Simple Models of EMI Filters for Low Frequency Range

Transcription

1 8 J. DŘÍOVSKÝ, J. SVAČA, Z. RADA, SMPE MODES OF EM FTERS FOR OW FREQECY RAGE Smple Models of EM Flters for ow Frequency Range Jří DŘÍOVSKÝ, Jří SVAČA, Zbyněk RADA Dept. of Rado Electroncs, Brno nversty of Technology, Purkyňova 8, 6 Brno, Czech Republc drno@feec.vutbr.cz, svacna@feec.vutbr.cz, rada@feec.vutbr.cz Abstract. Ths paper deals wth mathematcal smulatons of EM flters performance. These flters are commonly used for the suppressng of electromagnetc nterference whch penetrates through the power supply networks. The performance of these flters depends on termnatng mpedances whch are plugged to the nputs and outputs clamps of the EM flters. Ths paper descrbes the method by whch t s possble to calculate the nserton loss of the flters. The method s based on the modfed nodal voltage method. The crcutry of the EM flters s used for ther descrpton. The effect of spurous components s not taken nto account. The flter tself s descrbed by set of admttance parameters, whch makes the presented method more unversal. The calculated results were compared wth measured data of several flters for several mpedance combnatons. Dfferent test setups, lke asymmetrcal, symmetrcal, etc. were taken nto account. The smplcty and accuracy of the presented method s dscussed n the concluson. The acheved accuracy s on hgh level. The descrbed method s unversal, but for flters wth more than one current compensated nductor, the mentoned method s complcated. The sze of the fnal equaton for calculatng the nserton loss rapdly ncreases wth the number of current compensated nductors. Keywords Electromagnetc compatblty EMC, EM mans flter, nserton loss, mpedance termnaton, flter model, modfed nodal voltage method, current compensated nductors.. ntroducton The EM flters are one of most often used tools for suppresson of electromagnetc nterference (EM) whch penetrates through the power network. The foremost task of the EM flters s usually to ncrease the mmunty of an electronc equpment operated on power lne nputs and smultaneously to decrease the level of hgh frequency emssons suppled by the equpment nto outer power network. The most mportant characterstcs of the EM flters are values of nserton loss or RF attenuaton. The nserton loss of the EM flters depends on the current frequency of the nterference sgnal. The performance of the EM flter depends on the current mpedance termnatng the flter termnals, current load, etc. n general, there are lot of parameters, condtons and nfluences whch degrade the nserton loss of the EM flters. Coverng all these possbltes n measurng setup and n the data sheets n catalogues s nearly mpossble. The EM flters usually have more then two clamps at the nput and output sde, e.g. the sngle-phase EM flter has mnmally these three clamps (lve, neutral and earth) at the nput and also at the output sde. Ths stuaton complcates the correct presentaton of the EM flter performance data. Thus, t s possble to determne several confguratons by whch nterferng sgnals should penetrate through the power network. Ths stuaton s depcted n Fg.. Three dfferent measurements could be carred out a on snglephase EM flter. The most common presented data have been measured n asymmetrcal mode (a) n Fg. ). Several producers gve the data for the system b) (Fg. ), whch s called symmetrcal. t s also possble to determne the system, whch s depcted n Fg. c). For the purpose of ths paper, ths system wll be denoted as a non-symmetrcal system. The measurements tself were done accordng to the ČS CSPR 7 and as sgnal recevers the HP E744A and HP 35665A spectral analysers were used. The R&S SM3 and Aglent 33A were used for sgnal generatng. The producers of EM flters usually gve the nserton loss characterstcs for the asymmetrcal system and for the mpedance system 5 Ω/5 Ω [], []. The frst number means the value of the termnatng mpedance on the nput termnals of the EM flter. The second number refers to the output termnatng mpedance. Several producers sometmes gve the data also for the symmetrcal system for several termnatng mpedances, e.g.: 5 Ω/5 Ω;. Ω/ Ω and vce versa. The nserton loss measurng wth the termnatng mpedances. Ω/ Ω and vce versa s defned by the nternatonal standard CSPR 7 Methods of measurement of the suppresson characterstcs of passve rado nterference flters and suppresson components [3]. Schaffner company, whch s the wdely known producer of EM fl-

2 RADOEGEERG, VO. 7, O. 3, SEPTEMBER 8 9 u u u G 5 W G 5 W G Flter 5 W 5 W : : Flter Flter 5 W 5 W SA u SA 5 W SA 5 W Fg.. Possble nserton loss measurng systems: a) asymmetrcal; b) symmetrcal; c) nonsymmetrcal. ters, publshed applcaton note [4] where measurng defned by [3] s clearly descrbed. Smlar standard s defned by the Department of Defence of the nted States of Amerca [5] where smlar measurng setup lke n the CSPR standard [3] s descrbed. The systems wth the termnatng mpedances. Ω/ Ω and vce versa are recommended by [3] for testng of the worst-case. n other words, the EM flters wll have, n ths system, the lowest nserton loss. Thus ths system s called an approxmate worst-case system. Ths method wth the. Ω/ Ω (vce versa) termnatng mpedances s not so wdely used by flter s producers, probably because the usual sgnal generators and spectral analysers are matched to 5 Ω mpedance. For measurng n. Ω/ Ω and vce versa systems t s necessary to use two mpedance transformers. These transformers transform the 5 Ω to. Ω and Ω. The transformers could have two types of outputs: balanced or unbalanced.. Smple Models of EM Flters The comparson of several dfferent flters from several companes could not be easy, as t s wrtten n the ntroducton. The bggest problem s to obtan data for dfferent flters and n the same tme for several combnatons of mpedance termnatons. One possble way could be through the analyss of flters models. Ths approach has several advantages. The method based on the models s not so tme consumng as a lot of measurements wth dfferent mpedance termnatons. For makng the models, plenty of commercal software could be used (PSpce, Mcro-Cap, Ansoft Desgner R, etc). These all software u u a) b) c) systems are very specalzed for specfc tasks. The EM flters performance analyss should be very unversal and varable because the confguraton of each flter s very varable. The crcutry knowledge of the certan EM flter s other precondton, whch should be fulflled. For that reasons, the Matlab R was chosen for ths analyss. sng of the Matlab R brngs unversalty because the determnaton of the nserton loss relaton uses only general Y parameters. These parameters could be effectvely changed n relaton wth the crcutry of the concrete EM flter. The basc sngle-phase EM flter, whch dagram of connectons s depcted n Fg., could be descrbed by the followng equatons = Y + Y + Y 3 + Y 4, () = Y + Y + Y 3 + Y 4, () = Y 3 + Y 3 + Y 33 + Y 34, (3) = Y 4 + Y 4 + Y 43 + Y 44, (4) where x s the nput current for clamps or, x s one of the two output currents. n the same manner the nput and output voltages are determned as t s shown n Fg.. The Y xy s the admttance parameter of the tested EM flter. The sngle admttance parameters n admttance matrx Y could be easly calculated by the modfed nodal voltage method. By these admttance parameters t s possble to construct the admttance matrx Y. Equatons () to (4) could be rewrte n to the matrx form = Y (5) where s the vector of the unknown currents, and s the vector of the varable voltages. The equatons () to (4) exactly descrbe the propertes of an arbtrary EM flter, but for correct calculatons, t s necessary to add more equatons whch wll refer to confguratons of the mpedance network (Fg. ) and to the locaton of the source of the nterference sgnal. The nserton loss data are obtaned after calculatons of these several equatons. The real frequency on whch the nserton loss of the tested flter s calculated s also ncluded n each element of matrx Y. By ths method t s possble to determne the nserton loss data, e.g. ( n db). Flter Fg.. Tested flter and dstrbuton of currents and voltages. The usual EM flters nclude the current compensated nductors, whch are not easy to descrbe by the modfed

3 J. DŘÍOVSKÝ, J. SVAČA, Z. RADA, SMPE MODES OF EM FTERS FOR OW FREQECY RAGE nodal voltage method. Ths method fts well for descrpton of smple and lnear electronc crcuts. The above shown method has to be extended for correct determnaton of admttance matrx Y of the EM flters. The admttance matrx Y has to be enlarged by two columns and two lnes. The nfluence of the current compensated nductors s wrtten nto the added cells. By ths step the equatons () to (4) wll be added up by the followng two equatons ab = jω + jωm, (6) cd = jωm + jω (7) where the meanng of varables ab, cd, a to d s obvous from Fg. 3. The constants and M represent own and mutual coeffcents of nducton of the current compensated nductor. The relatonshp between these two quanttes s gven by M = k, (8) where k s the couplng coeffcent. The values of teh own coeffcents of nducton and are commonly the same for most of EM flters ( = ). a a b ab a b b M Fg. 3. Current compensated nductor wth the mutual coeffcents of nducton. The electromagnetc crcut gven n Fg. 3 could be descrbed by the followng equaton [6] a b c = d jω jωm d c d jωm jω d c cd c a b c d. (9) The admttance parameters have to be added nto the equaton (9). The nfluence of current compensated nductors s not taken nto account. The fnal obtaned matrx of the flter could be wrtten as followng Y Y Y 3 Y 4 Y Y Y 3 Y 4 Y 3 Y 3 Y 33 Y 34. Y 4 Y 4 Y 43 Y 44 jω jωm jωm jω Ths presented matrx s deduced for sngle-phase EM flter whch contans only one current compensated nductor. For the descrpton of flters wth more nductors t s necessary to create a bgger matrx. Ths fact rapdly reduces an effcency and degrades unversalty of ths analyss. More unversal method could be made by usng a frmly set of the matrx dmenson. The smallest dmenson of the matrx could be 4 4 because sngle-phase flters have usually nput and output clamps. Thus, t s possble to produce a unversal relaton for nserton loss calculaton, whch depends only on the admttance parameters of the flters (Y to Y 44 ). These parameters are defned for nput and output nodes (clamps) of the EM flter. The rest of the nodes has to be reduced onto the dmenson 4 4. For ths reducton t s possble to use the pvot condensaton [7]. The prncple of the reducton s possble to wrte down n ths mathematcal form M R = M E M E (M ) M E. () Matrces M E, M E, M and M E were created from the admttance matrx Y of the EM flter by the followng way [ ] [ ] [ ] M E M E M E M X X E = E, () where X and X E represent the nternal and external unknowns. The and E represent external sources. The matrx M R s the fnal reduced matrx after pvot condensaton. Ths matrx has the desred dmenson of 4 4. Each matrx element s afterwards establshed nto the relaton for calculatng the nserton loss data. Ths element depends on the frequency. By the descrbed method t s possble to calculate nserton loss of sngle-phase EM flters. The method could be modfed for mult-phase flters. Ths setup calculates only wth the data whch are wrtten n the data sheet. From ths condton t follows that calculatons of nserton loss are not possble on hgher frequences, because n ths setup, spurous propertes of real electronc parts and devces are not covered. The value of couplng coeffcent k should be set by measurng or by optmzaton. The measured nserton loss data, e.g. n 5 Ω/ 5 Ω, whch should be gven n data sheets could be used for ths optmzaton. The equaton for the calculaton of the nserton loss can be determned for each system shown n Fg.. The smlar setup can be used for the estmatng of the nserton loss wth dfferent mpedance termnatons [8].

4 RADOEGEERG, VO. 7, O. 3, SEPTEMBER 8 3. Results The method mentoned above was tested on seven flters: Schurter 5.33., Schaffner F 3 /5, F -6-6, F 7--6, Elfs EF6V, EF6VY- 4, and Fltana TS 8 6. The calculated nserton loss data were compared wth measured ones. Several results and examples wll be dscussed n the followng paragraphs. Frstly, the nserton loss of the Schurter EM flter, whose nner crcutry s depcted n Fg. 4, was tested. The data sheet of the flter could be found n [9], for comparng the results. The nserton loss performances of ths flter are shown n Fg. 5, 6 and 7. n these fgures, the measured data are compared wth the calculated nserton loss of ths flter. The flter was tested n several dfferent systems and also wth dfferent mpedance termnatons. The usual mpedance termnatons lke 5 Ω/5 Ω,. Ω/ Ω and / Ω/. Ω were tested, and several mpedance systems, whch are not so common lke. Ω/. Ω and Ω/ Ω were thought too. From the presented data depcted n Fg. 5, 6 and 7, t s possble to say that the performance of the descrbed method s really hgh..4 mh 47nF(X).4 mh. nf(y). nf(y) Fg. 4. Typcal sngle-phase flter s crcutry (Schurter 5.33.) k k k M M M G measured 5 / 5 calculated 5 / 5 measured. / calculated. / measured /. calculated /. measured. /. calculated. /. measured / calculated / Fg. 6. The nserton loss of the Schurter n symmetrcal systems. measured 5 / 5 - calculated 5 / 5 - measured. / - calculated. / - measured /. - calculated /. - measured. /. - calculated. /. - measured / - calculated / - k k k M M M G Fg. 7. The nserton loss of the Schurter n non-symmetrcal systems measured 5 / 5 calculated 5 / 5 measured. / calculated. / measured /. calculated /. measured. /. calculated. /. measured / calculated / The measured and calculated performance data are shown n Fg. 9,, and. f we compare the measured and calculated data, the great accuracy n the doman of low frequences can be seen. mh. mf MW. mf k k k M M M G. mf MW mh Fg. 5. The nserton loss of the Schurter n asymmetrcal systems. Fg. 8. The sngle-phase flter s crcutry of Elfs EF6V. For the second example, the Elfs EF6V flter was chosen. The data sheet wth the techncal data could be found n []. The basc crcutry of ths flter s depcted n Fg. 8. Ths schema s n fact the mrrored schema of the Schurter EM flter. The dschargng resstors are added up. The last example s gven for the Schaffner F EM flter. Ths flter contans two current compensated nductors. These two nductors mprove nserton loss of the flter, but the mathematcal descrpton of ths flter s more complcated. n practse, the equaton for calculat-

5 J. DŘÍOVSKÝ, J. SVAČA, Z. RADA, SMPE MODES OF EM FTERS FOR OW FREQECY RAGE measured 5 / 5 calculated 5 / 5 measured. / calculated. / measured /. calculated /. measured. /. calculated. /. measured / calculated / measured by the producer could be also found there. The crcutry of ths flter s gven n Fg.. The measured and calculated data are depcted n Fg. 3, 4, and 5. From Fg. 3, 4, and 5 s obvous that the performance of mathematcal models s very good n low frequency range. Ths s due to the neglectng of the spurous components of each electrcal components of the EM flter. 4.5 mh 4.5 mh - k k k M M M G Fg. 9. The nserton loss of the Elfs EF6V n asymmetrcal systems. mf (X) 4.5 mh kw mf (X) 4.7 nf (Y) 4.7 nf (Y) 4.5 mh measured 5 / 5 calculated 5 / 5 measured. / calculated. / measured /. calculated /. measured. /. calculated. /. measured / calculated / -3 k k k M M M G Fg.. The nserton loss of the Elfs EF6V n symmetrcal systems Fg.. Typcal sngle-phase flter s crcutry schema wth two current compensated nductors (Schaffner F 7--6). measured 5 / 5 calculated 5 / 5 measured. / calculated. / measured /. calculated /. measured. /. calculated. /. measured / calculated / k k k M M M G measured 5 / 5 - calculated 5 / 5 - measured. / - calculated. / - measured /. - calculated /. - measured. /. - calculated. /. - measured / - calculated / - Fg. 3. The nserton loss of the Schaffner F n asymmetrcal systems k k k M M M G Fg.. The nserton loss of the Elfs EF6V n non-symmetrcal systems. ng the nserton loss s ten tmes bgger than the equaton used for Schurter flter. The data sheet of the Schaffner F 7 could be found n []. The real data 4. Conclusons Ths paper deals wth the smple mathematcal models of sngle-phase EM flters. By these models t s possble to descrbe whchever EM flter, but the nner crcutry of the flter has to be known ncludng the quantty of each electronc component. The flter tself s descrbed by admttance parameters. The spurous components have not been taken nto account. The nserton loss of the EM flter was deduced by usng the modfed nodal voltage method. The bggest problem s wth the unversalty of the method because the EM flters have not the same number of nodes and the current compensated nductors added up several rows and columns nto the admttance matrx Y. For that reasons, the fnal admttance matrx should be reduced by the pvot

6 RADOEGEERG, VO. 7, O. 3, SEPTEMBER measured 5 / 5 calculated 5 / 5 measured. / calculated. / measured /. calculated /. measured. /. calculated. /. measured / calculated / -3 k k k M M M G Fg. 4. The nserton loss of the Schaffner F n symmetrcal systems. measured 5 / 5 - calculated 5 / 5 - measured. / - calculated. / - measured /. - calculated /. - measured. /. - calculated. /. - measured / - calculated / - - k k k M M M G Fg. 5. The nserton loss of the Schaffner F n non-symmetrcal systems. condensaton. The user has to be careful durng the condensaton and does not condensate the outer nodes where the flter s connected nto the outer system: power supply network at the nput of the EM flter and outputs for connectng the protected devce. After ths step the reduced admttance matrx Y s obtaned. Ths matrx has the dmenson of 4 4. The dmenson corresponds wth the outer clamps of the sngle-phase EM flters. The performance of the descrbed method was tested on seven sngle-phase EM flters from several producers wth a dfferent crcutres. There were: Schurter 5.33., Schaffner F 3 /5, F -6-6, F 7--6, Elfs EF6V, EF6VY-4, and Fltana TS 8 6. The testng was done n asymmetrcal, symmetrcal and non-symmetrcal confguratons. The followng combnatons of mpedance termnaton were used 5 Ω/5 Ω,. Ω/ Ω, Ω/. Ω,. Ω/. Ω and Ω/ Ω were used. The odd clamps n nonsymmetrcal mode were termnated by the 5 Ω n all cases. The nserton loss data were measured and calculated for these mpedance systems. These data for three flters are gven n Fg. 5, 6, 7, 9,,, 3, 4, and 5. After comparson of the calculated data wth the measured ones, t s possble to conclude. The smple models work qute well n the range of low frequences. Ths fact s caused by gnorng the spurous propertes of real electronc parts. But on the other hand, the accuracy n ths low frequency range s really good; under 3 db n the most cases. The advantage of the descrbed method s unversalty caused by usng the modfed nodal voltage method and by usng the admttance matrx for descrbng the EM flters propertes. The method also respects the real flter s crcutry and also works wth the current compensated nductors. The smple models are also applcable for dfferent testng confguratons, lke asymmetrcal or symmetrcal test setups. Dsadvantages are usually caused by naccuracy of electronc components used n the tested EM flters. n general, the smple models and also the descrbed method could be used for quck analysng and comparng several EM flters n dfferent mpedance systems. For obtanng the fnal equaton for calculatng nserton loss the Maple software was used. Fnal calculatons n the frequency doman were done n the Matlab R. The fnal equatons for EM flter s wth two current compensated nductors reached hundreds of pages n the Mcrosoft R Word edtor unformatted text. The measurements has been commanded by VEE Pro from Aglent Technologes c. The presented method could be modfed for quck estmaton of flter performance or estmaton of EM flter s worst-case. More detals could be found n [8]. These technques are avalable only for asymmetrcal and symmetrcal systems and relably dentfy the mentoned worstcase performance of the EM flter. Further work wll be focused on the mprovement of the performance and the accuracy of descrbed method manly n hgh frequency range. The accuracy could by mproved by mplementng the effects of spurous parts of electroncs components whch are ncluded nto the EM flters. The quanttes of the spurous components wll be estmated from the measured nserton loss data n 5 Ω/5 Ω systems. For estmatng of spurous parameters of ndvdual components, the Partcle swarm optmzaton could be used. Acknowledgement Ths work has been prepared as the part of the soluton of the grant no. /7/688 Advanced mcrowave structures on non-conventonal substrates of the Czech Scence Foundaton and wth support of the research plan MSM 6353 Advanced Electronc Communcaton Systems and Technologes (ECOM) and the research project E67 Advanced Electroncs and Communcaton Technologes - Adventure and nvtaton for Rsng Generaton of the Mnstry of Educaton, Youth and Sports of the Czech Republc.

7 4 J. DŘÍOVSKÝ, J. SVAČA, Z. RADA, SMPE MODES OF EM FTERS FOR OW FREQECY RAGE References [] Schaffner Holdng AG, Swtzerland. [onlne], 5-5-8, Avalable at: [] Schurter AG, Swtzerland. [onlne], 5-5-8, Avalable at: [3] ČS CSPR 7: Methods of Measurement of the Suppresson Characterstcs of Passve Rado nterference Flters and Suppresson Components. Czech Techncal Standard. Czech ormalzaton nsttute. Prague, ovember. 7 pages. (n Czech). [4] Schaffner Holdng AG, Swtzerland. CSPR 7 Measurements (applcaton note). pages. [onlne] -5-8, Avalable at: [5] M-STD-B: Test Method Standard - Method of nserton oss Measurement. Mltary Standard, Department of Defence, SA. June 4. 8 pages. [onlne] 4-5-8, Avalable at: STD-/std.pdf. [6] BOEK, D. Solvng of Electrcal Crcuts. Prague: BE, 4. 5 pages. SB X. (n Czech). [7] ČAJKA, J., KVAS, J. The Theory of near Crcuts. Prague: ST, pages. SB (n Czech). [8] DŘÍOVSKÝ, J. SVAČA, J. Estmaton of EM flter performance for the worst-case system. Radoengneerng (Part : Specal ssue:electromagnetc Compatblty), 6, vol. 5, no. 4, p. 6-, SS -5. [9] Schurter AG, Swtzerland. 5 EM flter (data sheet). 4 pages. [onlne] 3-6-8, Avalable at: 5.pdf. [] Elfs spol. s r. o. EM flters 3 A - 6 A (data sheet). pages. [onlne] 6-6-8, Avalable at: [] Schaffner Holdng AG, Swtzerland. -phase flters F 7 (data sheet). 4 pages. [onlne] 6-6-8, [avalable at] pdf/datasheet%f 7%e%55.pdf. About Authors... Jří DŘÍOVSKÝ was born n tomyšl, Czech Republc, n 979. He receved the M.Sc. and Ph.D. degrees n Electroncs and Communcaton from the Brno nversty of Technology, Brno, Czech Republc, n 3 and 7, respectvely. Hs Ph.D. thess was awarded by Eml Škoda Award n 7. Snce 6 he has been assstant professor n Electroncs and Communcaton at the Dept. of Rado Electroncs, Brno nversty of Technology. Hs research actvtes nclude selected topcs of EMC, EM measurements, and EMS testng. He s also nterested n specalzed problems of radofrequency and mcrowave measurements. Snce 8, he leads the course Radoelectronc measurement n master degree study program at the Faculty of Electrcal Engneerng and Communcaton, Brno nversty of Technology. He s a member of EEE. Jří SVAČA receved the M.Sc. and Ph.D. degrees from the Brno nversty of Technology, Brno, Czech Republc, n 97 and 978, respectvely. Snce 983 he has been Assoc. Professor, and snce 995 he has been Professor n Electroncs and Communcaton at the Dept. of Rado Electroncs, Brno nversty of Technology. Hs research nterests nclude theoretcal and mathematcal problems of specal planar structures for mcrowave ntegrated crcuts, and mcrowave measurements. He s nterested n specalzed problems of EMC, EM, and EMS. Prof. Svačna s a member of the Scentfc and Pedagogcal Boards of FEEC, Brno nversty Technology, and a member of the scentfc boards of FEC CT n Prague, and WB n Plsen, and of the Dept. of Rado Engneerng, Czech Academy of Scences n Prague. He s a Senor Member of EEE,.S.A., and a Fellow of EE,.K. Zbyněk RADA receved ng. (M.Sc.) and Dr. (Ph.D.) degrees from the Brno nversty of Technology (BT) n 99 and 994, respectvely. Snce 993, he has been wth the Dept. of Rado Electroncs of BT as the assstant professor (993 to 998), assocate professor (999 to 3), and professor (snce 4). From 996 to 997, he spent 6 months at the aboratore de Hyperfrequences, nverste Catholque de ouvan, Belgum as an ndependent researcher. Prof. Rada has authored or co-authored more than 8 papers n scentfc journals and conference proceedngs. Hs research has been focused on numercal modellng and optmzaton of electromagnetc structures, applcaton of neural networks to modellng and desgn of mcrowave structures, and on adaptve antennas. Prof. Rada s a member of the EEE Mcrowave Theory and Technques Socety. From to 3, he chared the MTT/AP/ED jont secton of the Czech-Slovak chapter of EEE. n 3, he became the Senor Member of EEE. Snce, Prof. Rada has been edtor-n-chef of the Radoengneerng journal (publcaton of Czech and Slovak Techncal nverstes and RS commttees).