Systematic Power Line EMI Filter Design for SMPS uttipon Tarateeraseth ollege of Data Storage Innovation King Mongkut's Institute of Technology Ladkrabang Bangkok Thailand ktvuttip@kmitl.ac.th Kye Yak See School of Electrical and Electronic Engineering Nanyang Technological University Singapore EKYSEE@ntu.edu.sg Abstract Based on a two-probe measurement approach noise source and noise termination impedances of a switched-mode power supply (SMPS under its normal operating condition are measured. With the known information of the noise source and noise termination impedances an electromagnetic interference (EMI filter can be designed systematically with good confidence. An example to design a power line EMI filter for a SMPS to meet a regulatory conducted EMI limit using the proposed procedure is demonstrated. Flavio G. anavero Dipartimento di Elettronica Politecnico di Torino Torino Italy flavio.canavero@polito.it (a Keywords- Electromagnetic interference (EMI EMI filter switched-mode power supplies noise source impedance line impedance stabilization network. I. INTRODUTION onducted electromagnetic interference (EMI is one of the major design concerns for switched-mode power supply (SMPS designs []. To comply with the international regulatory EMI requirements an EMI filter is necessary to lower a conducted EMI level of a SMPS below the limit [2-3]. Unlike the filter designs for communications and microwave applications where the source and termination impedances are well defined (usually specified at 50 the noise source and termination impedances of an EMI filter for a SMPS are unfortunately not readily available. Hence the design of an EMI filter without known source and termination impedances can be a challenging task [4-7]. In addition conducted EMI exists in two modes the common-mode ( and differentialmode ( emissions which further complicates a filter design process as an EMI filter requires to suppress both and emissions effectively. Some EMI filter design methods adopt a simplistic approach to design an EMI filter without taking into account the noise source and noise termination impedances [8-]. Other EMI filter design methods do take into account the noise source and the noise termination impedances but approximate them as purely resistive elements Figure. (a Two-probe in-circuit measurement setup Equivalent circuit of the measurement setup. [2-3]. As there are some approximations being made in these methods worst-case conditions (maximum or minimum possible and impedances have to be assumed to design an EMI filter [3]. These methods usually lead to overdesign or non-optimal design of an EMI filter. Without precise information (magnitude and phase of the noise source and noise termination impedances over a frequency range of interest it is difficult to decide an appropriate EMI filter configuration and to design an optimal filter for a SMPS to meet a specific conducted EMI limit. In this paper based on an in-circuit impedance measurement setup the amplitude and phase information of the noise source and noise termination impedances can be measured and extracted under its actual operating condition. With the known impedance information the limitations of previously mentioned methods can be overcome and a
(a (c (d Figure 2. Measured LISN and SMPS impedances (a magnitude phase (c magnitude (d phase. systematic EMI filter design procedure to achieve the desired filter attenuation performance becomes possible. II. IN-IRUIT IMPEDANE MEASUREMENT Fig. (a shows the basic setup to measure an unknown impedance x in a circuit loop where the loop can be active and carrying high voltage and/or high current. The setup requires an injecting current probe a receiving current probe and a vector network analyzer (NA. The two current probes couple to the circuit through inductive couplings without direct connection to an active circuit. By transferring the primary circuits of the injecting and receiving probes in the coupled circuit loop the equivalent circuit of the setup is given in Fig.. M and M2 are the equivalent impedances of the injecting and receiving probes respectively. M is the induced signal voltage in the circuit loop from port of the NA through the injecting probe. L w and r w are the loop inductance and resistance respectively. x is the impedance to be measured and all the other impedances present in the circuit loop are due to the measurement setup. Let setup M M 2 rw jlw the resultant current flowing in the circuit loop due to the injected signal through injecting probe is given by: I w M setup The induced voltage in the loop is given by: x. ( M j M p where M is the mutual inductance between the injecting probe and the coupling loop is the output source voltage of port of the NA is the output impedance of port and p is the input impedance of the injecting probe. On the other hand the received signal at port 2 of NA depends on the current I w measured by the receiving current probe that is: 2 T2 w (2 I (3 where T2 is the calibrated transfer impedance of the receiving probe given by the probe manufacturer. By substituting equations (2 and (3 into ( the unknown impedance to be measured can be determined as follows: (4 x k setup 2 jm T 2 where k is a frequency dependent coefficient. The p pre-measurement calibration process to obtain k and setup are described in details in [4] and will not be repeated here. Using the procedure outlined in [4] the and impedances of a LISN (noise termination and a SMPS (noise source can be extracted as follows.
respectively. Figs. 2 (c and (d show the magnitudes and phases of the measured LISN and SMPS impedances in. With the known magnitudes and phases of the noise source (SMPS and noise termination (LISN impedances systematic design of an EMI filter to meet a specific conducted EMI limit becomes relatively easy. Figure 3. onducted EMI measurement with a / discrimination network. (a III. EMI FTER DESIGN PROEDURE The same SMPS mentioned earlier is used to guide a reader through the design procedure. The intended conducted EMI limit to be met by the SMPS is ISPR 22 lass B limit [3]. The LISN specified by this Standard can only measure the total conducted emissions that consist of both and components. Therefore a discrimination network is needed to discriminate the and components from the LISN so that they can be measured separately [5] and used to set the required and filter insertion losses. The measurement setup is shown in Fig. 3 and a HP 8595E spectrum analyzer (9 khz 6.5 GHz is chosen for the conducted emission measurement. The conducted emissions of the SMPS without the filter are measured with the LISN alone. With the help of the discrimination network the required and filter insertion losses can be found by subtracting the Standard limit i.e. req[db] [dbμ] LIMIT [dbμ] (7 req[db] [dbμ] LIMIT [dbμ] (8 (c Figure 4. onducted EMI measurement with / discrimination network (a test setup filter (c filter. SMPS T LISN setup SMPS T LISN setup. As an example the SMPS (TM22WB 5W +2 D /0.75A 2 D /0.5A is powered through the LISN (Electro-Metrics M 5-25/2. The and impedances of the LISN (noise termination and the SMPS (noise source are determined. Figs. 2 (a and show the magnitudes and phases of the measured LISN and SMPS impedances in (5 (6 where = Measured emission from SMPS without the filter [ dbμ ] = Measured emission from SMPS without the filter dbμ LIMIT = ISPR 22 lass B conducted emission limit dbμ. The required and filter insertion losses are plotted in Figs. 5 (a and respectively. Obviously the SMPS can never meet the required EMI limit without an EMI filter. The adopted filter configuration is a conventional one as commonly proposed in the literature (e.g. [] and is illustrated in Fig. 4 (a. The filter is composed of one choke (L one capacitor ( X and two capacitors ( Y and Y2. Due to the leakage inductance of L it also doubles up as two inductors in the line and neutral lines. This particular configuration with the capacitors facing the SPMS side is needed in order to achieve the optimal filter attenuation since the and SMPS impedances are higher than the corresponding LISN impedances as clearly documented by Figs. 2 (a and (c. In fact the capacitor to be effective must be placed in parallel to high impedance and the inductor must be connected in series
(a (c (d Figure 5. EMI filter insertion losses (a required and designed insertion losses required and designed insertion losses (c measured conducted emission with filter (d measured conducted emission with filter. with a low impedance []. The and interpretations of Fig. 4 (a lead to two separate circuits shown in Figs. 4 and (c respectively. Fig. 4 presents the -suppressing part of the filter and is composed by L which is the inductance due to L and by XT which represents the effective capacitor ( Y and Y2 in series and then in parallel with X. Fig. 4 (c presents the -suppressing part of the filter and is composed by L which is the inductance due to L and by YT which is the effective capacitor ( Y and Y2 in parallel. From Figs. 4 and (c the expressions of and filter insertion losses can be evaluated according to [6-7]. L L estimate s s L 2 YT SMPS L YT SMPS LISN estimate 20log s s SMPS LISN SMPS LISN 2 XT SMPS XT SMPS LISN 20log SMPS LISN SMPS LISN where s = j2 f YT = Y// Y2 [F] XT = X // Y Y2 [F] L = inductance of the choke [H] (9 (0 L = Equivalent inductance of choke [H] = SMPS impedance [Ω] SMPS SMPS = SMPS impedance [Ω] LISN = LISN impedance [Ω] LISN = LISN impedance [Ω]. The capacitor is usually constraint by safety requirements and therefore the maximum capacitance connected to ground cannot exceed about 3000 pf for 220 A 50 Hz mains [4]. Hence Y and Y2 are chosen to be 000 pf each. Substituting the known LISN and SMPS impedances and the chosen capacitors into equation (0 and assuming that the filter elements are ideal the required inductance that could provide the filter insertion loss higher than req indicated in Fig. 5 is about 2 mh. Hence two 000 pf lass Y capacitors and a 2 mh choke (NE/TOKIN S-02-0A are chosen for the filter. Once the choke and the capacitors are selected we proceed with the filter design. For the capacitor since there is no safety issue the value can be chosen to be as large as possible but larger capacitor usually exhibits low self resonant frequency (SRF [6]. Substituting the known LISN and SMPS impedances and the measured inductance of the choke into equation (9 the required capacitor that could provide the necessary filter insertion loss higher than reqindicated in Fig. 5 (a is about.5 F. The impedance behavior of the filter components chosen according to the above considerations were measured by means of a HP 4396B impedance analyzer (00 khz.8 GHz. The actual insertion losses of and filters of Figs. 4 and (c can be worked out as an adaptation of (9 and (0 with the inclusion of the parasitic effects of the filter components [6-7].
designed EMI filter fulfills the requirement and their results indicate that the SMPS complies with the regulations limits. (a I. EFFET OF INORRET FTER ONFIGURATION To illustrate the effect of an incorrect filter configuration that leads to non-optimal performance of the EMI filter we discuss a new setup where the same filter already designed is mounted in a reverse configuration; this means that the inductor position is now shifted to the SMPS end. The new EMI filter configuration is shown in Fig. 6 (a with its and circuits shown in Figs. 6 and (c respectively. With the new filter configuration the actual and filter insertion losses become [6-7]: actual ( ( 20log ( LISN SMPS L LISN XT XT XT LISN SMPS (3 Figure 6. (a New EMI filter configuration filter (c filter. actual actual where L L YT (c ( ( XT L LISN SMPS XT L LISN 20log ( XT LISN SMPS ( ( YT L LISN SMPS YT L LISN 20log ( YT LISN SMPS = Effective impedance of the choke [Ω] = impedance of the choke [Ω] = Effective impedance of // 2 [Ω] = Effective impedance of // XT Y Y X Y Y2 [Ω]. ( (2 Using the measured impedance of all the chosen filter components as well as the LISN and SMPS impedance behavior the actual insertion losses of the and the filters (computed by means of ( and (2 are plotted in Figs. 5 (a and respectively. For comparison the required and filter insertion losses are also shown in the same figures. Both actual and filter insertion losses are higher than the required ones which indicate that the designed EMI filter is able to suppress the and conducted emissions below the ISPR 22 lass B limit. Fig. 5 (c and (d show the measured and conducted emissions after the designed EMI filter is inserted. Both the and conducted emissions are below the limit which verify that the actual ( ( YT LISN SMPS L YT LISN 20log (.(4 YT LISN SMPS Based on equations (3 and (4 the actual and filter insertion losses are plotted in Figs. 7 (a and respectively. By comparing with the required and filter insertion losses they show that the actual filter insertion loss deteriorates marginally at the low frequency end but the actual filter insertion loss is unable to meet the required filter insertion loss below.8 MHz. The and conducted emissions shown in Figs. 7 (c and (d respectively have confirmed that the conducted emission has exceeded the limit marginally at the low frequency end and the conducted emission has exceeded the limit badly below.8 MHz. As the conducted emission dominates below.8 MHz the SMPS does not comply with the limit.. ONLUSIONS Based on a two-probe measurement approach the noise source (SMPS and noise termination (LISN impedances under incircuit operating condition are measured. The information of noise source and termination impedances has been proven to be an useful input for designing an optimal EMI filter to meet a specific EMI limit in a systematic manner. Both the and filter insertion losses for any EMI filter can be determined accurately so that a designer has a very clear picture of the EMI filter insertion loss characteristics without any doubt. Using a SMPS as a practical design example we have demonstrated that with the same filter components but choosing an incorrect filter configuration can lead to nonoptimal filter attenuation which can cause the same SMPS to fail the required EMI limit badly. With the proposed EMI filter design methodology the optimal filter configuration and the correct filter component value can be chosen without guessing. Hence this prevents the tendency of over-designing an EMI filter with more components than necessary.
(a (c (d Figure 7. Non-optimized filter (a required and designed insertion losses required and designed insertion losses (c measured conducted emission (d measured conducted emission REFERENES []..R. Paul Introduction to Electromagnetic ompatibility 2nd Edition J. Wiley 2006 [2]. Specification for Radio Disturbance and Immunity Measuring Apparatus and Methods Part : Radio Disturbance and Immunity Measuring Apparatus ISPR 6-999. [3]. Limits and Methods of Measurement of Radio Interference haracteristics of Information Technology Equipment ISPR 22 2004. [4]. L. Tihanyi Electromagnetic ompatibility in Power Electronics IEEE Press 997. [5]. B. Garry and R. Nelson Effect of impedance and frequency variation on insertion loss for a typical power line filter in 998 Proc. IEEE EM Symposium pp. 69 695. [6]. B. Audone and L. Bolla Insertion Loss of Mismatched EMI Suppressors IEEE Trans. Electromagn. ompat. vol. 20 no. 3 pp 384-389 Sep. 978. [7]. S. M. akil A technique for determination of filter insertion loss as a function of arbitrary generator and load impedances IEEE Trans. Electromagn. ompat. vol. 20 no. 2 pp 273-278 Sep. 978. [8]. F.-Y. Shih and et al. A procedure for designing EMI filters for A line applications IEEE Trans. Power Electron. vol. pp. 70 8 Jan. 996. [9]. M. Kumar and. Agarwal Power line filter design for conducted electromagnetic interference using time-domain measurements IEEE Trans. Electromagn. ompat. vol. 48 no. pp 78-86 Feb. 2006. [0]. J. Biela and et al. Passive and active hybrid integrated EMI filters IEEE Trans. Power Electron. vol. 24 no. 5 pp. 340 349 May. 2009. []. T. Nussbaumer M. L. Heldwein and J. W. Kolar Differential Mode Input Filter Design for a Three-Phase Buck-Type PWM Rectifier Based on Modeling of the EM Test Receiver IEEE Trans.Ind. Electron. vol. 53 no. 5 pp. 649 66 Oct. 2006. [2]. M. J. Nave Power Line Filter Design for Switched Mode Power Supplies New York: an Nostrand Reinhold 99. [3]. Sheng Ye W. Eberle and Yan-Fie Liu A Novel EMI filter design method for Switching Power Supplies IEEE Trans. Power Electron. ol. 9 N. 6 Nov 2004 pp. 668-678. [4].. Tarateeraseth Hu Bo K. Y. See and F. anavero ``Accurate extraction of noise source impedance of SMPS under operating condition'' IEEE Trans. Power Electron. vol. 25 no. pp. -- 7 Jan 200. [5]. K. Y. See Network for conducted EMI diagnosis IEE Electronics Letters Aug. 999 ol. 35 No. 7 pp. 446-447. [6]. S. Wang F.. Lee and J.D. an Wyk A study of integration of parasitic cancellation techniques for EMI filter design with discrete components IEEE Trans. Power Electron. vol. 23 no. 6 Nov 2008 pp. 3094 302.