Research on the prediction method of conducted EMI of switching power supply

1 Introduction

With the increase of switching frequency and power density, the electromagnetic environment inside the switching power supply is becoming more and more complex. Its electromagnetic compatibility problem has become a major focus in power supply design, and it has also become a major difficulty in power supply design. In conventional design methods, EMC problems are handled by empirical design, and the final consideration of EMC problems can only be made after the prototype is built. The traditional EMC remedy can only add additional components, and the addition of components may affect the original control loop bandwidth, resulting in the worst case of redesigning the entire system, increasing the design cost. In order to avoid such a situation, it is necessary to consider EMC issues during the design process, analyze and predict the EMI of the switching power supply with a certain degree of accuracy, and improve the design according to the mechanism of interference generation and its distribution in each frequency band, reduce the EMI level, and thus reduce the design cost.

2 Switching Power Supply EMI Characteristics and Classification

To predict the conducted electromagnetic interference of the switching power supply, it is first necessary to clarify its generation mechanism and the characteristics of the noise source. Due to the high-speed switching action of the power switch tube, the voltage and current change rates are very high, and the rising and falling edges contain a wealth of high-order harmonics, so the electromagnetic interference intensity generated is large; the electromagnetic interference of the switching power supply is mainly concentrated near the diode, power switch device, and the heat sink and high-frequency transformer connected to it; because the switching frequency of the switch tube ranges from tens of kHz to several MHz, the interference form of the switching power supply is mainly conducted interference and near-field interference. Among them, conducted interference will be injected into the power grid through the noise propagation path, interfering with other equipment connected to the power grid.

Switching power supply conducted interference is divided into two categories.

1) Differential mode (DM) interference. DM noise is mainly caused by di/dt, which propagates in the loop between the live wire and the neutral wire through parasitic inductance and resistance, generating a current Idm between the two wires without forming a loop with the ground wire.

2) Common mode (CM) interference. CM noise is mainly caused by dv/dt, which propagates in the loop between the two power lines and the ground through the stray capacitance of the PCB. The interference invades between the line and the ground, and the interference current flows through half of each of the two lines, with the ground as the common loop. In the actual circuit, due to the unbalanced line impedance, the common mode signal interference will be converted into crosstalk interference that is difficult to eliminate.

3 Simulation Analysis of Switching Power Supply EMI

Theoretically speaking, whether it is time domain simulation or frequency domain simulation, as long as a reasonable analysis model is established, the simulation results can correctly reflect the EMI quantification degree of the system.

The time domain simulation method requires the establishment of a circuit model containing all the component parameters in the converter, the use of PSPICE or Saber software for simulation analysis, and the use of fast Fourier analysis tools to obtain the spectrum waveform of EMI. This method has been verified in the analysis of DM noise. However, the nonlinear characteristics and stray parameters of nonlinear components in switching power supplies, such as MOSFET, IGBT and other semiconductor devices, make the model very complex. At the same time, the circuit topology of the switching power supply circuit is constantly changing when it is working, resulting in non-convergence problems in the simulation. When studying CM noise, all parasitic component parameters must be included. Due to the influence of parasitic parameters, it is difficult for the FFT results to match the experimental results; switching power converters usually work in a large range of time constants, mainly including 3 groups of time constants: time constants related to the basic frequency of the output end (tens of ms); time constants related to the switching frequency of the switching element (tens of μs); time constants related to the rise time and fall time when the switching element is turned on or off (several ns).

For this reason, in time domain simulation, a very small calculation step must be used, and it takes a long time to complete the calculation. In addition, the results obtained by the time domain method often cannot clearly analyze the impact of various variables in the circuit on interference, cannot deeply explain the EMI behavior of the switching power supply, and lacks judgment on the EMI mechanism, and cannot provide a clear solution for reducing EMI.

Frequency domain simulation is an analysis method based on the noise source and propagation path impedance model. LISN is used to provide a standard load impedance for the noise source. As shown in Figure 1, from the perspective of LISN, the entire system can be simplified into a noise source, a noise path, and a noise receiver (LISN). The frequency domain method can greatly reduce the simulation calculation time, and generally there will be no situation where the calculation results do not converge.

Figure 1 Concept of noise source and propagation path

In Figure 1, the noise paths include PCB conduction, coupling paths, heat sink capacitance coupling paths, transformer coupling paths, etc.

4 SMPS equivalent circuit model based on frequency domain method

To perform frequency domain simulation on a switching power supply, you must first establish a frequency domain simulation model of the switching power supply. The focus of frequency domain prediction of switching power supply EMI is the modeling of the noise path, including: high-frequency models of passive components; extraction of PCB and structural parasitic parameters.

By considering the parasitic parameters of passive components, PCB and structures, a circuit model of concentrated parameters of the switching power supply is established. The impedance, resonance point, etc. of the circuit can be obtained by calculation or simulation, thus providing a strong basis for reducing EMI.

Since the propagation paths of differential mode noise and common mode noise are different, it is necessary to model the DM propagation path and the CM propagation path separately. This can better analyze the characteristics of various interferences and provide a strong basis for designing filters.

4.1 Noise source model establishment

Since DM noise and CM noise need to be analyzed separately, the DM noise source and CM noise source also need to be modeled separately. M. Nave proposed in the literature [3] to use a current source as a DM noise source and a voltage source as a CM noise source because DM noise is mainly caused by di/dt, while CM noise is mainly caused by dv/dt. On this basis, the literature [4] improved the CM noise source, taking into account voltage overshoot and undershoot, and used a DM current source and a voltage source to represent the CM noise source when the line impedance is approximately balanced (as shown in Figure 2).

Figure 2 Representation of common-mode noise sources

The literature basically uses trapezoidal waves to represent noise sources, but in reality, not every waveform of the switching device in the circuit can be well approximated by a trapezoidal wave. Figure 3 shows the current and voltage waveforms of a flyback power switch. In addition to the trapezoidal wave, there are also current spikes, voltage overshoot and undershoot, which will cause the spectrum of the noise source to be somewhat different from the trapezoidal wave. Therefore, the trapezoidal wave cannot be used blindly to characterize the noise source. Instead, the circuit needs to be analyzed or simulated to obtain the current or voltage waveform of the switching device, and the noise source is modeled based on this waveform, so that the electromagnetic interference of the switching power supply can be more accurately reflected.

Figure 3 Current and voltage waveforms of a flyback power switch tube

4.2 High-frequency models of passive components

Within the frequency range of EMI, commonly used passive components can no longer be considered ideal, as their parasitic parameters seriously affect their high-frequency characteristics.

Among various passive components, the high-frequency equivalent parasitic parameters of resistors, inductors and capacitors can be measured using a high-frequency impedance analyzer. Table 1 shows the ideal model and high-frequency equivalent model of various passive components.

Table 1 High-frequency equivalent models of resistors, capacitors, inductors and transformers

For high-frequency transformers, it is proposed that finite element analysis and experimental measurement methods can be used to obtain leakage inductance, primary-secondary self-capacitance, and primary-secondary mutual capacitance, which are the main parameters that cause circuit oscillation and increase conducted EMI. Using the Maxwell simulation software of ansoft, the geometric dimensions and electromagnetic parameters of the transformer windings and cores can be input, and the parasitic parameters can be obtained by finite element analysis. The overall idea of ​​the experimental measurement method is to derive the impedance characteristics of the transformer under different working conditions (such as different combinations of open and short circuits of the primary and secondary windings) based on the established model, and then measure the impedance under these conditions to obtain leakage inductance and parasitic capacitance.

4.3 Extraction of PCB and structural parasitic parameters

In addition to the selection of components, circuit and structure design, the layout, wiring design and circuit board processing of PCB will have a great impact on electromagnetic compatibility and is a very important design link. Since the PCB wiring of the switching power supply is basically arranged manually based on experience, it is very random, which increases the difficulty of extracting PCB distribution parameters. The parasitic parameters of the PCB will cause the impedance change of the noise propagation path of the switching power supply, affecting the control of the controller on the output voltage and current of the switching power supply. The unreasonable layout of the PCB will also form a path for the switching power supply to radiate electromagnetic interference outward, and at the same time absorb external electromagnetic interference through this path, thereby reducing the electromagnetic interference immunity of the switching power supply. Therefore, the layout and wiring of the PCB is an extremely important link in the EMC design of the switching power supply.

For conducted interference, the accuracy of parasitic parameter extraction is the key to effectively predict EMI levels through simulation. Although parasitic parameters are easy to calculate for components with simple structures, it is not so easy to obtain parasitic parameters for components in complex structures, such as multilayer boards and DC busbars.

In order to establish a high-frequency model of the switching power supply PCB, it is necessary to extract the structural parasitic parameters of the PCB. There are many methods to extract PCB parasitic parameters, among which the TDR (time domain reflection) method can extract parasitic inductance and parasitic capacitance without knowing the actual geometric shape, but the TDR (time domain reflection) method requires a time domain reflectometer, which is used after the prototype is built, which greatly increases the development cost, and the TDR method cannot find the coupling effect in the complex structure; however, the FEA (finite element analysis) method can overcome this shortcoming and be used before the prototype is built. The FEA tool can accurately obtain the parasitic parameters of the PCB and can consider the coupling of complex geometric structures.

There are many software for extracting parasitic parameters of PCB structures, such as InCa, SIwave, Q3D, etc., which use different methods to calculate and extract the parasitic parameters of PCB, such as partial element equivalent circuit method, finite element analysis method, finite element analysis method and moment method combined method, etc. Among them, InCa software can only calculate distributed inductance, not suitable for calculating distributed capacitance, and not suitable for processing simulation analysis of common mode interference; SIwave software extracts S parameters of the circuit, which cannot clearly reflect the coupling situation in PCB and its impact on switching power supply EMI; Q3D software uses FEA and MOM combined method to solve electromagnetic field, which can obtain PEEC partial element equivalent circuit, and can also obtain the mutual inductance and mutual capacitance of each conductor on PCB, which can clearly analyze the impact of PCB structure on switching power supply EMI under various conditions.

J. Ekman proposed a method for establishing an equivalent circuit based on a parasitic parameter matrix, which is to equate all mutual inductances and mutual capacitances to controlled voltage sources and connect them with self-inductance and self-capacitance (equivalent to equating the effects of all mutual inductances and mutual capacitances on the circuit to controlled voltage sources), thereby establishing an equivalent circuit model. Figure 4 shows an equivalent circuit model between any two nodes.

Figure 4 Equivalent circuit model between any two nodes

In Figure 4:

Where: Lpmn is the mutual inductance between the two wires m and n.

Although this can improve the accuracy of the simulation, it increases the amount of analysis calculations. The amount of calculations can be reduced by ignoring some mutual inductance and mutual capacitance that do not have a significant impact on the results.

There is a capacitive effect between the heat sink and the switching tube, and noise can be transmitted between the circuit and the ground through this effect. Reference [9] gives a detailed explanation of the effect of the heat sink on the conduction and radiation interference of the switching power supply.

There is also other noise that is transmitted to the receiver through inductive or capacitive coupling in space and cannot be ignored.

After the model is established, simulation software can be used to simulate the switching power supply EMI to obtain the spectrum waveform of the switching power supply conducted EMI. By analyzing the waveform, the problem of switching power supply EMI can be located, and then EMI can be reduced by solving the problem.

5 Design methods and strategies to reduce EMI

To reduce the EMI of the switching power supply, we need to start with the noise source and the propagation path. First, for the noise source, we can reduce the EMI level by adding an absorption circuit and reducing di/dt and dv/dt, but this will affect the efficiency of the switching power supply, so we need to make a certain trade-off between the two.

Then the propagation path is improved. The purpose of the improvement is to increase the impedance of the propagation path to the interference, blocking its propagation to the receiver, and to reduce the impedance to the power provided by the power grid, thereby increasing the working efficiency of the switching power supply.

When selecting components, it is necessary to select components with small parasitic parameter effects, such as the ESR and ESL of capacitors should be as small as possible, and the parasitic capacitance of inductors should be small, etc. In the design process of PCB and heat sink position, it is also necessary to increase the impedance of the interference propagation path as much as possible, so that as little noise as possible is transmitted to the receiver through the PCB path.

If all the above measures to reduce EMI have been completed and the EMC standard has not been met, the filter can be designed based on the waveforms of differential and common mode interference obtained from the previous simulation analysis. When designing the filter, attention should also be paid to the layout of the components and the influence of PCB parasitic parameters on the filter impedance. Its essence is to increase the impedance to the interference so that the interference cannot pass through the propagation path. The design process of the switching power supply is shown in Figure 5.

Figure 5 Switching power supply design process

6 Conclusion

In summary, there are two methods for predicting the conducted interference of switching power supplies: time domain method and frequency domain method. Since the time domain method requires a very small calculation step, it takes a long time to calculate, and the simulation results are prone to non-convergence. At the same time, the results obtained by time domain simulation often cannot clearly analyze the impact of various variables in the circuit on interference. The frequency domain simulation has a clear physical meaning and is easier to judge the impact of various parameters on EMI, which can provide a strong basis for reducing EMI. The key issue is to establish a reasonable frequency domain model of interference sources and propagation paths.

There are many software for extracting PCB parasitic parameters. These software are suitable for different fields and can be selected according to task requirements.

For high-frequency equivalent circuit models, some factors such as mutual inductance and mutual capacitance that have little effect on EMI can be ignored through circuit analysis methods, which can reduce the amount of calculation without reducing the calculation accuracy too much.

The main method to reduce EMI is to increase the impedance of the propagation path to electromagnetic interference, so that the electromagnetic interference passes through the propagation path as little as possible. The filter design can be based on the simulation results of DM noise and CM noise respectively, and special attention should be paid to the layout of filter components. A good layout can better suppress the propagation of noise.