A New EMI Filter Selection Method and a Comparison with the Existent Radiation-based Method

Transcription

1 ISSN AKAAF 48(3 4), 6 67 (007) Mako Podbešič, Vojko Matko A New EMI Filte Selection Method and a Compaison with the Existent Radiation-based Method UDK IFAC Oiginal scientific pape his aticle descibes a new method fo the selection of an appopiate signal line Electomagnetic Intefeence (EMI) filte. It descibes an EMI filte selection method based on the signal s FF measuement. his selection has poved to be efficient also fo quick enginee use. We optimize the EMI filte fo each single line sepaately. he method descibed in this aticle involving a Cental Pocesso Unit (CPU) module demonstates the compaison between new FF-based selection method and the existent adiation-based one. Key wods: EMC, EMI filte, citical line length, citical fequency INRODUCION hee ae two EMI filte selection methods. he fist one is based on adiation measuements. he adiation of the entie device is measued, which povides a discete fequency component with the maximum amplitude. he EMI filte with maximum insetion loss close to this discete fequency component is selected. he esulting filtes have the same fequency chaacteistics within the entie device. he second method has been developed and tested in the EMC Laboatoy of the ISKRAEMECO d.d. Company. It is based on the FF measuements on a single signal line. he main advantage of the new method is that a filte is selected fo each single line sepaately. his gives bette esults in tems of EMC poblem solving than the existent adiation-based method. It is vey impotant that the new method is suitable fo quick enginee use. he compaison between this new EMI filte selection method and the existent adiation-based method is showed in chapte 6. CRIICAL LINE LENGH An expession»citical line length«is known fom high speed tansmission-line theoy []. What is a long tansmission-line? hat depends on the tansmitted fequencies. A fequency spectum of a digital peiodic signal (a tapezoidally shaped wave) is pesented in Figue. Fig. Spectum of digital peiodic signal his signal can be descibed as seies spectal components A 0 f( t) = + [ Ancos( nω0t) + Bnsin( nω0t) ]. () n= he coefficients A 0, A n in B n ae found as ω 0 = π, () AUOMAIKA 48(007) 3 4,

2 A New EMI Filte Selection Method... M. Podbešič, V. Matko A B n n A 0 = = = f() t dt, f( t)cos( nω t) dt, f()si t n( n t) dt. (3) (4) (5) Whee ae: ω 0 basic angula fequency, peiod of the signal, t 0 time in the moment of obsevation. Fo the tapezoidal signal shown on Figue a, we find A A t n t h n t sin π sin π h n = (6) n t n t, h π π whee: A amplitude point to point, t h pulse width t ise time, t f fall time, peiod of the signal, n hamonic numbe. his spectum consists of discete fequency components f n = nf, whee f = /. When we daw asymptotes to this spectum, we find a hoizontal line up to the fist cone fequency (Figue b) f =, (7) π t h and fom thee a descending line at a ate of decease of 0 db/dec up to the second cone fequency (Figue b) f =. (8) π t Fo highe fequencies, the ate of decease is 40 db/dec (Figue b). fknee = (9) t. Fequency f knee is a pactical maximum which is about.5 times f. he citical line length is detemined by the fequency f knee. It is well known that electomagnetic 0 ω 0 emission inceases with fequency until a half the wavelength of the signal exactly fits the length of the tace. fknee = Hz, signal popety (0) t, λ knee = length of the line, m, line popety. he popagation on the line is v pop l = = t max, pd µε m/s. () () t pd is the popagation delay. he wavelength in equation (), can be witten as λ knee = v f pop knee (3) with equations (0) and () we find a citical line length l max vpop λknee fknee lmax = = = t, m. (4) µε At the citical line length the ise-time t exactly matches the two-way popagation delay time t pd (souce-load-souce). his means that the tansient phenomenon fomed by the low-to-high signal tansition pecisely fits the line length. Fo that eason, this distance is called the length of the ising edge []. We must stess that the citical line length l max means two-way popagation delay (souce-load-souce). o simplify equation (4), the eal value of popagation delay fo FR-4 cuently the most commonly used mateial in the poduction of pinted cicuit boads (PCBs) is used. With equations (5) and (6) the maximum electical line length befoe the equied line temination is detemined. he calculations below ae used fo the dielectic constant of FR-4 mateial (ε = 4.6). he latte vaies with signal fequency within the mateial. Most enginees geneally assume ε to be in the ange of 4.5 to 4.7. hese ae the values that ae published in vaious technical efeences [3]. m, l max = 9 t, ( in ns) (fo micostip topology in cm), (5) l max = 7 t, ( in ns) (fo stipline topology in cm). (6), 6 AUOMAIKA 48(007) 3 4, 6 67

3 M. Podbešič, V. Matko A New EMI Filte Selection Method... A line length equal to o longe than the citical length definitively behaves as a tansmission line. Chaacteistic impedance, delay and eflections should not be ignoed. At the same time, it also behaves as an efficient antenna and is eflected in consideable electomagnetic adiation and susceptibility poblem. Coupted signals ae usually ich on highe fequency components. 3 YPICAL FREQUENCY OBSERVED AS EMI ypical fequency obseved as EMI depends upon used logic elements and micocontolle. Moe pecisely, it depends upon the ise time of the signals tansmitted by these elements (7). Special attention should theefoe be paid to the selection of appopiate logic elements and micocontolle. If, fo instance, a faste High-speed CMOS with L inputs (HC) instead of a slowe Low-powe Schottky tansisto-tansisto logic (LS-L) is used, the electomagnetic emission can incease up to thee times. As might be expected, a discete fequency component at typical fequency stats to emit electomagnetic distubances at a cetain line length at which it becomes an efficient antenna [3]. f = 0 t max. π (7) hese typical fequencies obseved as EMI ae vey impotant when designing an electonic cicuit. o check the accuacy of the equation (7), the amplitude of the discete fequency component at the typical fequency obseved as EMI was measued [9]. he measuement equipment poved to be inadequate. We measued with fou channels oscilloscope ektonix DS744A with maximum digitizing ate GS/s and analog bandwidth 500 MHz. his was too low fo 74 Advanced CMOS (74AC) logic with a typical fequency of.6 GHz. Fo that eason the 74HC logic with a typical fequency of 70 MHz was chosen. Figue shows the output signal fom 74HC45 logic cicuit. he fequency spectum ises again at fequencies in the ange MHz. he peak with the amplitude of 8.64 db is located at the typical fequency of MHz. Fig. FF of an output signal fom 74HC45 logic cicuit We measued ise time t of the signal and we got ns. hen we calculated the typical fequency obseved as EMI fom equation (7). We obtained MHz. We obtained the same esult (8) as with the measuement. f 0 0 = = = MHz. πt π ns max. (8) he calculation of the typical fequency obseved as EMI is elatively simple. Its accuacy was also confimed by the measuement. 4 SELECION OF AN APPROPRIAE EMI FILER SRUCURE EMI filte insetion losses ae handled at 50 Ω input and output impedance [5]. Nomally, such Fig. 3 Selection of an appopiate EMI filte stuctue AUOMAIKA 48(007) 3 4,

4 A New EMI Filte Selection Method... M. Podbešič, V. Matko impedance does not exist in eal electonic cicuits. It is a well-known fact that EMI filte pefomance stongly depends on the input and output impedance. his is the impedance of the suounding electonic cicuit. Some geneal ules fo selection of an appopiate EMI filte stuctue should be consideed (Figue 3) [6]. he fact is that capacitos ae best used in high impedance cicuits, wheeas coils ae best used in low impedance cicuits. he eason lies in component chaacteistics at high fequencies [8]. A suitable EMI filte stuctue can be selected using the table in Figue 3. 5 SUIABLE EMI FILER SELECION he pocedue fo a pope EMI filte selection consists of the following steps: Define a signal ise time t fom the used logic elements; Calculation of the typical fequency obseved as EMI; Selection of a suitable EMI filte family with egad to the application equiements; Selection of an EMI filte fom the family which has the maximum insetion loss at the appoximate typical fequency. he need fo the use of EMI filtes depends upon the citical line length which has been veified in pactice [7]. If a two-way line length is shote than the peviously calculated l max (citical line length) and if thee ae no via connections on the line, the use of an EMI filte is not necessay. A pocedue of EMI filte selection method descibed above is pesented on Figue 4. 6 EXPERIMENAL WORK Some tests wee made on the CPU module. In the fist stage, we did not conside EMC developing ules. wo layes pinted cicuit boad was used. he density of lines was high. Radiated emission was measued. he esult of this measuement is pesented on Figue 5. he uninteupted hoizontal line illustates maximal allowed adiated emission accoding to the standad EN B. he second stage was followed. All EMC adiation emission eduction technics wee consideed in this stage. Eight layes pinted cicuit boad was used now. We did not use EMI filtes yet. Radiated emission was measued once again. he esult of this measuement is pesented on Figue 6. Fig. 4 A suitable EMI filte selection method In the thid stage, CPU module was equipped with EMI filtes. Filtes wee selected with the adiation-based method and the adiated emission was measue. It was below the allowed limit but still athe high (Figue 7). 64 AUOMAIKA 48(007) 3 4, 6 67

5 M. Podbešič, V. Matko A New EMI Filte Selection Method... Fig. 5 Radiated emission (two layes pinted cicuit boad) Fig. 6 Radiated emission (eight layes pinted cicuit boad) Finally, EMI filtes selected accoding to the new method wee installed in the same CPU module. As expected, the adiation level was significantly lowe (Figue 8). he tests pefomed on the CPU module clealy demonstate that the poposed FF-based EMI filte selection method is bette than the adiationbased one. AUOMAIKA 48(007) 3 4,

6 A New EMI Filte Selection Method... M. Podbešič, V. Matko Fig. 7 Radiated emission (EMI filtes, adiation-based method) Fig. 8 Radiated emission (EMI filtes, new method) 66 AUOMAIKA 48(007) 3 4, 6 67

7 M. Podbešič, V. Matko A New EMI Filte Selection Method... 7 CONCLUSION Geneally, a selection of suitable EMI filtes is vey impotant fo designing. We can not always solved EMC emission poblems without EMI filtes. he tests pefomed on the CPU module clealy demonstate that the poposed FF-based EMI filte selection method is bette than the adiation-based one. he main advantage of this new method is that it suitable fo quick enginee use and that solves EMC emission poblems sepaately fo each single signal line, which is why we believe it poduces bette esults. REFERENCES [] D. Campbell, H. Keidl, Solving EMC Issues. Motoola, Wokshop 36, EMV 03, pp. 6, Augsbug, 003. [] F. J. K. Buesink, homson CSF Signaal, Henglo, Nl, High Speed Digital Design opics fo Pinted Cicuit Boads. Wokshop 4, EMV 0, pp., Augsbug, 00. [3] M. I. Montose, EMC and the Pinted Cicuit Boad: Design, heoy, and Layout Made Simple. Fist edition, IEEE Pess Editoial Boad, vols, 3, 6, 7, pp., New Yok, 999. [4] S. Kazama, S. Shinohaa, R. Sato, Evaluation of Methods of Measuing Digital IC eminal Output. EMC Reseach Laboatoies Co., Ltd., pp , Sendai, Japan, 000. [5] V. Đ. Desnica, L. D. Živanov, O. S. Aleksić, M. D. Luković, M. D. Nimihte, Compaative Chaacteistics of hick-film Integated LC Filtes. IEEE tansactions on instumentation and measuement, vol. 5, no. 4, pp , august 00. [6]..., murata, Noise Suppession by EMI Filteing: Basics of EMI Filtes. No. E04EA-, pp. 4, 998. [7] M. Šegula, M. Podbešič, Reševanje EMC poblematike. ISKRAEMECO, d.d, pp. 56, Kanj, 00. [8] D. Andeson, L. Smith, J. Guszynski, S-Paamete echniques fo Faste, Moe Accuate Netwok Design. Hewlett-Packad Company, pp. 4 3, USA, Novi postupak odabia EMI filta i uspoedba s postojećim postupkom temeljenim na adijaciji. Ovaj članak opisuje novi postupak za odabi odgovaajućeg elektomagnetsko-intefeentnog (EMI) filta linije signala. Opisana je metoda odabia EMI filta temeljena na mjeenju spekta signala (FF). Ovakav odabi se, takođe, pokazao kao efikasan za bzu inženjesku pimjenu. Optimizia se EMI filta posebno za svaku liniju signala. Metoda opisana u ovom članku koisti CPU modul i pikazuje uspoedbu novog postupka odabia temeljenog na FF-u i postojećeg postupka temeljenog na adijaciji. Klju~ne ije~i: EMC, EMI filta, kitična duljina linije, kitična fekvencija AUHORS ADDRESSES d. Mako Podbešič Ministy of Defence, Administation fo Civil Potection and Disaste Relief Vojkova cesta 6,.000 Ljubljana, Slovenia. pof. d. Vojko Matko Univesity of Maibo, Faculty of Electical Engineeing and Compute Science Smetanova ulica 7,.000 Maibo, Slovenia Received: AUOMAIKA 48(007) 3 4,