SIMULATION of EMC PERFORMANCE of GRID CONNECTED PV INVERTERS Qin Jiang School of Communications & Informatics Victoria University P.O. Box 14428, Melbourne City MC 8001 Australia Email: jq@sci.vu.edu.au James Brown Power Solutions Australia Pty Ltd 3-6 Holloway Drive Bayswater, Vic 3153 Australia Email: james@powersolution.com.au Abstract This paper looks at the fundamental factors determining the EMI generated by Pulse Width Modulated (PWM) power inverters in the frequency range up to several MHz. It uses simple simulations of typical circuits to show the influence of several converter topologies and the effects of various components on EMI performance, particularly the common mode components which are responsible for many of the interference problems experienced. The objective is to compare the relative performance of different systems, not to achieve absolute accuracy. The paper is intended as a tutorial to help understanding and provide a perspective for both novices and practitioners. 1. INTRODUCTION Because of rapid changes in voltages and currents within a switching converter, power electronic equipment is a source of Electromagnetic Interference (EMI). Legislation and regulations require interference generated by equipment to be limited to relevant standards. However, particularly in domestic applications, unique installation factors can result in fully compliant products not meeting end users expectations. Coupling into AM radio, television, audio and telephone systems can occur and generate unacceptable interference. This applies not only to grid connected photovoltaic (PV) installations but also to inverters in stand alone remote area power systems (RAPS). Lots of research efforts have been invested in the Electromagnetic Compatibility (EMC) of AC drives [1-3] and switching mode power supply [4-5]. However, not much can be found in literature for the EMC performance of power inverters such as grid connected PV inverters and stand alone remote area power systems. Those power inverters are unique in several ways: They inherently involve high power levels, kilo Watt rather than milli Watts, Large levels of power are present in the range of 10k to 100kHz, though lower and higher switching frequencies are often used. They have two sets of external high power terminals (most end user equipment only has a high power input from the mains). The high power dc and ac wiring is inherently separated and becomes a very effective antenna to radiate into free space and to couple into other systems. There are large components such as transformers and filter components which are inherently required, but which can be exploited to optimise EMI. Because of the high power levels there are additional and significant EMI filter components where their size and cost must be optimised. In contrast in many non-power electronic systems EMI can be controlled largely by good PCB layout plus some lowcost low-power components. Inverters operates in all four quadrants but at different times over the low power frequency cycle. Simulation must include all modes, so generally a full cycle of operation must be simulated with data generated at a time resolution sufficient to accurately represent the highest frequencies of interest. In this paper, a voltage source PWM full bridge inverter as shown in Fig.1 is discussed. Fig.1 (a) shows an ideal full bridge with a load R L and an output filter L f /C f to attenuate differential mode (DM) noise at the output terminals. The circuit of Fig.1(b) is a practical inverter including the stray impedance (C2, L1, L2), the common mode impedance (Z 1, Z 2, C1), the shunt capacitor (C sh ) and the damping resistor R 1. In addition, an isolating transformer and a common mode (CM) choke, at the dc and ac side respectively, will be added to the full bridge (Fig.2). Their effect on the EMC performance of the inverter is investigated.
(a) (b) Fig.1: (a) An ideal full bridge inverter. (b) A realistic full bridge inverter. 2. BACKGROUND EMI is transmitted in two forms: radiated and conducted. Inverters connected to the power lines generate conducted noise into both ac and dc power wiring, which is usually several orders of magnitude higher (in terms of power) than the direct radiated noise. Metal cabinets normally used for housing power converters effectively shield the direct radiated component of the EMI. Differential and common mode noise limits in standards are very similar but because differential mode noise currents inherently flow in opposite directions in two closely coupled conductors the coupling into other systems is modest. In a real installation the common mode noise is of greater significance because it flows through separated conductors and through a much larger path which operates as a very effective antenna. The dominant component of EMI is the harmonics associated with switching of power devices at PWM carrier frequencies. For small converters carrier (modulation) frequencies are generally above 20 khz though in large converters frequencies down to several khz may be used. DM noise is inherently generated due to the PWM process. The desired output is the fundamental frequency used to modulate the carrier, the carrier and its harmonics must be filtered out to leave an acceptable fundamental waveform at the ac output. Additional noise is generated due to reverse recovery of diodes and other artefacts of switching. This can generate more power at high frequencies than the normal commutation processes. Various factors influence DM noise levels The attenuation in the range 10kHz to 500kHz is important because of the high levels of carrier harmonics in that band and the limited attenuation that can be achieved in realistic filters before the common limits start to apply, typically at 150kHz. The effective attenuation of typical filters increases with the frequency range at 40 db per decade but in a single stage, the attenuation is limited at high frequencies by stray shunt capacitance on inductors and series inductance of capacitors. Cascaded filters can always achieve the required attenuation but attention to component design, construction, placement, wiring and the use of shielding can improve maximum attenuation levels of single stage filters Ringing at high frequencies due to switching transients exciting parasitic resonances in the power circuits or the filters can produce spikes of energy up to tens of MHz. Differential mode noise is also present on the dc input and filtering is required here as well. For the inverter in Fig.1, the DM noise source is voltage source type seen from the ac side, hence the series not the parallel filter is effective. The attenuation of the DM noise by the series output filter L f and C f degrades by the stray shunt capacitor C sh on the output filter L f and the series inductor of C f as shown in Fig.1 (b).
CM choke Transformer CM choke R L Fig.2: A realistic full bridge inverter including a transformer and two CM chokes. Common-mode noise is generated through several mechanisms The pulse width modulation process can produce large common mode voltages. Consider a standard single phase bridge with normal modulation patterns. At the carrier frequency one of the ac output terminals is connected alternately to the positive and negative dc input terminals. This generates a peak to peak common mode input voltage which can be as large as the dc bus voltage. Shunt capacitance results in currents flowing due to rapidly changing voltages. The amplitude can be reduced by controlling stray capacitance and the coupling to the output by controlling the paths through which transient currents flow and the voltages generated. Series inductance results in large voltages appearing in paths which have rapidly changing currents. Again the impact of this can be controlled by manipulating the coupling to the output terminals. Unbalances in the circuit and stray components can convert (couple or transform) what appears to be a differential mode signal into a common mode signal. 3. THE SIMULATION METHOD In the lower frequency ranges (up to tens of MHz) modelling of power converter equipment using discrete components can effectively represent the EMI performance and give insights into the role of the factors determining EMI performance. In this region (as in all others) EMI performance is the result of specific factors which if understood can help achieve designed levels of performance rather than relying entirely on empirical methods. practical inverter of Fig.1(b). The circuit is used to describe three configurations concerned in this paper. In each case, only one of the three added components is present, the common mode voltages of the resulting system, and as shown, are simulated accordingly. Two stages of simulation are performed, at the circuit level in the time-domain using the PSIM software package to generate a set of waveforms covering a full cycle time with a time resolution of 0.1us allowing spectral components up to 5 MHz to be estimated. This is then used to generate a spectral analysis in the frequency-domain using MatLab to perform an FFT. There are several differences in the results generated from this simple approach Switching devices in the PSIM simulation are represented by ideal switches, no diode recovery or dead time is represented. Damping at high frequencies is negligible. The representation of the circuit particularly the damping of the various natural modes of ringing has a large effect on the energy actually output to the terminals, damping at low frequencies has been provided but reasonable representation of high frequency damping has not been included. This can be seen in the lightly damped ringing shown in some of the time responses below. The spectrum measured in a real conducted measurement test is determined by the response of the measuring instrument (its bandwidth and detector characteristics ) as well as the impedance presented by the line impedance stabilising network (LISN), which is specified in the relevant standards. In Fig.2, an isolating transformer and a CM choke at In this simulation the spectrum produced by the FFT, the dc and ac sides respectively are added to the which measures the energy in a very narrow
Fig.3: Spectra of case 1, no isolating transformer, no dc or ac common mode chokes. bandwidth is used as a measure. This is a fixed nontime variant voltage value so there is no issue relating to peak, quasi-peak or average measurement. The common mode and differential mode voltages are defined by the impedances of the circuit under examination which have been chosen to be typical for a practical system. Fig.4: Spectra of case 2, the ac choke present alone. 4. SIMULATION RESULTS In Fig.2 the dc supply voltage is 300V, the regular sampled PWM is used, with an modulation index of 0.8 and a carrier frequency of 5kHz, so that the fundamental ac output is around 240V. Other parameters of Fig.2 are given as follows: Stray impedance: L1=0.1µH, C2 = C sh = 100pF Output Filter: L f = 5mH, C f =1µF CM choke: L C =1mH, coupling coefficient 0.99 DC side: R d = 0.1 Ω, C d =5mF Damping resistor: R 1 = 1kΩ Measurement resistor: R 2 = 150Ω Transformer: Voltage ratio 1:1, ideal, C t =100pF CM impedance: Z 1 =1µF / 0.1µH, Z 2 =47nF/0.2µH, C 1 =100pF Where Z 1 and Z 2 are capacitors with an internal series inductance. Total four configurations are simulated as follows: Case 1: the practical inverter of Fig.1(b), Case 2: the practical inverter and a choke at ac side, Case 3: the practical inverter and a choke at dc side, Case 4: the practical inverter and a 50Hz transformer. The CM noise level spectra at the dc side,, and the ac side,, for the four configurations are presented in Figures 3 to 6. Fig.5: Spectra of case 3, the dc choke alone present Fig.6: Spectra of case 4, only the transformer is present.
I CM + - I CM + - Fig.7: The equivalent circuit of CM components of Fig.2 The vertical scale is 20log 10 (V/µV) db and the frequency range is10khz to 5MHz. Also shown are the limits set down in Australian EMC standards 1044 and 2064, extrapolated to lower frequencies than the normal 150kHz limit. Fig.3 shows the simulation spectra of case 1, it is used as a reference for other cases. It can be seen that CM noise level just met the limits above 150kHz, and exceeds the limits below 150kHz where the carrier multiples are dominant. Fig.4 shows the simulation results of case 2 when a choke is present at the ac side, the transformer and the CM choke at the dc side are removed from Fig.2. As a result, the CM noise at the ac terminal is suppressed effectively above 150kHz, but insignificant at the carrier multiples. The EMC improvement at the dc side as a whole is insignificant. Fig.5 shows the simulation results of case 3, where the dc side CM choke alone is present, the transformer and the CM choke at the ac terminal are removed from Fig.2. It can be seen that when the choke is shifted to the dc side, the CM noise at the dc side is attenuated effectively above 100kHz, while the attenuation at the ac side by the dc choke is insignificant. the transformer the carrier component in is up to the dc bus voltage of 300 V (upper trace), and it drops to 1.5 V (lower trace) when the transformer is installed. At the dc side of the inverter, however, the CM noise source is a high impedance current source type, hence the parallel filter is effective. The bridge shunt capacitor, C1, acts as a low pass filter and is effective in suppress the high frequency harmonics. Fig.9 shows the spectra of case 3 when a CM choke is present at the dc side and C1, the local dc bus bypass capacitors are increased from 100pF in Fig.3 to 0.1µF. The reduction of CM in is significant in the high frequency range, but marginal in the low frequency range. As expected, the attenuation is insignificant for all frequency ranges at the ac side. Finally, the spectra of DM noise is simulated for case 1 (no transformer) and case 4 (with transformer) as shown in Fig.10. It can be seen that unlike the CM noise, the addition of the transformer to the bridge inverter brings little improvement in the DM noise. Fig.6 shows the simulation results of case 4, where the isolating transformer alone is present. As the inter winding capacitance C t is in series with the common mode source impedance, it acts as a high pass filter, the low order carrier harmonics are attenuated effectively in both and. These simulation results can be explained by the equivalent circuit of the common mode components as shown in Fig.7. Basically, seen from the ac side the noise source is a low impedance voltage source type, hence the series filter is effective. The output filter, L f, in series with the CM choke, L C, acts as a low pass filter, reducing high frequency EMI. While the interwinding capacitance of the transformer acts as a high pass filter reducing the low frequency carrier components. For example, as shown in Fig. 8, without Fig.8: CM noise at the ac output.without isolating transformer (top) and with an isolating transformer (bottom)
Without T With T Fig.9: Spectra of case 3 at C1= 0.1µF. Fig. 10: Spectra of the DM noise (without/with T) 5. CONCLUSIONS It has been shown that simple simulation methods can give useful insight into the effect of inverter topology and possible configuration of EMI filters. Even converters without isolation and with large inherent common mode noise can provide satisfactorily EMI performance if adequate common mode filtering is provided. The attenuation of large common mode voltages in non-isolated configurations increase losses because some common mode current must flow and the energy is dissipated in the common mode filters. The addition of an isolating transformer either at power frequency on the output (or internally at high frequency, not shown) inserts a low value capacitor in series with the common mode source. This results in low levels of common mode at carrier frequencies and reduces the common mode filtering required at low frequencies. This avoids large low frequency common mode currents and the resultant losses. Because bridge dc busses can be bypassed to ground to reduce common mode voltage simple common mode filtering on the dc side can still achieve good EMI performance. Bypassing bridge dc busses to ground does have a slight negative effect on ac common mode but there are additional advantages in helping reduce current paths for other switching transients. These results and observations apply generally to other power converter applications including adjustable speed drives (ASDs) where common mode noise is known to produce problems. These simulations are easily performed with standard tools and should be part of the design of any power converter to help understand the options available to achieve EMI targets. 6. REFERENCES [1] E. Zhong & T. A. Lipo, Improvements in EMC Performance of Inverter-fed Motor Drives, IEEE Transactions on Industry Applications. Vol.31. No. 6. November/December 1995, pp.1247-1256. [2] A. Kempski, R. Smolenski, & R. Strzelecki, Common Mode Current Paths and Their Modeling in PWM Inverter-Fed Drives, Proceedings of 33 rd IEEE Power Electronics Specialists Conference, Cairns, Australia, 2002. [3] H. Akagi, H. Hasegawa, and T. Doumoto, Design and Performance of a Passive EMI Filter for Use with a Voltage-Source PWM Inverter Having Sinusoidal Output Voltage and Zero Common-Mode Voltage, 33 rd Proceedings of IEEE Power Electronics Specialists Conference, Cairns, Australia, 2002. [4] M.H.Nagrial and A.Hellany, EMI/EMC Issues in Switch Mode Power Supplies, Proceedings of IEE EMC Conference Youk 99, pp.180-185. [5] P.R.Mugur, J.Roudet, and J. Crebier, Power Electronic Converter EMC Analysis through State Variable Approach Techniques, IEEE Transactions on Electromagnetic Compatibility Vol.43, No.2, May 2001.