Computerized Conducted EMI Filter Design System Using LabVIEW and Its Application

Transcription

1 Proc. Natl. Sci. Counc. ROC(A) Vol. 25, No. 3, pp Computerized Conducted EMI Filter Design System Using LabVIEW and Its Application CHIA-NAN CHANG, HUI-KANG TENG, JUN-YUAN CHEN, AND HUANG-JEN CHIU Department of Electronic Engineering National Taiwan University of Science and Technology Taipei, Taiwan, R.O.C. (Received May 1, 2000; Accepted August 7, 2000) ABSTRACT A novel computerized system for obtaining the conducted EMI measurement and systematic filter design of a switched-mode power supply has been developed. The measurement, control and filter design of the conducted EMI noises of the tested device were obtained and integrated using an automatic data acquisition system and a comprehensive virtual instrument, which was developed using the LabVIEW application software program. As an application of this system, conducted EMI noise measurement and filter design of a boost ac-dc converter with PFC (94 khz, 100 W, 200 V) has been achieved while successfully satisfying the FCC Class A limit in the frequency range from 450 khz to 30 MHz, which confirms the validity of the developed computerized system. This computerized system has the advantages of providing accurate measurement of different conducted EMI noises and fast determination of the associated filter corner frequencies and component values. Modification and enhancement of this computerized system are also easy due to the block diagram form of the source code in the virtual instrument. Key Words: conducted EMI, accurate measurement, systematic filter design, LabVIEW software, application I. Introduction The advent and widespread use of switched-mode power supplies (SMPSs) has attracted many researchers to the study of conducted electromagnetic interference (EMI) conducted into power mains (Guo et al., 1993; Tihanyi, 1995; Chen and Lee, 1996; Chang et al., 1998; Ran et al., 1999). Most SMPSs that are intended to be marketed in many countries must satisfy some limits on conducted emission, such as the FCC limits in America and the V.D.E. limits in European countries. The frequency range of the conducted emission limits is from 450 khz to 30 MHz for the FCC (Class A) regulations and 10 khz to 30 MHz for the VDE regulations. In a conducted EMI emission test, the measured emission is, in general, a mixture of both differential-mode (DM) and common-mode (CM) noises. These two modes of noises come from different sources and must be dealt with separately in the conducted EMI filter. To minimize the trial-and-error guesswork involved in selecting the suitable components for a proper conducted EMI filter and to reduce the design cost, it is necessary to develop a systematic and accurate system for practical EMI filter design. Recently, two important advances in this regard have been reported. One is the development of a passive Noise Separator that is constructed mainly from a 0 degree power combiner and a 180 degree power combiner (Guo et al., 1993; Chen and Lee, 1996). By using such a Noise Separator, the noise spectrums of both the DM and CM modes can be separately measured and diagnosed. Another advance is a EMI filter design approach which does not consider the exact magnitude of the source impedance of both DM and CM conducted noises (Shih et al., 1996). LabVIEW, the abbreviation of Laboratory Virtual Engineering Workbench, is a program development system and is very similar to the commercial C or Basic development software programs. In contrast to a text-based language that uses lines of code in an application program, LabVIEW uses a graphical programming language, G, to create application programs in a block diagram form, which is easy to modify. Under Windows, LabVIEW contains both application specific libraries for data acquisition and instrument control and application-specific libraries for GBIP and serial instrument control; thus, LabVIEW has been widely used in various highlevel laboratories for both research and education (Buckman, 1997). An effective system for systematic conducted EMI filter design of SMPSs should have the following three characteristics: (1) accurate measurement of different conducted noises from the tested device with external perturbations reduced to a minimum, (2) a simple model for designing the conducted EMI filter that can neglect the effect of the complicated noise source impedance in the associated SMPS, and (3) efficient determination of EMI filter components from measured data without the need to analyze the measured noises artificially. In this work, a novel computerized system which uses LabVIEW for systematic conducted EMI filter design of SMPS is reported. This computerized system can provide both accurate noise 185

2 C.N. Chang et al. measurement and efficient determination of EMI filter components from measured data, and also adopts a simple conducted EMI filter design model that can ignore the effect of noise source impedance. In this paper, the configuration of this automatic system and the noise sources in the tested device will be described first. The related theory for designing the associated software and the flowchart and pictorial software for systematic conducted EMI filter design will then be described. Finally, the application of this computerized system to the EMI filter design of a boost ac-dc converter (94 khz, 100 W, 200 V) with power factor corrector (PFC) will be reported and discussed, based on which the advantages of the present system for conducted EMI noise measurement and filter design will be shown. II. Computerized System and Conducted EMI Noise Sources Figure 1 shows the setup of the present system for systematic EMI filter design of the device under test (DUT), which may be any kind of switched-mode power supply. In Fig. 1, the device LISN is a line impedance stabilizing network (HP M-3852/2) that can both provide ac power from Power Mains for DUT and keep Power Mains from contaminating the conducted EMI generated by DUT. The passive Noise Separator is employed to separate the CM noise and DM noise from the incoming conducted EMI noises, and is constructed mainly by means of a 0 degree RF power combiner (ZFSC ) and a 180 degree RF power combiner (ZFSC ) provided by Mini-Circuits. A selective switch is inserted in the Noise Separator for the purpose of choosing either DM noise or CM noise during measurement. Also, by employing a by-pass switch, the total EMI noises from LISN can fully pass through the Noise Separator. The Spectrum Analyzer (HP8591E) is used to carry out measurement of the conducted EMI noises coming from the Noise Separator, using an embedded software program that employs standards for measuring conducted EMI. The input impedance of both the Spectrum Analyzer and the Noise Separator is 50 ohms, which is necessary to keep the conducted EMI noises from reflecting during measurement. Through the combined action of both LabVIEW run on a 586 PC and the IEEE-488 interface between the associated instruments, the measured conducted EMI data from the Spectrum Analyzer can be both displayed on the screen of PC and saved in a data file for further analysis and EMI filter design. Fig. 1. Computerized system for conducted EMI measurement and systematic filter design. Fig. 2. Circuit diagram of the boost ac-dc converter under test in which dashed-line block is the place where the EMI filter is inserted. The ac-dc boost converter with PFC depicted in Fig. 2 is the device under test in the present application, where the dashed-line block is the place where a proper conducted EMI filter will be inserted. In Fig. 2, the full-wave diode bridge rectifies the ac commercial power wave-form and produces a pulsating dc wave-form, which is then smoothed by the bulk capacitor C. The sensing resistor R s initializes the forward current-feeding circuit for power factor correction using its active switch Ql (K2194) controlled by the integrated circuit UC3852. This integrated circuit implements zero-currentswitched boost conversion, thus producing a sinusoidal input current that is almost in phase with the rectified input voltage and, hence, raising the power factor of the device. This boost converter provides a constant output voltage for the load through the action of a feedback loop that consists of a voltage divider at the output circuit. Because of the variable switching frequency control of UC3852 and the asynchronous behavior between the PFC circuit and the switching power supply, the conducted EMI problem in this ac-dc boost converter is serious, and it is a challenge to design a proper filter to reduce the conducted EMI noise to a required low level. The main specifications of this boost converter are: input voltage 110 V, switching frequency 94 khz, power factor 0.95, output voltage 200 V, and output power 100 W. The main source of the CM noise in the ac-dc boost converter is the capacitance of the switch Ql (transistor: K2194) shown in Fig. 2. When this transistor is operating as a switch in the boost converter, the high frequency collector voltage swing in transistor causes charging and discharging of the transistor insulator capacitance, resulting in CM noise currents flowing out through live (L) and neutral (N) wires and returning via the ground (G) wire (see L, N, and G wire in Fig. 1). Those parts of the boost converter which are mounted on its chassis via an insulator (parasitic) capacitance and have an ac wave 186

3 Computerized EMI Filter Design System form flowing through them also contribute CM noise. The main source of the DM noise current in the ac-dc boost converter comes from the inductance of the input inductor Li (A shown in Fig. 2), which produces many switching harmonics currents (Chang et al., 1998). The unfiltered portion of the transistor (switch Q1) pulsation current flowing through the 50 ohm resistor of LISN contribute some of the DM noise currents, which flow out through L (N) wire and return via N (L) wire. In addition, the reverse recovery of the bridge diodes can also affect the magnitude and spectral content of the DM noise current (Nave, 1989). III. Related Theory and Procedure for Developing Software The time average DM noise currents, I dm, in the L and N wires are equal in magnitude and are directed in opposite directions, but the time average CM noise currents, I cm in the L wire and N wire are equal and are directed in the same direction. Thus, the output conducted noise voltages from LISN in Fig. l are V live = R(I cm + I dm ) = V cm + V dm, V neutral = R(I cm I dm ) = V cm V dm, (1) where R = 50 ohms is the common impedance between the L and G wires and between the N and G wires in LISN. If the input power from the L and N wires is denoted as P L and P N, respectively, then the output power P o of the Noise Separator shown in Fig. 1 from the o degree power combiner can be expressed as (Guo et al., 1993) P o = (P L + P N )/2 + P L P N. (2) By inserting P L = R (I cm + I dm ) 2 and P N = R (I cm I dm ) 2 into Eq. (2), the output power, after some management, becomes P o = 2RI cm 2 = ( 2V cm ) 2 /R. (3) Similarly, the output power of the Noise Separator from the passive 180 degree power combiner of LISN is P o = (P L + P N )/2 P L P N = ( 2 V dm ) 2 /R. (4) For conducted EMI measurement, micro-voltage (µv) is commonly used as the reference unit (Tihanyi, 1995). From Eqs. (3) and (4), the output conducted noise voltage from the Noise Separator and measured by the Spectrum Analyzer in relative units of dbµv with constant input impedance R is 3 db above the actual value, which must be subtracted from the measured data by the developed software. The first step that must be considered when designing software for systematic conducted EMI filter design is accurate measurement of the base-line (i.e., without filter) common-mode EMI noise spectrum,v cm, and differential-mode EMI noise spectrum,v dm, of the device under test by means of the measurement system shown in Fig. 1. The second step is to determine the required CM-noise attenuation (V req-cm ) db and DM-noise attenuation (V req-dm ) db at various sampled frequencies, which can be done by means of the computation loop in the software design using the following two equations: (V req-cm ) db = (V cm ) db (V lim ) db + 6 db, (V req-dm ) db = (V dm ) db (V lim ) db + 6 db, (5) where (V cm ) db and (V dm ) db are the base-line noise voltages from the first step, and (V lim ) db is the required conducted EMI limit specified by FCC or VDE. To avoid design error, +6 db is added to Eq. (5) because both the measured DM noise and CM noise using the present system are 3 db above the actual values, and because the measured CM and DM noise voltages may be in phase, which will cause a total error of 6 db in estimating the required attenuation. The third step in designing the software is to determine the corner frequencies of the required conducted L-C filter by searching for the minimum values of f c-cm and f c-dm from the required attenuation for all sampled frequencies using the following equations: (V req-cm ) db = 40 log 10 (f / f c-cm ), (V req-dm ) db = 40 log 10 (f / f c-dm ), (6) where (f c-cm ) and (f c-dm ) are the corner frequency variables of the CM and DM noise filter in the required conducted EMI filter, respectively, and (V req-cm ) db and (V req-dm ) db are the required CM and DM attenuation voltages in Eq. (5) at the sampled frequency f. Equation (6) will be explained at the end of this section. The final step in designing the software is to determine the inductor and capacitor component values (L CM, C CM ) and (L DM, C DM ) of the conducted EMI filter from the corner frequencies (f c-cm ) and (f c-dm ) found in the third step using the following two equations: f c cm = 1/(2π L CM C CM ), f c dm = 1/(2π L DM C DM ). (7) There is freedom in choosing the proper values of the four filter components L CM, C CM, L DM and C DM in Eq. (7). When developing the software, the small capacitance C CM in common-mode filter is chosen first due to the small leakage current limit on the grounded capacitor imposed by IEC regulation. The chosen value of C CM in the developed software 187

4 C.N. Chang et al. Fig. 3. (a) Conducted EMI filter topology adopted here, (b) CM noise equivalent circuit of (a), and (c) the DM noise equivalent circuit of (a). is 4700 pf, and the corresponding inductance L CM is then determined from the derived f c-cm. When the inductance L DM is related to the leakage inductance of the common-mode inductor, it is determined first in the differential-mode filter, and the corresponding C DM is then determined from the derived f c-dm. In practice, the predicted value of C DM can be adjusted so the resulting attenuation in the EMI filter will be a little higher than the required attenuation derived from Eq. (5) because of the need to obtain suitable commercial CM capacitors. Different types of conducted noise should be dealt with by different parts of a conducted EMI filter as indicated by Eq. (5). The conducted EMI filter configuration adopted to develop the present software is shown in Fig. 3(a), where L c is a common-mode choke and L d is a differential-mode choke, C x1 and C x2 are DM capacitors (called X capacitors), and C y is a common-mode capacitor (called a Y capacitor). In contrast to the two opposite windings in the differential-mode choke L d, the common-mode choke L c shown in Fig. 3(a) has two identical windings that are wound on the same core. It is found that the leakage inductance due to the coupling imperfection of a practical CM choke also has a filtering effect on the DM noise. In practice, the magnitude of leakage inductance in a CM choke is usually about 0.5% to 2% of the CM inductance (Nave, 1989; Ott and Nave, 1991). Shown in Fig. 3(b) and (c) are the EMI filter equivalent circuits for the CM noise and DM noise, respectively, which take into account the effect of the 50 ohms input impedance of the Spectrum Analyzer. In Fig. 3(b), the CM noise is affected only by the parallel effects of both Y capacitors and the two CM inductors. In Fig. 3(c), both the inductance L d of the DM choke and the leakage inductance L l of the CM choke can attenuate the DM noise: Although the two Y capacitors also affect the DM noise, their effect on DM noise attenuation is negligible in comparison with that of the two Y capacitors with large capacitance. To reduce the EMI filter design cost and size, the effect of the DM inductance L d shown in Fig. 3(c) can be totally replaced by the leakage inductance L l of the CM choke. Aside from the EMI filter components, the effectiveness of a conducted EMI filter also depends on the noise source impedance. For a switched-mode power supply, it has been observed (Ott and Nave, 1991) that the CM noise source can be modeled by a high impedance Z pc in parallel with a current source, and that the DM noise source can also be modeled by a high impedance Z pd in parallel with a current source when the rectifier diodes are on and by a low impedance Z sd in series with a voltage source when the diodes are off. The DM noise equivalent circuit fluctuates between these two models every 2x (line frequency); and it is difficult to distinguish each individual contribution to the total DM noise. Thus, it is troublesome to design a conducted EMI filter if the effect of the conducted noise source impedance is not negligible. In order to adopt a simple model for designing a conducted EMI filter that can ignore the effect of noise source impedance, the choice of the component value in the EMI filter in the present study must satisfy the following conditions: (1) for a CM filter equivalent circuit (Fig. 3(b)), 1/(2ωC y ) << Z pc, ω(l c + L d /2) >> 25 ohms; (8) (2) for a DM filter equivalent circuit with C x1 = C x2 = C x (Fig. 3(c)), (i) if the rectifier diodes are off: 100 Ω >> (1/ωC x ) >> Z sd ; (9) (ii) if the rectifier diodes are on: ω(l d ) >> 100 Ω, Z pd >> (l/ωc x ) >> 100 Ω, where ω is the angular frequency of the CM or DM noise. When the above impedance conditions are met, the equivalent filter circuits for both the CM and DM noises shown in Fig. 3(b) and (c) can be simplified to the two L-C filters depicted in Fig. 4(a) and (b) by using the circuit reciprocity theorem (Shih et al., 1996). The ratio of the output power P 0 from a noise source without an L-C filter to the output power P 0 from the same noise source with an L-C filter is defined as the attenuation (A) of this L-C filter. Since the output power is proportional to the square of the output voltage for the same load impedance, the attenuation A of an L-C filter in decibles (dbs) can be expressed as follows: A(dB) = 10 log 10 (P 0 /P 0 ) = 20 log 10 ( V 0 /V 0 ), (10) 188

5 Computerized EMI Filter Design System where V 0 and V 0 are the output noise voltage without and with an L-C filter, respectively. For high-frequency conducted noise with frequency f >> f c, where f c = 1/(2π LC ), it can be proved by means of circuit theory (Nave, 1991) that the attenuation A of a second order L-C filter in Eq. (10) can be simplified to A(dB) = 20 log 10 ( 1 (f / f c ) 2 ) 40 log 10 (f / f c ).(11) Equation (11) is the basis for using Eq. (6). When the required attenuation versus sampled frequency curves for both CM and DM noises are all below the 40 db/decade slope lines at frequencies far above their respective corner frequencies, f c-cm and f c-dm in Eq. (6), the total conducted noises of the switched-mode power supply under test can be attenuated to the required low level by the CM and DM filters. Thus, the two corner frequencies f c-cm and f c-dm in Eq. (6) correspond to the minimum intersections of the 40 db/decade slope line along the frequency axis, which can be computationally obtained through a searching loop when designing the software. IV. Software in LabVlEW Based on the theory and procedures described in Section III, a package of pictorial programs using LabVlEW for systematic conducted EMI filter design has been developed. The developed main program consists of a user interface (called the front panel) and a source code (called the block diagram). In Fig. 5, the flowchart of this main program that involves the actions of both the source code and user interface is depicted. Shown in Fig. 6 is the developed source code (block diagram) of the main program, and shown in Fig. 7(a) is its corresponding user interface (front panel) for common-mode EMI noise voltage filter design. Nodes, Terminals, and Wires are the three primary objects in Fig. 6. Nodes are the execution Fig. 5. Flowchart of the main program for systematic conducted EMI filter design. Fig. 4. Equivalent second order EMI filter circuit and resulting attenuation in the present design model for (a) CM noise and (b) DM noise. elements, such as functions, case statements (true or false cases), and for loops. Terminals are the entry and exit ports of execution elements. Data can exit from a source terminal and enter a sink (destination) terminal. Wires are the data paths between source and sink terminals, which are of different types for different kinds of flowing data. The five subroutines labeled as S 1, S 2, S 3, S 4, and S 5 on the left hand side of the source code are used to implement data acquisition of the baseline (i.e., without an EMI filter) noise voltages obtained from the measurement system. In order to reduce the errors caused by lightning or abnormal voltage surges, the measured noise voltage for each sampled frequency between 450 khz and 30 MHz is the average of ten measurements, which is indicated 189

6 C.N. Chang et al. Fig. 6. Source code developed using LabVlEW. by the ten arrow terminals and wires at the place labeled F 2 in the upper-left part of Fig. 6. The specified time interval for each 401 (the total number of sampled frequency) measurements is set to be 0.5 second using a 586 PC with 200 MHz Ram in this source code. The measured base-line noise voltage versus frequency result is displayed in the front panel (Fig. 7(a)) through the action of the wave-form graph function icon in the upper-right corner labeled F 4. The procedure used to obtain the required CM attenuation requirement (V req-cm ) db and DM attenuation requirement (V req-dm ) db for each sampled frequency is carried out by the block that lies between the symbols F 4 and F 6 on the right side of Fig. 6, in which the subtract function icon and the constant terminal (45 dbµv) are used to carry out the associated calculation specified by Eq. (5). The node of the for loop in the lower-left corner of Fig. 6 below the symbol F 6 provides the executive element for obtaining the corner frequency of a conducted EMI filter. Based on the inverse functions of Eq. (6), the minimum frequency f min among the 401 sampled frequencies in the chosen frequency range can be searched for through iterative comparison. By means of Eq. (7), the node block labeled F 8 in the upper-right corner of the source code is used to execute the procedure for determining the filter component values C x and L dm from the derived corner frequency of the CM filter under the False case structure. If the True case structure is chosen, another node block labeled F 7 that determines the filter component values C y and L cm for the CM filter will be present. Shown in Fig. 7(a), (b) and (c) are the user interfaces (front panels) of the main program corresponding to the baseline conducted noise measurement and filter analysis of CM, DM and total conducted EMI noise, respectively. These user interfaces are primarily combinations of controls and indicators. Controls simulate instrument input devices and supply data to the block diagram of the source code while indicators simulate instrument output devices that display data generated by the node blocks of the source code. The object in the upperleft corner of Fig. 7 is the serial instrument control that can determine the kind of conducted EMI noise being transferred from the source code and measurement system, and the object in the upper-right corner is the input control that can determine the kind of conducted EMI limits being chosen. The waveform graphs and the rectangular objects that display maximum and minimum noise voltages, corner frequencies, and filter component values in Fig. 7 are all output indicators that receive data from various parts of the source code shown in Fig. 6. V. Application Result and Discussion Application of the developed computerized system to the systematic conducted EMI filter design of the ac-dc boost converter shown in Fig. 2 will be described next. In this application, the FCC limit (class A) is adopted, which requires an upper-limit conducted noise voltage of 48 dbµv (V lim in Eq. (5)) in the frequency range from 450 khz to 30 MHz. Figure 7(a), (b) and (c) show the measured results and suitable filter components for the base-line ac-dc boost converter. The measured maximum and minimum values of the CM noise shown in Fig. 7(a) are dbµv and dbµv, respectively, and the corner frequency analyzed using the software is khz, based on which the component values C y = 4700 pf and L cm = 2.89 mh for the CM filter are taken from the source code. Similarly, the measured maximum and minimum values of the DM noise shown in Fig. 7(b) are dbµv and dbµv, respectively, and the corner frequency is khz, based on which the component values C x =

7 Computerized EMI Filter Design System Fig. 8. Common-mode EMI filter designed based on the CM noise data in Fig. 7(a). Fig. 7. (a) Front panel (user interface) of source code and measured result for base-line CM conducted noise, (b) measured result for baseline DM conducted noise, and (c) measured result for total baseline (CM + DM) conducted noise. µf and L dm = µh for the DM filter are chosen. Observation of the measured conducted noises reveals that the maximum and minimum noises in dbµv for the CM and DM noises are nearly equal in magnitude, and that the noise spectra below 10 MHz for the CM and DM noises are similar in both form and magnitude, which implies that there is not dominant noise mode in the base-line ac-dc boost converter; thus any change of both the CM and DM filter components would apparently affect the shape of the total noise spectrum. It is observed from Fig. 7(c) that this power supply can not satisfy the FCC limit without the insertion of a proper EMI filter. To reduce the amount of conducted noise, the CM filter is first added to the dashed-line block shown in Fig. 2 to investigate its insertion effect on the attenuation of the EMI noise. Figure 8 shows the circuit of the employed CM filter with C y = 4700 pf and L cm = 2.96 mf, in which L cm exhibits small deviation from the predicted value (2.89 mh) shown in Fig. 7(a) owing to the difficulty in choosing proper core material and wiring the CM choke. Figure 9 shows the measured spectra generated by the ac-dc boost converter when this CM filter is employed. Figure 9(a) reveals that adding the CM filter can effectively suppress the CM noise in order to meet the FCC limit. It is also observed from Fig. 9(b) that this CM filter can reduce the DM noises at higher frequencies, due to the effect of both the leakage inductance of the CM choke and the Y capacitance as expected from the DM equivalent circuit shown in Fig. 3(c). However, neither the DM noise spectrum shown in Fig. 9(b) nor the total noise spectrum shown in Fig. 9(c) satisfy the required FCC limit. Note that the total noise spectrum is now DM dominant, and that the need to add a DM filter to satisfy the conducted EMI regulation is apparent. Shown in Fig. 10 is the final EMI filter for the device under test when both the CM and DM filter components are combined together. In this filter, the two X capacitors have large capacitance of 2.2 µf as indicated in Fig. 7(b), and DM inductance is provided completely by the leakage inductance of the CM choke, which is on the order of nearly l3.3 µh. Figure 11(a), (b) and (c) show the measured CM noise, DM noise and total noise spectra, respectively, when the final EMI filter is employed. In contrast to Fig. 9(b) for the DM noise, it is observed from Fig. 11(b) that the addition of two X capacitors effectively reduces the DM noise at all frequencies below 15 MHz, but that the DM noises at higher frequencies 191

8 C.N. Chang et al. Fig. 10. (CM + DM) conducted EMI filter designed based on the CM and DM noise data shown in Fig. 7(a) and (b). 11 satisfy the FCC class A limit, which implies that the developed EMI filter is effective. Note that the impedance conditions indicated by Eqs. (8) and (9) are satisfied by all component values of the final EMI filter in the frequency range between 450 khz and 30 MHz. In the present example, both the CM and DM noises are nearly of the same order in magnitude, and it is necessary to use both CM and DM filter components to suppress the total noise in order to meet the FCC limit. If the measured base-line CM noise and DM noise spectra reveal that the total EMI noise is CM (or DM) dominant, then the addition of DM (or CM) filter components will have little effect on the reduction of total noise, and only the addition of CM (or DM) filter components will be necessary when designing a suitable conducted EMI filter. This is the reason why the CM noise spectrum and DM noise spectrum should be separately and accurately measured. The present computerized system for measurement and analysis of CM noise, DM noise, and total noise satisfies these requirements for systematic conducted EMI filter design. Vl. Conclusion Fig. 9. Measured conducted EMI noise spectra from the device under test when the CM filter depicted in Fig. 8 is inserted: (a) CM noise, (b) DM noise, and (c) total (CM + DM) noise. are almost unaffected by the insertion of the two X capacitors. As a whole, all three conducted noise spectra shown in Fig. A novel computerized system for systematic conducted EMI filter design of a switched-mode power supply has been constructed and described. In this system, measurement and analysis of the conducted EMI are accomplished and integrated by using an automatic data acquisition system and a comprehensive virtual (software) instrument, which was developed using LabVIEW application software. Application of the present system to an ac-dc boost converter with PFC to satisfy the FCC class A limit has been successfully carried out and confirms the validity of this computerized system. The three advantages of this novel system over the traditional one are: (1) accuracy in getting the base-line conducted EMI data because the measured conducted noise voltage for each sampled frequency is the average of many measurements, which can reduce deviations caused by voltage surges and lightning in measurement; (2) speed in evaluating the corner frequency of the associated EMI filter by employing a search-loop in the developed software and fast computation on a personal computer, which is efficient in comparison with its traditional artificial counterpart; (3) ease in modifying the action of the computerized system due to the block diagram form of the source code in the system. 192

9 Computerized EMI Filter Design System Acknowledgment The authors wish to thank Dr. Y.K. Lo for helpful discussion during the course of research. This paper was partly supported financially by the National Science Council, Republic of China, under grant NSC E References Buckman, A. B. (1997) An introductory electrical engineering laboratory based on virtual instrument. Conference Proceeding on Virtual Instrumentation in Education, pp , University of California at Berkeley, Berkeley, CA, U.S.A. Chang, C. N., C. L. Tsai, and J. Y. Chen (1998) EMI filter study of singlephase switch mode rectifier with PFC. Proceeding of the 19th Symposium on Electrical Power Engineer, R.O.C., pp , National Taiwan University of Science and Technology, Taipei, Taiwan, R.O.C. Chen, D. Y. and F. C. Lee (1996) Separation of the common-mode and differential-mode conducted EMI noise. IEEE Trans. on Power Electronics, 11, Guo, T., D. Y. Chen, and F. C. Lee (1993) Separation of common-mode and differential-mode conducted EMI. Proceeding of the Eleventh Annual VPEC Power Electronic Seminar, 11, pp , University of Virginia, Charlottesville, VA, U.S.A. Guo, T., D. Y. Chen., and F. C. Lee (1995) Diagnosis of power supply conducted EMI using a noise separator. IEEE Tenth Annual Applied Power Electronics Conference and Exposition Conference Proceedings, pp , Blacksburg, VA, U.S.A. Nave, M. J. (1989) The effect of duty cycle on SMPS common mode emissions: theory and experiment. IEEE Fourth Annual Applied Power Electronics Conference and Exposition Conference Proceedings, pp. 3-12, Seminole, FL, U.S.A. Ott, W. H. and M. J. Nave (1991) Power Line Filter Design for Switched- Mode Power Supplies, Chapter 5, pp Van Nostrand Reinhold Co., New York, NY, U.S.A. Ran, L., J. C. Clare, K. J. Bradley, and C. Christopoulos (1999) Measurement of conducted electromagnetic emissions in PWM motor drive system without the need for an LISN. IEEE Trans. on Electromagnetic Compatibility, 41, Shih, F. Y., D. Y. Chen, Y. P. Wu, and Y. T. Chen (1996) A procedure for designing EMI filters for AC line applications. IEEE Trans. on Power Electronics, 11, Tihanyi, L. (1995) Electromagnetic Compatibility in Power Electronics, pp IEEE Press, New York, NY, U.S.A. Fig. 11. Measured EMI noise spectra from the device under test when the conducted EMI filter depicted in Fig. 10 is inserted: (a) CM noise, and (c) total (CM + DM) noise. 193

10 C.N. Chang et al. LabVIEW EMI LabVIEW LabVIEW 100 W, 200 V, 94 khz 450 khz 30 MHz FCC class A 194