Research on the Principle and Design of Switching Power Supply EMI Filter

zbz0529

Research on the Principle and Design of Switching Power Supply EMI Filter [Copy link]

Abstract: In switching power supplies, EMI filters play a significant role in suppressing common-mode and differential-mode conducted noise. Based on the study of the filter principle, a method of independently analyzing and modeling common-mode and differential-mode signals is discussed, and finally a design procedure for EMI filters is proposed based on this.

Keywords: switching power supply; EMI filter; common mode; differential mode

　

0 Introduction

High-frequency switching power supplies have been widely used in various fields such as industry, national defense, and home appliances due to their many advantages in terms of volume, weight, power density, and efficiency. When switching power supplies are used in AC power grids, the rectifier circuit often causes the input current to be discontinuous, which not only greatly reduces the input power factor, but also increases a large number of high-order harmonics. At the same time, the high-speed switching action of the power switch tube in the switching power supply (from tens of kHz to several MHz) forms an EMI (electromagnetic interference) interference source. From the published switching power supply papers, it can be seen that the main forms of interference in switching power supplies are conducted interference and near-field radiation interference. Conducted interference will also be injected into the power grid, interfering with other devices connected to the grid.

There are many ways to reduce conducted interference, such as laying ground wires properly, adopting star-shaped grounding, avoiding ring ground wires, minimizing common impedance, designing reasonable buffer circuits, reducing circuit stray capacitance, etc. In addition, EMI filters can be used to attenuate the noise interference between the power grid and the switching power supply.

EMI disturbance is usually difficult to describe accurately, and the design of the filter is usually done through repeated iterations, calculations and production in order to gradually approach the design requirements. This article starts with the EMI filtering principle, analyzes its common mode and differential mode noise models, gives the method of designing filters in actual work, and gives a design example in steps.

1 EMI filter design principle

In a switching power supply, the main EMI disturbance source is the dv/dt and di/dt generated by the switching action of the power semiconductor device . Therefore , the electromagnetic emission EME (Electromagnetic Emission) is usually a broadband noise signal, and its frequency range is from the switching operating frequency to several MHz. Therefore, the measurement of the conducted electromagnetic environment (EME), as stipulated by many international and national standards, has a frequency range of 0.15 to 30MHz. The design of EMI filters is to give sufficient attenuation to the noise of the switching frequency and its higher harmonics. Based on the above standards, under normal circumstances, it is only necessary to consider attenuating the EME with a frequency higher than 150kHz to a reasonable range.

The low-pass filter concept generally recognized in the field of digital signal processing is also applicable to power electronic devices. In short, EMI filter design can be understood as meeting the following requirements:

1) Specify the required stopband frequency and stopband attenuation; (satisfying a specific frequency f _stop requires an attenuation of H _stop );

2) Low attenuation of grid frequency (meeting the specified passband frequency and passband low attenuation);

3) Low cost.

1.1 Commonly used low-pass filter models

EMI filters are usually placed at the front end of the switching power supply connected to the power grid. They are low-pass filters composed of series inductors and parallel capacitors. As shown in Figure 1, the equivalent impedance of the noise source is Z _source and the equivalent impedance of the power grid is Z _sink . The filter indicators ( f _stop and H _stop ) can be achieved by first-order, second-order or third-order low-pass filters. The calculation of the filter transfer function is usually approximated at high frequencies, that is, for n -order filters, all ω ^k- related terms are ignored (when k < n ), and only ω ^n- related terms are taken. Table 1 lists several common filter topologies and their transfer functions. It is particularly important to consider the impact of input and output impedance mismatch on the filter characteristics.

Figure 1 Filter design equivalent circuit

Table 1 Several filter models and transfer functions

1.2 EMI filter equivalent circuit

Conducted EMI noise includes common mode (CM) noise and differential mode (DM) noise. Common mode noise exists between all AC phase lines (L, N) and common mode ground (E), and its source is believed to be insulation leakage current between two electrical circuits and electromagnetic field coupling; differential mode noise exists between AC phase lines (L, N), and its source is pulsating current, ringing current of switching devices and reverse recovery characteristics of diodes. The sources of these two modes of conducted noise are different, and the conduction paths are also different, so common mode filters and differential mode filters should be designed separately.

Obviously, it is necessary to separate the two different modes of conducted noise and measure their actual levels separately. This will help determine which mode of noise accounts for the majority and reflect it accordingly in the corresponding filter design process to achieve parameter optimization. References [6] and [7] provide two noise separators for distinguishing common-mode and differential-mode noise. They can selectively attenuate common-mode or differential-mode noise by at least 50 dB, thus effectively measuring common-mode and differential-mode components. The principle and use of the separator are beyond the scope of this article. For details, see references [6] and [7].

Taking a commonly used filter topology (Figure 2(a)) as an example, the equivalent circuits of common-mode and differential-mode noise filters are analyzed respectively. Figure 2(b) and Figure 2(c) represent the equivalent circuits of common-mode attenuation and differential-mode attenuation of the filter respectively. From the circuit analysis, it can be seen that _{Cx1 and} Cx2 are only used to suppress differential-mode noise, and _the ideal common-mode choke inductor LC is only used to suppress common-mode noise. However, due to the asymmetry of _the actual LC winding, there is _a leakage inductance Lg between the two groups _{of LC} , which can also be used to suppress differential-mode noise. Cy can suppress both common-mode interference and differential-mode noise, but because the differential-mode suppression capacitor _{Cx2 is much larger than Cy, Cy's effect on differential-mode suppression can be ignored. Similarly, LD can} suppress both common _- mode _interference and differential -mode interference, but LD _is much smaller than LC , so its effect on suppressing common _- mode noise is relatively small.

(a) Common filter topologies

(b) Common-mode attenuation equivalent circuit

(c) Differential mode attenuation equivalent circuit

Figure 2 A commonly used filter topology

From Table 1 and Figure 2, it can be deduced that for the common-mode equivalent circuit, the filter model is a second-order LC low-pass filter. The equivalent common-mode inductance is recorded as L _CM , and the equivalent common-mode capacitance is recorded as C _CM , then

L _CM = L _C＋LD（1 _）

C _CM =2 C _y（2）

For the differential mode equivalent circuit, the filter model is a third-order CLC low-pass filter. The equivalent differential mode inductance is recorded as L _DM , and the equivalent differential mode capacitance is recorded as C _DM (let C _x1 = C _x2 and assume that Cy _/ 2<< C _x2 ), then

L _DM = 2 L _D＋L _g（3）

C _DM = C _x1 = C _x2 (4)

The cut-off frequency calculation formula of LC filter _is :

f _R,CM = (5)

Substituting equation (1) and equation (2) into equation (5), we have

f _R,CM = ≈ ( LC >> L _{D ) (} 6)

The cut-off frequency calculation formula of CLC filter is:

f _R,DM = (7)

Substituting equation (3) and equation (4) into equation (7), we have

f _R,DM = (8)

When the noise source impedance and grid impedance are determined and matched, the EMI filter's suppression effect on common-mode and differential-mode noise is shown in Figure 3.

Figure 3 Filter differential mode and common mode attenuation

2 Practical Methods for Designing EMI Filters

2.1 Some considerations in design

The effect of EMI filter depends not only on itself, but also on the noise source impedance and grid impedance. Grid impedance Z _sink is usually corrected by static impedance compensation network (LISN), which is connected between the filter and the grid, including inductor, capacitor and a 50Ω resistor, so as to ensure that the grid impedance can be calculated by known standards. The EMI source impedance depends on different converter topologies.

Take a typical flyback switching power supply as an example, as shown in Figure 4 (a), the current of its full-bridge rectifier circuit is in a discontinuous state, and the current and voltage waveforms are shown in Figure 5. For common-mode noise, Z _{source shown in Figure 4 (b) can be regarded as} a current source IS and a high impedance Z _P in parallel; for differential-mode noise in Figure 4 (c), depending on the on-off state of the rectifier bridge diodes, Z _source has two states: when any two diodes are turned on, Z _source is equivalent to a voltage source _VS connected _in series with a low-value impedance Z _S ; when all diodes are turned off, it is equivalent to a current source IS and a high impedance Z _P in parallel. Therefore, the noise source differential-mode equivalent impedance Z _source switches between the above two states at a frequency twice the power frequency [2].

(a) Typical flyback switching power supply

(b) Common-mode noise source equivalent circuit (c) Differential-mode noise source equivalent circuit

Figure 4 Typical flyback switching power supply and its noise source equivalent circuit

Figure 5 Voltage and current waveforms at the power input

In the above design process, EMI filter components (inductors, capacitors) are considered ideal. However, due to the existence of parasitic parameters in actual components, such as the parasitic inductance of capacitors, the parasitic capacitance between inductors, and the parasitic parameters of PCB board wiring, the actual high-frequency characteristics are often quite different from the ideal component simulation. This involves many issues such as EMC high-frequency modeling. The parameters of the model are often difficult to determine. Therefore, this article only considers the low-frequency suppression characteristics of the EMI filter, and the high-frequency modeling can refer to references [8] and so on. Therefore, the values of Z _S and Z _{P are related to these parasitic parameters such as parasitic capacitance, inductance, and equivalent capacitance of the rectifier bridge. It is difficult to directly solve the source impedance by using the solution based on circuit topology and parameter modeling. Therefore, actual measurement of} Z _source is often used in the design .

2.2 Practical Design Steps

EMI filter design often requires that the size of the filter itself should be as small as possible and the cost should be as low as possible while suppressing noise. At the same time, the filtering effect also depends on the actual noise level. The interference weights of common-mode and differential-mode noise are analyzed. Therefore, the following parameters are required to be determined before design to achieve design optimization.

1) Measure the equivalent impedance of the interference source Z _source and the equivalent impedance of the power grid. In the actual process, we often rely on the guidance of theory and experience to first make the PCB board of the power supply. This is because the common mode and differential mode noise sources and interference paths are different, and a slight difference in the circuit board routing may lead to a large EME change.

2) Measure the interference noise spectrum before adding the filter, and use the noise separator to separate the common mode noise V _MEASUREE,CM and the differential mode noise V _measure,CM to make the corresponding interference spectrum.

Then we can proceed to the actual design. Still taking the filter model proposed in this article as an example, the steps are as follows.

(1) According to formula (9), calculate the common mode and differential mode attenuation required by the filter, and draw the curves V _{measure, CM} - f and V _{measure, DM} - f , where V _{measure, CM} and V _{measure, DM} have been measured, and V _{standard, CM} and V _{standard, DM} can be set according to the national standard for conducted EMI interference. The reason for adding 3dB is that the measured value using the noise separator is 3dB larger than the actual value.

( V _req,CM )dB=( V _measure,CM )-( V _standard,CM )＋3dB

( V _req,DM )dB=( V _measure,DM )-( V _standard,DM )+3dB (9)

(2) As shown in Figure 3, the intersection of the two oblique lines with slopes of 40dB/dec and 60dB/dec with the frequency axis is f _R,CM and f _R,DM . Draw tangents to V _measure,CM - f and V _measure,DM - f. The slopes of the tangents are 40dB/dec and 60dB/dec respectively. By comparison, we can know that f _R,CM and f _R,DM can be obtained by measuring their intersections with the frequency axis. Figure 6 shows a schematic diagram.

(a) The solid line is the common mode target attenuation; the dotted line is the tangent line with a slope of 40dB/dec

(b) The solid line is the differential mode target attenuation; the dotted line is the tangent line with a slope of 60dB/dec

Fig.6 Determination of f _R,DM and f _R,CM

(3) Filter component parameter design

——Selection of common mode parameters Cy is connected between the phase line and the ground. If the capacitor capacity is too large, _the leakage current will be too large, which will reduce the safety. The smaller the leakage current requirement, the better. The safety standard is usually several hundred μA to several mA.

The calculation formula of EMI ground leakage current Iy is _:

I _y =2π fCV _c（10）

Where: f is the grid frequency.

In this example, V _c is the voltage drop across _{capacitor Cy} , f = 50Hz, C = 2 Cy , V _c = 220/2 = 110V, _then

Cy ₌ ( 11)

If the ground leakage current is set to 0.15 mA, Cy _≈ 2200 pF can be obtained. Substitute _Cy into step (2) to obtain the value of f _R,CM , and then substitute f _R,CM into formula (6) to obtain

L _c = (12)

——Selection of differential mode parameters From formula (8), it can be seen that _{there is no unique solution for the selection of} Cx1 , Cx2 , and LD , _which allows the designer to have a certain degree of freedom.

As shown in Figure 2, _the leakage inductance Lg of the common mode inductor Lc can also suppress differential mode noise. Sometimes, in order to simplify the filter, Ld can be omitted . _{Experience shows that} the leakage inductance _{Lg is} usually _{0.5 % to 2% of the Lc} value . Lg can be measured. At this time, _the corresponding Cx1 and Ccx2 values should be larger.

3 Conclusion

This article is based on the low-frequency model analysis of low-pass filters. Due to the influence of actual component parasitic parameters, especially in the high-frequency band, it is often necessary to repeatedly revise the parameters after the first determination, use capacitors with low ESR and ESL , optimize the materials and processes of winding the magnetic core, and gradually approach the required technical indicators.

Since it only involves the design of a single-stage filter, such as an LC filter with an attenuation of only 40dB/dec, when the attenuation is required to be above 60-80dB, a multi-stage filter is often required.

It is usually difficult to design a universal EMI filter. This is because the electromagnetic environment levels of different power converters vary greatly due to topology, component selection, PCB layout, etc., and the impedance matching problem greatly affects the versatility of the filter. Therefore, the design of the filter often needs to be targeted and gradually corrected during actual debugging.

　

About the Author

Wei Yingdong, male, master's student, is currently engaged in research related to power electronics topology and electromagnetic compatibility.

Wu Xiehua, female, professor, master's supervisor, is currently engaged in research on power electronics system integration and intelligent control.