Frequency Domain Prediction of Conducted EMI in Power Converters with. front-end Three-phase Diode-bridge

Transcription

1 Frequency Domain Prediction of Conducted EMI in Power Converters with front-end Junsheng Wei, Dieter Gerling Universitaet der Bundeswehr Muenchen Neubiberg, Germany Marek Galek Siemens Corporate Technology Munich, Germany Keywords: EMC/EMI, Three-phase System, Diode, Simulation Abstract This paper investigates the frequency domain prediction of conducted EMI within power converters supplied by front-end three-phase diode-bridge. Noise source, noise propagation path and noise receiver are identified. Difficulty due to the presence of diode-bridge is analyzed and a method is proposed and verified by measurement to take the variation of propagation path into account while still enable the simulation to be carried out in frequency domain. 1. Introduction Conducted EMI has been important concern during the design phase of power converters since certain limits are defined by international standards committee to ensure that the emission will not endanger other equipment in the vicinity. In order to adopt measures as early as possible in the design phase and preserve the possibility of optimally solving the EMI problem, prediction of the EMI performance of power converter is desired. Even though there exists behavioral modeling of the conducted EMI [1], to consider the EMI problem and test certain measures in the design phase, physical modeling is preferred. With the parameters extractions based on numerical methods [2] or measurement [3], the system can be simulated in time domain (TD) [2] or frequency domain (FD) [4] [5]. It is observed that the time required for TD simulation is still too long for optimization purpose and therefore the possibility to carry out the prediction in FD should be investigated. FD prediction has been investigated in literature for different types of converters. However, none of them deals with converters with three-phase diode-bridge as front-end while this type of converters plays also important role as they are widely applied in industry application due to the high power characteristic. As discussed in [6], the presence of three-phase diode-bridge implies time varying propagation path and could lead to difficulty in identifying noise appearing at the line impedance stabilization network (LISN), especially with the effects of different detector types. This paper is therefore dedicated to discuss the identification of noise source, noise propagation and noise receiver for FD simulation on a power electronics circuit which consists of three-phase diode-bridge as front-end and a boost converter to emulate constant load for power factor correction. EPE'13 ECCE Europe ISBN: and P.1

2 This paper is organized as following: in second chapter the analysis and identification of noise source, noise propagation path and noise receiver are carried out by investigating the typical measurement environment for conducted EMI; in third chapter, based on the analysis the modeling for separate parts are proposed, especially the propagation path is modeled with a novel method to take into account its characteristic and describe it in frequency domain; latter in forth chapter validation of used method is made by comparing predicted and measured results; at the end conclusions are given in last chapter. 2. Identification of components according to EMI mechanism To carry out prediction, especially based on frequency domain simulation, the noise source, noise propagation path and noise receiver must be firstly identified. With knowledge of different components in these parts, the final spectrum shown can be calculated as Eq. 1 in frequency domain. Here U represents the spectrum of voltage at receiver and S represents the spectrum of noise source while T represents the transfer function of propagation path. U ( f ) S( f ) T ( f ) = Eq. 1 The definition of components according to EMI mechanism can be done by analyzing the conducted EMI measurement environment as Fig. 1 shows. According to CISPR [7], the conducted EMI measurement consists of test receiver, line impedance stabilization network (LISN), device under test (DUT), load and ground plane. The EMI noise is generated by the DUT and more specifically by the high speed switching of semiconductors. Propagation path as indicated in mechanism in Fig. 1 includes inductive and capacitive paths and both provide channels for transferring EMI noise to the receiver. The LISN is typical device used to provide stable grid side impedance at different frequencies and it can also be considered as part of propagation path. The noise receiver is the test receiver, which has different detector functions implemented inside and senses the voltage in LISN. Fig. 1 Conducted EMI test environment and EMI mechanism 3. Modeling of three-phase power converter with diode-bridge as front-end Regarding the investigated object as shown in Fig. 2, there are two types of semiconductors inside: diode and MOSFET. The switching speed of MOSFET is much faster than that of diode since MOSFET is controlled by gate voltage while diode changes its state only according to the voltage across it, which is rather smooth. The main source of noise caused by the switching of the MOSFET EPE'13 ECCE Europe ISBN: and P.2

3 can be modeled by an equivalent voltage source representing the voltage across drain and source of the device. Therefore, if the noise source is identified as the MOSFET and by extracting it from the circuit, the remain part can be considered as propagation path which consists of boost inductor, boost diode, rectifier diode-bridge, output capacitor, output resistor, LISN as well as associated parasitic elements such as stray inductance due to PCB and stray capacitance from each part to earth. EMI receiver is the device to record noise level and should be modeled to represent detector effects. Fig. 2 Converter circuit with front-end three-phase diode-bridge 3.1 Modeling of noise source and receiver The voltage across drain and source points of MOSFET changes between voltages on the two DC buses. Trapezoidal waveform is a well-known practice to represent this voltage and together with the parasitic components like parasitic inductance of copper connection and output capacitance of the MOSFET, ringing is observed which influences the conducted EMI level in high frequency range [4]. With information about gate circuit and datasheet of MOSFET, the slope rate of trapezoidal waveform can be calculated. An illustration of trapezoidal waveform and the induced ringing due to parasitic parameters can be seen in Fig. 3. With trapezoidal waveform for the equivalent source an assumption of perfect filtering with the output capacitors is made. In order to take into account the imperfection of output capacitors, an additional noise source using the current source which has the current form as at inductor is fed into the output stage and the resulting current in the output capacitors is then coupled back to the input stage. Fig. 3 Modeling of noise source with trapezoidal waveform (left: without ringing; right: with ringing) In a standard EMI test environment, LISN and EMI receiver are used. They can be modeled with equivalent circuit in the propagation path and signal processing in, e.g. Matlab. The equivalent circuit of LISN is set according to definition in CISPR [7], see Fig. 4. Signal processing to emulate the detector procedure in EMI receiver is shown in Fig. 5 [8]. EPE'13 ECCE Europe ISBN: and P.3

4 Fig. 4 Equivalent circuit of LISN 3.2 Modeling of propagation path Fig. 5 Signal processing procedure to emulate detector effects Concerning the propagation path, the main difficulty is imposed by the existence of semiconductors in the circuit. The semiconductors introduce both nonlinearity and time-variation. The MOSFET has been replaced by the equivalent noise source and what has more influences on the propagation path is now the diode-bridge. For three-phase diode-bridge, depending on whether the diode is conducting or blocking, there are 12 different possibilities of conduction patterns as illustrated in Fig. 6. These possibilities appear in different spans in the time frame. Considering that for taking the detectors effects into account as explained with Fig. 7, the time domain information is also required. Therefore this variation can not be simply ignored. Fig. 6 Variation of conduction pattern caused by three-phase diode-bridge EPE'13 ECCE Europe ISBN: and P.4

5 Fig. 7 Different detector effects on signal Since FD simulation is, from its essence, only suitable for linear time invariant (LTI) system, method should be used to take the variation into account while enable the simulation to be carried out in FD with consideration of detector effects. As discussed in [9], approximations can be made so that the combined simulation in time and frequency domain approximates the real response of system. Such approximations introduce additional errors and the result is a compromise of simulation speed and accuracy. In order to include the characteristic accurately and enable real frequency domain simulation, in the following a new method is proposed. Firstly, the performance of diode is actually non-linear and the junction capacitance depends on the reverse voltage across. To enforce linearity, the diode is approximated as forward resistance when conducting and a constant capacitance when blocking. To model the time varying characteristic of propagation path, the system is reconstructed as in Fig. 8. It means that the response of time varying system is actually the sum of set of time invariant systems. Therefore, in time domain, the response can be calculated as Eq. 2 while transformed into frequency domain, Eq. 3 can be obtained. Np Fig. 8 Structure of time varying system represented by time invariant system Np Eq. 2 yt () = y () t = G{ ut ()} b() t p p p p= 0 p= 0 Np Np Eq. 3 Ys () = Y() s= [( G() s Us () B()] s p p p p= 0 p= 0 4. Validation of frequency domain prediction With the model described above, the spectrum received at the LISN is calculated. Fast Fourier transformation is used then to get time domain waveform and effects of detectors can be applied to get the output as would be shown on EMI receiver. Simulation result using this method is shown for peak detector together with simulation result from time domain simulation and measurement for EPE'13 ECCE Europe ISBN: and P.5

6 comparison in Fig. 9. A comparison of results with average detector between FD, TD simulation and measurement is provided in Fig. 10. These comparisons confirm the effectiveness of the used method. Compared to the TD simulation, in high frequency range the accuracy of FD simulation is significantly improved. This is because of the use of complex cable model in FD simulation, which is not possible to implement in TD simulation otherwise extreme long simulation time and convergence problem may occur. As an important advantage of FD simulation, the simulation time has been reduced drastically compared to that of TD simulation. The TD simulation time is in the order of several hours for the system because of the small time step used and lots of parasitic components inside the simulation model. The FD simulation time is about 20 minutes. The short simulation time of FD simulation makes it suitable to be used for test of possible measures to improve EMI performance. Fig. 9 Peak detector results of FD (top), TD (middle) simulation and measurement (bottom) EPE'13 ECCE Europe ISBN: and P.6

7 Frequency Domain Prediction of Conducted EMI in Power Converters with front-end Fig. 10 Average detector results of FD (top), TD (middle) simulation and measurement (bottom) 5. Conclusion Prediction of conducted EMI in power converters during developing process necessitates effective methodology. Due to the long simulation time required by TD simulation, attention is turned to FD simulation. This paper has presented the method to simulate conducted EMI accurately, based on analysis and identification of noise source, noise propagation path and noise receiver. A novel approach has been used to model the propagation path in FD by considering it as a linear-time-varying system and thus the variation brought by diode-bridge is taken into account. The method is applied for a test object consisting of three-phase diode-bridge and boost converter. Comparisons of conducted EMI levels using peak and average detector between FD simulation, TD simulation and measurement EPE'13 ECCE Europe ISBN: and P.7

8 validate the effectiveness of used method. The used method, compared to existing methods, helps to improve the accuracy of FD simulation. The fast simulation speed of FD simulation makes it also suitable to be used in developing process for optimization purpose. References [1] L. Qian, W. Fei, and D. Boroyevich, "Modular-Terminal-Behavioral (MTB) Model for Characterizing Switching Module Conducted EMI Generation in Converter Systems," Power Electronics, IEEE Transactions on, vol. 21, pp , [2] J. Wei and D. Gerling, "Prediction of Conducted EMI in Power Converters Using Numerical Methods," presented at the 15th International Power Electronics and Motion Control Conference, EPE-PEMC 2012 ECCE Europe, Novi Sad, Serbia, [3] Y. Liyu, L. Bing, D. Wei, L. Zhiguo, X. Ming, F. C. Lee, and W. G. Odendaal, "Modeling and characterization of a 1 KW CCM PFC converter for conducted EMI prediction," in Applied Power Electronics Conference and Exposition, APEC '04. Nineteenth Annual IEEE, 2004, pp vol.2. [4] L. Jih-Sheng, H. Xudong, E. Pepa, C. Shaotang, and T. W. Nehl, "Inverter EMI modeling and simulation methodologies," Industrial Electronics, IEEE Transactions on, vol. 53, pp , [5] W. J. E. Hoene, M. Michel, H. Reichl, "Evaluation and Prediction of Conducted Electromagnetic Interference Generated by High Power Density Inverters," in the EPE 2001, Graz, Austria, [6] W. Shen, F. Wang, and D. Boroyevich, "Conducted EMI characteristic and its implications to filter design in 3-phase diode front-end converters," in Industry Applications Conference, th IAS Annual Meeting. Conference Record of the 2004 IEEE, 2004, pp vol.3. [7] "CISPR 16-1: Specification for radio disturbance and immunity measuring apparatus and methods," [8] "Spectrum Analysis Basics," Application Note, Agilent Technologies. [9] J. Wei, D. Gerling, and M. Galek, "Prediction of Conducted EMI in Power Converter with front-end in Frequency Domain," in the PCIM Europe, Nuernberg, EPE'13 ECCE Europe ISBN: and P.8