EMI Filter Design - IEEE Xplore

EMI Filter Design

Part III: Selection of Filter Topology for Optimal Performance Vuttipon Tarateeraseth, Member, IEEE, Department of Electrical Engineering, Srinakharinwirot University, Thailand. E-mail: [email protected] Abstract – In the first two parts of the electromagnetic interference (EMI) filter design series, the conducted EMI generation mechanism and the method on the measurement of noise source impedances of a switched-mode power supply were described. In this final part, the selection of filter topology for optimal performance is explained. With good accuracy of both amplitude and phase of termination impedances, an EMI filter can be designed systematically to give optimal performance.

I. Introduction To comply with the conducted electromagnetic interference (EMI) limits, the passive EMI filter is still widely used in a switched-mode power supply (SMPS). However, one challenging task is to design the EMI filter systematically with a minimum guess. A few papers exist on systematic design procedures of EMI filters [1], [2]. The main difference between the two papers is that in [2] the noise source and load impedances are taken into account in the EMI filter design procedure, but they are not considered in [1]. Without the noise source and load impedances in the design process, it will lead to an over-designed EMI filter performance, as the filter performance is strongly dependent on the terminating impedances [3] - [5]. In this final part of EMI filter design series, a systematic and effective design procedure for the power line EMI filters to be used in SMPS applications is described [7] - [8]. The proposed procedure is validated and demonstrated by practical design examples.

II. Equivalent Circuit of EMI Filters The structure of a typical generic filter topology of a power supply filter which is composed of common mode choke (CM choke), differential mode choke (DM choke), X-capacitors and Y-capacitors, as shown in Fig. 1. It should be noted that the most common power supply filter topology is some version of the generic filter topology [9]. The main objective of an EMI filter is to suppress the conducted DM and CM emissions generated by the SMPS and denoted as currents ID and IC in Fig. 1. The noise currents measured at LISN and denoted as I′D and I′C should be lower than the conducted EMI limit line. Like other passive filter circuits, the passive EMI filters are made from passive components; specific types of resistor, inductor and capacitor are used. More details of those components and equivalent circuits of EMI filter are described in this section.

' ' '

Fig. 1: The structure of a typical generic filter topology of a power supply filter.

60

©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

Zs = high impedance

20 dB/decade

Zl = high impedance

Zs = low impedance

Zs

Zl

C-filter

vo′

vs

Load

Source

L-filter

40 dB/decade

Zl = low impedance

Zs = low impedance

40 dB/decade

Zl

CL-filter

vo′

vs

Load

Zl

Source

LC-filter

(c)

vo′

Load

(d)

60 dB/decade

Zl = high impedance

Zg = low impedance

Zs

60 dB/decade

Zl = low impedance

Zs

vs

Zl

Source

Zl = high impedance

Zs

vs

Zg = high impedance

Load

(b)

Zs

Source

vo′

Zl

(a)

Zs = high impedance

Zl = low impedance

Zs

vs

Source

20 dB/decade

π -filter

Load

vo′

vs

Zl

Source

T-filter

(e)

vo′

Load

(f)

Fig. 2: Basic filter configurations. (a) C-filter; (b) L-filter; (c) CL-filter; (d) LC-filter; (e) p-filter; (f) T-filter.

Fig. 2: Basic filter configurations. (a) C-filter; (b) L-filter; (c) CL-filter; (d) LC-filter; (e) π-filter; (f) Even though EMI filter circuits are diverse, they are composed of equivalent DM filters and equivalent CM filters that are based on basic T-filter. filtering circuits i.e. C-filter, L-filter, LC-filter, CL-filter, p-filter and T-filter as shown in Fig. 2 [10], [11]. The basic filtering circuits provide different insertion loss (IL) rates: 20 dB/decade for C- and L-filters, 40 dB/decade for CL- and LC-filters and 60 dB/decade for p- and T-filters. However, the IL rates of the filters are strongly dependent on the impedances seen at either ends of the filters. The filter behavior can be significantly affected, if the terminating impedances at both ends are not appropriate. The C-filter will provide 20 dB/decade IL rate with a high impedance system but it is ineffective if it is employed in a low impedance system. On the other hand, the L-filter is suitable for a low impedance system but not for a high impedance system. To gain higher IL rates, CL-, LC-, p- and T-filters need to be used. Again, to prevent the ineffectiveness of the filter performances, filter configurations should be chosen appropriately with their connecting impedances at both ends. The appropriate impedances for basic filter configurations are summarized in Fig. 2 (a) -(f). Fig. 3 shows a typical EMI filter. It is composed of two X-caps (CX) connecting between the line and neutral wires, two Y-caps (CY ) connecting between both line and ground and neutral and ground wires and a CM choke (LC). A discharging resistor is used only for discharging the (CX) and it does not affect the noise suppression performance of the filter; thus, it can be ignored. The given EMI filter shown in Fig. 3 can be converted to the schematic shown in Fig. 4 (a). Fig. 4 (a) can be further converted to equivalent DM and CM circuits as shown in Fig. 4 (b) and (c), respectively. Due to the DM and CM noise current directions, the CM choke and the Y-caps (CY1 and CY2) affect both CM and DM emissions while the X-caps (CX1 and CX2) affect the DM noise [12]. When the CM noise propagates through the EMI filter, the CM noise is suppressed by the ©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

61

Common mode choke

Discharging resistor

NOISE SOURCE

LISN

CX

CY

Fig. 3: Structure of a typical EMI filter. Fig. 3: Structure of a typical EMI filter.

equivalent CM filter. Since CY is placed by facing the noise source, the equivalent CM filter is a CL-configuration filter as shown in Fig. 4 (c). At the same time, the DM noise is eliminated by the equivalent DM filter, which is a p-configuration filter as shown in Fig. 4 (b). The equivalent CM filter comprises of the CM inductance (LCM) of the CM choke at the LISN side and the parallel of CY1 and CY2 (CYT ) at the SMPS side as shown in Fig. 4 (c). The equivalent DM filter resulted from the total leakage inductance of two windings of the CM choke (LDM = Llk,p +Llk,s), the CX2 at the LISN and the CXT at the SMPS as shown in Fig. 4 (b). As the DM noise passing through the Y-caps in series, the CXT is a parallel combination of CX1 and the series of CY1 and CY2. However, it is possible also to increase the DM noise reduction by adding the DM chokes (LD) in series with the CM choke; in that case, the LDM is a summation of the total leakage inductance and the LD.

III. Systematic EMI Filter Design There are many factors which can deteriorate the designed EMI filter performance e.g. the termination impedances [2], [11], [12], the layout of the EMI filter components [13], and the connection between the EMI filter and a LISN [14] - [15]. In addition, the use of an EMI filter with a SMPS might perturb the overall system stability, if a DM input impedance of SMPS is larger than a DM impedance of a EMI filter connected to it [16]. Conventional filter design procedure used in communication systems and microwave applications is well developed by assuming that the terminated source and load impedances are 50 W matched [12], [17]. This assumption cannot directly apply to the design of SMPS EMI filters, where the actual noise source and terminated load impedances are far from 50 W. In practice, EMI filters are influenced by the noise source impedance and by the terminated load impedance [4] - [5], [18] and the filter design by assuming 50 W noise source and termination impedances leads to deviation of EMI filter performance from the actual operating conditions. In this section, the systematic EMI filter design procedure is explained and its validation is demonstrated by the following design examples. As an example, the EMI filter is designed for the SMPS (VTM22WB, 15W, +12VDC/0.75A, -12Vdc/0.5A). The SMPS is powered through the LISN (Electro-Metrics MIL 5-25/2) following the CISPR 16-1 standard requirement. The output of the SMPS is connected to a resistive load at its full load condition. To separate the total noise into DM and CM components, the output ports of the LISN are connected to the noise discrimination network [19]. The DM and CM disturbances are monitored by the spectrum analyzer (HP 8595 E, 9 kHz - 6.5 GHz). As shown in Fig. 5, the EMI filter design procedure can be summarized as follows. First, the DM and CM disturbances of the SMPS without EMI filter are measured; the required insertion losses (ILDM,req and ILCM,req) can be determined by subtracting the emission limit from the measured DM and CM noise levels without filter. Next, the DM and CM impedances of the LISN (ZLISN,DM and ZLISN,CM) and of the SMPS (ZSMPS,DM and ZSMPS,CM) under its actual operating conditions must be considered. As shown in second part of EMI filter design series, the magnitude and phase of LISN and SMPS impedances were extracted and compared; as a result, the suitable equivalent DM and CM parts of the EMI filter can be chosen. Once all needed parameters are known, the equivalent DM and CM filter components can be determined using the IL formulas (ILDM,estimate and ILCM,estimate). Before assembling the EMI filter prototype, the actual insertion loss (ILDM,actual and ILCM,actual) are recalculated using the measured impedances of the designed EMI components. Additionally, the power cable between the LISN and the EMI filter can also degrade the EMI filter performance at high frequency [14] - [15]. Moreover, the filter resonances and the interaction between the filter capacitor and the noise source impedance can not only degrade the EMI filter performance at high frequency but can also cause a stability problem in the SMPS. In such a case, the filters must be damped [12], [26].

62


5

16

ICM1 IDM A. Determination of the required insertion losses L

LC

L

Base on the knowledge of the DM and CM imped-

The conducted emissions of the SMPS without the filter are measured theSMPS LISN alone. previously The ances of thewith LISN and as extracted in the second part of EMI filter design series, a step-

CY1 measurement setup is shown in CFigs. 6. The Cline-to-ground and neutral-to-ground conducted emissions X2 X1 by-step proposed procedure to design the EMI filter LISN

I

SMPS

I

is given in Section IV. Finally, the effect of incorrect

DM CY2 are shown in Figs. 7 CM2 (a) and (b), respectively. Obviously, the SMPS can never meet required EMIis filter configuration to the the EMI filter performance

N

N

verified in Section V.

limit without an EMI filter. With the help of the discrimination network proposed in [19], the DM and EMI FILTER

G

G

CM conducted emissions of the SMPS without the filter are measured and shown#in Figs. 7 (c) and (d), IV. Realization 1: Proposed EMI Filter ICM = ICM1 + ICM2

Design Procedures For SMPS

respectively. Thus, the required DM (a)and CM filter insertion losses can be found by subtracting from the measured DM and CM Lemissions, given LDM in Figs. 7 (e) and (f). ′ ZLISN, DM VO,DM

CX2 CXT

ZSMPS, DM

B. Selection of the DM/CM filter topologies basedVS,on termination DM N

Next step is to choose the proper EMI filter topology. From the (b)

The intended conducted EMI limit to be met by the SMPS is the CISPR 22 Class B limit [27]. The LISN specified by this standard can only measure the total conducted emissions consisting of both the DM and CM components. Therefore, a discrimination network impedances is needed to separate the DM and CM components from the LISN so that they can be measured sepaDM and CM impedances comparison rately [19].

between SMPS and LISN shown in second part of EMI filter design series, it can be seen that the SMPS L,N

LCM

A. Determination of the required insertion losses

ZSMPS, CM CYTthat of the DM impedance magnitude is higher than LISN by a few tens ohm to a few hundred ohm over ZLISN, CM VO,′ CM

The conducted emissions of the SMPS without the

the frequency range of measurements. The comparison shows that the CM impedance VS, CM of CM impedances filter are measured with the LISN alone. The meaG

surement setup is shown in Figs. 6. The line-to-

of the SMPS is much larger than that of the LISN one over the frequency range of interest. With this ground and neutral-to-ground conducted emissions (c)

are shown in Figs. 7 (a) and (b), respectively. Obvi-

information, the proper EMI filter circuit can be chosen. Since theously, capacitor, to be effective, must be Fig. 4: Equivalent circuit of typical EMI filter: (a) schematic of Fig. 3; (b) equivalent the SMPS can never meet the required EMI Fig. 4: Equivalent circuit of typical EMI filter: (a) schematic of Fig. 3; (b) equivalent DM filter; (c)

DM filter; (c) equivalent CM filter.

limit without an EMI filter. With the help of the dis-

CM filter. placed inequivalent parallel to a high impedance and the inductor must be connected in series with a low impedance crimination network proposed in [19], the DM and

CM conducted emissions of the SMPS without the filter are measured and shown in Figs. 7 (c) and (d), respectively. Thus, the required

[3], CL-filter configurations of both equivalent DM and CM filters are chosen where the X-caps and DM and CM filter insertion losses can be found by subtracting from the measured DM and CM emissions, given in Figs. 7 (e) and (f). Y-caps are at the SMPS side and the CM choke is at the LISN side as illustrated in Fig. 8 (a). B. Selection of the DM/CM filter topologies based on termination impedances

The EMI filter is composed of one CM choke (LC ), one DM capacitor (CX ), and two CM capacitors Next step is to choose the proper EMI filter topology. From the DM and CM impedances comparison between SMPS and LISN shown in

(CY 1 and C 2 ). ofDue to the leakage inductance of LC ,SMPS the CM choke also doubles up as two DM inductors secondYpart EMI filter design series, it can be seen that the DM impedance magnitude is higher than that of the LISN by a few tens ohm to a few hundred ohm over the frequency range of measurements. The comparison of CM impedances shows that the CM

in the impedance line and ofneutral lines. The DM and CM interpretation of Fig. 8 (a) leads to two separate circuits, the SMPS is much larger than that of the LISN one over the frequency range of interest. With this information, the proper circuit can be chosen. Since the capacitor, to be effective, must be placed in parallel to a high impedance and the inductor shownEMI in filter Figs. 8 (b) and (c), respectively. Fig. 8 (b) presents the DM-suppressing part of the filter and is must be connected in series with a low impedance [3], CL-filter configurations of both equivalent DM and CM filters are chosen where the X-caps and Y-caps are at the SMPS side and the CM choke is at the LISN side as illustrated in Fig. 8 (a). composed by LDM , which is the DM inductance due to LC , and by CXT , which represents the effective The EMI filter is composed of one CM choke (LC), one DM capacitor (CX), and two CM capacitors (CY1 and CY2). Due to the leakage inducDM capacitor (C 1 and C 2 indoubles seriesupand then in parallel with CX ).neutral Fig. lines. 8 (c)The presents the CM-suppressing chokeYalso as two DM inductors in the line and DM and CM interpretation of Fig. 8 (a) tance of L , theY CM C

two separate circuits, shown in Figs. 8 (b) and (c), respectively. Fig. 8 (b) presents the DM-suppressing part of the filter and is part ofleads thetofilter and is composed by LCMdue , which is the CM inductance due to LC capacitor and by (CCY and T , which to LC, and by CXT , which represents the effective DM CY2 in series composed by LDM, which is the DM inductance Y1 and then in parallel with CX). Fig. 8 (c) presents the CM-suppressing part of the filter and is composed by LCM, which is the CM inducis the effective CM capacitor (CY 1 and CY 2 in parallel). tance due to LC and by CYT, which is the effective CM capacitor (CY1 and CY2 in parallel).

From Figs. 8 (b) and (c), the expressions of DM and CM filter insertion losses are rewritten as From Figs. 8 (b) and (c), the expressions of DM and CM filter insertion losses are rewritten as

2 LDM CXT ZSM P S,DM LDM + CXT ZSM P S,DM ZLISN,DM ILDM,estimate = 20 log s +s + 1 , ZLISN,DM + ZSM P S,DM ZLISN,DM + ZSM P S,DM ©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

(1)

63

17

START

Measure DM and CM noise sources without EMI filter using discrimination network

Noise source and load impedances measurement using direct clamping two- probe approach

Determine the required IL for DM/CM noises comparing to the EMI regulation

Select the DM/CM filter topologies based on noise source and load impedances

Calculate the DM/CM filter components using IL formulas of the selected topologies comparing to the required IL

Calculated IL above required IL at low frequency?

Adjust the DM/CM filter components of the selected DM/CM filter topologies

No

Yes Recalculate the required IL using the measured impedances of the selected components

Recalculated IL above required IL at all frequency?

No

Improve the self resonant frequencies of the selected components

- Check the length (l) of the power cord between EMI filter and LISN

Yes

( l < λ 10 )

Assemble EMI filter prototype Yes PASS Regulation?

No

Yes END Fig. 5: Flow diagram of the proposed filterdiagram design procedure. Fig. 5:EMI Flow of the

64

Fail at low frequency?

- Check the radiation coupling exists between SMPS and EMI filter Minimize inductive No and capacitive couplings existing in EMI filter - Damp filter resonant - Damp source impedance-filter capacitor resonance

proposed EMI filter design procedure.


L

220 V 50 Hz

L

LISN

SMPS

N

N

G

G

VLISN(L-G) VLISN(N-G)

R

VDM CM/DM Discrimination VCM Network Spectrum Analyzer

Schematic of DM and CM noise measurements without EMI filter being inserted. Fig. 6: Schematic of DM andFig. CM6:noise measurements without EMI filter being inserted.

ILCM,estimate

where

2 LCM CY T ZSM P S,CM LCM + CY T ZSM P S,CM ZLISN,CM = 20 log s +s + 1 , ZLISN,CM + ZSM P S,CM ZLISN,CM + ZSM P S,CM

6

(2)

where

s = j2πf, s = j2pf,

f = frequency normally spanningspanning in the conducted EMIconducted range [Hz], EMI range [Hz], f = frequency normally in the

[F], Y1//CY2 CY T CYT== CC Y 1 //CY 2 [F], CXT = CX//(CY1 + CY2) [F],

CXT = CX //(CY 1 + CY 2 ) [F], LCM = CM inductance of CM choke [H],

LCM = CM inductance of CM choke [H], LDM = Equivalent DM inductance of CM choke [H],

LDM = Equivalent DM inductance of CM choke [H], ZSMPS,DM = DM SMPS impedance [W],

ZSM P S,DM == CM DM SMPS impedance [Ω], SMPS impedance [W], ZSMPS,CM

LISN impedance [W], ZLISN,DM ZSM P S,CM == DM CM SMPS impedance [Ω],

LISNLISN impedance [W]. ZLISN,CM ZLISN,DM == CM DM impedance [Ω],

Based on Eqs. (1) - (2), the EMI filter component values are ready to be determined.

ZLISN,CM = CM LISN impedance [Ω].

CM Eqs. filter design BasedC. on (1) - (2), the EMI filter component values are ready to be determined. For the equivalent CM filter design, as shown in Fig. 8 (c), there are two unknown parameters that are the CM inductance of CM choke and the capacitance of Y-caps. Since the size of the EMI filter is strongly dependent on the size of CM choke, the Y-caps capacitance be design chosen as large as possible, to minimize the CM inductance value. The Y-capacitors are usually constrained by safety requireC. CM must filter ments, e.g. EN 60335-1 Class I portable, and therefore the maximum capacitance connected to ground cannot exceed about 4700 pF on chosen to be there 1000 pF are each.two Substituting the known CM LISN and SMPS eachequivalent phase for 250 VAC Hz mains [10]. Hence, CY1 and CY2 For the CM50filter design, as shown inare Fig. 8 (c), unknown parameters that impedances and the chosen CM capacitors into (2) and assuming that the filter elements are ideal, the required CM inductance is choand the ILof is minimized which Hence, twofilter 1000 pFis a way that of the CM distance between thethe ILCM,estimate CM,req are the sen CMin such inductance choke and capacitance Y-caps. Since theis about size 2ofmH.the EMI Class Y capacitors and a 2 mH CM choke (NEC/TOKIN SC-02-10A1) are chosen for the CM filter.

strongly dependent on the size of CM choke, the Y-caps capacitance must be chosen as large as possible, ©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

to minimize the CM inductance value. The Y-capacitors are usually constrained by safety requirements,

65

Substituting the known DM LISN and SMPS impedances and the measured DM inductance of the CM choke into (1), the required X-cap capacitance is chosen in such a way that the distance between the D. DM filter design

ILDM,estimate and the ILDM,req is minimized: this leads to a capacitance value of about 1.5 µF.

Once the CM choke and the CM capacitors are selected, the equivalent DM filter design is addressed. The equivalent DM filter, as shown in Fig. 8 (b), has two unknown parameters, i.e. DM inductance and X-caps capacitances. Because the CM choke was chosen, its DM inductance (leakage inductance) is about 17.6 μH. With this information, only the X-caps capacitance must be determined.

E. Actual insertion losses and validations

For X-type capacitors, since there is no safety issue, the value can be chosen as large as possible but larger capacitors usually exhibit a low self resonant frequency (SRF) and high inrush current [29]. Substituting the known DM LISN and SMPS impedances and the meaBefore building the EMI filter, its real performance must be assessed, and the parasitic effects of the filter sured DM inductance of the CM choke into (1), the required X-cap capacitance is chosen in such a way that the distance between the ILDM,estimate and the ILDM,req is minimized: this leads to a capacitance value of about 1.5 μF.

components need to be considered. The impedance behavior of the filter components, chosen according

to the above considerations, were measured by means of a HP 4396B impedance analyzer (100 kHz E. Actual insertion losses and validations

1.8 GHz), and the results are shown in Figs. 9 (a) - (f). The actual insertion losses of the DM and CM

Before building the EMI filter, its real performance must be assessed, and the parasitic effects of the filter components need to be conThe impedance behavior of the filter components, chosen according to the above considerations, were measured by means of a filterssidered. of Figs. 8 (b) and (c) can be worked out as an adaptation of (1) and (2) with the inclusion of the HP 4396B impedance analyzer (100 kHz - 1.8 GHz), and the results are shown in Figs. 9 (a) - (f). The actual insertion losses of the DM and CM filters of Figs. 8 (b) and (c) can be worked out as an adaptation of (1) and (2) with the inclusion of the parasitic effects of the filter parasitic effects of the filter components: components:

ZCXT (ZLDM + ZLISN,DM ) + ZSMP S,DM (ZCXT + ZLDM + ZLISN,DM ) , = 20 log ZC (ZLISN,DM + ZSMP S,DM )

(3)

ZC (ZLCM + ZLISN,CM ) + ZSMP S,CM (ZCY T + ZLCM + ZLISN,CM ) ILCM,actual = 20 log Y T , ZCY T (ZLISN,CM + ZSMP S,CM )

(4)

ILDM,actual

XT

where where

ZLDM = Effective DM impedance of the CM choke [W],

ZLCM = CM impedance of the CM choke [W],

ZCYT = Effective CM impedance of CY1//CY2 [W],

ZCXT = Effective DM impedance of CX//(CY1 + CY2) [W].

ZLDM = Effective DM impedance of the CM choke [Ω], ZLCM = CM impedance of the CM choke [Ω], ZCY T = Effective CM impedance of CY 1 //CY 2 [Ω],

ZC

= Effective DM impedance of CX //(CY 1 + CY 2 ) [Ω].

XT Using the measured impedances of all the chosen filter components (shown in Fig. 9 (a) - (f)), as well as the LISN and SMPS impedance characteristics, the actual insertion losses of the DM and the CM filters (computed by means of (3) and (4)) are plotted in Figs. 11 (a) and (b), respectively. For comparison, the required DMthe and chosen CM filter insertion losses are also shown in the the actual DM Using the measured impedances of all filter components (shown insame Fig.figures. 9 (a) Both - (f)), as well and CM filter insertion losses are higher than the required ones, which indicate that the designed EMI filter is able to suppress the DM CM conducted emissions below the CISPR 22 Class B limit. as theandLISN and SMPS impedance characteristics, the actual insertion losses of the DM and the CM filters

Finally, to verify the noise reduction performance of designed EMI filter, conducted EMI (DM, CM, L-G and N-G noise voltages) are measured when the designed EMI filter is added in. The schematic of the measurement setup are shown in Fig. 10. Figs. 11 (c) and (d) show the measured DM and CM conducted emissions after the designed EMI filter is inserted. Both the DM and CM conducted emissions are below the limit, confirming that the designed EMI filter fulfills the requirement. As a final compliance check, Figs. 11 (e) and (f) show the line-to-ground and the neutral-to-ground conducted emissions, respectively, and their results indicate that the filtered SMPS complies with the regulations limits.

V. Realization # 2: Effect of Incorrect Filter Configuration To illustrate the effect of an incorrect filter configuration that leads to non-optimal performance of the EMI filter, the same EMI filter already designed in realization #1 is mounted in a reverse configuration; this means that the CM choke position is now shifted to the SMPS end. The new EMI filter configuration is shown in Fig. 12 (a). The equivalent DM and CM filters are LC-configuration filters as shown in Figs. 12 (b) and (c), respectively. The actual DM and CM filter insertion losses of the LC-configuration filters can be rewritten as follows,

66


19

200

Total noise: line−to−ground V

L−G

180

Magnitude [dBuV]

120 100 80

120 100 80 60

40

40 6

7

N−G

Limit line

140

60

20 5 10

8

10 10 Frequency [Hz]

V

160

140

20 5 10

Total noise: neutral−to−ground

180

Limit line

160 Magnitude [dBuV]

200

10

6

(a) 200

Differential Mode Emission DM

Limit line

120 100 80

120 100 80 60

40

40 7

CM

Limit line

140

60

6

V

160

140

20 5 10

8


Common Mode Emission


Magnitude [dBuV]

200

V

160

10

6

200

ILDM,req

Common Mode Insertion Loss Requirement ILCM,req

150 Magnitude [dB]

Magnitude [dB]

8

10

(d)

Differential Mode Insertion Loss Requirement

150 100 50 0 −50 5 10

7


(c) 200

8

10

(b)

180

20 5 10

7


100 50 0

6

7


8

10

−50 5 10

6

7


(e)

8

10

(f)

Fig. Measured conducted emission without filter. (a) Line-to-ground; (c) DM; Fig. 7: 7: Measured conducted emission without filter. (a) Line-to-ground; (b) neutral-to-ground; (c) DM; (d)(b) CM;neutral-to-ground; (e) required DM filter insertion loss; (f) required CM filter insertion loss.

(d) CM; (e) required DM filter insertion loss; (f) required CM filter insertion loss. ©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

67

20

L

L

LC

8

CY1and (b), respectively. For comparison, the (computed by means of (3) and (4)) are plotted in Figs. 11 CX (a) LISN SMPS required DM and CM filter insertion losses are also shown inCY2 the same figures. Both theRactual DM and N

N

CM filter insertion losses are higher than the required ones, which indicate that the designed EMI filter EMI FILTER

G

G

is able to suppress the DM and CM conducted emissions below the CISPR 22 Class B limit. Finally, to verify the noise reduction performance of designed EMI filter, conducted EMI (DM, CM, L-G and N-G noise voltages) are

VDM CM/DM measured the designed VLISN(N-G) when Discrimination VCMEMI Network VLISN(L-G)

filter is added in. The schematic of the

measurement setup are shown in Fig. 10. Figs. 11 (c) and (d) show the measured DM and CM conducted Spectrum Analyzer

(a) DM and CM conducted emissions are below emissions after the designed EMI filter is inserted. Both the

the limit, confirming that the designed EMI filter fulfills LDM the requirement. As a final compliance check, L ZSMPS, DMemissions, respectively, CXT Figs. 11 (e) and (f) show the line-to-ground and the neutral-to-ground conducted ZLISN, DM VDM and their results indicate that the filtered SMPS complies with the regulations limits. VSMPS, DM N V. R EALIZATION # 2: E FFECT OF I NCORRECT F ILTER C ONFIGURATION (b)

To illustrate the effect of an incorrect filter configuration that leads to non-optimal performance of the LCM L//N EMI filter, the same EMI filter already designed in realization #1 is mounted ZSMPS,inCMa reverse configuration; CYT ZLISN, CM this means that the CM choke position is now shifted to the SMPS end. The new EMI filter configuration VCM VSMPS, CM is shown in Fig. 12 (a). The equivalent DM and CM filters are LC-configuration filters as shown in Figs. G 12 (b) and (c), respectively. (c)

The actual DM and CM filter insertion losses of the LC-configuration filters can be rewritten as follows, 8: Conducted measurement infilter. the presence of the EMI filter. discrimination (a) Test setupnetwork; with CM/DM Fig. 8: Conducted EMI Fig. measurement in theEMI presence of the EMI (a) Test setup with CM/DM (b) DM part of the filter; (c) CM part of the filter.discrimination network; (b) DM part of the filter; (c) CM part of the filter.

ILDM,actual =

ILCM,actual =

(Z CXT 2 20 log (Z CY T 20 log

+ ZLISN,DM )(ZSM P S,DM + ZLDM ) + ZCXT 2 ZLISN,DM , ZCXT 2 (ZLISN,DM + ZSM P S,DM )

+ ZLISN,CM )(ZSM P S,CM + ZLCM ) + ZCY T ZLISN,CM . ZC (ZLISN,CM + ZSM P S,CM )

(5)

YT

(6)

Based ononequations - the (6),actual the actual DMfilter and CM filter plotted in Figs. By 13comparing (a) and Based equations (5)(5) - (6), DM and CM insertion losses insertion are plotted inlosses Figs. 13are (a) and (b), respectively.

with the required DM and CM filter insertion losses, they show that the actual DM filter insertion loss deteriorates marginally at low fre-

quency, but theBy actual CM filter insertion unable to meet theand required filter insertion losslosses, below 1.8they MHz. The DM that and CM (b), respectively. comparing withloss theis required DM CMCMfilter insertion show theconducted emissions, shown in Figs. 13 (c) and (d) respectively, confirm that the DM conducted emission exceeds the limit marginally at low

and insertion the CM conducted emission exceeds the limit badly MHz. actual frequency DM filter loss deteriorates marginally atbelow low 1.8 frequency, but the actual CM filter insertion

the line-to-ground neutral-to-ground conducted emissions measured shownThe in Figs. 13 (e) and CM (f), respectively. As the loss isFinally, unable to meet theand required CM filter insertion lossarebelow 1.8and MHz. DM and conducted ©2012 Electromagnetic Compatibility Magazine 1 – Quarter 2 68 emissions, shown in Figs. 13 (c) andIEEE(d) respectively, confirm that– Volume the DM conducted emission exceeds

21

CM choke impedance: NEC/TOKIN

16000

Z

Z

L

14000

Z

L

CM

L

12000

Z

10000 8000 6000 4000

CM

L

50

DM

Phase [degree]

Magnitude [Ohm]

CM choke impedance: NEC/TOKIN

100

DM

0

−50

2000 0 5 10

6

7

−100 5 10

8


10

6

(a) Z

C

8

10

(b)

capacitor impedance

ZC capacitor impedance

X

10

7


X

100

ZC

ZC

X

X

50 Phase [degree]

Magnitude [Ohm]

8 6 4

0

−50

2 0 5 10

6

7

−100 5 10

8


10

6

(c) Z

C

600

capacitor impedance (C

Y1

=C )

Z

Y2

ZC

Y1

C

−76

= ZC

Y2

capacitor impedance (C

Y1

Y

400 300 200 100

=C ) Y2

ZC

Y1

−78 Phase [degree]

500 Magnitude [Ohm]

8

10

(d)

Y

0 5 10

7


= ZC

Y2

−80 −82 −84 −86

6

7


8

10

−88 5 10

6

7


(e)

8

10

(f)

Measured impedance of the chosen (a) and (b) refer to the CM choke (NEC/TOKIN); (d) refer the C X = choke Fig. Fig. 9: 9:Measured impedance of filter thecomponents. chosen Plots filter components. Plots (a) and (b)(c) and refer to tothe CM 1.5 μF capacitance; (e) and (f) refer to the CY1 = CY2 = 1000 pF capacitances.

(NEC/TOKIN); (c) and (d) refer to the CX = 1.5 µF capacitance; (e) and (f) refer to the CY 1 = CY 2 = 1000 pF capacitances.


69

L

220 V 50 Hz

L

EMI FILTER

LISN

SMPS

N

N

G

G


R


Fig.noise 10: Schematic of DMwith and EMI CM noise with EMI filter being inserted. Fig. 10: Schematic of DM and CM measurements filtermeasurements being inserted.

For Fig. 11 – see page 71

CM conducted emission dominates below 1.8 MHz, the SMPS does not comply with the limit at frequencies below 1.8 MHz.

24

L

LC

VI. Conclusion

CY1

In this final article of EMI filter design series, we focus on systematic EMI filter design procedure. With known amplitude and phase information of LISN and SMPS impedances, EMI filters can be designed systematically. Using a SMPS as a practical design example, it has been demonstrated that • The optimal filter configuration and the correct filter component values can be chosen.

LISN

L

CX SMPS

CY2

N

R

N EMI FILTER

G


G


• The incorrect filter configuration leads to non-optimal filter attenuation, thus causing the same SMPS to fail the required EMI limit badly.

(a)

LDM

L CXT

ZSMPS, DM

ZLISN, DM VDM Nevertheless, in the proposed design procedure, it is not only specific for a single-stage passive EMI filter but factors like component volVSMPS, DM ume and component price are not taken into account. It might be difN ficult to choose between a CL-configuration of equivalent DM filter (b) with one huge CX (CX = 1.5μF) and a pi-configuration with two smaller CX (CX1 = CX2 = 0.47μF) since both can comply with regulations. OptiLCM L//N ZSMPS, CM mization might be applied to take into account physical and economiCYT ZLISN, CM VCM cal parameters of the filter. The possible further applications of the EMI filter design procedure are for diverse types of systems, like VSMPS, CM three-phase AC industrial motor drives, large uninterruptible power G supplies (UPS), wind turbine power stations, etc. and DC supplies, (c) e.g. photovoltaic inverters-whose input impedances need to be Fig. (a) 12: (a) Realization #4: reversed EMI filter configuration. (b) DM filter; (c)(b) CM DM filter. filter; Fig. 12: Realization #4: reversed EMI filter configuration. determined.

(c) CM filter.

VII. Acknowledgment The author would like to acknowledge the useful suggestions and comments from Assoc. Prof. See Kye Yak and Prof. Flavio G. Canavero on the content of this series of articles.

70


23

200

Differential Mode Insertion Loss Comparison

200

IL

Common Mode Insertion Loss Comparison IL

DM,req

100 50 0 −50 5 10

ILCM,actual

150 Magnitude [dB]

150 Magnitude [dB]

CM,req

ILDM,actual

100 50 0

6

7

−50 5 10

8


10

6

(a) 70

Differential Mode Emission VDM

Limit line

Common Mode Emission VCM

Limit line

60 Magnitude [dBuV]

Magnitude [dBuV]

70

50 40 30

50 40 30

6

7

20 5 10

8


10

6

Total noise: line−to−ground

70

VL−G

Limit line

Total noise: neutral−to−ground VN−G

Limit line

60 Magnitude [dBuV]

Magnitude [dBuV]

8

10

(d)

60 50 40 30 20 5 10

7


(c) 70

8

10

(b)

60

20 5 10

7


50 40 30

6

7


(e)

8

10

20 5 10

6

7


8

10

(f)

Fig. 11: Effect of the insertion of the EMI filter. (a) Required and actual DM insertion losses; (b) required and actual CM insertion losses; (c) DM conducted with filter; (d) measured CM conducted emission with (e) measured total conducted emissions line-toFig.measured 11: Effect of theemission insertion of the EMI filter. (a) Required andfilter; actual DM insertion losses; (b)forrequired ground; (f) measured total conducted emission for neutral-to-ground.

and actual CM insertion losses; (c) measured DM conducted emission with filter; (d) measured CM ©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

71

25

200

Differential Mode Insertion Loss Comparison

200

IL

Common Mode Insertion Loss Comparison IL

DM,req

100 50 0 −50 5 10

ILCM,actual

150 Magnitude [dB]

150 Magnitude [dB]

CM,req

ILDM,actual

100 50 0

6

7

−50 5 10

8


10

6

(a) 120

7

Differential Mode Emission

120

V

Common Mode Emission V

Limit line


Magnitude [dBuV]

CM

Limit line

100 80 60 40

80 60 40

6

7

20 5 10

8


10

6

7

Total noise: line−to−ground

120

V

Total noise: neutral−to−ground V

N−G

Limit line

Limit line



10

(d)

L−G

80 60 40 20 5 10

8


(c) 120

10

(b)

DM

20 5 10

8


80 60 40

6

7


8

10

20 5 10

6

7


(e)

8

10

(f)

Fig. 13: 13: Non-optimized filter. (a) Required DM insertion losses; (b) required and designed CM insertion losses; (c) measured DM conFig. Non-optimized filter. and (a)designed Required and designed DM insertion losses; (b) required and designed ducted emission; (d) measured CM conducted emission; (e) measured total conducted emission for line-to-ground; (f) measured total conducted CM insertion losses; (c) measured DM conducted emission; (d) measured CM conducted emission; (e) emission for neutralto-ground.

measured total conducted emission for line-to-ground; (f) measured total conducted emission for neutral-

72

to-ground. ©2012 IEEE Electromagnetic Compatibility Magazine – Volume 1 – Quarter 2

References

[22] H. J. Chiu, T. F. Pan, C. J. Yao and Y. K. Lo, “Automatic EMI Measurement and Filter Design System for Telecom Power Supplies,” IEEE Trans. Instrum. Meas., vol. 56, no. 6, pp. 2254–2261, Dec 2007. [23] S.Wang, R. Chen, J.D. vanWyk, F. C. Lee, and W.G.Odendaal, “Developing parasitic cancellation technologies to improve EMI filter performance for switching mode power supplies,” IEEE Trans. Electromagn. Compat., vol. 47, no. 4, pp. 921–929, Nov 2005. [24] T. C. Neugebauer and D. J. Perreault, “Parasitic capacitance cancellation in filter inductors,” IEEE Trans. Power Electron., vol. 21, no. 1, pp. 282–288, Jan 2006. [25] T. C. Neugebauer and D. J. Perreault, “Filters and components with inductance cancellation,” IEEE Trans. Ind. Appl., vol. 40, no. 2, pp. 483–490, Mar./ Apr. 2004. [26] D. M. Mitchell, DC-DC Switching Regulator Analysis, McGraw-Hil, 1988, ch. 4. [27] Limits and Methods of Measurement of Radio Interference Characteristics of Information Technology Equipment, CISPR 22, 2004. [28] R. L. Ozenbaugh, EMI Filter Design, Marcel Dekker, second edition, 2001. [29] S. Wang, F. C. Lee, and J.D. Van Wyk, “A study of integration of parasitic cancellation techniques for EMI filter design with discrete components,” IEEE Trans. Power Electron., vol. 23, no. 6, pp. 3094–3102, Nov 2008. [30] D. Zhang, D. Y. Chen, M. J. Nave12 and D. Sable, “Measurement of noise source impedance of off-line converters,” IEEE Trans. Power Electron., vol. 15, no. 5, pp. 820–824, Sep 2000.

[1] F .-Y. Shih and et al., “A procedure for designing EMI filters for AC line applications,” IEEE Trans. Power Electron., vol. 11, no. 1, pp. 170–181, Jan 1996. [2] S. Ye, W. Eberle and Y.F. Liu, “A novel EMI filter design method for switching power supplies,” IEEE Trans. Power Electron., vol. 19, no. 6, pp. 1668–1678, Nov 2004. [3] J. J. Goedbloed, Electromagnetic Compatibility, Prentice Hall, 1990. [4] B. Audone and L. Bolla, “Insertion loss of mismatched EMI suppressors,” IEEE Trans. Electromagn. Compat., vol. 20, no. 3, pp. 384–389, Sep 1978. [5] S. M. Vakil, “A technique for determination of filter insertion loss as a function of arbitrary generator and load impedances,” IEEE Trans. Electromagn. Compat., vol. 20, no. 2, pp. 273–278, Sep 1978. [6] V. Tarateeraseth, Hu Bo, K. Y. See and F. Canavero, “Accurate extraction of noise source impedance of SMPS under operating condition,” IEEE Trans. Power Electron., vol. 25, no. 1, pp. 111–117, Jan 2010. [7] V. Tarateeraseth, K. Y. See, F. Canavero and R.W.Y. Chang, “Systematic electromagnetic interference filter design based on information from in-circuit impedance measurement,” IEEE Trans. Electromagn. Compat., vol. 52, no. 3, pp. 588–598, Aug 2010. [8] V. Tarateeraseth, K. Y. See, L.B. Wang and F.G. Canavero, “Systematic Power Line EMI Filter Design for SMPS,” Proc. of 10th International Symposium on EMC (EMC Europe 2011), York (UK), pp. 586-591, September 26 30, 2011. [9] Clayton R. Paul, Introduction to Electromagnetic Compatibility, John Wiley & BIOGRAPHIES Sons, second edition, 2006. [10] Tim Williams, EMC for product designers, Newnes, forth edition, 2007. M. J. Nave, Power Line Filter for Switched Modehonors) Power Supplies, New Biography eeraseth[11]received the B. Eng.Design (second-class York: Van Nostrand Reinhold, 1991. [12] L. Tihanyi, Electromagnetic Compatibility in Power Electronics, IEEE Press, Vuttipon Tarateeraseth received the e both in electrical engineering from King’s Mongkut 1997. [13] S. Wang, F. C. Lee, D. Y. Chen and W. G. Odendaal, “Effects of parasitic B. Eng. (second-class honors) and parameters on(KMITL), EMI filter performance,” IEEE Trans. Power Electron., vol. 19, ology Ladkrabang Thailand, and Ph.D. M. Eng. degree both in electrical engino. 3, pp. 869–877, May 2004. neering from King’s Mongkut Institute of [14] S. A. Pignari and A. Orlandi, “Long-cable effects on conducted emissions communications engineering from vol. Politecnico di Feb 2003. Technology Ladkrabang (KMITL), Thailevels,” IEEE Trans. Electromagn. Compat., 45, no. 1, pp. 43–54, [15] S. Wang, J. D. Van Wyk and F. C. Lee, “Effects of interactions between filter land, and Ph.D. in electronics and comand power interconnects on EMI filter e 2011, he parasitic is a lecturer at Department of performance,” ElectricalIEEE Trans. munications engineering from Politecnico Power Electron., vol. 54, no. 6, pp. 3344–3352, Dec 2007. di Torino, Italy. Since 2011, he is a lectur[16] T. Nussbaumer, M. L. Heldwein and J. W. Kolar, “Differential mode input filter kharinwirotdesign University, Thailand. From 2010-2011, er at Department of Electrical Engineerfor a three-phase buck-type PWM rectifier based on modeling of the EMC test receiver,” IEEE Trans.Ind. Electron., vol. 53, no. 5, pp. 1649–1661, Oct ing, Srinakharinwirot University, Thailand. 2006. t College of Data Storage Innovation, KMITL. Prior From 2010-2011, he was a lecturer at Col[17] J. C. Shifman, “A Graphical Method for the Analysis and Synthesis of Electrolege of Data Storage Innovation, KMITL. Prior to 2007, he worked Interference Filters,” IEEE Trans. Electromagn. Compat., vol. 7, no. 3, as a head magnetic of297–318, environment pp. Sep 1965. testing laboratory at Delta as a head of environment testing laboratory at Delta Electronics [18] B. Garry and R. Nelson, “Effect of impedance and frequency variation on (Thailand) for three years. He also worked as an EMC engineer insertion loss for aHe typical powerworked line filter,” inas 1998an Proc.EMC IEEE EMC Sympoand) for three years. also for 2 years under the Joint Development of Teaching Materials to sium, pp. 691–695. Improve EMC Skills of Academic Staff and Postgraduate Elec[19] K. Y. See, “Network for conducted EMI diagnosis,” IEE Electronics Letters, vol. ars under the Joint Development of Teaching Materials to Improve EMC Skills of 35, no. 17, pp. 1446–1447, Aug 1999. tronic Designers Project funded by European Commission. His [20] M. Kumar and V. Agarwal, “Power line filter design for conducted electromagresearch interests are mainly in the fields of EMI reduction technetic interference using time-domain measurements,” Trans. Electrond Postgraduate Electronic Designers Project IEEE funded by European Commission. His niques, EMC/EMI modeling, EMC instruments and measurements magn. Compat., vol. 48, no. 1, pp. 178–186, Feb 2006. [21] Yuang-Shung Lee, Yu-Lin Liang, and Ming-Wang Cheng, “Time Domain Mea- and EMI filter design. He is author or coauthor of more than 40 are mainly in the fields of EMI reduction techniques, EMC/EMI modeling, EMC surement System for Conducted EMI and CM/DM Noise Signal Separation,” technical papers published in international journals and conferInternational Conference on Power Electronics and Drives System, vol. 2, proceedings one book chapter. easurements2005, and filter design. He is author or coauthor ofence more than 40 and technical pp. EMI 1640–1645.

n international journals and conference proceedings and one book chapter.


73