Interaction between Passive Common Mode Noise

Cancellation and Conservative Passive Filtering Martin Schmidt, Jiirgen Stahl, Manfred Albach

Chair of Electromagnetic Fields, Friedrich-Alexander-University Erlangen-Nuremberg 91058 Erlangen, Germany

{martin.schmidtljuergen.stah1Imanfred.albach}@emf.eei.uni-erlangen.de

Abstract- Passive cancellation of common mode (cm) noise is a

very promising alternative to conservative cm filtering. The basic

idea has already been described in literature. A thorough

analysis of parasitic effects as well as the effects of a finite

differential mode (dm) filter shows that a significant reduction of

cm noise is achievable. However, in a lot of applications the

electromagnetic compatibility (EMC) standard has not yet been

fulfilled and an additional compact conservative cm filter is

necessary. These two filters interact. Hence, a detailed

investigation of the resulting circuit with both filters is essential.

An analysis based on the frequency-response allows quantifying

the influences of different design parameters and optimizing the

circuit. Finally, the theoretical results are verified in a test set-up.

I. INTRODUCTION

Every switch mode power supply (SMPS) has to fulfil EMC standards. Generally passive filters are used in order to reduce the noise level below the allowed limits. The influence

of a passive filter on size and cost is quite relevant and not

negligible at all. In this paper passive cancellation, a very

promising method to reduce the cm noise is analysed in a

more realistic setting than in previous papers [1,2,3,4]. There the method was analysed as a stand-alone filter. In [4] at least the effects of a finite dm filter are considered. The

parasitic effects of real components prevent an infinite

attenuation due to perfect compensation. Hence, in a lot of

applications, the EMC standard has not yet been fulfilled and

an additional conservative cm filter

theoretical results are compared to measurements. The whole

principle is exemplarily shown using a conventional dc-dc

boost converter with the design data of Table I.

II. BASIC IDEA

Cm noise is caused by displacement current. Fig. 1 depicts

this issue in a boost converter. The switching time of the

MOSFET is typically small and as a consequence the voltage VI is nearly rectangular. Hence, the potential of the orange marked area as well as the voltage vCp are changing rapidly.

The noise caused by the current icp = Cp dvc/dt is measured

with a line impedance stabilization network (USN) at both line impedances Z. If the dm filter is ideal, the voltage across the boost inductor Vr.ll is equal to the ac value of VI' By adding a second winding to the boost inductor a new area (marked

blue in Fig. I) with another jumping potential is created. The direction of the second winding should be chosen in such a

way that the new current iccomp compensates ic'p at the line

impedance Z. This method is already described in [1,2,3,4].

[1,3,4] show some measurements and [3,4] deal with parasitic effects. An additional conservative cm filter is not considered in any of these papers.

III. WITHOUT ANY FILTER

The circuit can be analysed best in the frequency domain.

The according model is depicted in Fig. 2. A validation of the

is necessary. However, the

conservative cm filter and the

cancellation filter interact with each

other. Hence, a thorough

...................................... 'LiSN .... j

investigation is necessary. At first the basic principle of the compensation method is shown

briefly in chapter two. The initial

cm level is presented in chapter three. Afterwards, a conservative

cm filter is analysed. In chapter five the compensation method is investigated by its frequency­

response and limiting effects are

discussed. Subsequently, a thorough

investigation of a circuit with both

filters is shown in chapter six and an optimization strategy is exposed in chapter seven. Finally, the

978-1-4577-1559-4/12/$26.00 ©2012 IEEE

low

pass

dm and cm

filter

, I

'., /Cwm, vc" t ; c,' ic:comp

Fig. 1 Boost converter with passive cancellation filter connected to the LlSN

121

model is shown in [4]. Tn this chapter the conservative cm filter is removed. This means that point A is connected to A' and point B is connected to B'. The design data are shown in Table 1. The cm noise

VZl +vZ2 Vem = 2

(1)

for the chosen design and the EMC standard [5] are visible in Fig. 3. The cm noise level is approximately constant in the relevant frequency range from 150kHz to 30M Hz. Hence, an attenuation of around 30dB is necessary in order to fulfil the standard.

IV. CONSERVATIVE COMMON MODE FILTER

Next, a conservative cm filter is analysed. Fig. 2 depicts the frequency domain model. Design data of the filter capacitors are shown in Table I. In Fig. 4 the attenuation

_ -201 (Vcm_with_conse.rvatiVejilter _Fig2 : Gconservative - og 1 0 Vem wtthout filter

(2)

is shown for different values of the cm inductor Lem. The values of Lem are all quite small and the achieved attenuation is not sufficient to fulfil the EMC standard. The filter resonance frequencies are approximately 80kHz, 230kHz and 2MHz. The amplitudes of the resonances depend on the ratio

,.-.. > :::!. oo

:2-

r --------.--, conservatIve , : cm filter ,

......--A, • A' 'Z'

iO z •

90

85

80

75

Cy2T I -=- I ___________ J

Fig. 2 Frequency domain model for a boost converter

*, ** *

t-

�� 70

65

60

55 105

(Hz) Fig_ 3 Results of the frequency domain model

Parameter Value Vn ISV V",a (= VI) 30V

j; SOkHz Mode CCM L" 136J.lH

TABLE I DESIGN DATA

Description Input voltage Output voltage Switching frequency Continous conduction mode Inductance of primary winding

Cp 22SpF Parasitic capacitor (artificially increased) Cdlll IJ.lF Om-capacitor Z son Line impedance (measurement: Z(j) Cd 2_022nF Y -capacitor in the conservative cm filter Cv2 2_022nF Y -capacitor in the conservative cm filter L22 8_SJ.lH Inductance of secondary winding k 0.99 Coupling factor

of Lem and the Y -capacitors. At higher frequencies the increase of the attenuation is 40dB/decade.

V. PASSIVE CANCELLATION FILTER

A thorough stand-alone analysis of the passive cancellation method is shown in [4]. Fig. 5 shows the circuit and the design data are given in Table I. In this chapter the conservative cm filter is removed. Hence, point A is connected to A' and point B is connected to B'. The design constraints

Cp JF,11 Ccomp =-- - - and Cgnd = Cp +Ccomp k L22

for Ccomp and Cgnd are derived in [4]. The attenuation

_ -20 I (vcm with eance.llatiOn filter _FigS] acaneellation - ogl o V cm_ wlthoutjilter

(3)

(4)

is plotted in Fig. 6 with a solid line. The decrease with 40dB/decade and the resonance at 16MHz is due to the parasitic leakage inductance [4]. Cm noise is even increased at the resonance frequency. An additional resistor Rcomp in series with Ceomp yields an attenuation of the resonance peak. However, the yielded compensation at lower frequencies is reduced. The corresponding curves are also plotted in Fig. 6.

100

80 --- Lern = 2 x 1.5�IH

iii' 60 :2---- Lern = 2 x 1l0J.lH

� "' 40 .

� 20 2 "

0

-20 +---------- to conservative filter

105 «Hz)

Fig. 4 Attenuation of cm noise due to a conservative cm filter

122

VI. PASSIVE CANCELLATION FILTER AND CONSERVATIVE

FILTER

In the next step, both filters are taken into account. The circuit is shown in Fig.

5. The attenuations acancellalion and

points. However, the resonance peak at high frequency points is more crucial when both filters are combined. Fig. 7 depicts the total attenuation.

- -201 ( Vcm with hothJiiters_FigS ] (5) acancellation add - oglo - Vem with _ conservative filter _ Fig2

Hence, a resistor K'omp in the compensation path is more important than before. Consequently, the total attenuation alota!

is reduced at low frequency points. The curves are shown in Fig. 8.

will be equal at frequencies below the resonance frequency due to the cancellation method if the capacitors Cy1l2 are replaced by

Ccomp + Cgnd C yl!2_reduced = C y1l2 - 2 = InF . (6)

This is an important step and it is of worthy note that the sum of all capacitors to protective earth is the same in Fig.

2 and

Fig. 5. With this modification the total attenuation

- -201 (Vcm_wilh_h(.JthJiite.rs_Figs] alotal - og I 0 V cm wahout .{tlter (7)

IS the addition of acancellalion and acanServalive at low frequency

r--------:--I

conservatIve I I cm filter :

I I

-=- I

Cy1 redlle:I.i Lem :

I I I I

: Cy2 redlle�'d : I I I I I -=- I -----------

cancellation filter •

E L22 Rcomp

M-k � Cco:r - "1/ L11 L22 "'l -- ---------------�--

I

Ygnd i I

__ : ___ --�

Fig. 5 Boost converter with passive cancellation and conservative cm filter

80

ii3' 60 :2, " � 40 " � "

20 S '" 0

-20 10

5 10

6

((Hz)

--- without R comp --------- R = 3Q comp

Fig. 6 Attenuation of cm noise due to the cancellation filter. Different values of Reomp

VII. OPTIMIZATION

In a lot of applications the cm choke Lem is the most crucial component. Hence, a small Lem yields best results in cost and size. A stand-alone optimization for the cancellation filter is shown in [4].

An independent optimization of Lem and Reomp is not possible. Therefore, a strategy to find the best K'omp for a given Lem is proposed in the next paragraph.

The two critical frequency points in the total filter attenuation atota! are the resonance frequencies due to the conservative filter and due to the cancellation filter. The remaining alola! at the resonance due to the cancellation filter is directly influenced by Reomp- A higher Reomp yields a higher remaining alalal at this frequency. In contrast, the amplitude of the resonance due to the conservative filter is not influenced directly. Nevertheless, a higher Reomp results in a lower acancellalion at low frequencies. Hence, a higher Rcamp yields a lower remaining alalal at lower frequencies. These effects are

123

ii3' :2, 'Ii Q '"

100

80

60

40

20 --- Lem = 2 x 1.5!lH

--- Lem = 2 x 11O!lH 0 --- L = 2 x 1mH em filter

-20 10

5 10

6 10

7

((Hz)

Fig. 7 Attenuation of cm noise due to both filters. Without Reomp

100 r---������--�--�����--�-'

,"' '' ... -......... ... ................ -_ .. -........ , �

20 '.,' '\

- - -; - -

•• _.-._ •.•• �" .�:/ o reSCinallQeS' due· to· the·. .resonances. du.e. to. ........ .

conservative filter the canCellation filtet

-20 �--�������--������----�� 10

5

((Hz)

Fig. 8 Attenuation of cm noise due to both filters. Different values of Rco",p and Iem. The legend for the colour and the line style is the same as in Fig. 6 and Fig. 7.

shown in Fig. 8. Therefore, for a given Lem an optimization of Reomp is necessary.

The lowest possible value of Lem is found, when the optimized K'ol1lp yields a cm noise level that just fulfils the standard at both mentioned resonance frequencies.

VIII. MEASUREMENTS

Finally, a boost converter was built in order to compare the theoretical results to measurements. The design data are shown in Table I. A photo of the boost converter is depicted in Fig. 9. The parasitic capacitance Cp is increased artificially to 22SpF with an additional capacitor to protective earth in order to get a noise level that is comparable to the noise level of a

boost converter used in a power factor corrector circuit. Using the measurement equipment that is characterized in

[6] a separate measurement of cm and dm noise is possible. The cm noise level without any filter is almost constant at around 87dBflV. The measured cm noise level with a conservative filter is plotted in Fig. 10. The resonance frequencies and the attenuation at high frequencies fit to the theoretical prediction. In order to fulfil the EMC standard a Lem of around 2 x 2mH would be necessary.

Adding the cancellation filter without Reomp yields both a high attenuation at low frequency points and a significant resonance at around 16MHz. This is already predicted in chapter VI. The corresponding curves are shown in Fig. 1 I. The blue curve easily fulfils the EMC standard and the used value of Lem is far too high.

The results with an additional resistor Rcomp are presented in Fig. 12. The achieved attenuations confirm the prediction in

Fig. 8. The best value of KOl1lp is 3.30 (red curve in Fig. 12). At the low resonance frequency the noise level is just below the standard. However, at the high resonance frequency the standard is not fulfilled. Hence, the inductance of Lem has to be increased by a few per cents. Nevertheless, the cancellation filter allows reducing the value of Lem by a factor of thousand.

IX. CONCLUSION

The reduction of cm noise via passive compensation works very well at lower frequencies. In contrast, a conservative passive filter is more effective in the higher frequency range. Hence, a combined filter yields very good results. The interaction at low frequency points is negligible. Therefore, the total attenuation at these points is the addition of the attenuations of the two filters. A resonance at high frequency

points occurs and is caused by the cancellation filter. Furthermore, it is

increased by the conservative filter. An optimized Reol1lp prevents this disadvantage and a significant reduction of the value of the cm choke Lem is possible. All theoretical results are

Fig. 9 Photo of the boost converter validated experimentally.

124

� �

� '"

� � ""

[1]

[2]

[3]

[4]

[5]

[6]

�

100

--- Lem = 2 x 1.5flH

80 --- Lem =2 x l10flH

--- DIN EN 55022 [5]

60

40

20 105 10

6

(Hz)

Fig. 10 Cm noise with a conservative cm filter

---LC/II=2x 1.5�IH

80 --- Le/ll = 2 x II0flH --- DIN EN 55022 [5]

60

40

20 t!!!II .......... IW.&I .......... .-

105

106

(Hz)

Fig. 11 Cm noise with both filters. Without Reomp

100 r;:::==========:r::::========:;--T----,

--- R"Oml) = 3.30, Lem = 2 x 1.5flH

80 --- RtI)ml) = 100, Lem = 2 x I.SflH --- DI E 55022 [5]

60

40

20 '-""" ..... -..-.-.. .. 105 10

6

(Hz)

Fig. 12 Cm noise with both filters. Different values of Reomp

REFERENCES

D. Cochrane, D. Y. Chen, D. Boroyevic, "Passive cancellation of common-mode noise in power electronic circuits", IEEE Transactions

on Power Electronics, vol. 18, Nr. 3, S. 756-763,2003 M. E. Jacobs, R. L. Taylor, "A New, Low-Cost Common Mode RFI Suppression Technique for Switch Mode Power Supplies", in: Proc. INTELEC '86, S. 475-478,1986 Wu Xin, M. H. Pong, Z. Y. Lu, Z. M. Qian, "Novel boost PFC with low common mode EM!: modeling and design", in: Proc. APEC '00, Bd. 1, S. l78-181, 2000 M. Schmidt, J. Stahl, M. Albach, "Influence of Parasitic Effects on Passive Cancellation of Common Mode Noise in a Boost Converter", in: Proc. APEC '12, in press Einrichtungen der Informationstechnik - FunkstOreigenschajten -

Grenzwerte und Messverfahren, NORM DIN EN 55022, 2003 1. Stahl, D. Kuebrich, A. Bucher, 1. Duerbaum, "Characterization of a modified LlSN for effective separated measurements of common mode and differential mode EMI noise", in: Proc. ECCE 'la, S. 935-941, 2010