4380 ieee transactions on power electronics, vol. 27,...

4380 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 27, NO. 11, NOVEMBER 2012

Design of a Hybrid Integrated EMC Filter for aDC–DC Power Converter

Marwan Ali, Eric Laboure, Francois Costa, Member, IEEE, and Bertrand Revol

Abstract—This study focuses on the design and the realizationof an integrated EMI filter embedded in an aircraft power supply.The first part of the study is dedicated to the modeling of the dc–dcpower, based on a “black box” representation: the converter is rep-resented by an association of sources and impedances that reflectits external EMI behavior. This initial step enables us to define theelectrical structure of the filter and the component values; a hybridsolution (active plus passive) can be effective for optimal filteringwhile minimizing the size and mass. The second part of the studyis devoted to its realization, based on the integration of capacitiveand inductive components inside a printed circuit board structure.This technology is named hybrid integration. Some technologicaldifficulties are highlighted regarding the geometry of the compo-nents and their manufacture using this technology. Finally, filterarchitecture is proposed and a functional prototype is realized. Ex-periments show that significant gains in volume are achieved whilecompliance with the DO 160 standard (aeronautics) is obtained.

Index Terms—Common and differential mode interferences,electromagnetic compatibility, electromagnetic interference (EMI)hybrid filter, integration of passive components.

I. INTRODUCTION

THE use of semiconductor components (SiC/GaN) in powerelectronics can reduce the volume of equipment, increase

their operating temperature, and improve their energy effi-ciency [1], [2]. However, due to very fast changes in voltagesand currents in the switching components (dV/dt, dI/dt), con-ducted and radiated electromagnetic disturbances occur at highfrequencies [3], [4]. To protect the network from these emis-sions and meet the standards for electromagnetic compatibility(EMC) including the aeronautic standard “DO160F (150 KHz,30 MHz),” an EMC filter is absolutely necessary [5], [6].

The EMC filter requires a specific design for the two modesof propagation of parasitic currents, i.e., common mode (CM)and differential mode (DM). The design is based on the use ofan accurate electromagnetic model of the converter. The behav-ioral model has proved its efficiency in representing accuratelyelectromagnetic interferences [7], [8]. This kind of model is

Manuscript received May 4, 2011; revised November 3, 2011 and December21, 2011; accepted December 29, 2011. Date of current version June 20, 2012.This work was supported by Fondation de Recherche pour l’Aeronautique etl’Espace (FRAE) under the Project Filtrage Electromagnetiques et Materiauxpour l’INtegration en Aeronautique (FEMINA). Recommended for publicationby Associate Editor M. Harke.

M. Ali, F. Costa, and B. Revol are with Laboratoire SATIE, ENSC, Cachan94235, France (e-mail: [email protected]; [email protected];[email protected]).

E. Laboure is with Paris-Sud 11 University, Orsay 91400, France (e-mail:[email protected]).

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TPEL.2012.2183387

TABLE ISPECIFICATION OF CONVERTER UNDER TEST

constituted by disturbance equivalent sources and CM and DMequivalent impedances. In recent years, several studies haveshown the interest but also the limitations of EMC active filtersin a frequency range up to several megahertz [9], [10]. Otherresearch has show a reduction of the efficiency of EMC filtersdue to their parasitic components and to imperfections in mate-rials [11]. Thus, to exploit the performances of each technology,a hybrid optimized EMC filter, that is composed of an activepart for low frequencies and a passive part for high frequencies,is the most compact solution [12]–[14]. The integration of thepassive part of this hybrid filter will be presented in this paper.

First, the electromagnetic characterization of the converterunder test is presented. In order to increase the effectiveness ofthe passive part of a hybrid filter and to reduce its volume, afull integration of this part in the printed circuit board (PCB) isproposed. Its implementation requires the choice of a magneticmaterial which can be integrated and the use of high permittivitydielectric materials in order to achieve high value of planarcapacitances. Taking into account the brittleness of magneticmaterials (ferrites), several tests for their integration into a PCBhave been done. A planar geometry that meets the specificationsis then chosen. After characterization of the materials that areused in this new architecture, its modeling by the finite elementmethod (FEM) is presented. Measurements of the attenuationsin CM and DM of the integrated passive filter are then comparedwith those of identical discrete ones.

II. CHARACTERIZATION AND MODEL OF ELECTROMAGNETIC

INTERFERENCES GENERATED BY THE POWER CONVERTER

The power converter considered in this paper is a dc–dc powersupply designed to equip a civil aircraft. Aeronautic standardsmust be therefore be respected and in particular the DO160Fstandard, which defines the maximum allowable levels for elec-tromagnetic interferences. The specifications of the power con-verter are given in Table I

The intrinsic level of conducted electromagnetic interferencesgenerated by a power converter can be identified and separatedinto two contributions, i.e., CM and DM. In this section, an

0885-8993/$31.00 © 2012 IEEE

ALI et al.: DESIGN OF A HYBRID INTEGRATED EMC FILTER FOR A DC–DC POWER CONVERTER 4381

Fig. 1. Converter global EMI model (CM and DM).

Fig. 2. Input voltages and currents of converter under test.

equivalent behavioral EMC model is proposed with a limitednumber of parameters that can be identified by analytical calcu-lations, simulations, or electrical measurements [15], [16]. Thismodel includes two parts: an equivalent impedances circuit [17]associated with equivalent sources of voltage (CM) and current(DM) (see Fig. 1). This model allows the calculation of theemission level of the converter regardless of the configurationof the EMC filter placed at the input. Note that the identificationof impedances and sources is realized without EMC filter and,in this model, the equivalent noise sources are placed accordingto their mode of propagation [18], [19]. In the proposed method,the power converter is considered a “black box” with two inputports, respectively; between line 1 and mass, and between line 2and mass (see Fig. 2). The development of this equivalent modelwill be discussed in the following sections.

A. Impedance Modeling

In this section, the identification procedure of the EMC blackbox impedances is described. The system to be identified has twoports. Therefore, there are four impedances to define, namelyZ11 , Z22 , Z12 , and Z21 , i.e., the four values of the impedancematrix

V = ZI ⇒[

v1v2

]=

[Z11 Z12Z21 Z22

] [i1i2

]. (1)

In order to perform both the direct measurements of impedanceand transimpedance of the system under test, an impedanceanalyzer HP4194a configured for transfer mode measurement

Fig. 3. Measured and calculated input impedances: (a), (b), (c), and (d)correspond to Z11 , Z22 , Z12 = Z21 and ZDM , respectively. 〈〈ZdBohm s =20log(Zohm s )〉〉.

(T/R) is used. A current is, therefore, injected by the sourcegenerator of the analyzer in one of the input ports of the off-state converter board. The resulting voltage is then measured inport T. The injected current is measured using a current sensorwith a large bandwidth (Tektronix CT2: 1 mV/1 mA, loaded with50 Ω) and connected to port R. This protocol requires correctionsthat depend on imperfections due to connections and transferfunctions of the current and voltage probes used.

The equipment under test is shielded to avoid the influence ofexternal fields. This measurement bench also guarantees a stableconfiguration of connections, and thus, the parasitic impedancesof connections are almost invariant (see Fig. 2).

Using this test bench, three impedances are measured, i.e., theimpedance between line 1 and mass, between line 2 and mass,and between line 1 and line 2 corresponding, respectively, to Z11 ,Z22 , and the differential impedance ZDM . These impedances areshown in Fig. 3.

From these measurements, the impedances Z11 and Z22 areclearly identical leading to the same CM impedance. In addition,the transimpedances Z12 and Z21 are also identical, which allowsus to propose a simple electrical model with only three localizedimpedances (see Fig. 4)

The proposed model consists basically of two parasitic in-ductances of equal value, i.e., L1 and L2 placed on both lines inseries with two parasitic resistances RLS . This model is supple-mented by two identical CM capacitances (C1 and C2) placedbetween each line and the mass. The DM impedance Zdiff isrepresented by a capacitance C0 connected in parallel with aresistor R0 , and parasitic components Rps and Lps .

Measurements and identifications led to determination of thevalue of each element of the model [20]. These values are shownin Table II.

The theoretical impedances corresponding to each measure-ment configuration have been calculated from this model. Theexperimental results presented in Fig. 3 are clearly in goodagreement with simulated results.


Fig. 4. Converter equivalent model of impedance for EMI analysis.

TABLE IIIMPEDANCE OF EMI MODEL COMPONENTS

B. CM and DM Sources Model

The EMC “black box” model must be completed by equiv-alent CM and DM sources. The symmetry of the DM and CMimpedances limits the phenomena of mode conversion. Thissimplifies the extraction of the equivalent sources.

In order to realize this extraction, the power converter isfed through two identical line impedance stabilization network(LISN) placed on each line. They meet the recommendations ofstandard DO160F and they stabilize the impedances of the powersource. According to the complete model given in Fig. 5, theequivalent sources can be extracted by measuring the two linecurrents. The equivalent CM voltage source and the differentialcurrent source can be determined by

VCM =(

ZLISN + ZL + ZC

2

)(I1 + I2) (2)

IDM = (I1 − I2)(

ZLISN + ZL

2ZC+

ZLISN + ZL

ZDM+

12

). (3)

In order to improve the accuracy in the calculation of the twoequivalent sources, currents “I1 + I2” and “I1−I2” have beendirectly measured on the test bench using two identical cur-rent probes. The acquisitions at this step are performed us-ing an oscilloscope (Tektronix DPO 7104 “1 GHz”). The dataare then processed using MATLAB. In this software, the cur-rent sensor voltages are first corrected by applying the inversetransfer function of the sensors, leading to the current values,

Fig. 5. CM and DM sources associated with the power converter equivalentmodel of impedance.

Fig. 6. Calculated spectra of CM and DM sources [fast Fourier transform(FFT)] in the EMI model of the power converter.

i.e., “I1 + I2” and “I1−I2”. The CM and DM sources are thencalculated using (2) and (3).

The calculated spectra corresponding to CM and DM sourcescan be seen in Fig. 6.

C. EMC Model Validation

To verify that the proposed model can correctly representthe levels of electromagnetic interference (EMI) generated bythe power converter, the LISN voltages have been calculatedfrom this model. Experimental and calculated LISN voltagespectra corresponding to VRrsil1 and VRrsil2 are shown in Fig. 7.Experimental spectra are measured using an HP4195A spectrumanalyzer. They are in good agreement with the simulated levels.This agreement proves that the proposed model as well as themethod of identification of its parameters are correct.

This model provides a relevant way to calculate and to opti-mize the EMC filter. In the next section, it will be shown howthe required attenuations can be calculated to meet the DO160Fstandard required levels.


Fig. 7. Calculated and measured LISN voltages.

III. FILTER DESIGN

The first step of the design of the EMC filter is to set the re-quired attenuation levels for CM and DM. The previous modelin Fig. 5 can be split into two configurations corresponding toDM and CM as presented in Fig. 8. Because the DO160F stan-dard gives the limit levels in terms of current rather than voltage,it is better to reason with current sources of disturbances. Thus,the model of CM can be rearranged to bring up an equivalentsource of CM current, denoted ICM .

The theoretical required levels of attenuation can now becalculated from the previous models by determining the levelsof emission associated with the CM current source only, thenthose related to the DM current source.

From the spectra of IDM and ICM (see Fig. 9), the requiredattenuation level can be determined on the DO160F frequencyrange by

AttDM = IDM(d B μ A ) −(

Limit (DO160F)(dBμA) − 6 dB)(4)

AttCM = ICM(d B μ A ) −(

Limit (DO160F)(dBμA) − 6 dB)

.

(5)An extra 6 dB reduction of the standard level is justified becausethe overall level of disturbances of CM and DM may add up inthe worst case. This gives, for the same amplitude of the twocontributions, twice the level of each emission.

The attenuation provided by an EMC filter is a smooth con-tinuous function; thus, the attenuations required are given bythe envelopes of the calculated attenuations. They are plotted bythe continuous line in Fig. 10.

Without a filter, the attenuations provided by the struc-tural elements of the power converter (impedances networkof the “black box”) for both modes are calculated using the

Fig. 8. Electrical model configuration with a passive filter for parametricsimulation on (a) CM and (b) DM.

Fig. 9. Comparison of CM and DM current sources with levels defined by theDO160F standard.

Fig. 10. (a) CM and (b) DM attenuations to meet the DO160 requirements.


Fig. 11. (a) Schematic of a single-stage passive EMI filter. (b) Magneticbehavior of coupled inductors for CM and DM currents.

electrical model given in Fig. 8. The corresponding attenuationsare plotted by the dotted line in Fig. 10. It can be seen from thisfigure that the attenuation introduced by the structural passiveelements does not achieve the required attenuations. For fre-quencies above 1 MHz, these attenuations will be provided bythe passive filter and the values of the passive filter componentscan be calculated to ensure these required attenuations [21].

In this paper, the necessary attenuation on the frequency bandof the DO160F standard is obtained, thanks to a hybrid filter.Such hybrid architecture is known to be the best approach tooptimizing EMI filtering [12], [22]. For one part of the fre-quency band, an active filter compensates parasitic signals, butthe limitations of the active filter begin to be significant beyonda few megahertz. To cover the entire frequency band, the activefilter must be associated with a passive filter giving the requiredattenuation in the HF frequency range, for example [2–30 MHz].

In a hybrid filtering structure, the architecture of the passivefilter can be simpler than in a purely passive structure. A single-stage structure, as shown in Fig. 11, is sufficient in this case. TheCM filter consists of two symmetrical capacitors between eachline and ground, associated with two coupled inductors. Theleakage inductance of the coupled inductor constitutes, withthe capacitor added between the two lines, a DM filter. In anintegrated view of this passive filter, values of capacitive ormagnetic components depend on their geometry and they canbe determined from the required theoretical attenuation. Thistheoretical attenuation can be calculated using the “black box”model.

Electric models that represent the filter are added to the pre-vious DM and CM models (see Fig. 12). In these models, theposition of the capacitances C(CM) and C(DM) can be modi-fied: they can be placed at one end or the other of the coupled

Fig. 12. Electrical model with a passive filter for parametric simulation for(a) CM and (b) DM.

Fig. 13. (a) CM and (b) DM attenuations with a one-stage passive filter.

inductor. These DM and CM capacitances are, however, definedby the limits of the chosen technology for integration in PCB.In our case, capacitances are made of thin layers of high-k ma-terial embedded in the PCB, leading to C ≈ 1.7 nF/cm2 . For anavailable area of 4 cm × 5 cm for the PCB, the capacitances ofthe proposed design will be limited to 20 nF for DM and 10 nFfor CM, leading to an equivalent 20 nF capacitance in CM and25 nF capacitance in DM.

Based on the previous values of DM and CM capacitances,the parametric calculation of the attenuations can be achieved byvarying the inductance values for both modes. The attenuationscalculated in Fig. 13 show the existence of optimum values forLCM and LDM , namely LCM = 3 μH and LDM = 3 μH, achievingthe requirements of DO 160 over the entire frequency range for


TABLE IIICOMPONENT VALUES OF THE REQUIRED PASSIVE FILTER

CM and in the range [2.5–30 MHz] for DM. In this last case,the associated active filter will have to effectively reduce theDM disturbances from 150 kHz to 2.5 MHz. This is reasonablypossible with suitable electronics.

Therefore, for the converter considered in this paper and thestructure chosen for the passive filter, the theoretical requiredvalues are given in Table III.

However, the effectiveness of the passive filter is often de-graded by imperfections of the passive components i.e., par-asitic elements, interconnections, and materials properties. Tocontrol and manage these effects, the integration of the passivecomponents within the PCB is proposed in the next section.

IV. INTEGRATION OF THE PASSIVE FILTER

Integration of passive components for power electronic con-verters and in particular of EMC filters is the main focus ofthis section. In recent years, several integration technologies[23], [24] and multilayer structures embedded in PCB havebeen studied [25]–[27]. In order to meet the application con-straints (EMC, volume, mass, mechanics, environment, etc.),planar passive components embedded within a PCB will beproposed to use.

The passive EMI filter structure, proposed in Fig. 14, consistsof an inductive part, i.e., a coupled inductor, and two capacitiveparts (CCM and CDM ). In this section, the technological feasibil-ity of each part will be discussed separately. The design of thisstructure requires a global approach to achieve the specificationsand, at the same time, minimize mass and volume.

A. Impedance Modeling

1) Proposed Design: In the proposed filter, the CM and DMcapacitors are made from a material known as C-Ply [28]. Thismaterial is sandwiched between two copper plates [see the crosssection of this assembly in Fig. 15(b)]. It enables the design ofplanar capacitors compatible with PCB technologies. The onlyconstraint in the implementation of such a material is to placethese layers symmetrically in the PCB stacking. In our design,there are two capacitive layers. One layer is dedicated to the DMcapacitor and the other to the CM capacitors. As a consequenceof this design, the total area dedicated to the DM capacitor istwice that of the CM capacitors leading to an equal equivalentcapacitance in CM and in DM.

In Fig. 15(a), a PCB prototype can be seen including twocapacitive layers with, on the top, six connections including twocentral connections to ground on both sides of the filter and four

Fig. 14. (a) Realization of the integrated EMI filter. (b) Integrated filter exter-nal dimensions. (c) Equivalent electrical circuit.

input and output connections. The total area of this integratedfilter is 20 cm2 . The dielectric material used for the capacitorsis 16 μm thick and its main characteristics are summarized inTable IV.

2) Main Characteristics of the Integrated Capacitors: A ca-pacitor is usually modeled by a capacitance in series with an


Fig. 15. (a) Photograph of the integrated passive filter. (b) Transversal crosssection of the capacitive layer.

TABLE IVCAPACITIVE LAYER MAIN PROPERTIES

Fig. 16. Equivalent circuit of the integrated capacitor.

equivalent series resistor (ESR) and an equivalent series induc-tance (ESL) [29]. If the capacitor is connected using the four-point transmission line connection method, like in Fig. 16, theECLs will add up to CM and DM inductances and the filteringefficiency will not be degraded by the parasitic elements [23].The impedance measurements of CM and DM capacitors areshown in Fig. 17. From these measurements, the equivalentelectrical model can be identified leading to CCM = 9 nF, CDM= 18 nF, ESR = 3 mΩ, ESL = 0.78 nH.

B. Inductive Part of the Integrated EMC Filter

Regarding the CM passive filtering, two coupled inductorsusing a common magnetic material with a high permeability

Fig. 17. CM and DM capacitor impedances.

Fig. 18. First prototype integrating a magnetic component within a PCB.

and a high saturation level are designed. In the configurationof Fig. 11(b), the leakage flux determines the DM inductances.These inductances must be controlled by correct design of themagnetic core.

1) Choice of Magnetic Material and Core Design: Magneticmaterials often used for filtering EM interferences are ferritematerials. Their magnetic properties vary with frequency [30].For EMC filtering applications, materials with high losses athigh frequencies have been sought.

The first design for testing integration of a magnetic compo-nent in the PCB led us to integrate a component that containstwo magnetic plates and two vertical legs (see the cross sectionin Fig. 18). In this magnetic structure, windings are integratedin the PCB around the two vertical legs leading to the requiredmagnetic structure, i.e., two coupled inductors. But this firstdesign also highlighted some technological problems namelydifficulties in ensuring proper contact between the top and bot-tom parts of the magnetic core and the presence of numerousfractures near the overhanging portions.

After this initial feedback, a new design of magnetic compo-nent has been proposed. In this design, the magnetic core is aplanar ring core (see Fig. 19). This design lowers the constraintsarising from the need to ensure a contact between different por-tions of the magnetic circuit and reduces the risk of fractures inoverhanging parts.

For such a magnetic core, the windings have to be designedon two copper layers connected by vias. This core has beenobtained from a ferrite material dedicated to high-frequencyfiltering applications and the dimensions have been defined tomeet the MC and MD attenuation requirements.

2) Characterization of Material Proprieties: From (6) andthe impedance spectrum measurement (realized with an


Fig. 19. New core design to be integrated (length = 3 cm, width = 3 cm, andthickness = 1.5 mm).

Fig. 20. Complex permeability as a function of frequency.

Fig. 21. Photograph of the integrated coupled inductor corresponding to thenew design (volume = 4.5∗4∗0.25 = 4.5 cm3 ).

HP4195A impedance analyzer) of a N-turn winding wound onthe magnetic core, the complex permeability of the magneticmaterial can be determined (see Fig. 20)

ZS = jωL = jωN 2 μ0Ae

le(μ′

S − jμ′′S ) . (6)

From these values, it is then possible to calculate the coupledinductors CM and DM impedances throughout the frequencyrange of the DO160 standard.

3) Design and Finite Element Modeling: To validate the de-sign principle of the inductors, two coils are etched on two400-μm-thick PCBs. The turns of the two windings are spacedby 0.8 mm to minimize the stray capacitance (EPC) betweenturns [31] (see Fig. 21). The ceramic magnetic core is integratedwithin a PCB, and the layers connecting vias are performed withmetalized holes.

The inductances corresponding to CM and DM are measuredaccording to the configurations shown in Fig. 22. The resultscorresponding to these two configurations are given in Fig. 23.

Fig. 22. (a) Measurement configurations for CM impedance and (b) DMimpedance [32].

Fig. 23. CM and DM inductances of the planar coupled inductors.

From these measurements, it was found that the proposeddesign, reducing the turn-to-turn parasitic capacitance as wellas the interwinding capacitances, leads to maintenance of theinductive behavior of the magnetic component on the entirefrequency range of the DO160 standard.

The behavior of the magnetic component has also been cal-culated using a simulation software based on the FEM (COM-SOL 3-D). Fig. 24(a) and (b) shows the simulated behavior ofthe magnetic component in the two current configurations inFig. 22.

It can be seen in these figures that the field distributionsare very different in the two configurations. For CM currents,flux density flows entirely within the magnetic circuit; for DMcurrents, magnetic field lines are closed in the air, outside thecore. These behaviors are consistent with our expectations.

C. Performances of the Integrated Passive Filter

The performances of a filter can be determined, for a definedsource and load impedance, by measuring the attenuation intro-duced by the filter in this configuration. This attenuation is givenas the ratio between the filter input voltage and the filter out-put voltage. This measurement is performed using an HP4195Afrequency analyzer used in transfer mode configuration. Themeasurement configurations are shown in Fig. 25(a) and (b) forCM and DM, respectively.


Fig. 24. (a) CM and (b) DM: finite-element simulation of the integratedinductance.

To demonstrate the benefits realized by the integration of anEMC filter, an equivalent discrete filter with identical compo-nent values at low frequencies was constructed [see Fig. 25(c)].In Fig. 26, the attenuations obtained, respectively, with the dis-crete and integrated filters are compared. For frequencies be-tween 500 kHz and 10 MHz, attenuations for DM and CM arehigher with the discrete filter. This relative lack of performancein this frequency band is linked, for the integrated passive filter,to the reduction of inductance with frequency due to the re-duction of the real part of relative permeability with frequency(see Fig. 20). Therefore, these lower performances are due tomaterial properties and are not due to design. Nevertheless, themain advantages of integration can be seen beyond 10 MHz ascontrolled and lower inductive and capacitive parasitic effectslead to much higher resonant frequencies. Thus, attenuations ob-tained at high frequencies are higher for the integrated passivefilter.

The interconnection parasitic inductances Lvia between theintegrated capacitors are minimized due to the use of via withlow lengths [see Fig. 27(a)]. In contrast, the parasitic inductances

Fig. 25. Configuration for measurement of (a) CM attenuation and (b) DMattenuation. (c) Photograph of the discrete EMC filter (volume = 3.5∗3.2∗1.2= 13.44 cm3 ).

Fig. 26. CM and DM attenuation measurement for (a) discrete passive filterand (b) passive integrated filter.


Fig. 27. Electrical representation (a) via interconnecting the CM and DMcapacitors and (b) parasitic capacitor of the integrated winding.

Fig. 28. CM and DM currents with a passive integrated EMI filter.

of the discrete capacitors and the interconnection wires make theparasitic inductive effects important in the discrete EMC filter.The parasitic capacitive effect in an integrated winding is shownin Fig. 27(b). This effect can be optimized by a calculation ofthe wire-to-wire spacing D and the wire-to-core spacing d. Theminimal controlled value of “D and d,” in the current integrationprocess, is 200 μm. Moreover, the high resistivity of the ferritelowers the effect of the wire-to-core capacitance. Due to thethinness of the copper, the global ESR of the embedded chokeincreases for the dc current. To overcome this disadvantage, anadditional metallization is recommended.

The resulting embedded choke volume is 63% lower thanthe equivalent discrete choke. It can also be emphasized thatthe total volume of the integrated filter is 58% lower than theequivalent discrete filter. The proposed integrated structure isalso shielded by the two ground planes located at the top andbottom of the PCB stacking.

The global behavior of the integrated filter is, therefore, betterthan its equivalent using discrete components.

The integrated filter is now set up to reduce the interferencesgenerated by the converter used to support our study. It can beshown in Fig. 28 that, as expected, the integrated filter reducesCM and DM interferences in the frequency range that extendsfrom 2.5 to 30 MHz.

An active filter must be added to reduce EMI interferencesdue to switching in the frequency band up to 2.5 MHz. Possibletopologies of active filters are proposed in [33] and [34]. For ourapplication, the chosen topology uses a voltage compensationeffect generated by an auxiliary winding added to the coupledinductor. This topology will be described in a future paper.

V. CONCLUSION

Reducing the volume and improving the efficiency of EMCfilters for power electronics devices are strategic issues partic-ularly in the case of complex systems such as those found inthe aerospace or automotive industries. In this paper, a char-acterization procedure for a power supply has been presented.This method allows constructing an electromagnetic behavioralmodel of the converter. From this model, it is possible to de-fine the characteristics of an EMC filter optimized to meet thespecifications given by a standard. The proposed EMC filter isdesigned to be integrated into the thickness of a PCB. One ofthe main challenges of such a design is the integration of mag-netic components. The proposed design reduces the volumeof the filter by 58% and improves the CM and DM attenua-tions at high frequency compared to the same filter made ofdiscrete components. The proposed structure optimized for fil-tering high-frequency interferences (beyond 2.5 MHz) must besupplemented by a low-frequency active filter. It is easy to usethe PCB board incorporating the passive filter as a support to thisnew role. This architecture, thus ensures maximum efficiencyand optimum compactness for the EMC filter.

ACKNOWLEDGMENT

The authors would like to thank the industrial partners ofthe project, namely AIRBUS France, EADS France InnovationWorks, and HISPANO-SUIZA. In addition, they would like toexpress their thanks to CIREP the company that supports themanufacture of prototypes.

REFERENCES

[1] I. Omura, W. Saito, T. Domon, and K. Tsuda, “Gallium nitride powerHEMT for high switching frequency power electronics,” in Proc. Int.Workshop Physics of Semiconductor Devices, 2007, pp. 781–786.

[2] M. Besacier, M. Coyaud, J. Schanen, J. Roudet, and B. Rivet, “Hybrid Si-SiC fast switching cell modelling and characterisation including parasiticenvironment description by PEEC method,” in Proc. IEEE 33rd Annu.Power Electron. Spec. Conf., 2002, pp. 1753–1757.

[3] X. Pei, J. Xiong, Y. Kang, and J. Chen, “Conducted EMI emission inPWM inverter,” in Proc. 4th Int. Conf. Power Electron. Motion Control,2004, vol. 2, pp. 630–635.

[4] T. Ninomiya, M. Shoyama, C.-F. Jin, and G. Li, “EMI issues in switchingpower converters,” in Proc. 13th Int. Symp. Power Semicond. Devices ICs,2001, pp. 81–86.

[5] H. Akagi and T. Shimizu, “Attenuation of conducted EMI emissions froman inverter-driven motor,” IEEE Trans. Power Electron., vol. 23, no. 1,pp. 282–290, Jan. 2008.

[6] K. Raggl, T. Nussbaumer, and J. Kolar, “Guideline for a simplifieddifferential-mode EMI filter design,” IEEE Trans. Ind. Electron., vol. 57,no. 3, pp. 1031–1040, Mar. 2010.

[7] A. C. Baisden, D. Boroyevich, and F. Wang, “Generalized terminal mod-eling of electromagnetic interference,” IEEE Trans. Ind. Appl., vol. 46,no. 5, pp. 2068–2079, Sep./Oct. 2010.

[8] P.-S. Chen and Y.-S. Lai, “Effective EMI filter design method for three-phase inverter based upon software noise separation,” IEEE Trans. PowerElectron., vol. 25, no. 11, pp. 2797–2806, Nov. 2010.


[9] L. LaWhite and M. Schlecht, “Design of active ripple filters for powercircuits operating in the 1–10 MHz range,” IEEE Trans. Power Electron.,vol. 3, no. 3, pp. 310–317, Jul. 1988.

[10] M. Di Piazza, G. Tine, and G. Vitale, “An improved active common-modevoltage compensation device for induction motor drives,” IEEE Trans.Ind. Electron., vol. 55, no. 4, pp. 1823–1834, Apr. 2008.

[11] S. Wang, J. van Wyk, and F. Lee, “Effects of interactions betweenfilter parasitics and power interconnects on EMI filter performance,”IEEE Trans. Ind. Electron., vol. 54, no. 6, pp. 3344–3352, Dec.2007.

[12] J. Biela, A. Wirthmueller, R. Waespe, M. Heldwein, K. Raggl, and J. Kolar,“Passive and active hybrid integrated EMI filters,” IEEE Trans. PowerElectron., vol. 24, no. 5, pp. 1340–1349, May 2009.

[13] P. Pairodamonchai, S. Suwankawin, and S. Sangwongwanich, “Design andimplementation of a hybrid output EMI filter for high-frequency common-mode voltage compensation in PWM inverters,” IEEE Trans. Ind. Appl.,vol. 45, no. 5, pp. 1647–1659, Sep./Oct. 2009.

[14] S. Wang, Y. Maillet, F. Wang, D. Boroyevich, and R. Burgos, “Investiga-tion of hybrid EMI filters for common-mode EMI suppression in a motordrive system,” IEEE Trans. Power Electron., vol. 25, no. 4, pp. 1034–1045, Apr. 2010.

[15] F. Giezendanner, J. Biela, J. Kolar, and S. Zudrell-Koch, “EMI noiseprediction for electronic ballasts,” IEEE Trans. Power Electron., vol. 25,no. 8, pp. 2133–2141, Aug. 2010.

[16] A. Farhadi and A. Jalilian, “Modeling and simulation of electromagneticconducted emission due to power electronics converters,” in Proc. Int.Conf. Power Electron., Drives Energy Syst., 2006, pp. 1–6.

[17] Y. Panov and M. Jovanovic, “Practical issues of input/output impedancemeasurements in switching power supplies and application of measureddata to stability analysis,” in Proc. 20th Annu. IEEE Appl. Power Electron.Conf. Expo., 2005, vol. 2, pp. 1339–1345.

[18] J. Meng, W. Ma, Q. Pan, L. Zhang, and Z. Zhao, “Multiple slope switchingwaveform approximation to improve conducted EMI spectral analysis ofpower converters,” IEEE Trans. Electromagn. Compat., vol. 48, no. 4,pp. 742–751, Nov. 2006.

[19] F. Costa, C. Vollaire, and R. Meuret, “Modeling of conducted commonmode perturbations in variable-speed drive systems,” IEEE Trans. Elec-tromagn. Compat., vol. 47, no. 4, pp. 1012–1021, Nov. 2005.

[20] V. Tarateeraseth, Bo Hu, Kye Yak See, and F. Canavero, “Accurate extrac-tion of noise source impedance of an SMPS under operating conditions,”IEEE Trans. Power Electron., vol. 25, no. 1, pp. 111–117, Jan. 2010.

[21] T. Nussbaumer, M. Heldwein, and J. Kolar, “Differential mode input filterdesign for a three-phase buck-type PWM rectifier based on modelingof the EMC test receiver,” IEEE Trans. Ind. Electron., vol. 53, no. 5,pp. 1649–1661, Oct. 2006.

[22] W. Ho, C. Lee, X. Liu, P. Chan, S. Hui, and et Y. Lee, “A hybrid EMIfilter with ultra-wide bandwidth,” in Proc. 23rd Annu. IEEE Appl. PowerElectron. Conf. Expo., 2008, pp. 676–681.

[23] R. Chen, J. van Wyk, S. Wang, and W. Odendaal, “Technologies andcharacteristics of integrated EMI filters for switch mode power supplies,”in Proc. IEEE 35th Annu. Power Electron. Spec. Conf., 2004, vol. 6,pp. 4873–4880.

[24] B.-C. Tseng, L.-C. Liao, and J.-Y. Ling, “Bandwidth improvement ofmultilayer common mode filter by time domain method,” in Proc. IEEEInt. Symp. Electromagn. Compat., 2009, pp. 171–175.

[25] R. Chen, F. Canales, B. Yang, P. Barbosa, J. van Wyk, and F. Lee, “Inte-gration of electromagnetic passive components in DPS front-end DC/DCconverter—A comparative study of different integration steps,” in Proc.18th Annu. IEEE Appl. Power Electron. Conf. Expo., Miami Beach, FL,pp. 1137–1142.

[26] E. Waffenschmidt, B. Ackermann, and J. Ferreira, “Design method andmaterial technologies for passives in printed circuit board embedded cir-cuits,” IEEE Trans. Power Electron., vol. 20, no. 3, pp. 576–584, May2005.

[27] Q. Chen, Z. Gong, X. Yang, Z. Wang, and L. Zhang, “Design consid-erations for passive substrate with ferrite materials embedded in printedcircuit board (PCB),” in Proc. IEEE Power Electron. Spec. Conf., 2007,pp. 1043–1047.

[28] [Online]. Available: http://www.faradflex.com/[29] S. Wang, F. Lee, and J. van Wyk, “A study of integration of para-

sitic cancellation techniques for EMI filter design with discrete compo-nents,” IEEE Trans. Power Electron., vol. 23, no. 6, pp. 3094–3102, Nov.2008.

[30] R. Huang, D. Zhang, and K. Tseng, “Determination of dimension-independent magnetic and dielectric properties for Mn–Zn ferrite cores

and its EMI applications,” IEEE Trans. Electromagn. Compat., vol. 50,no. 3, pp. 597–602, Aug. 2008.

[31] S. Wang, F. Lee, and J. van Wyk, “Inductor winding capacitance cancel-lation using mutual capacitance concept for noise reduction application,”IEEE Trans. Electromagn. Compat., vol. 48, no. 2, pp. 311–318, May2006.

[32] J. Kotny, X. Margueron, and N. Idir, “High frequency modeling method ofEMI filters,” in Proc. IEEE Energy Convers. Congr. Expo., 2009, pp. 1671–1678.

[33] M. Heldwein, H. Ertl, J. Biela, and J. Kolar, “Implementation of atransformer-less common mode active filter for off-line converter sys-tems,” in Proc. 21st Annu. IEEE Appl. Power Electron. Conf. Expo., 2006,p. 7.

[34] K. Mainali and R. Oruganti, “Design of a current-sense voltage-feedbackcommon mode EMI filter for an off-line power converter,” in Proc. IEEEPower Electron. Spec. Conf., 2008, pp. 1632–1638.

Marwan Ali was born in Lebanon in 1985. Hereceived the Master I degree in electronics fromLebanese University, Beyrouth, Lebanon, in 2007,and the Master II degree in electromagnetic compati-bility for industrial systems from Universite BlaisePascal, Clermont Ferrand, France, in 2008. He iscurrently working toward the Ph.D. degree in elec-trical engineering at Laboratoire Systemes et Ap-plications des Technologies de l’Information et del’Energie (SATIE), Ecole Normale Superieure deCachan (ENSC), Cachan, France.

His research interests include the EMI/EMC in power electronics systems,design of integrated passives for power electronics applications, active filtersperformance, and optimization of EMI hybrid filters.

Eric Laboure received the electrical engineering de-gree from INSA-Lyon, Lyon, France, in 1989, and thePh.D. degree in electrical engineering from the EcoleNormale Superieure de Cachan, Cachan, France, in1995.

Prior to 2008, he was an Assistant Professor atENS de Cachan in the team IPEM of LaboratorySATIE. Since 2008, he has been a Professor at Paris-Sud 11 University, Orsay, France. He is currentlyworking in the team ICHAMS of Laboratory LGEP.His research interests include power electronics, in-

tegration in power electronics, electromagnetism, and electromagnetic compat-ibility.

Francois Costa (M’09) received the Ph.D. degree inelectrical engineering from the University of Paris-Sud, Orsay, France, in 1992.

He is a Full Professor at the Institut Universi-taire de Formation des Maıtres, University Paris EstCreteil, France, where he is responsible for the masterin teaching sciences and technology. Since 1999, hehas been the Leader of the Power Electronics Teamin SATIE Laboratory (SATIE, CNRS UMR), Cachan,France. His research interests include high-frequencymedium-power converters, EMI issues and their mod-

eling, HF instrumentation, integration in power electronics, and piezoelectricconverters.

Bertrand Revol received the Engineer degree of EN-SIEG from the Institut National Polytechnique deGrenoble, Grenoble, France, in 2000, and the Ph.D.degree in electrical engineering from the UniversityJoseph Fourier, Grenoble, France, in 2003.

Since September 2004, he has been an AssistantProfessor at the Ecole Normale de Cachan (ENSC),Cachan, France, where he carries out research in thelaboratory of SATIE and works on improvement ofthe EMI issues and modeling for optimization andintegration in power electronics. His principal fields

of research relate to the EMC of the power electronics converters.

4380 ieee transactions on power electronics, vol. 27,...

Documents