1826

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 5, MAY 2011

A Novel Hybrid Common-Mode EMI Filter With Active Impedance Multiplication

Wenjie Chen, Member, IEEE, Xu Yang, Member, IEEE, and Zhaoan Wang, Senior Member, IEEE
AbstractThis paper presents a novel hybrid common-mode (CM) electromagnetic interference lter for a military power supply. The proposed lter is composed of an active impedance multiplication circuit cascaded with a benchmark CM inductor. The active impedance multiplication circuit is very effective in boosting the impedances of both the benchmark CM inductor and the noise source. Thus, the total equivalent impedance of the lter is enlarged by the active booster to a great extent. An important advantage of the approach is that only a small-size CM choke is used as the passive lter instead of a traditional CLC lter. Furthermore, the noise source is turned to be an inevitable part of the hybrid lter by the impedance multiplication effect. Experimental results were carried out according to MIL-STD-461. It is shown that the proposed approach can acquire good noise attenuation performance within a wide frequency range. Index TermsActive lter, common mode (CM), electromagnetic interference (EMI), hybrid lter, impedance multiplication.

Fig. 1.

Typical three-order CM EMI lter.

I. I NTRODUCTION LECTROMAGNETIC INTERFERENCE (EMI) lters have been widely used for years to solve conducted EMI problems for many electronic applications. A typical threeorder common-mode (CM) EMI lters is shown in Fig. 1. It is acknowledged that minimizing shunt-path impedance and maximizing series-path impedance at high frequencies are an effective way to achieve good noise attenuation. However, CM noise cannot be reduced by using an arbitrarily large value of CY , since this capacitor is usually limited by the safety standard that species the maximum leakage current allowed owing to the ground. Therefore, in most instances, designers have to resort to a larger CM inductance. For this reason, the size and weight of PEFs or even the whole equipment are very huge. Hybrid EMI lters (HEFs) are proposed to overcome these problems [1][9]. A planar passive lter with an active lter is reported in [5]. A HEF designed for a motor drive system is introduced in [3], where active lters were used to optimize the passive lter performance. Detailed analysis and quantitative calculation of the hybrid lter parameter are reported in [2].

Manuscript received March 23, 2010; revised May 5, 2010; accepted May 24, 2010. Date of publication June 21, 2010; date of current version April 13, 2011. This work was supported in part by the Key Projects in the National Science and Technology of China under Grant 2007BAA12B01 and in part by the National Natural Science Foundation of China under Project 50707024. The authors are with the State Key Laboratory of Electrical Insulation and Power Equipment, Xian Jiaotong University, Xian 710049, China (e-mail: cwj@mail.xjtu.edu.cn; yangxu@mail.xjtu.edu.cn; zawang@mai.xjtu.edu.cn). Digital Object Identier 10.1109/TIE.2010.2053341

Moreover, the impedance feature and the connection between passive and active lters are studied in [8]. What is more, Zhu et al. [6] proposed a hybrid ripple lter in dc/dc converters. Furthermore, an intensive study of the hybrid lter has just been made by Ostroznik et al. recently [1]. Nevertheless, in the authors opinion, there are still two major problems that need to be paid more attention to. First, for the hybrid lter system, active EMI lters (AEFs) and PEFs are often treated as two separated parts in most reported papers. The following questions are therefore raised: How will these two lter types (AEFs and PEFs) perform when placed in one hybrid lter system? Is there any cooperation or interrelation between them? Second, when selecting a power line lter, the designer must be aware that the performance of the lter is dependent on the source and load impedance to a great extent. Another question also comes into mind. If HEFs have also such problems, then are there any approaches to maximize the HEF noise attenuation in a real environment? In an effort to obtain insight into these questions, investigations are carried out in this paper. A novel approach that uses an active circuit to increase the impedances of both the inductor and the noise source is proposed. A distinguished advantage of this approach is that the equivalent CM impedance of the passive component, as well as that of the noise source, is boosted by the active circuit entirely. Thus, only a small-size CM inductor but not a traditional CM lter (shown in Fig. 1) is used in the novel hybrid lter system. In addition, the increased noise impedance can also act as a necessary part of the whole lter system. Section II explains the principle of the active impedance multiplication method. The design considerations of the proposed HEF for a military power supply is presented in Section III. Section IV gives experimental results. Conclusions are presented in Section V.

0278-0046/$26.00 2010 IEEE

CHEN et al.: NOVEL HYBRID COMMON-MODE EMI FILTER WITH ACTIVE IMPEDANCE MULTIPLICATION

1827

Fig. 2. Proposed approach to boost the impedance with the active multiplication method. (a) Benchmark impedance without active multiplication. (b) Conguration of the active multiplication method. (c) Thevenins equivalent translation. (d) Equivalent impedance of port AB with active multiplication.

Fig. 3.

Proposed HEF with the active impedance multiplication method. (a) Topology of the hybrid lter. (b) Equivalent impedance model of the hybrid lter.

II. P RINCIPLE OF THE ACTIVE I MPEDANCE M ULTIPLICATION M ETHOD A. Basic Active Multiplication Structure The principle of the active impedance multiplication method is shown in Fig. 2. It is a current-controlled current source (CCCS) Ai in parallel with a benchmark impedance Z. An important feature of this structure lies in the controlling current i that is not on the impedance branch but on the input port AB. According to Thevenins theorem, the input impedance Zin of port AB can be calculated as follows: Zin = Z(A + 1)i u = = Z + AZ. i i (1)

Thus, if the original input impedance of port AB is Z, then, by introducing an active circuit in parallel with it, the impedance of Z can be boosted by a factor of A, where A is the current gain of the CCCS. This approach is called the active impedance multiplication method in this paper. B. Proposed Filter Topology With Active Impedance Multiplication Method Based upon that idea, a topology of a HEF for CM noise reduction is proposed in Fig. 3. R represents the CM impedance of the loada standard line impedance stabilization network (LISN). Zn is the equivalent impedance of the noise source or equipment under test (E.U.T.). To achieve the function of the active multiplication method, the controlling current i of the CCCS should be sampled on the load branch, and the CCCS should also be placed beside it. A single CM inductor sL, instead of a traditional three-order lter (as shown in Fig. 1), is used in the proposed system. It is because there is a CCCS to boost up the passive impedance, so only a small passive lter component is used here. Additionally, since the size of

the CM inductor is usually much larger than that of the CM capacitor, reducing the passive lter size would be best to start from reducing the CM inductor size. According to the active impedance multiplication method, it can be clearly observed that the impedances of both the inductor sL and the noise source Zn are all increased by A times. The equivalent impedance model is shown in Fig. 3(b). The physical meaning of this circuit can be explained as adding a boosted inductor AsL and a boosted source impedance AZn between the load R and the noise source Zn . Therefore, the larger the multiplication factor A, the greater the series component value AsL and AZn , and the better the noise reduction effect. In a word, it is only by means of the correlation between the active multiplication circuit and the passive component followed that the lter impedance can be increased to a large extent. Moreover, the increased noise impedance AZn also acts as a necessary part of the whole lter system. It is noticed in Fig. 3(b) that there is always a certain part of noise voltage damped on the equivalent component AZn . Therefore, the noise source impedance, whether it may be large or not, will be a useful part of the lter. C. Quantitative Comparison of Filter With and Without Impedance Multiplication Insertion loss (IL) is generally used to characterize an EMI lter. IL is dened as the ratio of the signal level in a test conguration without the lter installed (U1 ) to the signal level with the lter installed (U2 ). In the proposed scheme, the IL of the single inductor is ILP =
RZn in U1 = R+Zn RZn U2 R+Zn +sL in R + Zn + sL sL = =1+ . R + Zn R + Zn

(2)

1828

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 5, MAY 2011

Fig. 5.

Basic structure of the proposed EMI lter.

Fig. 4. Simulated insertion loss of the lter with and without active impedance multiplication under different noise source conditions.

By means of impedance multiplication, the IL of the novel lter system is described as follows: ILh = = U1 = U2
RZn R+Zn in RZn R+Zn +AsL+sL+AZn in

3) Although, for a xed-value inductor, larger noise impedance will result in smaller insertion loss, this bad characteristic can be weakened in the hybrid lter. It is simulated under low-noise-impedance conditions, and the hybrid lter still maintains a decent attenuation decibel by the same lter parameters. This is because the insertion loss of the hybrid structure mainly comes from the boosted inductor and the boosted source impedance. For this reason, the noise attenuation of the proposed one is not only dependent on the value of passive lter component sL but also highly dependent on the gain A of the active multiplication circuit. III. BASIC S TRUCTURE OF THE P ROPOSED EMI F ILTER

R + Zn + AsL + sL + AZn R + Zn AsL sL AZn =1 + + + . R + Zn R + Zn R + Zn

(3)

It can be seen from (2) and (3) that an additional insertion loss is introduced by cooperation among impedance multiplication Ai, inductor sL, and noise source Zn IL = AsL + AZn . R + Zn (4)

D. Simulation Analysis of Different Filter Types Simulations are carried out rst under ideal conditions without the knowledge of noise source. The passive component value L is chosen as 50 mH. The factor of the active multiplication circuit is dened as a simplied open-loop transfer function A(s), which will be discussed in detail later. Furthermore, the source impedance is modeled as constant Zn1 = 300 and Zn2 = 30 k. The simulated insertion loss of the CM inductor, the hybrid lter with an active booster, and the entire active circuit are shown in Fig. 4. It consists of the following features. 1) The noise attenuation of the hybrid lter with impedance multiplication is always higher than that of the single passive inductor or the active booster. 2) The insertion loss of the inductor increases with the increase of frequency, while for the active booster, the insertion loss curve is comparatively attened. Thus, the hybrid lter with impedance multiplication combines the advantage of both of these lter types. At lowfrequency range, the boosted impedance can provide a comparatively high insertion loss decibel, while at highfrequency range, the intrinsic high-attenuation feature of the inductor will be further amplied.

The basic structure of the proposed EMI lter is shown in Fig. 5. To be as a benchmark of the active impedance booster, the CM inductor must be placed near the E.U.T.; otherwise, the multiplication approach will not be accomplished. The active circuit consists of a current-transformer (CT)-based sensor, an operational amplier block, and a current injection branch. The active circuit plays two kinds of roles. On the one hand, it is used to accomplish the active multiplication method together with the followed inductor and the noise source so as to increase the impedance of these parts. On the other hand, it is constructed as an active lter. The CM noise is detected by the CT and then amplied and injected by the op-amp and capacitors, respectively. In general, both feedback and feedforward control can be used in the active circuit. Although most reported lters use feedforward control, the reason why feedback control is chosen here comes from the following considerations. 1) The feedback control structure obtains the noise current at the process output, but the feedforward one obtains data at the process input. As a result, feedback control takes into account any possible disturbance that might cause EMI noise at the LISN terminal. 2) Feedback control does not require detailed knowledge about the noise source. Although the frequency range of EMI noise is wide, the feedback control architecture ensures the desired performance by altering the inputs immediately once deviations are observed regardless of what caused the disturbance. 3) In order to maximize the effect of impedance multiplication, the control factor A of the CCCS is ideally innite. Since the active circuit is essentially a single-loop system, it is comparatively easy for the op-amp to achieve

CHEN et al.: NOVEL HYBRID COMMON-MODE EMI FILTER WITH ACTIVE IMPEDANCE MULTIPLICATION

1829

Fig. 6.

Congurations of the experimental setup.

large gain in feedback control. While for feedforward control, A = 10 means that the op-amp gain will be 0.1. In addition, if A = 100, the op-amp gain will be 0.01. In most cases, such a high accuracy is difcult to implement and maintain for the op-amp. 4) Feedback control is a closed-loop system, which is essentially different from open-loop feedforward control. Therefore, by careful design and analysis, unstable processes may be stabilized through feedback control. IV. P RACTICAL D ESIGN C ONSIDERATION OF THE H YBRID F ILTER In this section, a detailed consideration of several practical issues related to the design of the hybrid lter is presented. To acquire an accurate result, the CM noise impedance of the E.U.T. is tested rst. Then, the value of inductor LCM is determined by taking into account the noise impedance, the forthcoming multiplication effect, and the limit of EMI regulation. Finally, the selection and specication of the components used in the active circuit are studied in detail. A. Description of the CM Noise Source Used in the Experiment It has long been recognized that noise impedance has a signicant impact on the attenuation performance of the EMI lter, no matter what kind of lter it may be. Thus, in order to effectively attenuate the EMI noise over the frequency range of interest, the EMI lter must be designed to match the noise impedance. A 1-kW power supply for military aviation is used as the noise source in our experiment. The rated output current is 5 A, and the output frequency is 152000 Hz. The system setup is shown in Fig. 6. According to the requirements for the control of the EMI characteristics of the power supply, a standard GJB152 is chosen for this application. It is dened from 10 kHz to 10 MHz for the conducted EMI noise and can be equivalently exchanged with MIL-STD-461 CE102. The EMI noise detected by LISN will be separated into CM and DM by a noise separator. Since the CM noise, as well as the CM lter, is the main concern of this paper, only CM equivalent impedance is tested here. An insertion loss method proposed in [10] is used to measure the noise impedance. The equivalent CM impedance curve is shown in Fig. 7. The test results showed that the CM impedance of the E.U.T is within several kiloohms.

Fig. 7. Measured equivalent CM noise impedance curve of the E.U.T.

Fig. 8. CM noise spectrum of the E.U.T. without an EMI lter installed.

B. Parameter Study on Inductor LCM of the Hybrid Filter In the whole hybrid lter system, the passive lter component is mainly used as a basic benchmark that needs further increase through the impedance multiplication circuit. Because of this, an appropriate value of LCM will play an important role in the whole hybrid lter system. Fig. 8 shows the measured spectrum of the E.U.T. when no lter is installed. It is observed that the E.U.T. cannot pass the

1830

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 5, MAY 2011

Fig. 9. Simulation circuit for the CM inductor.

MIL-STD-461 limit over a wide frequency range from 10 kHz to 8 MHz. In particular, at a frequency near 40 kHz, the noise spectrum is higher than 120 dB V. It is expected that, after applying the proposed EMI lter, at least 40-dB attenuation should be reached so as to pass the MIL-STD-461 limit. As the noise impedance is known in Fig. 7, and according to (2), the value of inductor LCM can be determined by LCM = 2freq 10
Attenuation requirement 20

Fig. 10. Simulated curves for the CM inductor (L1 = 12 mH, Cp1 = 20 pF, Rp1 = 100 k, L2 = 50 mH, Cp2 = 80 pF, Rp2 = 100 k, L3 = 550 mH, Cp2 = 800 pF, and Rp2 = 100 k).

1 (R + Zn )

(5)

where freq is the frequency requirement of the lter. In this example, freq is 40 kHz, Zn is 1400 , R is 25 , and the attenuation requirement is 40 dB. Thus, it is calculated that a huge inductor, which is as large as 550 mH, might satisfy that requirement. Consequently, in order to minimize the passive lter size, it is motivated to decrease the CM inductor value as small as possible in the proposed lter. To do that, three times (10-dB) attenuation at a frequency of 40 kHz is considered as a benchmark value of the passive lter. Based on (5), the minimum value of the CM inductor is about 12 mH. Specically, this value is not the nal value of the CM choke. If the active multiplication factor A cannot be high enough, for instance, for the reason of instability, then the inductor value should be increased. Therefore, it is also calculated that, to acquire 20-dB attenuation, the inductance should be about 50 mH or so. Fig. 9 shows a classic high-frequency simulated circuit of the passive lter. The simulated impedance curves and the insertion loss curves are shown in Fig. 10. Moreover, to decrease the passive component size further, a nanocrystalline core is chosen in this paper as the best suitable material for the inductor due to its high permeability, high saturation ux density, and high Curie temperature. C. Parameter Study on Detecting and Injecting Part of the Active Multiplication Circuit The motivation of this design is to try to make the boost factor A as large as possible in order to maximize the impedance multiplication effect. To do that, the gain of the active circuit is analyzed rst. The transfer functions of the detecting CT, the op-amp circuit, and the driving capacitor network

Fig. 11. Schematic of the detection circuit using a CT and of the injection circuit using a capacitor. (a) CT. (b) Injection network.

are determined as G1 (s), G2 (s), and G3 (s), respectively. The open-loop transfer function of the active multiplication circuit is the product of them A(s) = G1 (s) G2 (s) G3 (s). (6)

A high-frequency model of the CT and the injection network is shown in Fig. 11. Based upon that, the transfer function of G1 (s) is shown in (7) at the bottom of the page, and G3 (s) may be determined to be G3 (s) = sC s2 ESLC + s(R4 + ESR)C + 1 (8)

G1 (s) =

sL s4 Cp Cs R L + s3 Cp L + s2 R Cp (Lp + L ) + Cs (Ls + L ) + s (Ls + L ) + R

(7)

CHEN et al.: NOVEL HYBRID COMMON-MODE EMI FILTER WITH ACTIVE IMPEDANCE MULTIPLICATION

1831

Fig. 13. Frequency response plots of the injection network. (R4 = 10 ; when C1 = C2 = 3.3 nF, ESL = 20 H, and ESR = 1 m; when C1 = C2 = 6.6 nF, ESL = 40 H, and ESR = 3 m; and when C1 = C2 = 10 nF, ESL = 60 H, and ESR = 5 m). TABLE I SPECIFICATIONS OF THE CM CURRENT TRANSFORMER

Fig. 12. Frequency response plots of the transfer function of CT. (a) Transfer function versus magnetizing transformer inductance L . (b) Transfer function versus load resistor R. (c) Transfer function versus transformer ratio n1 : n2 .

Moreover, the magnitudes of the frequency responses of G1 (s) and G3 (s) are shown in Figs. 12 and 13, respectively. It can be seen in Fig. 12 that whether we change the magnetizing transformer inductance L , the load resistor R, or the transformer ratio n, only the low-frequency cutoff fL changed, whereas the passband gain of the CT remains unchanged. Therefore, the gain of the CT has little contribution to the whole gain A(s) in the active multiplication circuit. Furthermore, the bandwidth of the CT should be enlarged in order to sense the high-frequency EMI noise. Table I summarizes the specications of the CT used in this paper. In Fig. 13, although it is observed that larger values of a driving capacitor can achieve higher gain, the earth leakage current limitation restricts the value of the capacitor. Thus, according to GJB 152 and the safety margin, the value of the capacitor is selected as 10 nF in our experiment. As a result, it is known that the gain of G1 (s) is nearly zero and that the gain of G3 (s) cannot be improved signicantly, so the most effective way to increase the gain A(s) is to increase the op-amp gain G2 (s). For this reason, a high-performance video op-amp AD811 is used in the experiment. D. Parameter Study on the Op-Amp Part of the Active Multiplication Circuit As discussed earlier, a large active multiplication factor A might result to an instability problem in the feedback system. Thus, the goal of this design is try to nd a tradeoff between the high gain and stabilization of the active booster. The

where Ls = n2 /n2 Ls 1 2 L = Lp Ls + L Ls + L Lp Cp = n2 /n2 (Cp + Cm /2) 1 2 Cs = n2 /n2 (Cs + Cm /2) 1 2 R = n2 /n2 R. 1 2

1832

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 5, MAY 2011

Fig. 14. Active multiplication gain A(s) versus feedback proportion Rf /R1 (Cf is open).

Fig. 16. Active multiplication gain A(s) versus feedback resistor Rf (Cf = 10 pF; R1 = 50 ).

Fig. 15. Active multiplication gain A(s) versus feedback capacitor Cf (Rf is open; R1 = 50 ).

Fig. 17.

Simulated curves for each part of the proposed lter.

stabilization and gain of the op-amp circuit may be inuenced by the value of feedback resistor Rf , feedback capacitor Cf , or negative input resistor R1 . Fig. 14 shows that increasing the feedback proportion Rf /R1 from 10 to 1000 (when Cf = 0) can increase the magnitude of the active multiplication gain A(s) from 0 dB Rf /R1 to about 40 dB. Moreover, the bandwidth is also broadened with the increase of Rf /R1 . However, according to the Bode stabilization criterion, when Rf /R1 is 1000, the system is unstable. Furthermore, when the feedback ratio is smaller than ten, the gain A(s) is lower than 0 dB within the whole frequency range. Thus, the minimum ratio of Rf /R1 should be larger than ten. Fig. 15 shows that reducing the value of feedback capacitor Cf from 1000 to 1 pF (when Rf is open) can increase the gain A(s) from 10 to 50 dB. However, it is found that an op-amp with Cf = 1 pF is actually an unstable system. Thus, in order to have a maximum gain A(s), the optimum value of Cf is around 10 pF. Furthermore, another unexpected phenomenon in Fig. 16

is that low-frequency noise (for example, frequency lower than 1 kHz) is also amplied. That will result in the amplication of 50-Hz power frequency signal. Thus, the resistor Rf should not be open. Fig. 16 shows that increasing the value of Rf from 300 to 30 k (when Cf = 10 pF and R1 = 50 ) can increase the magnitude of A(s) signicantly, while further increasing Rf to 3 M will broaden the bandwidth of A(s) without changing the stabilization of the system. Thus, the suitable magnitude of A(s) is around 30 dB. V. S IMULATION AND E XPERIMENTAL R ESULTS A. Simulation Based upon that analysis, six envelope curves of the simulated lter are shown in Fig. 17. In the simulation, Rf is 30 k, Cf is 10 pF, R1 is 50 , LCM is 22 mH, and C1 and C2 are 10 nF. The magnetizing inductance L of the CT is 5 mH. The dashed curves are the transfer functions of each part of the active multiplication circuit. It can be seen that the gains

CHEN et al.: NOVEL HYBRID COMMON-MODE EMI FILTER WITH ACTIVE IMPEDANCE MULTIPLICATION

1833

Fig. 18. only.

Conducted EMI CM noise spectrum result with the passive inductor

Fig. 19. only.

Conducted EMI CM noise spectrum result with the active circuit

of the injection network and CT are very small, which will give little help for improving the total gain A(s). However, the curve of the op-amp circuit maintains a high gain level within a wide frequency range, which will play a very signicant role in the whole lter system. The solid lines are the insertion losses of the CM inductor, the active circuit, and the hybrid lter. It can be seen that the CM noise is greatly reduced by the hybrid lter because of the active multiplication effect. There is a 2030-dB improvement in the whole frequency range compared with the single CM inductor. At high frequency, it should be noted that the hybrid lter performance is not as good as that at low frequency. It is mainly because the benchmark impedance of the passive inductor decreases with the increase of frequency. That is, although the inductor is boosted by A times, the multiplication impedance cannot be too large enough. The parasitic cancellation method reported in [11] can be used to optimize the high-frequency performance of the inductor. B. Experiment In our experiment, an Agilent E7401A spectrum analyzer is used to measure the CM noise. First, measurements were conducted with a pure passive inductor and a pure active circuit. Fig. 18 shows the noise spectrum when only CM inductor LCM is installed. Although a 22-mH inductor is not a small-size component, dominant peaks can still be seen at the fundamental switching frequency (40 kHz). A reduction of about 1015 dB is achieved by this entirely passive inductor when compared with the original CM noise. The experimental result of the active multiplication circuit is shown in Fig. 19. It can be seen that 2030-dB attenuation is achieved, as predicted by the model. In practical design, special attention has to be paid for the stability of the closed-loop system. Another set of measurements were taken with the HEF. It can be clearly observed from Fig. 20 that the HEF successfully attenuates the CM noise within a wide frequency range. Table II compares the attenuation accomplished by the passive lter and the hybrid lter. It shows that the average attenuation of the hybrid lter is about 3540 dB or so, but at some frequency, it cannot reach such a high value.

Fig. 20. Conducted EMI CM noise spectrum result with the combination of passive inductor and active circuit. TABLE II PERFORMANCE COMPARISON OF THE PASSIVE INDUCTOR AND THE HYBRID FILTER

In order to provide a comparison to the traditional CM lter, a three-order CLC CM lter shown in Fig. 1 is also built. The component values of the CLC lter are selected as CY = 4700 pF and LCM = 50 mH. Fig. 21 shows the measured spectrum for the CLC lter. It can be seen that, at a frequency near 40 kHz, the noise spectrum still cannot pass the EMI limit. In order to fulll this acquirement, the value of the CM

1834

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 5, MAY 2011

[6] M. Zhu, D. J. Perreault, V. Caliskan, T. C. Neugebauer, S. Guttowski, and J. G. Kassakian, Design and evaluation of feedforward active ripple lters, IEEE Trans. Power Electron., vol. 20, no. 2, pp. 276285, Mar. 2005. [7] D. Hamza and P. K. Jain, Conducted EMI noise mitigation in DCDC converters using active ltering method, in Proc. IEEE Power Electron. Spec. Conf., Jun. 1519, 2008, pp. 188194. [8] N. Mortenson and G. Venkataramanan, An active common mode EMI lter for switching converters, in Conf. Rec. IEEE IAS Annu. Meeting, Oct. 59, 2008, pp. 17. [9] M. L. Heldwein, H. Ertl, J. Biela, and J. W. Kolar, Implementation of a transformerless common-mode active lter for ofine converter systems, IEEE Trans. Ind. Electron., vol. 57, no. 5, pp. 17721786, May 2010. [10] D. Zhang, D. Y. Chen, M. J. Nave, and D. Sable, Measurement of noise source impedance of off-line converters, IEEE Trans. Power Electron., vol. 15, no. 5, pp. 820825, Sep. 2000. [11] S. Wang, J. D. van Wyk, and F. C. Lee, Effects of interactions between lter parasitics and power interconnects on EMI lter performance, IEEE Trans. Ind. Electron., vol. 54, no. 6, pp. 33443352, Dec. 2007. Fig. 21. lter. Conducted EMI CM noise spectrum result with the traditional CLC Wenjie Chen (S06M08) was born in Xian, China, in 1974. She received the B.S., M.S., and Ph.D. degrees in electrical engineering from Xian Jiaotong University, Xian, in 1996, 2002, and 2006, respectively. Since 2002, she has been a Member of the faculty of the School of Electrical Engineering, Xian Jiaotong University, where she is currently an Associate Professor and is also with the State Key Laboratory of Electrical Insulation and Power Equipment. Her main research interests include soft-switching dc/dc converters and active lters, and power electronic integration.

inductor should be enlarged further. However, the diameter of the 50-mH CM inductor is larger than 6 cm, although it is made of nanoscale crystalline cores. Thus, the advantage of the proposed hybrid lter over the traditional EMI lter lies not only in noise attenuation effect but also in size. VI. C ONCLUSION In this paper, a HEF with a benchmark CM inductor and an active booster that operate through feedback control has been built based on the active impedance multiplication approach. The design and compensation performance of the hybrid lter have been discussed in detail. Quantitative analysis and simulations of both the passive element and active components have been carried out, and the limitations and design tradeoffs have also been addressed. The experimental results for a military power supply have been presented according to MIL-STD-461. It has been proven that the proposed active impedance multiplication approach is able to boost the impedances of both the inductor and the noise source simultaneously. Thus, the cascade of passive component and active booster allows for better CM noise attenuation performance. R EFERENCES
[1] S. Ostroznik, P. Bajec, and P. Zajec, A study of a hybrid lter, IEEE Trans. Ind. Electron., vol. 57, no. 3, pp. 935942, Mar. 2010. [2] N. He, D. G. Xu, and L. N. Huang, The application of particle swarm optimization to passive and hybrid active power lter design, IEEE Trans. Ind. Electron., vol. 56, no. 8, pp. 28412851, Aug. 2009. [3] M. C. Di Piazza, G. Tin, and G. Vitale, An improved active commonmode voltage compensation device for induction motor drives, IEEE Trans. Ind. Electron., vol. 55, no. 4, pp. 18231834, Apr. 2008. [4] P. Pairodamonchai, S. Suwankawin, and S. Sangwongwanich, Design and implementation of a hybrid output EMI lter for high-frequency common-mode voltage compensation in PWM inverters, IEEE Trans. Ind. Appl., vol. 45, no. 5, pp. 16471659, Sep. 2009. [5] J. Biela, A. Wirthmueller, R. Waespe, M. L. Heldwein, K. Raggl, and J. W. Kolar, Passive and active hybrid integrated EMI lters, IEEE Trans. Power Electron., vol. 24, no. 5, pp. 13401349, May 2009.

Xu Yang (M02) was born in China in 1972. He received the B.S. and Ph.D. degrees in electrical engineering from Xian Jiaotong University, Xian, China, in 1994 and 1999, respectively. Since 1999, he has been a Member of the faculty of the School of Electrical Engineering, Xian Jiaotong University, where he is currently a Professor and is also with the State Key Laboratory of Electrical Insulation and Power Equipment. From November 2004 to November 2005, he was with the Center for Power Electronics Systems, Virginia Polytechnic Institute and State University, Blacksburg, as a Visiting Scholar. He then came back to Xian Jiaotong University and engaged in teaching and research works in the power electronics and industrial automation areas. His research interests include soft-switching topologies, pulsewidth modulation control techniques and power electronic integration, and packaging technologies.

Zhaoan Wang (SM98) was born in Xian, China, on June 9, 1945. He received the B.S. and M.S. degrees from Xian Jiaotong University, Xian, in 1970 and 1982, respectively, and the Ph.D. degree from Osaka University, Osaka, Japan, in 1989. From 1970 to 1979, he was an Engineer with Xian Rectier Factory. In 1982, he became a Lecturer with Xian Jiaotong University, where he is currently a Professor and is also with the State Key Laboratory of Electrical Insulation and Power Equipment. He is engaged in research on power conversion system, harmonic suppression, reactive-power compensation and power electronic integration, and active power lters. He has published over 150 technical papers and has led numerous government- and industrysponsored projects in the areas of power and industrial electronics.