Methods of 100 dB RF Performance Verification of Facility Filters Thomas D. Geist Kermit 0. Phipps Power ElectronicsApplication Center Knoxville, Tennessee Power Electronics Application Center Knoxville, Tennessee EMI/RFI facility filters specified to 100 dB are usedin governmentand commercialinstallationsto prevent unwantedsignalson power lines. The main standardusedto measurefilter insertion loss (synonymousfor matched impedanceattenuation),is MIL-STD-220A. But becauseMILSTD-220Awas developedonly for matchedimpedance communicationsystems,it has beenlabeledunrealistic. Since then,otherstandardsand methodshavebeendevelopedthat attemptto model “real world” attenuationby requiring varying sourceand load impedance.To date,however,noneof these measuringmethodshave replacedMIL-STD-220A. This paper supportsthe continueduseof the much-beleaguered MIL-STD220A as the preferredchoicefor high frequencyperformance characterization,Additionally, this paperillustrateswhy none of the existing methodsare sufficient to fully verify 100 dB attenuation.Furthermore,in recognizingMlL-STD-220A’s shortcomings,this paper suggeststhe developmentof a new facility filter standardthat addressespower quality issues, critical RF performancefactors,and filter/shieldedroom equipmentincompatibility issuesthat previously have not beenconsidered. h&IMODS Abstract: BACKGROUND Since 1995,the EPRI Power ElectronicsApplications Center (PEAC) hasbeenengagedin evaluatingfilter standardsand methodsof filter characterization.Theseefforts are directedat bringing aboutstandardizedtest protocolsfor the effective measurementand specificationof facility filters. This paper reportsthe results of the project’s investigationinto facility filter RF performancefor filters rated at 100 dB. After conductinga literaturesearch,PEAC selectedfour standards and/ormethodsthat apply to 100 dB RF performance characterization, They are: 1. MIL-STD-220A 2. CISPR 17 3. ARP-4244 4. Current Injection ProbeMethod Although the Current Injection ProbeMethod appearsas part of ARP-4244, it is not itself a standard.The use of AC power differentiatesthe Current Injection ProbeMethod enoughthat it warrantsseparateattention. 0-7SO3-j015-4/9Sl$lO.~~0 1998IEEE MIL-STD-220A Implementedin 1959 by the United StatesArmy Signal corps, MIL-STD-220A, Method of Insertion Loss Measurement[ 11, was designedto rate Radio FrequencyInterference(RFI) capacitors,which are usedwith some mobile radio transceivers when they experienceradio interferencefrom the automobile’s ignition system.MIL-STD-220A test setup is shown in Figure 1. However, MIL-STD-220A, basedon 1952’sMILSTD-220and on previous filter capacitorstandardsdating from the beginningof military radio communications,is commonly usedtoday to rate industrial and commercialfacility filters. Figure 2 shows a low-frequency plot of a facility filter performancecurve as measuredusing MIL-STD-220A. The test methodspecifiedin MIL-STD-220A was developedfrom techniquesfor measuringinsertion loss in matchedimpedance communicationsystems,typically 50 or 75 ohms. However, most applicationsof filters are in areaswhere impedanceis not matched.Thus, measurementsmade using MIL-STD-220A test methodswill yield different insertion-lossvalueswhen made in an unmatchedsystem,particularly at low frequencies-100 kHz and below. Even though this problem has beenrecognizedfor more than thirty years,filter manufacturersand userscontinueto rely on MIL-STD-220A test methodsto specify performance. 159 I I I I I- -1 Tracking Generator Spectrum Analyzer Unit Under Test /////////////////////////// Ground Plane Figure 1. MIL-STD-220A Test Setup. achievean appreciablefrequencyrange-10 kHz to 10 MHz-a compromisemust be reachedbetweena high turns ratio, resultingin reducedbandwidth, and a large voltagedivision, resulting in high insertion loss. The impedancetransformers, therefore,have a 50 dB insertion loss [4]. G iven realistic measurementequipment,CISPR 17 providesgood resultsfor about 60 to SOdB of dynamic measurementrange.However, the transformerinsertion loss effectively restrictsthis method from useon 100 dB filters (seeFigure 4). 100 50-bhm insertion loss - 100,000 10,000 1,000 Frequency (Hz) Figure using CISPR 2. Typical performance MIL-STD-220A. curve as measured 17 Answeringthe needto addressunmatchedimpedance,CISPR 17 specifiedproceduresfor measuringfilters underconditions of mismatchedload and sourceimpedance.It provideda means of performingmeasurementsin both symmetrical(common) and asymmetrical(differential) modes[3] [4]. Createdin the mid 197Os,The InternationalElectrotechnicalCommission developedCISPR 17, Methods of Measurement of Suppression Characteristics of Passive Radio Intelference Filters and Suppression Components [2], to provide a universalstandardthat would allow variouslaboratoriesand manufacturersto comparefilter measurementresults.It is basedon the conceptof combining extreme-valuetheory of mismatchwith a statisticaldata baseof load and source impedance[3] [4]. Becausethe method,shownin Figure 3, measuresfilter performanceusing extremevaluesof sourceand load impedance,it is usually called the worst-case,or extremevalue, test method.After initial measurements are made,the TrackingGeneratorand SpectrumAnalyzerarereversed. BecauseImpedancetransformersare not readily available commercially,and becausethe methodis limited in dynamic range,CISPR17hasneverbeen widely used.In order to Figure 4. CISPR 17 Measurement showing dynamic range due to impedance transformer insertion loss. ARP-4244 As anothermore recentattempt to rectify the shortcomingsof MIL-STD-220A, a committee of the Society of Automotive Engineers(SAE), AE-4P, developedthe Aerospace Recommended PracticeARP-4244standard.In this standard, sourceimpedanceis changedby addingor removinga LineImpedanceStabilizationNetwork (LISN), and load impedance is changedthrough an adjustableresistivebank [5]. In the setup,shown in Figure 5, R.Fcurrent is injected into the systemby an RF current injection probe (CIP) and the measurementof attenuationis made with an W-receiving currenttransformer(CT). This CIP methodis an enhancement of the buffer networksusedin MIL-STD-220A for load testing andprovidesa wide-frequencymeasurement range. Measurements are taken ,c----.-----.-.-----,---. in this configuration __- Ground Plane Figure 3. CEPR limited 17 Test Setup. RF Port Terminated with 50 Ohm 16” 10 pfh Isolated DC Power SUPPlY <, 50 pH LEN + /\ 30 mf short as practical and two inches above copper groundplane, except at terminations, where different height is mandated. Figure 5. ARP-4244 Filter t Generator RF Port Terminated with 50 Ohm Setup for Differential The objective of testing a filter under load is to determine if the seriesinductor saturates. However, as the ARP-4244 standardpoints out, it is not necessaryto use the CIP method to test for saturationabove the inductor resonantfrequency, because at that frequency,the inductors are effectively capacitors,as shown in Figure 6. Therefore, above the resonantfrequency, which is typically less than 5 MHz, the impedanceseenby a noise source is actually the line impedanceand not the seriesinductor. As per the standard,line impedanceabove 5 MHz is generally between 40-300 ohms. Thus, the use of the 50-ohm impedancetest method, MILSTD-22OA,is a reasonableprocedureabove 5 MHz. Pad RF Performance With 3 dB Pad Measurement. However, ARP-4244 is vague in its requirement for testing differential-modeperformance. The test procedurerequires adjusting the resistive load to produce the dc peak equivalent of rated current. Unfortunately, this current requirement is open to interpretation.Nonlinear loads can produce either less peak current or more peak current than a resistive load dependingon the design of the filter series inductor. If, for example, the inductor is designedfor nonlinear current, the inrush due to nonlinear loading will cause voltage flat topping that will limit the peak current. If, on the other hand, the inductor is not designedto support the nonlinear current, the inductor will saturateand the peak current will be higher than the resistive load. In both cases,however, the current peak will be higher than the peak current for a sinusoidal-shapedcurrent waveform. Figures 7 and 8 show the voltage and current wave shapesof a thirty-amp 100 dB facility filter under twenty-five percent switch-modepower supply loading (nonlinear). Given the uncertainty as to what the rated current, or more correctly proper “measurement”current should be, more work is needed to define a method that measuresRF performanceunder “real world” loading conditions. 5 MHz Frequency Figure 6. Typical frequency. curve showing impedance versus 1Gl 200 Current ‘--- Injection Probe ‘Technique Another method of eliminating the low-frequency matched impedanceproblem of MIL-STD-220A is to use current probes for both the injection and reception of RF signals. A typical setup is shown in Figure 9 (A good description of the method can be found in [6], [7] and [8]). As in the SAE ARP-4244 method, the CIP method allows for sourceimpedancevariation by adding an LISN or an air-core inductor in serieswith the feed-throughcapacitor (not shown). 150 --ml2 -100 9\ -150 I -200 Figure 7. Flat-topped with 25% nonlinear voltage loading. of a facility 20 ---- filter --- - 15-----A10 - / I \\ Crest Factor = 2.05 \ --_ -20 Figure 8. Current wave-shape with 25% nonlinear loading. of a facility filter While not a new idea-its beginnings can be traced back to at least the mid 1960s-a survey of filter manufacturersindicates that its use has been very limited, perhapsdue to the difficulty in obtaining a high dynamic range. A 100 dB measurement, for example,requiresa CIP capableof transferringappreciable amountsof power from the signal source to the test circuit. This then leads to difficulties in shielding the current injection probe so that it does not radiate and couple to the filter output. Additionally challenging is the elimination of erroneous measurementsdue to sneak ground paths, unintentional radiatedcoupling, and other undesiredpaths. On the other hand, the advantageof the CIP method is that it allows variation of filter sourceand load impedancesto any desiredvalue so that actual installation conditions can be simulated. This simulation means that testing under AC loading using both linear and nonlinear loads is possible. Figure 10 shows that a difference in measuredfilter performancemay exist between resistive (30 Amp R) and nonlinear loading (30 Amp PC, as noted in Figure 10) for two different filters. The difference in measuredperformance, however, is a result of different filter design quality. 162 Figure 9. Typical CIP measurement setup shortcomings.They each fall short in either the frequency rangeover which measurementsare madeor becauseof an 70 80 100,000 inability to test under realistic load conditions. However, Table 1 illustratesthat when used togetherthere are two methodsthat can accommodateall the needsof RF performance measurements for facility filters rated at 100 dB: (1) the CIP methodallows for realistic load testing at lower frequencies (down to 10 kHz and below), and (2) MLSTD-220A allows for measurementof RF performanceat high frequencies(above 5 MHz). 1 ,ooo.ooo Frequency(tk) i Figure 10. CIP measurement results for linear nonlinear loading, for two similar filters with different manufacturers. Which method should Characterization? be used for and RF The backgroundresearchdescribedin this paperalong with extensivetesting conductedby PEAC shows that no existing facility-filter RF performancetest standardor method is completelyadequate.Each standardor methodhasone or more WHAT’S NEXT The researchrepresentedin this paperis part of the ongoing effort to provide a standardizedmeasurementset and realistic specificationfor facility filters. The creation of a new standard that will provide meansfor addressingfacility-filter power quality issues,critical RF performancefactors, and filter/shieldedroom equipmentincompatibility issuesis a logical next step to this research.What is neededis a standard with broadindustry recognition.The existing standardsfall short. MIL-STD-220A specifies only 50-ohm impedanceand, therefore,does not accuratelymeasureperformanceat frequenciesbelow 5 MHz. Rather than changing MIL-STD220A, a new non-military standardis neededfor facility filter RF performance verification, but as yet, none exists. CISPR 17 is a European standard,and is also not well-known in the United States.Similarly, becauseSAE ARP-4244 is an automotive standard,other industries may not be aware of it. Table 1. Comparison of facility Frequency Range 5 MHz - 22 GHz AC Loading MIL-STD-220A CEPR17 ARP-4244 CIP Method Yes Yes Yes Yes Yes No No No No No No Yes RF1 Corporation, Sy Rubin and Robbert Hassast Fil-Coil, Lester Hammer, Paul Alessandrini and Romesh Rajput 6 Elmag Division l Mike Leone MTK. Gil Geiger l LCR Electronics, of Lindgren, Terry Forsyte and Nissin Isankof and Jerry Womer REFERENCES [I] RF performance Frequency Range lOkHz-5MHz We would like to thank the following filter manufacturesfor their time, cooperation, efforts and full support in this project. l filter Standard or Method ACKNOWLEDGMENTS l Given the broad range of filter uses-medical, information technology, government, and others-the need exists for international and national standardssupportedby both IEEE and ANSI. Once this standardis developedand adopted,filter userswill no longer need to rely on inadequatestandardswhen specifying and purchasing facility filters. Instead, filter users will have a new standardthat addressesthe needsof today and tomorrow. MlL-STD-220A, Method of Insertion Loss Measurement, 15 December 1959. [2] CISPR 17, Methods of measurement of suppression characteristics of passive radio interferencefilters and suppression components, Geneva: IEC, 1981, Section 4.2.2.2. methods Varying Source and Load Impedance No Yes Yes Yes [3] H. Weidmann and W. J. McMartin, Two Worst-Case Insertion Loss Test Methods for Passive Power-Line Interference Filters, [4] A. Tucker, 100/O. 1 Ohm Filter Terminations and Other Extreme-Valued Affects, 1989 International Symposium on EMC, 8-10 Sept., 1989, Nagoya, Japan, p. 828-833, Vol. 2 [5] AerospaceRecommendedPracticeARP-4244, Recommended Insertion Loss Test Methods for EMI Power Line Filters, Preparedby SAE Committee AE-4P, 9 July 1996. [6] J.R. Fischer et at., Evaluating jilters in situ under heavy load currents and normal working impedances, presentedat the IEEE Electromagnetic Compatibility Symposium, Washington, DC., July 1967. [7] Robert E. Hassett,EM Filter Characteristics and Measurement Techniques, InterferenceTechnology Engineers’ Master (ITEM) 1997. [8] S. Eisbruck and F. Giordano, A Survey of Power-Line Filter Measurement Techniques, IEEE Transactionson Electromagnetic Compatibility, Vol. EMC-10, No. 2 June 1968. 16-c