Measurement of the Source Impedance of Conducted Emission Using Mode Separable LISN: Conducted Emission of a Switching Power SupplyJUNICHI MIY ASHITA,1 MASAYUKI MITSUZAW A,1 TOSHIYUKI KARUBE,1KIYOHITO Y AMASAW A,2 and TOSHIRO SA TO21Precision Technology Research Institute of Nagano Prefecture, Japan2Shinshu University, JapanSUMMARYIn the procedure for reducing conducted emissions, it is helpful to know the noise source impedance. This paper presents a method of measuring noise source complex impedances of common and differential mode separately. We propose a line impedance stabilization network (LISN) to measure common and differential mode noise separately without changing LISN impedances of each mode. With this LISN, conducted emissions of each mode are measured inserting appropriate impedances at the equipment under test (EUT) terminal of the LISN. Noise source complex impedances of switching power supply are well calculated from measured results. © 2002 Scripta Technica, Electr Eng Jpn, 139(2): 72 78, 2002; DOI 10.1002/eej.1154Key words:Conducted emission; noise terminal voltage; noise source impedance; line impedance stabiliza-tion network (LISN); EMI.1. IntroductionSwitching power supplies are employed widely in various devices. High-speed on/off operation is accompa-nied by harmonic noise that may cause electromagnetic interference (EMI) with communication devices and other equipment. To prevent the interference, methods of meas-urement and limit values have been set for conducted noise (~30 MHz) and radiated noise (30 to 1000 MHz). Much time and effort are required to contain the noise within the limit values; hence, the efficiency of noise removal tech-niques is an urgent social problem. Understanding of the mechanism behind noise generation and propagation is necessary in order to develop efficient measures. In particu-lar, the propagation of conducted noise must be investi-gated.Modeling and analysis of equivalent circuits have been carried out in order to investigate conducted noise caused by switching [1, 2]. However, the stray capacitance and other circuit parameters of each device must be known in order to develop an equivalent circuit, which is not practicable in the field of noise removal. On the other hand, noise filters and other noise-removal devices do not actually provide the expected effect [3, 4], which is explained by the difference between the static characteristics measured at an impedance of 50 Ω, and the actual impedance. Thus, it is necessary to know the noise source impedance in order to analyze the conducted noise.Regulations on the measurement of noise terminal voltage [5] suggest using LISN; in particular, the vector sum (absolute voltage) of two propagation modes, namely, common mode and differential mode, is measured in terms of the frequency spectrum. Such a measurement, however, does not provide phase data, and propagation modes cannot be separated; therefore, the noise source impedance cannot be derived easily. There are publications dealing with the calculation of the noise source impedance; for example, common mode is only considered as the principal mode, and the absolute value of the noise source impedance for the common mode is found from the ground wire current and ungrounded voltage [6], or mode-separated measure-ment is performed by discrimination between grounded and ungrounded devices [7]. However, measurement of the ground wire current is impossible in the case of domestic single-phase two-line devices. The complex impedance can be found using an impedance analyzer in the nonoperating state, but its value may be different for the operating state. Thus, there is no simple and accurate method of measuring source noise impedance as a complex impedance.© 2002 Scripta TechnicaElectrical Engineering in Japan, V ol. 139, No. 2, 2002Translated from Denki Gakkai Ronbunshi, V ol. 120-D, No. 11, November 2000, pp. 1376 1381The authors assumed that the noise source impedance could be found easily using only a spectrum analyzer, provided that the noise could be measured separately for each mode, and the LISN impedance could be varied. For this purpose, a LISN with a balun transformer was devel-oped to ensure noise measurement, with the common mode and differential mode strictly separated. An appropriate known impedance is inserted at the EUT (equipment under test) terminals, and the noise source impedance is found from the variation of the noise level. This method was used to measure the conducted noise of a switching power sup-ply, and it was confirmed that the noise source impedance could be measured as a complex impedance independently for each mode. Thus, significant information for noiseremoval and propagation mode analysis was acquired.This paper presents a new method of measuring the noise source impedance of conducted emission using mode-separable LISN.2. Separate Measurement for Common Mode andDifferential ModeThe conventional single-phase LISN circuit for measurement of the noise terminal voltage is shown in Fig.1. The power supply is provided with high impedance by a 50-µH reactor, and a meter with an input impedance of 50Ω is connected between one line and the ground via a high-pass capacitor, and another line is terminated by 50 Ω. Thus, the LISN impedance as seen at the EUT is 100 Ω in the differential mode, and 25 Ω in the common mode. The measured value is the vector sum of both modes, and the noise must be found separately in order to find the noise source impedance for each mode. There is LISN with Y-to-delta switching to provide mode separation [8], but its impedance is 150 Ω, giving rise to a problem of data compatibility with 50-Ω LISN. Thus, a new mode-separa-ble LISN was developed as shown in Fig.2. The circuit is identical to that in Fig. 1 from the power supply through the high-pass capacitor. Switching of the connection pattern ensures measurement with one line of the balun transformer terminated by 50 Ω, and another line connected to the meter.In Fig. 2, the secondary side of the 2:1 balun trans-former is terminated by 50 Ω, while the primary side has 200 Ω; in the differential mode, the impedance (line-to-line) is 100 Ω since 200 Ω at the high-pass capacitor is connected in parallel. With the switch set at D, the meter is connected to the secondary side of the balun transformer. The voltage is one-half that of the line-to-line voltage, and measurement is performed in the standard way.The common mode current flows from both sides of the balun transformer via the middle tap to the 50-Ω termi-nal. The currents in the windings are antiphase, and no voltage is generated at the secondary side. Therefore, the impedance of the primary side is the terminal resistance of the tap. Since this impedance is connected in parallel to 50Ω (two 100 Ω in parallel) at the high-pass capacitor, the impedance between the common line and ground is 25 Ω. With the switch set at C, the meter is connected to the middle tap of the balun transformer, and the common-mode voltage is the line-to-ground voltage.3. Measurement of Noise Source Impedance3.1 Measurement circuit and calculationThough the propagation routes are different in the two modes, propagation from the noise source to the LISN can be represented in a simplified way as shown in Fig. 3. In the initial measurement, the load impedance Z L is the LISN impedance. Z L can be varied by inserting a knownimpedance at the EUT terminals. Consider three load im-Fig. 1. Standard 50-Ω/50-µH LISN.Fig. 2.Mode-separable LISN.Fig. 3. Schematic circuit of noise propagation.pedances, namely, LISN only and LISN with two different impedances inserted, Z L 1(R 1 + jX 1), Z L 2(R 2 + jX 2), andZ L 3(R 3+ jX 3). Using the values I 1, I 2, I 3 (scalars) measured in the three cases, Z 0(R 0 + jX 0) is found. Since V 0 = |Z L | × I ,the following expressions can be derived:From the above,Here a , b , and c are as follows:Substituting Eq. (2) into Eq. (1), the following quadratic equation for R 0 is obtained:Thus, R 0 and X 0 have two solutions each. The series of frequency points with positive R 0 is taken as the noise source impedance.3.2 Method of measurementAn impedance is inserted at the EUT terminals in order to measure the noise source impedance in the LISN as seen at the EUT. As shown in Fig. 4, the impedance is inserted so as to vary only the impedance in the mode under consideration, thus preventing an influence on the imped-ance in the other mode. In the diagram, V m is the voltage at the meter connected to the LISN, while the input impedance of the meter (50 Ω) is represented by the parallel resistance.Since parameters of both the LISN and the inserted imped-ance are known, the noise current I can be calculated from V m . Now Z 0 is calculated for each mode from the measured data obtained while varying Z L , by using Eqs. (2) and (3).With the differential mode shown in Fig. 4(a), CR is inserted between the two lines, thus varying the load im-pedance Z L . In the differential mode, Z 0 is assumed to be a low impedance, and hence the inserted impedance exerts a significant effect on the measured value. For this reason, 1Ω/0.47 µF and 0 Ω/0.1 µF were inserted, which are rather small compared to the LISN impedance.The measurement of the common mode shown in Fig.4(b) employs common-mode chokes that basically have no impedance in the differential mode. The common-mode chokes are provided with a secondary winding (ratio 1:1),so that the impedance at the secondary side can be varied.In the common mode, Z 0 is assumed to have a particularly high impedance in the low-frequency band. For this reason,5.1 k Ω and 100 pF were used as the secondary load for the common-mode choke to obtain a high inserted impedance.The measured data for the inserted impedance in the case of resistive and capacitive loads are presented in Fig. 5. The impedance of the common-mode choke includes its own inductance and the secondary load. In the case of a capaci-tive load, the resonance point is around 200 kHz; at higher frequencies, the impedance becomes capacitive.A single-phase two-line switching power supply (an ac adapter for a PC with an input of ac 100 V , a rated power of 45 W, and PWM switching at 73 kHz) was used as the EUT, and the rated load resistance was connected at the dcside. Filters were used for both the common and differential(1)modes, except for the case in which one common-mode choke was removed, in order to obtain the high noise level required for analysis. Both the EUT and the loads had conventional commercial ratings, and were placed 40 cm above a metal ground plate; the power cord was fixed.4. Measurement Results and Discussion The results of conventional measurement as well as common-mode and differential-mode measurement for the LISN without inserted impedance are shown in Fig. 6. The measurements were performed in the range of 150 kHz through 30 MHz, divided into three bands, using a spectrum analyzer with frequency linear sweep. Time-variable data were measured at their highest levels using the Max Hold function of the spectrum analyzer, and only the peak values were employed for calculation of Z 0. For this purpose, the values measured in every frequency band were subjected to the FFT, and all harmonics higher than the fundamental frequency were removed. The data were smoothed, and about 10 peak points were detected in every frequency band. In addition, only those peaks that were stronger than the meter s background noise by at least 6 dB were consid-ered.The results in Figs. 6(b) and 6(c) pertain to the LISN only; the level would vary with inserted impedance. The noise source impedance for both modes calculated from the measured data (using triple measurement) is given in Figs.7 and 9, respectively. The bold and dashed lines pertain to data acquired with the impedance analyzer at the EUT power plug, with the EUT not in operation. With the differ-ential mode, there were no high-frequency components, as shown in Fig. 6(b), and hence the impedance is calculated only for significant low-frequency peaks.The noise source impedance in differential mode can be represented schematically as in Fig. 8. The noise sourceimpedance is equal to the impedance between the LISNFig. 5.Inserted impedance in common mode.Fig. 6. Measured results of standard, differential-mode,and common-mode.Fig. 7. Noise source impedance for differential mode.terminals when the noise source is short-circuited. With switching power supplies, filtering is usually performed by a capacitor of 0.1 to 1 µF inserted between the lines. Since the impedance of the power cord is small in the measured frequency range, one may assume that the impedance as seen at the LISN is low, and that the phase changes from capacitive toward inductive as with the measured static characteristics. However, in the case of the given EUT, a nonlinear resistor was inserted between the power cord and the filter as shown in Fig. 8, and hence the impedance is rather high in the nonoperating state. In addition, there are rectifying diodes on the propagation route, but they do not conduct at the measurement voltage of the impedance ana-lyzer. The noise levels show considerable variation at 120Hz, which corresponds to the on/off frequency of the recti-fying diodes; however, only the peak values are measured and then used for calculation, and hence the impedance obtained by the proposed method is considered to pertain to the conductive state. For this reason, the results do not agree well with static characteristics. Thus, the impedance in the operating state cannot be measured in the differential mode.On the other hand, the measured data for |Z 0| in common mode agree well with the static characteristics, as shown in Fig. 9. The phase, too, exhibits a similar variation,although the scatter is rather large. The resistive part of three load impedances and Z 0 may be presented in a simplified way as in Fig. 10. From Eq. (1), the following is true for R 2,R 3, and Z 0:The distance ratio from Z 0 to R 3 and R 2 on the R X plane that satisfies this equation is I 2:I 3, which corresponds to a circle with radius r as in Eq. (4), with the center lying on the line R 3R 2:Similar circles for R 1 and R 2 are also shown in the diagram.When Z 0 and the load impedances lie on one line, the twocircles have a common point. Equation (4) indicates that if I 3 increases slightly, the outer circle becomes bigger, and the two circles do not adjoin. On the other hand, when the outer circle becomes smaller, the two circles intersect at two points, and X 0 varies more strongly than R 0. In practice, the difference in noise level due to the inserted impedance may drop below 1 dB at some frequencies, so that the solution for Z 0 becomes unavailable because of the scatter, or the phase scatters too much. The measurement accuracy is governed by the difference in noise level, and thus the inserted impedance should have a large enough variation compared to the measurement scatter; in addition, there should be a phase difference so that the two circles are not aligned, as in Fig. 10.Figures 7 and 9 pertain to one of the solutions of Eq.(3) with larger R 0. Here R 0 is not necessarily positive and the other solution is not necessarily negative. The two solutions may be basically discriminated from the fre-quency response and other characteristics, but other inser-tion data are employed for the sake of accuracy.Fig. 8. Equivalent circuit of differential-mode noisesource impedance.(4)Fig. 9.Noise source impedance for common mode.Fig. 10. Load impedances and Z 0 on R X plane.Figure 11 compares the measured data and calculated data for the variation of noise level due to insertion of a commercially available common-mode choke, with the cal-culation based on the results of Fig. 9 and the impedance of the common-mode choke. As is evident, the calculation agrees well with the measured values. On the other hand, a considerable discrepancy was confirmed for the other solu-tion. The noise source impedance found as explained above is accurate enough to predict the filtering effect.The noise source resistance in the common mode can be represented as in Fig. 12. Here Z 1 is the stray capacitance between the internal circuit and the case, and Z 2 is the stray capacitance between the case and the ground plate (or in the case of the ground wire, the impedance of the wire). The common-mode noise source impedance for a single-phase two-line EUT is primarily Z 2, becoming capacitive at low frequencies. Since the EUT is equipped with a filter, the influence of the primary rectifying diodes is not related to common-mode, and hence the data measured by the pro-posed method are very close to the static characteristics.However, this is not necessarily true in the case of a grounded line (Z 2 short-circuited) with no filter installed.In addition, here the full impedance as seen at the LISN is found; in practice, however, a filter or Z 1 is employed to suppress noise. Therefore, the impedance of the power cord is required as well as Z 1 and Z 2 in order to analyze the filtering effect. The impedance of the power cord or grounded wire can be easily determined by measurement or calculation. In our experiments without ground, the impedance is very close to Z 2; on the other hand, Z 1 might be measured by grounding the case and removing the filter (Fig. 12), and then used to analyze the filtering effect between the case and the lines. However, noise propagation in the inner circuit must be further investigated in order to estimate the noise-suppressing efficiency of Z 1.5. ConclusionsA new mode-separable LISN is proposed that sup-ports noise measurement without changing the impedance depending on the mode. The proposed LISN ensures accu-rate measurement for each mode, thus supporting imped-ance analysis.With the proposed LISN, an appropriate impedance is inserted at the EUT terminals, and the noise impedance can be found as a complex impedance, just as simply as with conventional measurement of the noise terminal voltage.The value of the inserted impedance must be chosen prop-erly in order to determine the phase accurately. The pro-posed method ensures sufficient accuracy not only to investigate noise propagation and design efficient counter-measures, but also to predict the filtering effect. The pro-posed technique can supply important data for future analysis of noise generation and propagation in switching power supplies.REFERENCES1.Matsuda H et al. Analysis of common-mode noise in switching power supplies. NEC Tech Rep 1998;51:60 65.2.Ogasawara S et al. Modeling and analysis of high-frequency leak currents generated by voltage-fed PWM inverter. Trans IEE Japan 1995;115-D:77 83.3.Iwasaki M, Ikeda T. Evaluation of noise filters for power supply. Tech Rep IEICE EMCJ 1999;90:1 6.4.Kamita M, Toyama K. A study on attenuation char-acteristics of power filters. Tech Rep IEICE EMCJ 1996;96:45 50.rmation technology equipment Radio distur-bance characteristics Limits and method of meas-urement. CISPR 22, 1997.Fig. 11. V ariation of noise level due to insertion ofanother impedance (measured and calculated data).Fig. 12. Equivalent circuit of common-mode noisesource impedance.6.K amita M, Oka N. Calculation of common-mode noise output impedance during operation. Tech Rep IEICE EMCJ 1998;98:59 65.7.Ran L, Clare C, Bradley K J, Chriistoopoulos C.Measurement of conducted electromagnetic emis-sions in PWM motor drive without the need for an LISN. IEEE Trans EMC 1999;41:50 55.8.Specification for radio disturbance and immunity measuring apparatus and method Part 1: Radio dis-turbance and immunity measuring apparatus. CISPR 16-1, 1993.AUTHORS (from left to right)Junichi Miyashita (member) graduated from Tohoku University in 1981 and joined the Precision Technology Research Institute of Nagano Prefecture. His research interests are EMC measurement and prevention. He is a member of IEICE.Masayuki Mitsuzawa (nonmember) graduated from Nagoya University in 1984 and joined the Precision Technology Research Institute of Nagano Prefecture. His research interests are EMC measurement and prevention. He is a member of JIEP .Toshiyuki Karube (nonmember) graduated from Waseda University in 1991 and joined the Precision Technology Research Institute of Nagano Prefecture. His research interests are EMC measurement and prevention. He is a member of IEICE and JIEP .Kiyohito Yamasawa (member) completed the M.E. program at Tohoku University in 1970. He has been a professor at Shinshu University since 1993. His research interests are magnetic device integration, microswitching power units, and microwave sensors. He holds a D.Eng. degree and is a member of IEICE, SICE, the Magnetics Society of Japan, the Japan AEM Society, and IEEE.Toshiro Sato (member) completed his doctorate at Chiba University in 1989 and joined Toshiba Research Institute. He has been an associate professor at Shinshu University since 1996. His research interests are magnetic thin-film devices. He received a 1994 IEE Japan Paper Award and a 1999 Japan Society of Applied Magnetism Paper Award. He holds a D.Sc. degree,and is a member of IEE Japan, IEICE, and the Magnetics Society of Japan.。