Conducted EMI Simulation of Switched Mode Power Supply

Size: px

Start display at page:

Download "Conducted EMI Simulation of Switched Mode Power Supply"

Deirdre Caldwell
5 years ago
Views:

1 Conducted EMI Simulation of Switched Mode Power Supply Hongyu Li #1, David Pommerenke #2, Weifeng Pan #3, Shuai Xu *4, Huasheng Ren *5, Fantao Meng *6, Xinghai Zhang *7 # EMC Laboratory, Missouri University of Science and Technology 4 Enterprise Dr. Rolla, MO, USA 1 hlfm4, 2 davidjp, 3 wpqk7@mst.edu 4 * Huawei Technologies CO., LTD. Huawei Industrial Base, Bantian Longgang, Shenzhen, P. R. China xushuai, 5 renhuasheng, 6 mengfantao, 7 hwzhangxh@huawei.com Abstract This paper introduces an efficient method to predict the conducted EMI of switched mode power supply (SMPS) through time domain SPICE simulation, which can be used to design or optimize the filter circuit and quantify the suppression degree of the filter. The SMPS modelling method permits modelling of the SMPS as a noise signal source even when its internal structure is unknown. The method is verified by comparing predicted and measured noise signals. Key words: SMPS, DC/DC converter, Conducted EMI, SPICE, Filter I. INTRODUCTION Because of its superior efficiency, switched mode power supply (SMPS) is widely applied in modern electronic systems, [1] and there are many kinds of SMPS modules on the market, including DC/DC converter modules. However, SMPS generates electrical noise due to its high dv/dt and di/dt from switching. Typically, SMPS generates three types of EMI: low frequency conducted noise (<3 MHz), broadband noise (5 3 MHz), and high frequency noise (> 3 MHz). [2] This paper focuses on suppression of low frequency conducted noise. Many international agencies specify conducted and radiated emissions limits for electronic products. Included among these are CISPR, FCC, VCCI, and the new CE specifications. Most agency-conducted noise limits apply only to noise currents induced onto the AC power lines in finished products. European Telecommunication Standard Instructions (ETSI) are an exception, applying CE requirements to DC supplies with cables over three meters long. Although not required to do so by agency standards, some system designers apply the conducted emissions requirements to subassemblies within the product to reduce internal interference among subsystems and to reduce the difficulty of meeting overall system requirements. [3] High-density SMPS modules are usually designed to operate at a high switching frequency to reduce the size of the internal filter components. The small EMI filters internal to the modules are often inadequate to meet the stringent international EMI requirements. To meet these requirements, external filtering of the power module is often necessary. [3] Traditional filter design depends heavily on the designers experiences and experiments on prototype boards, and the effect of the filter is difficult to quantify. Therefore, an efficient method to predict the conducted EMI could permit filters design through simulations rather than experimental procedures and iterations. This paper introduces a method of simulating the conducted EMI of SMPS through the use of time-domain SPICE simulation. Fig. 1 shows the typical conducted EMI test system. input LISN Spectrum Analyzer Filter Circuit SMPS Module output Fig. 1. Block diagram of a conducted EMI test system To simulate conducted EMI, the entire test system must be modelled including the SMPS module, the filter circuit, and the LISN. The modelling method for each part will be introduced separately in sections II, III, and IV below. Research results show that near-field coupling significantly influences the conducted EMI; these results will be introduced in section V. Finally, all the models will be combined to form one system module to predict the conducted EMI. By comparing predicted and measured signals, the models and the method are fully verified. This simulation method is illustrated with the example of a DC/DC converter module. This is a full-brick package, flyback style, 48V to 28V, 7W output-module, as shown in Fig. 2. II. SMPS MODULE MODEL If the internal components and PCB layout of the SMPS module are known, a common method of modelling the SMPS module is to obtain the main and parasitic parameters of the components from datasheets and measurements, and then extract the parasitic parameters of the PCB using electromagnetic simulation tools and measurements. In many situations, however, the internal structure of the SMPS module is unknown, so this common method is inadequate. Furthermore, if many types of SMPS modules are used, it is not efficient to derive the models of each module using this common method. Therefore, an efficient method is needed to ZL /9/$ IEEE 155

2 model the SMPS without knowing the internal structure. One possibility is to take the SMPS module as a noise source and obtain the source impedance and source current or voltage by measurements and calculation, as illustrated in Fig. 2. Fig. 2. SMPS module as noise source A. Source Impedance In passive mode, the impedance can be directly measured by a network analyzer or an impedance analyzer. However the source impedance in active mode is unknown. To address this problem, this research presented here developed a method to measure the impedance in active mode, which involves high DC voltage and a very noisy environment where the usual equipment is not suitable. The basic idea of this method is to inject frequency excitations to the SMPS, measure the response, and then calculate the impedance based on the measured voltages and currents. The measurement setup is illustrated in Fig. 2, and the excitation is input by a current clamp. This work developed a single automatic measurement system controlled by Matlab code. Several key points of this system require explanation: 1) Differential and common mode: There are differential and common mode noises. [4] It is better to build the models for the two modes separately, because it is easier to locate the problem when filters are designed using the models. In the setup shown in Fig. 2, the current clamp injects common mode excitation signals. For differential mode impedance measurement, differential signal should be injected by putting the two cables in opposite directions in the current clamp. 2) Magnitude and frequency of the excitation: The excited signals should not disturb the normal operation of the SMPS; therefore, the magnitude of the excitation signal should be as small as possible as long as it can be detected by the measurement equipment. So that the noise will not be unintentionally detected as the response signal, the frequency of the excitation signal should not overlap with the noise. 3) Correction: The measured signals must be corrected by the frequency response characteristic and the time delay of the probes. 4) Synchronization: The excitation equipment (e.g., the signal generator), and signal measurement equipment (e.g., the oscilloscope) must run synchronously. Even slight differences between the time sequences of the two instruments will cause errors and reduce the accuracy of the measurement. In this research, the measured differential impedance in active mode of the DC/DC converter is shown as a blue line in Fig. 3. The impedance was measured under a driving 2W DC load. For SMPS modules generally, variation in the DC load does not change the internal structure and the coupling path; therefore, in active mode, impedance should remain constant regardless of DC load. This conclusion was verified by comparing impedances measured under various DC load conditions. These measurements were compared with those taken in passive mode using a network analyzer (NWA), an impedance analyser, and the automatic measurement system. Impedance in active mode was found to be the same as that in passive mode for this particular DC/DC converter module. Slight differences among measurements in the high frequency range (shown in Fig. 3) were caused by variations in the measurements setup. Fig. 3. Measured differential impedance of the DC/DC converter From the measured impedance curve, the equivalent circuit (EQC) can be obtained by calculating the capacitances or inductances in ranges in which the slope is decreasing or increasing by 2 db/decade, and the resistances at the resonant points. If the curve is complex and manual calculations to obtain the EQC become difficult, commercial software packages are available for this type of conversion. The EQC of the differential impedance of the DC/DC converter module used in this research is shown in Fig. 1, and the simulated impedance is compared to the measured impedance in Fig. 3. B. Source Current or Voltage According to the Thevenin s Theorem and Norton s Theorem, either current or voltage source may be used. Since the impedance of the source is already known, the source currents or voltages can be calculated from the measured external noise currents and voltages when the SMPS is 156

3 operating. This research used current source, and the calculation process is shown in Fig. 4. Perform FFT on the measured external voltages and currents Identify the peak value and frequency of the harmonics Calculate the internal source current at each harmonics in the frequency domain Transform the calculated frequency domain source currents into a time domain signal which can be imported in PSpice Fig. 4. Source current or voltage calculation process C. Verification Next, it was necessary to verify that an SMPS can be taken as a current or voltage source. The primary characteristic of a source is that the same source impedance and the current or voltage can be measured and calculated with different loads. For most SMPS modules, the external load does not change the internal structure, the high dv/dt and di/dt, or the coupling path. Furthermore, the field coupling from the outside to the interior of the module is negligible compared to the conducted coupling. Therefore, it is reasonable to suppose that the SMPS impedance and the source currents or voltages are essentially constant. This conclusion was verified by comparing the measured impedance with various filter circuits that are the loads of the AC noise source. Here, the verification method is illustrated by the example of the source current. The external currents and voltages were measured from filter and filter1. Comparison of the measured signals shows the significant differences between these two filters, as shown in Fig. 5. Voltage [v] 2 Comparison of measured Vip of filter and filter2 Time domain From filter Time [us] impedance and source current or voltage as a noise signal source. Current [A] Magnitude [dbua] Time [us] Frequency domain 1 5 Common mode source comparison Time domain From filter Fig. 6. Comparison of the calculated current sources From filter III. FILTER CIRCUIT MODEL For modelling the conducted noise signals, not only the main parameters, but also the parasitic parameters of the components and the PCB are critical to the accuracy of the simulation. A. Components For simple components such as capacitors, the parameters can be taken from the datasheets. Sometimes, the DC bias and temperature characteristics must also be considered. For the complex components such as the common mode choke, some parasitic parameters must be obtained from measurements. The method to model the common mode choke is shown in Fig. 7. [4][5] Magnitude [dbuv] Frequency domain From filter Fig. 5. Comparison of voltage signals measured from different loads Fig. 6 demonstrates that different measured signals derived almost the same source current signals, indicating that the noise of the SMPS has the primary characteristic of a source. It is reasonable, therefore, to measure and derive its Lcm: Common mode inductance; Llk: Leakage inductance; Cp1: Parasitic capacitances in the winding; Cp2: Parasitic capacitances between two windings; Rw: Winding conductor resistance; Rm: Core resistance. B. PCB Fig. 7. Common mode choke modelling method 157

The PCB parameters can be extracted from measurement or using tools such as the Ansoft Q3D Extractor, and both of the two methods were applied in this research to verify the results.

4 The PCB parameters can be extracted from measurement or using tools such as the Ansoft Q3D Extractor, and both of the two methods were applied in this research to verify the results. The filter circuit model of the DC/DC converter used in this research is shown in Fig. 1. IV.LISN MODEL International standards have defined the reference impedance on which the conducted EMI should be measured. This impedance is guaranteed by a Line Impedance Stabilization Network (LISN), and precisely defined by the CISPR16 document. Fig. 8 illustrates the internal structure of the device. The LISN offers a 5 impedance over the frequency of interest (e.g., 15 khz 3 MHz for CISPR22) and shields the measurement against unwanted incoming noises. [6]-[8] Fig. 8. LISN The SPICE model of LISN can be derived from the schematic of this device, and the model used in this experiment is shown in Fig. 1. V. NEAR FIELD COUPLING The near field coupling was found to influence significantly the conducted EMI. Some of the coupled signal will bypass the filter and propagate to the LISN directly. All the PCB loop and sensitive components can couple the near field. On the test board used in this research, the common mode choke was not optimally placed for the research purpose; therefore it s near magnetic field coupling is quite strong. The coupling is apparent from the near magnetic field scanning results shown in Fig. 9. The three subfigures show the magnetic field distribution of directions (x, y, and z). The cross indicates the position where the common mode choke is mounted. Modelling the coupling effect with SPICE is difficult, and the present SPICE model does not yet include this effect. Fig. 9. Near magnetic field scanning results, measured using a 5cmx5cm square loop H field probe [9] VI.SYSTEM MODEL All of the part models described here were combined to form a system model, as shown in Fig. 1. The system model can be used to predict the conducted EMI, and the predicted LISN voltage signals are compared with the measured signals in Fig. 11. Since near field coupling influences the conducted EMI, and since this effect is not taken into account in the SPICE model, the signal was measured after the sensitive components were shielded. As shown in Fig. 11, the predicted signals closely match the measured signals. VII. CONCLUSION This paper introduces an efficient method of predicting the conducted EMI of an SMPS through SPICE simulation that takes the SMPS as a noise source to model. The research results indicate that this method of SMPS modelling is reasonable and efficient. The combined system model accurately predicts the conducted EMI. This simulation method can be helpful to designers in several respects, including filter design and optimization of external filters for SMPS, quantification of the filter s suppression effect, and analysis of conducted EMI problems through simulations rather than experimental procedures and iterations. This method is especially suitable for situations in which many types of SMPS modules are used but the internal structures are unknown. This work also investigated the influence of near-field coupling on the conducted EMI. The method of modelling this effect in SPICE, however, requires further development. X Y Z 158

5 SMPS LISN Filter Fig. 1. System model Diff LISN lvdfvoltage Magnitude [dbuv] 1 5 measured shielded simulated Magnitude [dbuv] Comm LISN lvcm voltage measured shielded simulated Fig. 11. Comparison of predicted and measured LISN voltage signals 159

6 REFERENCES [1] Abraham I. Pressman, Switching Power Supply Design, 2nd ed., McGraw-Hill, [2] Zhe Li, Pommerenke, D., EMI specifics of synchronous DC-DC buck converters, Electromagnetic Compatibility, 25 International Symposium, on Volume 3, 8-12 Aug. 25 Page(s): Vol. 3. [3] FLTR1V2 Filter Module data sheet, Tyco Electronics Power Systems, Inc., TX, USA. [4] Zhe Li, Pommerenke, D., Shimoshio, Y., Common-mode and differential-mode analysis of common-mode chokes, Electromagnetic Compatibility, 23 IEEE International Symposium, on Volume 1, Aug. 23 Page(s): vol.1. [5] Liu Dehong, Jiang Xanguo, High frequency model of common mode inductor for EMI analysis based on measurements, Electromagnetic Compatibility, 2 3rd International Symposium, on May 22 Page(s): [6] Christophe Basso, Conducted EMI Filter Design for the NCP12, AND832/D, ON Semiconductor, USA. [7] C. Basso, Spice predicts differential conducted EMI from switching power supplies, EDN February 3, 1997 [8] Christophe P. Basso, Switch-Mode Power Supply SPICE Cookbook, McGraw-Hill. [9] Measured with the help of Amber Precision Instruments, 16

Characterization of Conducted Electromagnetic Interference (EMI) Generated by Switch Mode Power Supply (SMPS)

Revue des Sciences et de la Technologie - RST- Volume 5 N 1 / janvier 2014 Characterization of Conducted Electromagnetic Interference (EMI) Generated by Switch Mode Power Supply (SMPS) M. Miloudi*, A.