical, respectively. The contact force is about log and the velocity of contact separation is about O.lm/s. The length of contact gap is Imm. Break-make cycle rate is 15 times/min. Contact materials are Ag, Ag-20%Pd, Ag-40%Pd, and Pd, where the percentage represents the weight percent (wt.%). In the right-hand side of Fig-l, shown is an apparatus for measuring the electromagnetic radiation (EMR). In this experiments, two types of the electromagnetic field measuring apparatus and those of the quasi-peak value meters were used. The electromagnetic field measuring apparatus have the following specifications; one of them has the frequency range 0.15-IOMHz, IF Quasi-peak value meter Fig.1 Experimental , setup - bandwidth 4kHz (3dB down), and nonshielded loop antenna, and the other has 25-23OMHz, 120kHz+lO% (6dB down), and loop antenna. The quasi-peak value meters have also the following specifications; overload coefficient of beyond 12dB and 43.5dB, charge time constant of detector of 1+0.5ms and Ims, and discharge time constant of detector of 600+120ms and 55Oms, respectively. The distance between the contacts and the loop antenna of the apparatus is 3 meters. The floor of experimental room was covered with the metal grounding screen. The height of the contacts above the ground was kept at 0.5 meters and that of the antenna at about 1 meter. Measured EMR Caused Results by Normal of EMR Arc Frequency Distribution of EMR. A typical normal arc of Ag contacts is shown in Fig.2. Figure 3 shows the frequency distribution of the EMR caused by the normal arc in breaking contacts. Using Eq.(l), we Convert the value E(dB ) of EMR shown in Fig.3 into e(uV/m), E=2Ologg Fig.2 Typical 618 where e. is a standard field strength of luV/m. We then obtain the following relations in the frequency range from 0.1 to about 1OMHz. e oc f-u where f is a frequency and a is a constant inherent to the contact materials. (X=0.75-1.49 for Ag, u=O.95-1.6 for Ag-20%Pd, (3) for Ag-4O%Pd, and a=O.86-1.6 a=0.73-1.18 for Pd. The above values are applicable to the case where source voltage is d.c.50 volts and the circuit current region from 0.5 to 7 amperes. Besides, although the experimental results of Au, Ag-Au, and Ag-Cu contacts were omitted in this paper, c1 of these materials have similar tendency to those shown in (3) 121. On the other hand, in the frequency range from 25 to 20OMHz, E(dB) reachs the maximum at the frequency f=60-70 MHz, as shown in Fig.3. It is also found that there are some differences in the level of EMR by the alloy composition. ( .I [dBl waveform of normal (2) arc voltage. Circuit Current Dependence of EMR. In order to know the circuit current dependence of EMR, experiments are performed in source voltage V=5OV and circuit current 1=0.5-7A. Figure 4(a) shows E(dB)-I characteristics for Ag at the frequencies 0.15, 1, 5, and 50 MHz. The maximum of EMR appears at 2-3A which is markable in case of 5 and 50 MHz. The E(dB)-I characteristics at 5 MHz for Ag, Pd, and Ag-Pd alloy are shown in Fig.4(b). Excepting pure Pd, the maximum of EMR appears at 2-3A. With regard to those effects, we shall discuss in the following chapter. 80 I=2A Ag T i: -. i -L'** .-c .. ;;. Ag-208Pd 0.5 1 5 Frequency Fig.3 Frequency distribution 10 50 '100 f(MHz) of electromagnetic radiation in normal arc 619 - (a) (b) 0.15MHz 70 113 s4 f=5MHz,V=50V 50 -50 8 w 30 1 50MHz 10 Circuit v=5ov I L I I 3 1 Circuit EMR Caused I I I 5 current by Showering 3 I Fig.4 Relation 7 5 current between I(A) E(dB) and I in 7 I(A) ASI Arc v=25v 1=90mA L=550mH C=O.O026pF 200V/div. lOOp.s/viv. Frequency Distribution of EMR. A typical oscillograhic trace of a showering arc for Ag contacts is shown in Fig.5. Figure 6 shows the frequency distribution of EMR caused by showering arc which has similar tendency to that of the normal arc for Ag. That is to say, the EMR is roughly inverse proportion to the frequency in the frequency range O.l-about IOMHZ, while in 25-20OMHz, the maximum of EMR are seen at about 7OMHz. The difference of the EMR level owing to alloy composition is not recognized; in normal arc, the &iR level depends on alloy composition ( refer Fig.3). Fig.5 Typical waveform of showering arc voltage. limited up to lOOmA. Hence the only two cases of 50 and 9OmA have been tested, and the EMR from contact showering arc were measured under the conditions where the source voltage V=25V, inductive load (coil of relay)L=550mH, and parallel capacitance connecting between the contacts C=O.O026pF. Figure 7(a) shows the measured results of E(dB)-I characteristics for Ag contacts at the frequencies 0.2, 1, 5, and 70MH2, and Fig.7(b) shows those for Ag-Pd alloys Circuit Current Dependence of EMR. A coil in hinge type electromagnetic relay was used as an inductive load in our first experiment. In this case, the current flows through coil of relay was 80 V=25V 1=9omA L==550mH C=O.O026uF 2 .: u-l83 CG : (H" &I 2 .5 40 . I : Ag T 20 h : Ag-20%Pd 1 : Ag-40%Pd _ 1 : Pd . I . .; rd $ 0 0.1 II Ill I 0.5 I III 1 Frequency Fig.6 Frequency distribution I 5 I I III 50 10 I 100 f(MHz) of electromagnetic radiation in showering arc. - 620 IIntermittent (a) 1: - 7 ~ 4. - @Z:z 20 - arc voltages 0.2MHz Ag v=5ov I=2.5A lOV/div. lms/div. f=lMHz w v=25v L=550mH Fig.8 Normal C=O.O026pF arc voltage and EMR voltage 100 50 Circuit current I(mA) (b) - I: f=70MHz ACJ -r ': Ag-20%Pd L ;;i 2 40 E _$I/=4 W t: Ag-QO%Pd v=25v L=550mH 20 c 1 1: Pd C=O.O026pF I t I I I.1 1 50 100 Circuit current I(mA) 60 I : Ag I : Pd (c) 40 wave (a) (b). Discussion 01 60 wave /z f I P Z w / f=SOMHz 20 Normal Arc In order to clarify the effect of circuit current on the EMR, the waveforms of both normal arc voltage and EMR voltage were simultaneously observed by using a 2 channel oscilloscope. The EMR voltage was observed at the output of the intermediate frequency amplifier in the electromagnetic field measuring apparatus. Those observed waveforms are shown in Fiq.8. The source voltage is d.c.50V and the circuit current is 2.5A for Aq contacts. The larger EMR voltage is seen at the portion (lower trace) corresponding to the lager amplitude voltage fluctuations in the waveform of normal arc voltaqe (upper trace). The intermittent arc voltage fluctuations, which is shown in the upper trace of Fig.8, are governed by the transition of arc phase from a metallic phase to a gaseous phase, and occur at the circuit current 2-3A [3]. On the other hand, Figure 9 shows the relationship between arc duration Ta and circuit current I. In cases of Ag and Aq-Pd alloys, Ta chanqes discontinuously at 2-3A. It has been said that this discontinuous is also due to V=lOOV L=lOmH Ag-20%Pd C=O.OlpF I 0 10 50 Circuit Fig.7 Relation I 100 between I 500 1000 current case of showering I III A=' I(mA) E(dB) and I in arc. at 70MHz. Second experiment has also carried out under the conditions where V=lOOV, L=lOmH, and C=O.OlpF, the E(dB) -1 characteristics of Aq and Pd contacts were measured at the frequency 50MHz. These results were shown in Fig.7(c). As we can see from Fiq.7(a)-(c), the EMR increases with the circuit current. Although the E(dB)-I characteristic is independent on the alloy composition, it depends upon the circuit condition, such as V, I, L and C. 1 3 5 Circuit current Arc duration current. versus 7 I(A) circuit 621 t;le transition of arc phase from the metallic phase to the gaseous phase. As understood from the above two the contact arc has an descriptions, unstable condition in a current range of 2-3A. In this case, high frequency oscillation with large amplitude appears in the arc, which results in the electromagnetic radiation (EMR). Such an EMR can be detected as higher electromagnetic field intensity E(dB). On the contrary, in the case where the circuit current is higher than 2-311, the arc discharge becomes stable, so that E(dB) can be considered to take a relative low value. Showerinq Arc The showerinu_. arc is characterized by large, rapid voltage fluctuations appearing across the contacts [41. When the voltage across the contacts exceeds the dielectric breakdown voltage, the showering arc occurs. The voltage across contacts and EMR voltage for Ag contacts were observed simultaneously, under the conditions of V=25V, I=90mA, L=550mH, and C=O.O026pF. Typical examples are shown in Fig.10. The EMR voltage appears at the portion (lower trace) corresponding to the dielectric breakdown across the contacts (upper trace). The dielectric breakdown across the contacts during the contact break are consi.dered to occur due to the following three kinds of causes: (I) field emission breakdown, (2) breakdown occuring at the constant voltage VB2, regardless of the length of contact gap 151, and (3) air breakdown which is governed by Paschen's law. Figure 11 shows the times of the showering arc in which the three main mechanisms operate. According to Germer [5],the dielectric breakdown voltage VBZ for both fixed Ag and Pd electrodes are equal to 340V. Hence VB2 of Ag-Pd alloys is considered to be 340V. The dielectric breakdown voltage VBl in region (1) is approximated by [61: VBl = F u t f K(1) (4) where F = 2~10~ V/m for Ag, F = 3x108 v/m for Pd, u =constant separation velocity in m/set, and K =constant relating circuit current. Using VBl=340V, u=O.lm/s, F=2xlO*V/m for Ag and 3x108V/m for Pd, we obtain the relations of Eq.(5) as the duration of region (1) in Fig.11. t1 < 17u.s for Ag, and (5) t1 < llus for Pd. On the other hand, according to our experiments, the start time t of region (3) is approximately 2$ Ops, which is independent of the circuit condition and the kind of contact material. Hence, the regions (2) and (3) are nearly independent of the contact material, that is, alloy compo- 113 s4 - Ag v=25v 1=90mA L=550mH c=O.O0261;1F 2OOVldiv. 1 OOus/div. f=50MHz (a) (b) Fig.10 Showering arc voltage and EMR voltage wave wave (a) (b). VE2 -~Lzztp--_I I-----Fig.11 Schematic ( t Tsdrawing of showering arc. sition. Therefore the duration tl of the region (I), which depends on the contact material, is only a few percent of the total duration of showering arc Ts (see Fig-II). From this reason, it is considered that, since the showering arc voltage waveform does not nearly depend upon the alloy composition, the EMR characteristics become also roughly independent of the alloy composition. 1" the meantime, the voltage across the contacts begins to rise according to: dV (6) E t=() = ; where I=circuit current, and C=capacitance of circuit. From Eq.(6), if the circuit current I increases, dV/dt will become large. As the results, the dielectric breakdown occurs from just after contact break, so that the number of dielectric breakdown Nb results to increase (see Fig.11 ); this thing is also supported by Mills [4]. Eventually when Nb, that is, I increases, it can be considered that the indication of EMR in the electromagnetic field measuring apparatus becomes large. Furthermore, we can propose the method for suppressing the EMR from Eq. (6). That is to say, if we increase the value of C, Nb can be expected to - decrease [4], so that the Z:MR can be expected to decrease. From the above consideration, we tried to connect the C-r series circuit between the contacts in parallel. The result is shown in Fig.12, and it is found that the C-r series circuit is effective in suppressing the EMR from showering arc. In case of the normal arc, we already stated that the C-r series circuit was effective for suppressing the EMR, as well [2]. \ 60 z \ non C-r \ 44 40 -_ -1, ACJ \ \ \ \T i\ 20 C-r C=0.33pF 0 I III 0.1 0.5 Fig.12 Frequency \ T. r=14851 I III I I 5 1 Frequency current I char3. In the E(dB)-circuit acteristics, excepting pure Pd, the maximum of EMR appears at 2-3A in case of the normal arc. On the contrary, the EMR in the case of showering arc increases with the increasing of I. The reason for the difference of the above things was discussed from arc generation machanisms: the EMK of normal arc is due to the transition of arc phase, and The EMR of showering arc is due to three kinds of dielectric breakdown and depends upon dV/dt) t=O4. We briefly discussed that the EMR caused by showering arc was fairly suppressed by connecting a capacitance C between the contacts in parallel, which we confirmed by our experiments. The authors arc grateful to Associate Professor H.Echigo of Tohoku University for his useful advice, and Mr. S. Ogata , Mr. Y. Kosaka, Mr. H. Matuo and Mr. T. Matuura (Kumamoto University) for their useful assistances. References 10 f(MHz) distribution - Acknowledgment \ 3 w 622 of EMR in case with C-r circuit. Conclusion The results of this study are summarized as follows. 1. The EMR caused by normal arc depends upon the alloy composition of the contacts, while the EMR caused by showering arc is independent of the alloy composition of the contacts. 2. In the normal arc, the EMR is roughly inversely proportional to the frequency in the range of O.l-IOMHz, while in 25-20OMHz, the maximum of EMR appears at about 70MHz. In the showering arc, the frequency characteristics of EMR are similar to those in normal arc. [II T.R.Long and K.F.Bradford: "60Pd-40Ag as an Electrical Contact Material to Replace Palladium", Proc. of the 8th ICECP, pp. 56-61 (1976). r21 T.Aida, K.Uchimura, T.Noguchi and S,Ogata: "Effect of Alloying Elements on The Radio Noise Characteristics of Silver Based Alloy Contacts", Proc. of the Int, Symp. on EMC (Tokyo) (1984) (to be published). 131 P.J.Boddy and T.Utsumi: "Fluctuation of Arc Potential Caused by Metal-Vapor Diffusion in Arcs in Air", J. Appl. Phys., 42, 9, pp. 3369-3373 (Aug. 1971). 141 G.W.Mills: "The Mechanisms of the Showering AX”, IEEE Trans, Part, Materials and Packaging, PMP-5, 1, pp. 47-55 (Mar. 1969). r51 L.H.Germer: "Electrical Breakdown between Close Electrodes in Air", J. Appl. Phys., 30, 1, pp. 46-51 (Jan. 1959). [61 K.Uchimura and T,Aida: "Extinction Critical Curves of Showering Arc in Breaking Contacts at DC Source Voltage Less than lOOV", Trans, IEE of Japan (in Japanese), 103-A, 8, pp. 421-428 (Aug. 1983). - 623 114 - s5 ON THE CHARACTERISTICSOF THE ELECTROMAGNETIC FIELD GENERATED BY VIDEO DISPLAY UNITS W. van Eck, .T. Neessen and P. Rijsdijk Dr. Neher Laboratories Leidschendam Abstract The electromagnetic field generated by video is investigated at units (VDUs) display frequencies above 30 MHz. This type of data generates a radiated equipment processing frequency spectrum consisting of two major emission in the Narrowband components. frequency spectrum of the radiated field is predominantly generated by the VDU’s clock whereas the most powerful broadband circuitry, emission originates from the circuits for video From the broadband emission signal processing. of a VDU a TV receiver will reproduce the image ow screen. displayed on a VDU screen on -its This phenomenon is more severe if a coherence exists between the broadband and narrowband emission. 1. Introduction and assumption The use of square wave signals and fast circuits in digital equipment leads switching fields electromagnetic radiation of to containing frequency components up to far above As the radiation resistance of the VHF region. inside the equipment leads interconnecting increasing frequency, the high increases at frequency parts of the signals in digital efficiently, quite equipment may be radiated with broadcast reception. and may interfere generated by The electromagnetic field equipment generally consists of a digital narrowband isolated combination of linear distances on the frequency signals at regular from the axis, and ‘random’ noise originating in the signals ‘random’ binary various equipment. We now confine ourselves to the interference that may be produced in TV reception due to In general there will these types of signal. be no relation between a broadcasted TV signal When a video and the interfering signal. display unit (VDU) is incorporated in the digital equipment this may no longer be true. In a VDU the information is displayed on the screen using the same techniques as in a TV receiver. Therefore there is a rather good similarity between the video signal in a VDU and that in a TV receiver. In most VDUs even the same horizontal and vertical synchronused as in TV frequencies are ization As the video signal in a VDU is a receivers. each harmonic of impulses, stream of square this signal shows a remarkable similarity with a broadcasted TV signal. PTT - Netherlands Due to the amplification of the video signal in a VDU to several hundreds of volts before it is fed into the cathode ray tube (CAT), the radiation originating from the higher harmonics of the video signal will be stronger than radiation originating from other broadband signals inside the VDU. A television receiver picking up one of these higher harmonics will demodulate the signal - by envelope detection and thus reproduce the image displayed on the VDU on its own screen. This means the interference from video display units may be more annoying than interference from other types of digital equipment. 2. Theory The power spectral density of the video signal The screen of a video displav unit is built up from small dots, calleh pixels. These pixels are arranged in horizontal lines which are scanned by the electron beam in the CRT, as in a normal TV receiver. On the analogy of European black-and-white television, the refresh rate of the display is generally chosen equal to 50 Hz, and the line synchronization is about 16 kHz. In most VDUs no interlace is used, in contrast to TV reception. In figure 1 an overview is given of the build-up of an average VDU system, in which the generation of the video signal is shown. To build up the display in pixels the current of the electron beam in the CRT is on-off modulated. Thus the video signal in a VDU is a digital signal, in which a logical ‘one’ will evoke a ‘white’ spot on the screen, whereas a logical ‘zero’ will inhibit this. To obtain the required resolution on the figure 1 Overview of the VDU’s buildup 624 display, the bit duration in the video signal is shorter than 50 ns in most VDUs. Between two Success1 ve bits the video signal is generally equal to zero. This is accomplished by multiplying the initial video signal v(t) with bit duration Tb (< 100 ns) - with a square wave of period ‘I,,, using a logic AND. This square wave is called the video-dot-clock. The video-dot-clock is already available in the VDD, because it is used for synchronization of the parallel to serial conversion to obtain the initial video signal v_(t). Assuming the information displayed is non-repetitive the initial video signal v(t) may be approximated by a random binary signal. Now let the probability af v(t) having an then consequently amplitude equal to A be p, the probability of the amplitude being equal to For a VDU screen full of 0 is equal to l-p. text about 10% of the pixels are white, thus p is generally smaller than 0.1. The autocorrelation function Rxx(7) of the after final video signal x(t) - obtained multiplication of v(t) and the video-dot-clock An expression for the is given in figure 2. power spectral density of the signal x(t) can be obtained by computation of the Fourier transform of Rxx(~) according to the scheme R,,(T) is decomposed into given in figure 2. the sum of two auto correlation functions. The transform of both functions sum of the Fourier is the power spectral density S,,(f) of x(t). 2 sin(nfTb/2) S&(f )=;A’ Tb(p-$)+pzk> 6 (f-k/T,,) i.= W&l:! )[ I In the above expression we may respect to p, because p < 0.1. From the obtained expression neglect p and knowledge in - the way in which x(t) is generated it is about of- AM demodulation one evident that in case of S (f) provides enough information sidelobe Reconstruction to reconstruct the signal x(t). of the initial signal y(t) is already possible only half this sTdelobe of S,,(f) is if since v(t) has a bit duration T and available, x(t) has a bit duration Tb/2. Envelope detection of the signal filtered at wider than l/Tb will give the LF a bandwidth part of the signal v(t). whatever the central This is of the detection filter is. frequency If, however, a narrowband shown in figure 3. frequencies k/T, lies inside the component at filter bandwidth a better reconstruction of v(t) is possible since the narrowband component serves as a carrier. Reconstruction by means of a TV receiver with a As a TV receiver is equipped vestigial sideband demodulator over a detection bandwidth of approximately 5 MHz, reproduction of the LF part of the signal v(t) is possible without interference whenever intersymbol Tb>lOOns. In most VDUs the bit duration Tb is than 100 ns, thus causing slightly short,er intersymbol interference during detection with a TV receiver. Since a TV receiver is not equipped with a sampling detector and subsequent clock recovery envelope includes the receiver circuit and further analog signal processing detection - the intersymbol interference will manifest itself as a lengthening of the pixels on the TV This leads to a slight decrease in screen. readability. The LF signal beyond the envelope detector the TV receiver can be made to look like a in digital signal by adjusting the contrast level The clipping threshold to the maximum value. by means as shown in figure 2 can be adjusted of the brightness control in the TV receiver. 3. Verification figure 2 Srheme for c’omp~lf;~tion of F&f) Measurement set-up For verification of the reconstruction of TV information by a normal the radiated according to set-up receiver a measurement figure 4 was realised. (Singer NM 3’7/57, The measuring receiver 30-1000 MHz) is set to an IF banduidth of 1 MHz and the IF signal is fed into a TV broadcast after conversion of the frequency to receiver, an arbitrary VHF channel. possible In this set-up measurements are the TV broadcast bands and the ambient outside fieldstrength can be compared to the received The effective bandwidth of the image quality. 1 MHz. Thus the TV receiver is reduced to readability of the received image is reduced. Measurements proved that the picture is still the detection bandwidth is readable when of the pixels reduced to 1 MHz. The stretching on the TV screen can easily be compensated for by adjusting the clipping threshold, as shown in figure 3. As the video signal in a VDU does not contain synchronization information the horizontal and vertical synchronization signals have to be brought into the TV receiver seperately to carry out the measurements. In this set-up the synchronization signa Is are reconstructed from the magnetic field in t,he the VI%. ‘II-II c’ field is gener;,tt.tA,i II I cinity of 114 -L s5 black white -__-- ehreshold _- -et black white -L +t frequency domain of I impulse figure 3 Reconstruction representarion of 4th order Butterworth filtering and threshold setting the video by the high voltage transformer, which normally operates at the line synchronization frequency. A simple phase locked loop, a digital frequency divider (+ number of display lines) and two one-shot circuits transform the received signal into horizontal and vertical synchronization signals. Both signals are combined and brought into the synchronization separator of the TV receiver by means of an optical fibre. signal 4 Overview of the measurement setup Measurement results Reconstruction of the displayed information the received interference from signal is generally possible in the following frequency regions: - at almost any frequency between 30 and 300 MBz, - at almost any harmonic of l/Tb between 300 and 500 MHz, and - at some frequencies above 500 MHz. The distance between the VDU and the receiving biconical antenna may be as large as reconstruction of 50 metres before the information is impossible. displayed This refers to measurements carried figure out within the TV broadcast bands, using only the TV receiver for reception. For a VDU in metal covering this distance is generally reduced to about 5 metres. The fieldstrength measured at 1 MHz bandwidth is independent from the number of characters displayed on the VDU screen, in contrast with the results expected from theory Additionally it was p and p*). (S,,(f) :: noticed from the fieldstrength measurements that there are more narrowband components that these could not be however, expected, large distinguished very well due to the video signal in a TV receiver detection bandwidth (I MHz). Screening of the circuits for video signal and disconnecting the CRT had no processing drastic effect on the measured fieldstrength the reconstructability of on the nor information. To explain the origin of the unexpected measurement results a more detailed analysis of the interference from a VDU was performed. 4. Explanation figure TV receiver of results Measured fieldstrength as a function of display contents In the theoretical contemplation of chapter 2 it was assumed that the video signal x(t) is the most powerful source of radiation yn the assumption proves to be inadequate VDU. This for explanation of the measurement results obtained. The radiation originating from the video be very powerful due to the signal may amplification of more than 26 dB, however, the total power is spread over the entire frequency axis. In comparison to other broadband signals it will be the most powerful source in most Narrowband cases. signals such as the video-dot-clock and the microprocessor system clock (see figure I) may be radiated as well. but less These signals can still strongly. cause high levels in the measured power spectral density of the radiated field at discrete frequencies, since the total power is concentrated in equidistant spectral lines. This could perfectly explain why narrowband spectral components are measured at small frequency intervals, the level of which is independent of the number of charachters on the VDU screen. To verify this assumption measurements of the interference produced by a one single VDU The video4ot-clock of the were carried out. VDU chosen is equal to 11.004 MHz. The system clock frequency is equal to 1.57 MHz, so may be expected in the narrowband components radiated spectrum at 1.57 MHz intervals. Two types of measurement were carried out: interference available - The maximum on The mains power cord was power using the CISPR absorbing measured clamp. - The electric field radiated by the VDU in the direction of maximum radiation - 626 - 6’ I i f--I L :! I 1% 3Q 40 60 80 140 IQ0 290 220 240 figure hb 260 Fieldstrength radiation at Bandwidth frequency I 30 40 60 80 140 100 figure 6 , 160 280 in direction of maximum pal) I meterdisfance.(hor 10 km, VDU screen empty in MHz - I 180 200 Fieldstrength in the direction of maximum radiation at 1 m distance (horizontal pal) 220 240 300 260 280 I - 628 was measured with a biconical antenna according to MILL-STD-4G1/462 in 1 metre distance. In this set-up the mains power cord was shielded. Measurements were taken with a HP 8586 A spectrum analyser in the frequency range 30-300 MHz, at a detection bandwidth of 10 kHz and the function ‘MAX HOLD’ selected. The results are given in figures 5 and 6 respectively. Figures 5a and 6a give the measurement results for a VDU screen full of text and fb and 6b those for a cleared VDU display (only a cursor on the screen). From these results can be seen that: - The level of broadband interference is largely dependent on the number of characters displayed on the screen. - The level of narrowband interference is independent of the contents of the display and individual narrowband components are determined by the VDU system clock and the video-dot-clock. Evidently the theoretical assumptions made above may be considered true. The effect of shielding of the video circuitry In the theory developed in chapter 2 it was assumed that the signal v(t) is a random binary Its amplitude, however, is known to be signal. electron beam is to zero when the equal inbetween two characters. Assuming that the width of each character on screen is n pixels and the spacing between the the video signal two characters is m pixels, as the product of the x(t) can be described random signal z(t) and a square wave z(t) with period (n+m)Th and duty cycle n/(n+m).lOO%. The autocorrelation function RZZ(7) of z(t) triangle. The power spectral is a periodical density of z(t) is obtained through computation of the Fourier transform of R ZZ(r): 2 S,,(f) is obtained through convolution of S,,(f) and S,,(f): It can easily be seen that the video signal only modulated on the narrowband is not components but to all cOmpOnetItS k/Tb , i / (n+m )Tn (which includes all components k/T, ) . Because of the structure of the video signal density of the broadband the power spectral simple the obey will not components sine-function assumed in chapter 2, as can also be seen from figures 5 and 6. In most VDUs a square wave with period to clock out the video (n+m)Tb is necessary successive memory to obtain the coding for the arrays to be changed into a video signal pixel by parallel to serial conversion (see also This parallel to serial conversion figure 1). is synchronized by the video-dot-clock signal. Since the signal with period (nun)Tb is in many VDUs in the range l-2 MHz it is often also used incorporated as the system clock for the Therefore the narrowsystem. microprocessor band components in the radiated electromagnetic with the radiated video field are coherent signal. exists it has a large If this coherence effect upon the demodulation of the received broadband video signal by a TV receiver: In terms of modulation the narrowband components are responsible for the modulation index of the AM signal received by the TV receiver. As the narrowband components in the radiaited electromagnetic field are powerful, they mainly determine the total power received. Shielding of the circuits for video signal processing will reduce the broadband interferwhereas the narrowband interference ence level, level remains unchanged. Thus the modulation whereas index of the received signal decreases the total power received remains unaffected. experiment showed that the TV A concise produce a purely receiver used can still black-and-white picture when the modulation index of the received signal is 0.5%. average Figures 5a and 6a show that the difference between the narrowband and broadband level is 20 dB. As the measurements of figures 5 and 6 were taken at 10 kHz bandwidth, the power of the broadband signal measured at 1 MHz bandwidth will be about 20 dB higher. Thus the signal received by the TV receiver in our measuring set-up is a 100% AM modulated signal, video processing circuitry is not when the shielded. To decrease the modulation index to less than 0.5% the shield must reduce the emission of the video signal more than 46 dB. This explains that the reception quality will not when the video processing change circuitry is shielded less effectively. 5. Conclusions Narrowband components of the electromagnetic interference radiated by video display units in 30-1000 MHz are mainly the frequency range originating from the digital clock signals in the equipment. Broadband components in the frequency range 30-1000 MHz will generally originate from the The level video signal processing circuitry. of broadband emission is therefore dependent on the contents of the video display. A normal TV receiver picking up the interference signal from a VDU will reproduce the displayed information on its own screen, if synchronization signals are provided to the TV receiver. This may lead to a rather annoying type of when interference TV reception, in synchronization information is already received from a broadcast TV signal. This phenomenon may also lead to unwanted reconstruction of or privacy-sensitive information confidential at a large distance from the VDU. If the narrowband components and the broadband components in the radiated emission mentioned above are are coherent the effects more severe and the detection by a TV receiver cannot be prevented adequately by shielding the video processing circuitry in the VDU only. Reference Papoulis, A. : and stochastic Probability, random processes, McGraw-Hill, variables 1965. - OVERVOLTAGE 629 115 - PROTECTION S6 CIRCUITS W. Biichler Meteolabor 8620 Wetzikon, This paper deals with protection measures against overvoltages in electrical conductors resulting from NEMP or lightning. Measurements recorded for discharge elements and protection circuits under NEMP conditions will be presented in addition principles. to generally applicable Switzerland During the first phase the discharge element is converted from a high impedance condition to a low impedance condition. This is purely and simply a voltage problem. Important parameters are: du/dtmax, umax and lu - dt. basic of Overvoltage given to the protection following circuits is the combinations name of primary and secondary protection elements which have at least one longitudinal element for de-coupling various voltage-limiting elements. The scope of this paper is limited to arrangements which as far as possible are universally applicable. Basic principles Protection conductors of overvoltage against overvoltage is achieved the parameters are: sibly li2qdt. di/dtmax, Requirements of overvoltages a imax, not time during destroyed appreciably Residual which and /i .d,t and for discharge in electrical by dischar- phase does current to flow. Important Discharge elements are to the following criteria: as the elements assessed a 1: are is not changed. voltage clear Data on the maximum energy absorption capadity (J) are not useful criteria as it is not I to absorb energy, rather to prevent its conduction, i.e. to reflect it. The ideal protective element (short circuit) absorbs no energy whatsoever. is made between two phases ,“max I I ) t Fig. function element values most important task of a protective Equally useless are certain data over delays' by protective elements. 1 according certain discharge its electrical pos- The voltage across the discharge element during each phase of the overvoltage to be expected. the u second begin I). differentiation (fig. of during the significance Discharge capacity The flow of current in principle discharge Only any protection ging energy pulses to chassis (Faraday cage) and ultimately to earth. Discharge elements such as gas-filled surge protectors, varistors, protective diodes etc. are used for this purpose. In AG Voltage protector and / imax flow 'spark- The manufacturers of protective diodes often give a ‘switching time’ of less than IO-12 seconds. This value applies only to the semiconductor chip. Zt has practically no significance for a protective element as a whole. - 2.Phase current on element in t a gas-filled surge In many practical applications the protec-" tive value of a discharge element is not detkrmined by the theoretical characteristics of the element, rather by the skill with which it is installed. Therefore one should pay as much attention to this aspect as to the evaluatian, of a component. - 630 discharge manufacturers of Unfortunately elements issue little or no useful data on the performance of their products under NEMP condi- - - Each ways manufacturer's 230V surge protector alhad the lowest dynamic sparkover vol90V or 900V surge protectors have tage. higher tions, i.e. with extremely high du/dtanddi/dt. In order to rectify this situation we have carried out tests on the dynamicsparkover filled surge protectors. of gas- - The values. dynamic after sparkover powerful voltage discharge in attenuates the specified range. Gas-filled surge protectors under NEMP conditions - Gas-filled minimum Make Fig. 2: Coaxial pulse voltage for generator a du/dt of 4kV/ns A sure coaxial was specimen to spark used the maximum as gap under switching sparkover steepness normal gap. voltage du/dt gap On air the prestest in a voltage protectors values lay do between Specif. discharge cap. (8/20) kA not always Table 1: Dynamic sparkover __- covering dard range 60 1.4kV and 3.7kV. Sparkover voltage max. X min. (kV) 1.4 2.0 2.1 0.24 0.14 10 20 IO I II III was adjusted of 2-3kV. A commercially available coaxial holder was used for the test specimen. The bandwidth was 400MHz which is equal to a rise time of approximately 0.9ns. Fig. 3 shows a typical voltage curve for a particularly high sparkover level, surge sparkover at the same voltage under uniform test conditions (this depends on the momentary ionization level of the gas). Table 1 shows the results of 60 dynamic sparkovers using 230V surge protectors. Maximum and The dynamic sparkover voltage was measured in the test arrangement shown in fig. 2. 2.3 2.5 3.7 voltage sparkovers. range X: S 1.8 2.4 3.1 (du/dt mean 0.32 = value, 4kV/ns) s: stan- fluctuation. The most important finding of this investithat gas-filled surge protectors gation was have an extremely rapid sparkover under NEMP conditions. The remng dynamic sparkover voltages lie within a range which, as regards insulation, pose no difficulties. Although the dynamic sparkover voltage is higher than the static sparkover voltage (23OV) by more than a factor of ten, the voltage-time integral across the surge protector (and therefore the residual side energy when which voltage smaller than lightning). with Overvoltage Overvoltage Fig. 3: Voltage small curve div., t: for a 9ov Ins/small surge protector. u: 5oov/ div. principle makes in detail, as follows: the results can be summarized protected - NEMP) overvoltages protection - protection of such is (e.g. circuits circuits are a combi- coerse protection Fig. 1 C F of The longitudinal to Ooutput I 4: De-coupling rent through and enables an arrangement. ZL input 0 di/dt value Altogether 110 specimens from three ceramic 8x8mm series (static sparkover voltage: 9OV, 23OV, 9OOV) from three different manufacturers were tested. Between the initial and final sparkover tests, pulse tests were carried out which loaded the surge protectors at the highest discharge impulse current specified by the manufacturers. Without examining the individual slower the rapidly nation of primary (coarse) and secondary (fine) protection elements. Fig. 4 showstheinvariable A voltage rise and drop of approximately 4kV/ns was recorded. We can assume that with a higher bandwidth a somewhat higher peak voltage would have been measured. The second, smaller pulse is due to the inductive voltage drop at the surge protector. The maximum recorded here was over lOOkA/vs; the peak f at 300A. reaches increases primary and element fine protection secondary protection 2~ limits the cur- the secondary protection element the increase in input voltage up sparkover,ifa gas-filled used at the input. surge protector is - Fig. 5 shows the equivalent circuit for a protective circuit with a 9.1V protective diode of the 1.5kW type in conductive condition. 631 - 115 These components are designed units in a Faraday cage which tion protection. The protected sides are Secondary Bipolar thus optimally 5: Equivalent 9.1V circuit protective LL largely of diode a in protection: protective circuit conductive diode with fail-safe = Secondary Capacitor RL must than be the at frequency inductance least of the so large that table protection: . Q-d unprotected 0 1 diode. even protected 0 1 1 23ovprotector with a slow increase in input voltage the protective diode is not destroyed up to just below the static sparkover voltage of the surgeprotector. Normally, RL is designed for a pulse of I-IOms The time-to-half-width value of a duration. (see lightning current is approximately Ims also protected condition high Fig. 7: Diagrams wiring to show diagrams Dstand off Ppeak (Ims) 90V a bipolar RL min surge 230V protectors n V W 6.63 12.9 2.1 3.6 6.5 0.83 1.45 surge protectors n 26.8 6.63 12.9 600 600 600 1500 1500 IN6051 26.8 1500 2.6 7.5 5KP7.0 7.0 13.0 5000 5KP13 0.22 0.40 0.58 1.1 5KP28 28.0 0.69 2.1 various 5000 5000 longitudinal 2: Minimum primary NEMP protection Fig. 6 tection some principle commercial and block protection cir- 2). P6KE8.2 PbKE16 P6KE33 IN6037 IN6044 Table operating of cuits Diode-versions Prot. diode characteristic ;3J" a be- larger the with haviour of the protection circuit and above all the inductive voltage drop at the protective diode. At the highest frequencies (50-IOOMHz) attenuation is achieved by voltage division between LL and LF. LL must be at least 1000 times determines protection as feed-through provides radiaand unprotected de-coupled. unprotected Fig S6 shows circuit and resistances secondary circuits a for 3.9 using RLmin protection for signal commercial 5.3 9.6 18.7 2.1 elements lines single-wire universal application pro[I], ting (6/12/24V, protective voltage diode of +6.6V, see fig. 7) for a maximum +12.9V, +26.8V have operaas re- quired. These models are designed for the protection of signal inputs and outputs without For inputs and outputs electrical isolation. with electrical isolation the protection circuit with capacitor output (C-type) is recommended. The low-pass cut-off frequency (3db drop) lies at approximately 18OkHz. Because of the large gap between the probable useful frefrequency range) and quency range (e.g. voice the cut-off frequency, the protection circuit has no useful circuit influence on line termination frequency range. Thus the is universally applicable. in the protection Fig. 8 shows the internal construction of a 12V (diode-type) protection circuit. The protected and unprotected sides are ideally decoupled by the earth plate. Even with discharge currents of more than 50kA (waveform 8/20) no uncontrollable tected Faraday side. cage voltage surges The radiation is not affected occur on the attenuation by installing pro- of a such a component. surge protector younding disk cted case Fig. 6: Internal protected Fig. 6: Protection circuit (series USSl) coupled. construction and of unprotected a protection sides are circuit. The ideally de- - 632 9 shows the frequency response curve Fig. (50Q insertion transmission loss) of a C-type protection circuit. The uniform curve over the whole cut-off range is due to the good shielding between the input and output, the low parallel capacitance of the impedance coil. - In both With slow voltages traces of really high put voltage rates-of-rise in general higher. are circuit OdE cases frequen- cy voltages at the outputs were practically imperceptible (the upper cut-off frequency of the oscillographs was 400MHz). this not can in the connected approximately worst to earth) 2 3 the With case attain output C-type (filter out- a value of 550V. [VI 600 50dE 500 400 300 lo-500MHz 200 100 OdB n 1 "0 Fig. 11: C-type du/dt With 50dB input 5051 insertion type protection Fig. 10 set-up Fig. transmission shows conditions. It to that output of voltage 5 for loss of a commercial C- circuit the was shown output measured in fig. voltage under NEMP in a similar test 2. WI different input of the less low-pass unhindered. The than 25OV/vs filter output the can pass voltage peak value is equal to the dynamic sparkover of the gas-filled surge protector. 0,5-2,4GHz 9: the rates-of-rise signal practically Fig. circuit, at 4 voltage With models using diodes the output voltage can, in the worst case, attain twice the maximum permissible operating voltage. If voltage increases slowly to just below sparkover of the surge jected protector the to the greatest protective diodes are sub- stress. The discharge capacity of these types of protection circuit is 25kA (waveform 8/2Ops). With a single pulse of 40kA (waveform 8/2Ops), lasting changes to data can arise, although basic operation (normal operation and switching functions) is not affected. 10a Conclusion Fig. Gas-filled surge protectors are highly suitable as primary protection elements even under NEMP conditions. Unlike previous concepts with separately mounted surge protectors and filmodern protection circuits in which ters, coarse-protection,de-coupling and fine-protec- lob tion are optimally adjusted to each combined in a mechanical unit, permit petitively-priced and space-saving other and more comsolutions. Bibliography under Fig. voltage output Fig.: lob: output NEMP voltage conditions for diode-type (24V) circuit conditions for C-type circuit under NEMP Biichler, W. Bosshard: "Blitzschutz elektronischer Get-ate und Anlagen", Eigenverlag Meteolabor AG, 8620 Wetzikon [II w. [2] Meteolabor AG: "UeberspannungsDatenblstter spannungsschutz", Schutzschaltungen und Starkauflicher - 633 116 ST - Iteratlve InterferenceCompensator Simulation for the Division of TWO FM Sisnals Bykhovaky,M.A., Gurianov,G.G. Ministerstvo avjazi SSSR, Mogkva, SSSR Abstract The results of computer simulation study of phyrsicalprocesses taking place Sn the iterative interference canceller (IIC) are reported. The canceller irrdescribed and arimulation resulta are given in the form of threahold CUTVC(J.The data obtained prove8 an IIC capability to separate reliably two FM signalersharing the aame bandwidth. Introduction The development of devicea for separating two 'FMsignal8 sharing a common bandwidth would significantly improve electromagnetic compatibility (EXC) of the existing FM communication Prystems[I and lead to novel communication ayai ems transmitting two FM signalerover a common bandwidth with high noise immunity retained. Iterative interference canceller is one of much promising devioos. The theoretical investigation at S/N exceeding the IIC threslholdlevel iarpresented in 12 . The theoretical analyaim of the I4C operation near a threahold area is rather complicated, and because of thiar,the computer simulation of phylricalprocearee occurring in IIC waa chosen for our study. This method is very attractive, Princeit provide@ a complete analysis of the IIC operation near and above a threlrholdlevel. IIC Description A block-diagram of the IIC under study is ahom in Fig.1. The canceller includea a number of iteration atepa. Two demodulatora of the 1-th step, DW and DEM21, extraot merxrages and weak FM signalar. from?atrong has two outputs, a desired extracted at the low mesaage frequency (LF) output, while an FM signal replica paaaringto the next DJ3M irrextracted at the high fre$&@noy (HF) output. may be realized on the basfa &$ any of the conventional FM demodulators. Its block-diagram, (l?ig.l),includes a limiter (LIM), a frequency discriminator (FD), a low-paisla filter (LPF) and a voltagecontrolled oarcillator(VCO) uaed in tandem. The IIC delay line (DL) HF output provides at the DEM equalization of l?b! Qgnal replica delays resulting from LPF with rerrpeot to FM &gnaln coming to the IIC input and to the input of the next aubtracter, where an interfering FM aignal ir cancelled. Each Dmi;S Simulation alnorithm Basic relationship which are neC)e#aary for the IIC lemulation are given below. A arignalat the IIC input h'arr the form W(t)=Im(A,exp[j(U,t+a,+Y,))+ A2eXP[j(W2t+a2+~2)]+n(t)exp(jqt~],(1) A rsignslat the FD output of DEMil is dencribed by (2) where y (t) isre distortion of phase of e&tracted signal, which ia due to an interfering FM eignal and noise. The voltage at the VCO output af DEMil ia a# followr - 634 - W~~ft)=Im(Aiexp[j(Wit+ai+Q_+ “9;r +Eil)l) 9 wherenW=W2-U,, k=A2/A,< 1, (3) where &.- linear distortionn of mesaage& at the LPI?output, which 3 e negligible, when it% parameters arc properly chosen; E - the renult of filtering '#' (t) inithe LPI?. According d?(1)-(3), yi,(t> ie determined by y,,(t)=Im[ln[l+k exp[j(&.&+ti+&ti] + Y/2l(t)"Im(ln{l~ l (4) l [-j2Sin $!z$exp[-j(AS+Aq- q,J3+ Aa=a2’al, Av=(P2 -$$ When simulating the IIC operation, all physical procesreesare arimulated by the use of software. Sampler from a (t) and n(t) were obtained by mean8 of faat Fourier transformation on the ba&s of frequenoy samples obtained according to energy srpectrumform. y/ (t) filtering (the calculation ofi&l(tN may be suitably written akl w&re F(f), F"(fj-- direct and inverse PO ier transformationa of f function;+"TD(w)f=H(U)D(@) - filtering of a procearawith D(w) spectrum by LPF with frequency reaponne H(U). According to (51, filtering rsimulation may be represented by the following aeauenoe of calculationa of: 1)"a file ofyil(t) samplea calculated bs (411 2) 3) - ted by a formula S (w)=H(W)$,(eJ); 4) a file of c. (t> samples calculated using f&&,Fourier transformation Eil' _F {S&)j. A FORTRAN-program for simulating the IIC operation was developed on the basis of (4)-(5). The algorithm takes account for changes in the following parameters: I) message bandwidth ai(fbi; 21 effective index of &d modulation mj,; shift of central frequencies 3) between FM signals -AN; 4) ratio between signal amplitudes k; 5) ratio between power of the l-at PM signal and noise power in fbi bandwidth - p. Provisions for variations in LPF frequency response are also made. The validity of simulating physical processes in IIC by means of the developed algorithm was confirmed by the comparison of an average value of threshold pulses at the DEM,, output when receiving a single FM signal, with theoretical data calculated by the Rice equation, which is known to have a good agreement with experimental data. I.*.._.“.___” 635 - 116 ST Simulation results The program developed was used to obtain the relationship,E. (p), between a mean power of in?.&modulation product, resulting from the evaluation of the i-th signal at the LF output of the IIC 1-th step and S/N-p at the IIC input. The dependence, &. (p), agrees well with the theoretic&& results at the area above the threshold level (21. naln and IIC. Fig 2 presents this dependence for the case of an ideal LPI?,in IIC,when f =fb2, m,=m2, Fig.2, that kxO.1. It is meen &om the separation of two FM signals takes place,when p>pn . The value ofp, may be called a threshold for IIC. The analysis of curves for strong FM signals reveals that from the second step of iteration there are two values of p (p,, pn; with ), above which the signal rec&$8$% is markedly improved. This results from reaching a threshold in the first (p,) and in the second (p ),frequency discriminator, It may b& shown, that the IIC threshold may be estimated by the following approximation formulae,(dB) 636 - P”=-20 lg k -I10+15, when k (0.5 , when k)0.5 Pn =-20 lg(l-k) This estimation is supported by the curves of the figures presented. Fig.2 show&that extraction of desired messages in XIC becomes greatly improved (by lo+15 dl3)at p>>l, when l-2, rather than l=l. However, the use of l--3instead of 1~2 does not significantly increase noise immunity at the mentioned values ofp . This results from the fact, that after the second step of iteration, the level of interfering signals is lower than the level of thermal noise. Because of this, at the subsequent steps of iteration the ;;;;z $&lo;;;; $;i~~;;~;o;Yb~e reduced by increasing the number of iteration steps. Consequently, an additional step of iteration is effieient,only if at its input the level of an interfering signal is higher than that of thermal noise. Simulation results reveal an improvement in the IX! noise immunity at p>>l and rise in m. This is conditioned by the widening of the bandwidth occupied byY/. (t). If the L3?Fpassband ia fixed, %&en this leads to 3 reduction. SimulHtion of IIC at ml=ms=md has shown that two FM signa s not be separated at any modulation indices, but only if mb2.5. When m<2.5, the aeparation of two FM signals does not occur in IIC. It is interesting to note, that simulation results revealed the possibility of separating two FM signals of approximately the same levels (ksO.99, Fig.3). It follows from Figa. and 3, that when extracting a strong l?Msignal, IIC may provide a significant gain in noise power which is equal to IO-30 dB. Conclusion The results given above prove a significant potential of IIC to separate reliably two FM signals. This opens the way for enhancement in EMC of FM communication systems as well in their capacity without widening a common bandwidth. References Bykhovsky M.A.: Odnokanalnie kompensatori pomekh v aistemakh svyazi. (Single-channel Intcrference cancellers used in Communication Systems), Radiotekhnika, No.11, 1981 Bykhoveky,Y.A.: Razdelenie dvukh ChM signalov 8~ pomorrh.iu iterazionnogo kompensatora; (Separation of two FM Signals by means of Iterative Interference Canceller), NIIR papers, No.2, 1982