IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO.

4, JULY/AUGUST 2011

1741

Overvoltage Mitigation of Inverter-Driven Motors With Long Cables of Different Lengths

Hirofumi Akagi, Fellow, IEEE, and Itaru Matsumura
AbstractThis paper presents an intensive discussion on an overvoltage appearing at the motor terminals of an adjustablespeed drive that consists of a voltage-source pulsewidthmodulated inverter, an induction motor, and a three-phase symmetric long cable. It describes a design procedure for a simple passive electromagnetic-interference lter based on parallel connection of an inductor and a resistor, which is intended for mitigating the overvoltage at the motor terminals. The modeling of the cable and the motor is characterized by focusing on the natural frequency or ringing frequency inherent in the cable used, which is inversely proportional to cable length. Impedance mismatch at the inverter and motor terminals causes the overvoltage that may reach double the inverter dc-link voltage. The effectiveness and validity of the simple design procedure of the lter are conrmed on the 400-V 15-kW experimental system with either a 100- or 200-m-long cable, as well as on computer simulation based on the modeling of the cable and the motor. Index TermsElectromagnetic interference (EMI), long cables, motor drives, overvoltage, pulsewidth-modulated (PWM) inverters.

I. I NTRODUCTION IGNIFICANT developments in power electronics have brought high performance, low cost, compactness, and/or energy savings to adjustable-speed motor drives. However, as the switching or carrier frequency of voltage-source pulsewidth-modulated (PWM) inverters has got higher, application issues related to electromagnetic interference (EMI) have emerged from the motor drives, and solutions to the issues have been presented in the literature [1][20]. Industrial factories, mills, and plants have often used long cables for delivering electric power from inverters to motors, and their lengths have ranged over several hundreds of meters. Such an adjustable-speed drive with a long cable suffers from ground leakage current and conducted/radiated EMI emissions due to large stray capacitances inherent in the long cable. Moreover, the long cable acts as a distributed-parameter circuit in a frequency range of 100 kHz or higher. Impedance misManuscript received October 6, 2010; revised December 23, 2010; accepted February 27, 2011. Date of publication May 12, 2011; date of current version July 20, 2011. Paper 2010-IDC-394.R1, presented at the 2010 IEEE Energy Conversion Congress and Exposition, Atlanta, GA, September 1216, and approved for publication in the IEEE T RANSACTIONS ON I NDUSTRY A PPLI CATIONS by the Industrial Drives Committee of the IEEE Industry Applications Society. H. Akagi is with the Department of Electrical and Electronic Engineering, Tokyo Institute of Technology, Tokyo 152-8552, Japan (e-mail: akagi@ ee.titech.ac.jp). I. Matsumura was with the Department of Electrical and Electronic Engineering, Tokyo Institute of Technology, Tokyo 152-8552, Japan. He is now with the Railway Technical Research Institute, Tokyo 185-8540, Japan. Digital Object Identier 10.1109/TIA.2011.2154294

match is accompanied by reection at both inverter and motor terminals. This reection causes an overvoltage at the motor terminals, which reaches double the inverter dc-link voltage. The overvoltage may damage the motor-winding insulation and may cause it to break down in the worst case. Comprehensive research has been conducted not only on modeling and analysis of overvoltage generation but also on passive or active EMI lters for overvoltage mitigation [3][5], [7], [9][11], [13][17]. These lters can be classied from the point of installation into the following two groups; one is on the inverter terminals, and the other is on the motor terminals [13]. Generally, application engineers prefer the former to the latter due to easy maintenance. This paper describes a small-rated passive EMI lter consisting of parallel connection of an inductor and a resistor for overvoltage mitigation. This lter is connected in series with each of three-phase power lines at the inverter terminals. The lter conguration itself has been found in the literature [4], [7], [8], [10]. However, few papers have been presented or published on a simple and practical design procedure of the lter for inverterdriven motors with long cables of different lengths. The reason seems that the lter design requires a reasonable compromise between overvoltage mitigation and lter cost/volume in different cable lengths. This paper conducts the modeling of an adjustable-speed motor drive system, leading to a lter design. This design is characterized by not considering a broad range of frequencies but focusing on the ringing frequency of the cable used. As a result, the focus only on the ringing frequency makes the lter design simple and efcient. The effectiveness and validity of the lter design are conrmed not only by computer simulation based on the modeling but also by experiment using the 400-V 15-kW laboratory system with either a 100- or 200-m-long symmetric cable including a grounding conductor.

II. S YSTEM C ONFIGURATION A. Experimental System Fig. 1 shows the experimental system conguration with either a 100- or 200-m-long cable. A three-phase deltastar transformer changes the 200-V 50-Hz ac mains with V-phase grounding in Japan into the worldwide 400-V 50-Hz ac mains with neutral grounding. The 400-V 15-kW four-pole induction motor is driven at 20 Hz by a three-phase two-level voltagesource PWM inverter with a carrier frequency of 7 kHz. No mechanical load is coupled with the motor because load conditions produce little effect on the PWM waveform of the voltagesource inverter. Fig. 1 uses a seven-conductor symmetric

0093-9994/$26.00 2011 IEEE

1742

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

Fig. 1. Experimental system rated at 400 V and 15 kW where a 100- or 200-m-long seven-conductor symmetric cable was used.

Fig. 2. Cross section of the seven-conductor symmetric cable used for the experiment.

cable with the 600-V polyvinyl chloride insulated and sheathed. The cross-sectional area of each conductor is 3.5 mm2 . Fig. 2 shows the cross section of the cable, in which the center conductor is not connected (NC), three conductors jointed together are used as the grounding line (G), and the other three conductors are used as the three-phase power lines (U, V, and W). Thus, the cable acts as a three-phase symmetric cable with a single grounding conductor. The motor frame is grounded via the grounding conductor, as shown in Fig. 1. A small-rated passive lter is installed at the ac terminals of the inverter. It consists of parallel connection of the inductor LD and the resistor RD in each phase. A general asymmetric four-conductor cable results in making the modeling complicated. In an actual system, however, it is desirable to use a symmetric cable in terms of reducing both leakage currents and conductive EMI emissions. B. Overvoltage Mitigation When a long cable is used to connect a voltage-source PWM inverter and an induction motor, the cable acts as a distributedparameter circuit in a frequency range of higher than 100 kHz. Impedance mismatch occurs at the inverter terminals because the inverter can be considered as a voltage source with a negligible internal impedance, whereas the long cable has a characteristic impedance ranging from 30 to 70 . Moreover, impedance mismatch occurs at the motor terminals because the motor exhibits a high impedance in a frequency range of higher than 100 kHz. The so-called reection resulting from the impedance mismatching causes an overvoltage at the motor terminals, which may reach double the inverter dc-link voltage. This overvoltage is accompanied by an oscillating or ringing voltage with the same frequency as the natural frequency of the cable. The ringing frequency is determined by the cable length and the velocity of the incident and reection waves traveling

Fig. 3. Experimental line-to-line voltage waveforms at the motor terminals. (a) When the 100-m-long cable was used. (b) When the 200-m-long cable was used.

inside the cable. It is known that the ringing frequency is in inverse proportion to the cable length [17]. Fig. 3 shows the experimental waveforms of vUV at the motor terminals when either the 100- or the 200-m-long cable was used, where the passive lter was disconnected from Fig. 1. The overvoltages were observed as 1050 V in the 100-m-long cable and as 1020 V in the 200-m-long cable, which were almost double the inverter dc-link voltage (= 560 V). The observed ringing frequencies were 300 kHz in the 100-m-long cable and 150 kHz in the 200-m-long cable. These experimental results suggest that the following relation exists between the ringing frequency fr (in hertz) and the cable length l (in meters): 3.0 107 . fr = l (1)

A small-rated passive lter consisting of the inductor LD and the resistor RD in each phase is installed at the inverter terminals. When the corner or cutoff frequency of the lter is designed to be lower than the ringing frequency, the inverterfrequency (0 to 60 Hz) component contained in the inverter output current ows in the lter inductor LD , whereas the ringing-frequency component ows in the lter resistor RD . This consideration allows one to assign the resistance value of RD so as to realize the impedance matching between the inverter and the cable. This makes it possible to mitigate the overvoltage appearing at the motor terminals. Three pure resistors, each of which was rated at 50 W, were used to reduce stray inductances. III. M ODELING AT THE R INGING F REQUENCY This paper considers computer simulation as a supplementary tool for power electronics engineers who design a passive

AKAGI AND MATSUMURA: OVERVOLTAGE MITIGATION OF MOTORS WITH LONG CABLES OF DIFFERENT LENGTHS

1743

Fig. 5. Parameter-extracting methods for the 20-m-long cable. (a) Differential mode. (b) Common mode.

Fig. 4. Cable and motor models for the simulation. (a) Distributed-parameter model per unit length for the long cable. (b) Lumped-parameter model for the motor.

lter for overvoltage mitigation. Therefore, the modeling used for the simulation gives priority to practicality and convenience rather than accuracy and exactness. Fig. 4(a) shows a distributed-parameter model per unit length for the long cable, and Fig. 4(b) shows a lumped-parameter model for the motor. As shown in Fig. 2, the center conductor is not used in the seven-conductor cable. Thus, the six-conductor cable model is treated as cascade connection of 20 ladder circuits, each of which is based on a lossy distributed-parameter model per unit length. It includes the line inductance and resistance per unit length L and R, the adjacent line-to-line capacitance and conductance per unit length C1 and G, and the difference-mode capacitance per unit length C2 . A. Impedance Measurement of a 20-m-Long Cable The long cable used in this experiment has a frequency dependence due to the so-called skin effect and dielectric loss. It is therefore desirable to measure the cable parameters at the ringing frequency of overvoltage, i.e., at the natural frequency of the cable. When a 100-m-long cable is used, the cable parameters should be measured at 300 kHz, as derived from (1). However, the occurrence of resonance at that frequency makes it difcult to achieve accurate measurement because the frequency for measurement coincides with the ringing frequency. To overcome the difculty, the open/short method [21] was applied to a 20-m-long cable for measurement at 300 kHz because the ringing frequency of the cable is around 1.5 MHz, which is far from 300 kHz. Fig. 5(a) shows the circuit conguration of impedance measurement for differential mode. The terminals of the V-phase conductor and the W-phase conductor were jointed together at both ends, and the terminals of the three grounding conductors were also jointed together at both ends. An LCR meter was connected between the U-phase terminal and the jointed terminal at one end, and the two terminals at the other end were
Fig. 6. Cable parameters in differential mode when the 20-m-long cable was used. (a) RDif . (b) LDif . (c) CDif .

opened or shorted. Impedance measurement was carried out in a frequency range from 100 kHz to 1 MHz. Fig. 5(b) shows the circuit conguration of impedance measurement for common mode. The three-phase conductors were jointed together as if they were considered as a single power line, and the three grounding conductors were jointed together as if they were considered as a single grounding line. In the same way as the differential mode, impedance measurement was carried out in a frequency range from 100 kHz to 1 MHz. The results from both impedance measurements made it possible to calculate the distributed parameters of the lossy long cable as a set of resistance, inductance, conductance, and capacitance per unit length. Figs. 6 and 7 show the cable parameters for differential mode and common mode, respectively. Both per-unit inductance and capacitance can be considered as constant values, irrespective of frequency. On the other hand, the per-unit resistance can be approximated to be proportional to a square root of frequency

1744

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

Fig. 9. Extraction of per-unit-length constants for the simulation. (a) R and L. (b) G and C1 . (c) C2 .

A. Cable Constants The per-unit resistance R and the per-unit inductance L of the cable can be obtained from the measured results in differential mode. Fig. 9(a) shows the circuit model of R and L per unit length in differential mode. The total resistance and inductance of the circuit model correspond to RDif and LDif . Thus, the perunit resistance R and the per-unit inductance L in each phase are given by R= 2 RDif 3 L= 2 LDif . 3

Fig. 7. Cable parameters in common mode when the 20-m-long cable was used. (a) GCom . (b) CCom .

Fig. 8. Parameter measurement for the motor.

due to the skin effect, and the per-unit conductance can be considered to be proportional to frequency due to the dielectric loss. The characteristic impedance for differential mode was 67.2 at 300 kHz, which was almost constant in a frequency range of lower than 1 MHz. Connecting a matching impedance of 67.2 for the differential-mode circuit is equivalent to connecting a matching impedance of 44.8 to each phase. Thus, this paper denes the characteristic impedance of the three-phase cable as ZCable = 67.2 2 = 44.8 . 3

The per-unit conductance G and the per-unit capacitance C1 can be obtained from the measured results for common mode. Fig. 9(b) shows the circuit model for common mode, where six sets of G and C1 per unit length are connected in parallel, thus leading to G= 1 GCom 6 C1 = 1 CCom . 6

Fig. 9(c) shows that C2 can be calculated from CDif , existing between the U-phase and the jointed V/W-phase, and C1 as follows: C2 = 1 2 CDif C1 . 2 3

B. Measurement of Motor Parameters It is possible to bring out the neutral of three-phase starconnected stator windings from the induction motor. This allows one to treat the motor as a two-port circuit with the following two pairs of terminals; one pair is the motor terminal (U, V, and W) and the motor frame (Frame), and the other pair is the motor neutral (N) and the motor frame (Frame). Fig. 8 shows the two-port circuit in which the three-phase motor terminals are jointed together. Applying the open/short method [21] to the two-port circuit made it possible to conduct impedance measurement at 300 kHz. The two-port or F parameters obtained from the measured results enable one to calculate ZA , ZB , and ZC . IV. C IRCUIT C ONSTANTS U SED FOR THE S IMULATION This section extracts cable and motor constants from the cable and motor parameters measured in the previous section.

The ringing frequency of the 100-m-long cable is 300 kHz. Therefore, the 20-m-long-cable constants R, L, G, C1 , and C2 , which are obtained from the impedance measurement at 300 kHz, should be referred to the following general cable constants including the cable length l: R = 9.98 102 L = 0.31 H/m 1.08 103 G= S/m l C1 = 65.0 pF/m C2 = 8.1 pF/m. 30 /m l (2) (3) (4) (5) (6)

The measured results in Figs. 6(a) and 7(a) indicate that, due to skin effect, R is approximated to be proportional to f , while due to dielectric loss, G is approximated to be proportional to f . With the help (1), R and G are considered to be in inverse of proportion to l and to l, respectively. Table I summarizes the 100-m-long-cable constants obtained by the aforementioned procedure and the 200-m-long-cable constants obtained from (2)(6) at a ringing frequency of 150 kHz.

AKAGI AND MATSUMURA: OVERVOLTAGE MITIGATION OF MOTORS WITH LONG CABLES OF DIFFERENT LENGTHS

1745

TABLE I C ABLE C ONSTANTS P ER U NIT L ENGTH

TABLE II M OTOR C ONSTANTS

Fig. 11. Comparisons between simulated and experimental waveforms. (a) When the 100-m-long cable was used. (b) When the 200-m-long cable was used.

waveforms, thus conrming the exactness of the modeling for simulation, as well as the validity of the constant extraction. Fig. 11 concludes that the simulation is practical and effective in designing a passive lter for mitigating the overvoltage appearing at the motor terminals.
Fig. 10. Experimental voltage waveforms at the inverter output terminals. (a) Rising edge. (b) Falling edge.

V. D ISCUSSION BY S IMULATION A. Goal of Overvoltage Mitigation A compromise or tradeoff exists between the cost, volume, weight, and/or loss of the passive lter and the effect of the lter on overvoltage mitigation. In other words, as the lter is designed to be more effective in mitigating the overvoltage, the lter inductor LD is more costly and bulky, and the lter resistor RD is more lossy and bulky. Enameled copper wires are used for the randomly wound stator coils of the 400-V 15-kW induction motor. The partialdischarge voltage of the enamel coating with a thickness of 30 m is about 800 V [22]. The overvoltage appearing, for example, between phases U and V should be shared equally among the coils forming the corresponding stator windings. However, the highest voltage is applied to the rst coil that is closest to the U- or V-phase motor terminal. This suggests that reducing the overvoltage to 800 V can prevent the enameled copper wires from partially discharging. Thus, this paper aims to reduce the overvoltage to 800 V as a reasonable goal for the 400-V 15-kW induction motor in Fig. 1. B. Simulated Results Computer simulation was conducted to conrm the effectiveness of the passive lter when the 100-m-long cable is used in Fig. 1. Fig. 12 shows a trend in the simulated maximal overvoltage as a function of the lter resistor RD with a parameter of the lter inductor LD . A resistance value of RD = 47 , which is the closest to the cable characteristic impedance of 44.8 , results in the minimal overvoltage. This means that impedance

B. Motor Constants Table II summarizes the motor constants RIM , LIM , RS1 , CS1 , RS2 , and CS2 . They are calculated from ZA , ZB , and ZC on the basis of the impedance measurement at 300 kHz for Fig. 4(b). Note that no frequency dependence is considered in the motor constants because the impedance of the motor is much higher than that of the cable. C. Inverter Modeling Fig. 10 shows the actual voltage waveforms observed at the inverter output terminals with focus on the rising and falling edges. These waveforms conclude that the rising and falling times are almost equal to 0.18 s. Exactly speaking, both times depend slightly on switching conditions of the inverter. Thus, the inverter model used in this simulation is treated as a trapezoidal voltage source with an amplitude of 560 V and the same rising and falling times as 0.18 s. D. Comparison Between Experiment and Simulation Computer simulation was carried out on the software package PSpice, using the cable and motor models shown in Fig. 4 and the cable and motor constants. Note that the unit lengths are 5 m for the 100-m-long cable and 10 m for the 200-m-long cable because the cable model is based on cascade connection of 20 ladder circuits. Fig. 11 compares the waveforms of vUV in either the 100or the 200-m-long cable between experiment and simulation. The simulated waveforms agree well with the experimental

1746

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

inverter and the motor. Note that the cable was installed to hold a distance of more than 1 m between a part and the adjacent part for the purpose of avoiding the so-called proximity effect. A. Experimental Waveforms Fig. 14 shows the experimental and simulated waveforms of uUV and iRD when the 100-m-long cable was used with a passive lter of RD = 33 and LD = 30 H. Here, uUV is the line-to-line voltage between phases U and V at the motor terminals, and iRD is the current owing into RD . The overvoltage was 805 V in the experiment while it was 778 V in the simulation. The peak value of iRD was 4.9 A in the experiment while it was 4.7 A in the simulation. Fig. 15 shows the experimental and simulated waveforms when the 200-m-long cable was used with a passive lter of RD = 33 and LD = 70 H (0.21% on a 3, 400-V, 15-kW, and 50-Hz base). Note that three inductors of LD = 70 H were available for this experiment, so that their inductance value was not exactly double the inductance value of 30 H. The overvoltage was 780 V in the experiment while it was 755 V in the simulation. The peak value of iRD was 4.3 A in the experiment while it was 4.8 A in the simulation. These experimental results are in good agreement with the simulated ones. Figs. 14 and 15 conclude that the overvoltage observed in the 100-m-long cable is nearly equal to that in the 200-m-long cable. This means that the same mitigation of the overvoltage is achievable to long cables of different lengths when the inductance value of LD is proportional to the cable length. It is clear from the waveforms of uUV in Figs. 14 and 15 that neither undershoot nor ringing follows the rst overshoot. This implies that the passive lter provides good overvoltage mitigation, as well as good damping performance. B. Optimal Resistance Value of RD Table III summarizes three-phase measured power losses in RD when the inductor of LD = 30 H is used for the 100-mlong cable and that of LD = 70 H is used for the 200-mlong cable. The maximal overvoltage observed at RD = 33 was almost the same as that at RD = 47 . This is in good agreement with the results obtained from Figs. 12 and 13. Table III, along with the aforementioned consideration, suggests that the resistor of RD = 33 is smaller in power loss than that of RD = 47 without producing any bad effect on overvoltage mitigation. VII. D ESIGN P ROCEDURE OF RD AND LD The simulated and experimental results in Sections IV and V conclude the following. 1) The resistance value of RD should be close to but lower slightly than the characteristic impedance of the cable used. Generally, the characteristic impedance of a cable is almost constant, independent of cable lengths. 2) The inductance value of LD should be proportional to the cable length.

Fig. 12. Simulated maximal overvoltages when the 100-m-long cable was used, where the inverter is assumed as a trapezoidal voltage source with an amplitude of 560 V and the same rising and falling times as 0.18 s.

Fig. 13. Simulated maximal overvoltages when the 200-m-long cable was used, where the inverter is assumed as a trapezoidal voltage source with an amplitude of 560 V and the same rising and falling times as 0.18 s.

matching between the lter resistor and the cable is effective in mitigating the overvoltage as expected. The combination of RD = 47 with LD = 30 H (0.09% on a 3, 400-V, 15-kW, and 50-Hz base) makes the overvoltage lower than 800 V. The power loss dissipated in RD is calculated as 42.8 W, altogether in three phases, which is 0.29% of 15 kW (the motor rating). Fig. 13 shows another trend in the simulated maximal overvoltage when the 200-m-long cable is used. It is conrmed that the same overvoltage mitigation as that in Fig. 12 is achievable when the lter inductance in the 200-m-long cable is designed double the lter inductance in the 100-m-long cable. The reason is that the ringing frequency of the 200-m-long cable is half the ringing frequency of the 100-m-long cable. The combination of RD = 47 with LD = 60 H (0.18% on a 3, 400-V, 15-kW, and 50-Hz base) enables the lter to mitigate the maximal overvoltage to 744 V. The power loss of RD is calculated as 78.7 W (0.52%), altogether in three phases, which is nearly proportional to the cable length. VI. E XPERIMENTAL V ERIFICATION OF OVERVOLTAGE M ITIGATION Experiment was conducted on an overvoltage at the motor terminals and a power loss in RD , in which either the 100- or the 200-m-long cable was connected back and forth between the

AKAGI AND MATSUMURA: OVERVOLTAGE MITIGATION OF MOTORS WITH LONG CABLES OF DIFFERENT LENGTHS

1747

Fig. 14. Waveforms when the 100-m-long cable was used. (a) Experiment. (b) Simulation.

Fig. 15. Waveforms when the 200-m-long cable was used. (a) Experiment. (b) Simulation. TABLE III M EASURED L OSS D ISSIPATED IN RD

As an example, the parallel connection of RD = 33 and LD = 30 H has the capability of reducing the maximal overvoltage to a target voltage of 800 V when the 100-m-long cable is used in Fig. 1. Since the ringing frequency of the 100-m-long cable is 300 kHz, the argument of the lter impedance Zlter can be calculated as follows: Zlter = j2fr LD RD RD + j2fr LD = 30

Fig. 16. Experimental and simulated per-unit overshoot voltage at the motor terminals as a function of the sine of the lter argument.

(7) at the motor terminals as a function of the sine of the lter argument. Here, the lter resistor RD is xed at 33 , and the per-unit overshoot is based on the inverter dc-link voltage of 560 V. An overvoltage of 800 V corresponds to a per-unit overshoot of 43%. It is interesting that the sine of the lter argument is proportional to the per-unit overshoot occurring on the motor line-to-line voltage when the lter resistor RD is close to but lower slightly than the characteristic impedance of the cable used. VIII. C ONCLUSION This paper has discussed the modeling of an adjustable-speed drive including a three-phase symmetric long cable, with focus on the ringing frequency of the overvoltage appearing at the motor terminals. A small-rated passive lter, which consists of parallel connection of LD and RD in each phase, has been designed to mitigate the overvoltage. Computer simulation based on the modeling was carried out to make a compromise between lter design and overvoltage mitigation. Experiment was conducted to verify the viability and effectiveness of the

where RD = 33 , LD = 30 H, and fr = 300 kHz. Equation (7) yields the following equations: (RD + j2fr LD ) = 60 2fr LD = tan 60 = 3. RD (8) (9)

The lter resistor is assumed to be xed at RD = 33 , irrespective of cable lengths. Substituting (1) and RD = 33 into (9) leads to the following equation: LD = l 9.5 107 . (10)

This provides a relation between the cable length l (in meters) and the lter inductor LD (in henries), thus resulting in designing the lter inductor LD for the seven-conductor symmetric long cables of different lengths. Fig. 16 shows the experimental and simulated results of a perunit overshoot voltage superimposed on the line-to-line voltage

1748

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 47, NO. 4, JULY/AUGUST 2011

design procedure of the lter when either a 100- or 200-m-long cable was used. The essentials of the design procedure can be summarized as follows. 1) The lter resistor RD should be designed to be close to but lower slightly than the characteristic impedance of the cable used. 2) The lter inductor LD should be designed to hold a relation of Zlter 30 at the ringing frequency of the cable. 3) A three-phase symmetric cable with a lower characteristic impedance is effective in bringing further reductions in cost, volume, weight, and power loss to the lter. Although the experiment has used a seven-conductor symmetric long cable, both the modeling and the lter design would be applicable to other symmetric long cables. It has been well known that it is desirable to use a three-phase symmetric long cable in terms of reducing leakage currents and conducted/ radiated EMI emissions. However, asymmetric long cables have often been used in industrial low-voltage motor drives. Modeling of asymmetric long cables is left for future work. R EFERENCES
[1] Y. Murai, T. Kubota, and Y. Kawase, Leakage current reduction for a high-frequency carrier inverter feeding an induction motor, IEEE Trans. Ind. Appl., vol. 28, no. 4, pp. 858863, Jul./Aug. 1992. [2] S. Ogasawara and H. Akagi, Modeling and damping of high-frequency leakage currents in PWM inverter-fed ac motor drive systems, IEEE Trans. Ind. Appl., vol. 32, no. 5, pp. 11051114, Sep./Oct. 1996. [3] A. von Jouanne, P. Enjeti, and W. Gray, Application issues for PWM adjustable speed ac motor drives, IEEE Ind. Appl. Mag., vol. 2, no. 5, pp. 1018, Sep./Oct. 1996. [4] S. Ogasawara, H. Ayano, and H. Akagi, Measurement and reduction of EMI radiated by a PWM inverter-fed ac motor drive system, IEEE Trans. Ind. Appl., vol. 33, no. 4, pp. 10191026, Jul./Aug. 1997. [5] A. von Jouanne and P. Enjeti, Design considerations for an inverter output lter to mitigate the effects of long motor leads in ASD applications, IEEE Trans. Ind. Appl., vol. 33, no. 5, pp. 11381145, Sep./Oct. 1997. [6] S. Ogasawara, H. Ayano, and H. Akagi, An active circuit for cancellation of common-mode voltage generated by a PWM inverter, IEEE Trans. Power Electron., vol. 13, no. 5, pp. 835841, Sep. 1998. [7] N. Aoki, K. Sato, and A. Nabae, Damping circuit to suppress motor terminal overvoltage and ringing in PWM inverter-fed ac motor drive systems with long motor leads, IEEE Trans. Ind. Appl., vol. 35, no. 5, pp. 10141020, Sep./Oct. 1999. [8] G. L. Skibinski and H. B. Nielsen, Apparatus for eliminating motor voltage reections and reducing EMI currents, U.S. Patent 5 990 654, Nov. 23, 1999. [9] G. L. Skibinski, R. J. Kerkman, and D. W. Schlegel, EMI emissions of modern PWM ac drives, IEEE Ind. Appl. Mag., vol. 5, no. 6, pp. 4780, Nov./Dec. 1999. [10] A. Fratta, G. M. Pellegrino, F. Scapino, and F. Villata, A three-phase PWM VSI line termination net for load overvoltage cancellation, in Proc. IEEE ISIE, Jul. 1999, pp. 762767. [11] A. de Lima, H. W. Dommel, and R. M. Stephan, Modeling adjustablespeed drives with long feeders, IEEE Trans. Ind. Electron., vol. 47, no. 3, pp. 549556, Jun. 2000. [12] T. G. Habetler, R. Naik, and T. A. Nondahl, Design and implementation of an inverter output LC lter used for dv/dt reduction, IEEE Trans. Power Electron., vol. 17, no. 3, pp. 327331, May 2002. [13] A. F. Moreira, T. A. Lipo, G. Venkataramanan, and S. Bernet, Highfrequency modeling for cable and induction motor overvoltage studies in long cable drives, IEEE Trans. Ind. Appl., vol. 38, no. 5, pp. 12971306, Sep./Oct. 2002. [14] A. Kempski, R. Smolenski, and R. Strzelecki, Common mode current paths and their modeling in PWM inverter-fed drives, in Proc. IEEE PESC, Oct. 2002, pp. 15511556. [15] A. F. Moreira, P. M. Santos, T. A. Lipo, and G. Venkataramanan, Filter networks for long cable drives and their inuence on motor voltage distri-

[16]

[17]

[18] [19]

[20] [21] [22]

bution and common-mode currents, IEEE Trans. Ind. Electron., vol. 52, no. 2, pp. 515522, Apr. 2005. N. Hanigovszki, J. Landkildehus, G. Spiazzi, and F. Blaabjerg, An EMC evaluation of the use of unshielded motor cables in ac adjustable speed drive application, IEEE Trans. Power Electron., vol. 21, no. 1, pp. 273 281, Jan. 2006. G. Skibinski, R. Tallam, R. Reese, B. Buchholz, and R. Lukaszewski, Common mode and differential mode analysis of three phase cables for PWM ac drives, in Conf. Rec. IEEE IAS Annu. Meeting, Oct. 2006, pp. 880888. H. Akagi and T. Shimizu, Attenuation of conducted EMI emissions from an inverter-driven motor, IEEE Trans. Power Electron., vol. 23, no. 1, pp. 282290, Jan. 2008. H. Akagi and T. Oe, A specic lter for eliminating high-frequency leakage current from the grounded heat sink in a motor drive with an active front end, IEEE Trans. Power Electron., vol. 23, no. 2, pp. 763770, Mar. 2008. U. T. Shami and H. Akagi, Experimental discussions on a shaft end-toend voltage appearing in an inverter-driven motor, IEEE Trans. Power Electron., vol. 24, no. 6, pp. 15321540, Jun. 2009. Impedance Measurement Handbook, Agilent Technol., Santa Clara, CA, 2003. H. Kikuchi, Y. Tetsu, and S. Itonaga, Inverter surge voltage resistance enameled wire for high reliability inverter-fed motor, Hitachi Densen, no. 21, pp. 8592, Jan. 2002, (in Japanese).

Hirofumi Akagi (M87SM94F96) was born in Okayama, Japan, in 1951. He received the B.S. degree in electrical engineering from the Nagoya Institute of Technology, Nagoya, Japan, in 1974, and the M.S. and Ph.D. degrees in electrical engineering from the Tokyo Institute of Technology, Tokyo, Japan, in 1976 and 1979, respectively. In 1979, he joined the Nagaoka University of Technology, Nagaoka, Japan, as an Assistant Professor and then an Associate Professor in the Department of Electrical Engineering. In 1987, he was a Visiting Scientist at the Massachusetts Institute of Technology (MIT), Cambridge, for ten months. From 1991 to 1999, he was a Professor with the Department of Electrical and Electronic Engineering, Okayama University, Okayama. From March to August of 1996, he was a Visiting Professor at the University of Wisconsin, Madison, and then MIT. Since January 2000, he has been a Professor with the Department of Electrical and Electronic Engineering, Tokyo Institute of Technology. His research interests include power conversion systems, motor drives, active and passive electromagnetic-interference lters, high-frequency resonant inverters for induction heating and corona discharge treatment processes, and utility applications of power electronics such as active lters, self-commutated back-to-back systems, and exible ac transmission system devices. He has published more than 90 IEEE Transactions papers and two invited papers published in the P ROCEEDINGS OF THE IEEE in 2001 and 2004. He has made presentations many times as a keynote or invited speaker internationally. According to Google Scholar, the total citation index for all his papers is more than 12 000. Dr. Akagi was the President of the IEEE Power Electronics Society for 20072008 and is currently the Senior Past President. He was elected as a Distinguished Lecturer of the IEEE Power Electronics Society and the IEEE Industry Applications Society for 19981999. He was the recipient of three IEEE Industry Applications Society Transactions Prize Paper Awards in 1991, 2004, and 2009 and two IEEE Power Electronics Society Transactions Prize Paper Awards in 1999 and 2003, as well as nine IEEE Industry Applications Society Committee Prize Paper Awards. He was also the recipient of the 2001 IEEE William E. Newell Power Electronics Award, the 2004 IEEE Industry Applications Society Outstanding Achievement Award, and the 2008 IEEE Richard H. Kaufmann Technical Field Award. Itaru Matsumura was born in Tokyo, Japan, in 1984. He received the B.E. and M.S. degrees in electrical engineering from the Tokyo Institute of Technology, Tokyo, in 2007 and 2009, respectively. Since April 2009, he has been with the Railway Technical Research Institute, Tokyo.