IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 38, NO.

5, SEPTEMBER/OCTOBER 2002 1297

High-Frequency Modeling for Cable and Induction

Motor Overvoltage Studies in Long Cable Drives
Alessandro F. Moreira, Member, IEEE, Thomas A. Lipo, Fellow, IEEE, Giri Venkataramanan, and
Steffen Bernet, Member, IEEE

Abstract—High-frequency simulation models for power ca- Frequency of the first pole in the frequency
bles and motors are key tools to aid a better understanding of response of the phase-to-neutral motor
the overvoltage problem in pulsewidth modulation drives with impedance.
long feeders. In this paper, frequency responses of the cable
characteristic and the motor input impedances are obtained Frequency of the first zero in the frequency
experimentally and suitable models are developed to match the response of the phase-to-neutral motor
experimental results. Several lumped segments incorporating impedance.
a lossy representation of the line are used to model the cable.
The cable and induction motor models may be implemented I. INTRODUCTION
using a computational tool such as MATLAB, thereby providing
a convenient method to analyze the overvoltage phenomena.
Simulation and experimental results are presented for a typical
3-hp induction motor, showing the suitability of the developed
simulation models. The most promising filter networks
O VERVOLTAGE problems in long cable drives due to
steep voltage pulse rise time have become an important
research area during the last decade. The overvoltage phe-
are also investigated through simulation analysis, and a design nomenon is usually described using the traveling-wave and
approach based on a tradeoff between filter losses and motor peak reflection phenomena: a voltage pulse, initiated at the inverter,
voltage is proposed. Experimental results of an filter placed being reflected at the motor terminals due to a mismatch
at the motor terminals demonstrate the validity of the simulation
models.
between the cable characteristic impedance and the motor input
impedance. The magnitude of the overvoltage depends on the
Index Terms—Adjustable-speed drives, converter machine pulse rise time and on the characteristics and length of the cable
interactions.
[1]–[4]. The motivation for the work presented in this paper
is the lack of good simulation approaches that can be used to
LIST OF SYMBOLS AND ABBREVIATIONS accurately investigate the overvoltage phenomena. Therefore,
, Parameters of the high-frequency model of the objective of this paper is to study the motor overvoltage
the cable per-unit length. phenomena in a definitive manner by developing accurate and
Cable characteristic, short- and open-circuit fast simulation models for power cables and motors that allow
impedances. a better understanding of the overvoltage problem. The models
, Parameters of the high-frequency model of provide a convenient tool that can be utilized to benchmark the
the motor per phase. best filter solution for a particular drive.
Phase-to-neutral and phase-to-ground motor For the power cable, it is well known that distributed-param-
impedances. eter representation provides more accurate results in the study
Lowest and highest test frequencies in the of high-frequency transients than the lumped-parameter models
impedance measurements. [5]. However, lumped-parameter representation of the transmis-
sion line can be successfully used to analyze the overvoltage
phenomena if an adequate number of segments is used in the
calculation [4]. With regard to the properties of the line, early
Paper IPCSD 02–026, presented at the 2001 Industry Applications Society papers have proposed the use of lossless characteristics [6]–[8],
Annual Meeting, Chicago, IL, September 30–October 5, and approved for pub- which leads to a considerable amount of inaccuracy. The use of
lication in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Indus- a distortionless line representation was proposed in [9]. Prelim-
trial Drives Committee of the IEEE Industry Applications Society. Manuscript
submitted for review October 15, 2001 and released for publication June 1, 2002. inary investigations indicate it is still not sufficiently accurate
This work was supported by ABB Corporate R&D, Germany, WEMPEC at to investigate the overvoltage problem, especially for very long
the University of Wisconsin, Madison, and the Universidade Federal de Minas
Gerais and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior,
cable drives. Distortion has been demonstrated to be important
Brazil. for an accurate analysis of the overvoltage [4]. These issues are
A. F. Moreira, T. A. Lipo, and G. Venkataramanan are with the Deparment addressed in Section II, wherein a practical multiple segment
of Electrical and Computer Engineering, University of Wisconsin, Madison,
WI 53706 USA (e-mail: moreira@cpdee.ufmg.br; lipo@engr.wisc.edu; lumped-parameter model is proposed for modeling the power
giri@engr.wisc.edu). cable.
S. Bernet is with Faculty IV, Institute of Energy and Automation Technology, In Section III, a high-frequency representation for the induc-
TU-Berlin Sekr. E2, D-10587 Berlin, Germany (e-mail: steffen.bernet@TU-
Berlin.de). tion motor is developed. Regarding the induction motor model,
Publisher Item Identifier 10.1109/TIA.2002.802920. a simple – circuit has been used by most of the papers in
0093-9994/02$17.00 © 2002 IEEE
1298 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 38, NO. 5, SEPTEMBER/OCTOBER 2002

Fig. 1. Schematic of one of six possible active switch positions in a drive.

Phases A and B are in high potential and phase C is in low potential.

the literature [6]–[8]. This model is unable to capture all the

high-frequency content that is present in a pulsewidth modula- Fig. 2. Short-circuit impedance frequency response. Solid line: simulation
results; stars: experimental results.
tion (PWM) voltage pulse. Some references have shown the ne-
cessity to have a different model for the induction motor other
than a simple – circuit [4], [9]–[11]. Among them, [4], [10],
and [11] have successfully presented a high-frequency induction
motor model to calculate the overvoltage in long cable drives.
In this paper, an improved motor model is also developed based
on the model from [4], [10], and [11], which gives better results
in verifying the overvoltage phenomena in long cable drives.
The suggested high-frequency models for power cables and
motors are applied for specific case studies in the subsequent
sections. These models are implemented to perform computer
simulations using MATLAB. The simulation program is very
useful for the overvoltage analysis and provides a convenient
tool for a more reasonable design of the filters. The most
important filter networks are investigated and design
equations for each topology are presented. Through simulation
and analysis, an alternative design is suggested based on the
Fig. 3. Open-circuit impedance frequency response. Solid line: simulation
comparison between filter losses and motor terminal peak results. Stars: experimental results.
voltage. Experimental results of the overvoltage and of the
application of the filter are presented showing the validity cable characteristic impedance is calculated from short- and
of the simulation models. open-circuit impedances according to (1) [13]:

(1)
II. HIGH-FREQUENCY MODEL OF THE POWER CABLE
The impedance measurements have been carried out for an
An adequate estimation of the power cable parameters is unshielded four-wire cable (three phases one ground). The ex-
clearly needed in order to have an accurate computation of periments were conducted using 1-m samples of the following
the overvoltage phenomenon. Previous authors have suggested cable gauges: #6, #8, #10, #12, and #14 AWG. The frequency-
estimating the cable parameters by using equations related to dependent nature of the power cable impedances can be readily
the geometrical configuration of the cable. These approxima- verified from the frequency response plots. Figs. 2–4 show the
tions were found to be vastly different than actual parameter variation of the short-circuit, open-circuit, and characteristic im-
values because they do not include the frequency dependency pedances as a function of frequency for #6 AWG cable, re-
in the calculation. Herein, the cable parameters are estimated spectively. Similar plots have been obtained for the other cable
through experimental analysis by checking the frequency gauges. In these plots, the dots represent the frequency response
response of the power cable characteristic impedance. Fig. 1 obtained experimentally and the continuous line is the frequency
shows a schematic of one of six possible active switch positions response simulated in MATLAB using the proposed per-phase
in a typical inverter-fed motor drive. As can be seen, the high-frequency model per-unit length (Fig. 5).
connection between inverter and motor will have two phases in Since the overvoltage phenomena must be evaluated over the
parallel and the other phase as returning cable. Therefore, it is frequency spectrum of the voltage pulse, it would be incorrect
suggested to obtain the power cable characteristic impedance to assign a unique value for the cable parameters, which could
in such configuration. Two types of tests have been carried out: be associated to the voltage pulse dominant high frequency
measurement of short-circuit and open-circuit impedances. The [9]. Instead, the cable parameters can be estimated from the
MOREIRA et al.: CABLE AND INDUCTION MOTOR OVERVOLTAGE STUDIES IN LONG CABLE DRIVES 1299

TABLE I
PARAMETERS OF THE HIGH-FREQUENCY MODEL FOR VARIOUS CABLE GAUGES

Fig. 4. Cable characteristic impedance frequency response. Solid line:

simulation results. Stars: experimental results.

Fig. 6. Measurement of the phase-to-neutral impedance Z .

The frequency response calculated from the high-frequency

model is in good agreement with the measured characteristics
and the proposed lumped cable model per-unit length may be
used in the simulation analysis of the overvoltage phenomenon.
Table I summarizes the high-frequency model parameters
Fig. 5. Per-phase high-frequency model of the power cable per-unit length.
calculated for each cable gauge using the aforementioned
equations.
high-frequency model for which the calculated frequency re-
sponse matches the response obtained experimentally. The
parameters of the cable model are, thus, calculated by ana- III. HIGH-FREQUENCY MODEL OF THE AC MACHINE
lyzing the behavior of the short-circuit impedance ( ) and Another key factor for an accurate overvoltage analysis is the
open-circuit impedance ( ) over a broad range of frequency. high-frequency representation of the ac motor input impedance,
The cable model series parameters ( and ) are associated which must also be valid over the frequency range of the voltage
to the behavior of the short-circuit impedance, while the par- pulse. It is not necessary to verify how voltage distributes inside
allel parameters ( , and ) are associated to the ac machine winding in order to calculate the overvoltage.
the behavior of the open-circuit impedances. Equations (2)–(7) It is important, rather, to know the value of the ac motor input
are suggested to estimate the parameters of the cable model impedance and how it varies as a function of frequency.
per-unit length, in which and are the lowest (100 Hz) The proposed model for the ac motor input impedance is
and the highest (2 MHz) test frequencies in the impedance based on the high-frequency model suggested in [10] and [11],
measurements, respectively which has been successfully used in calculating the overvoltage
[4]. This section analyzes in detail the high-frequency ac motor
(2)
model with newly introduced modifications. The parameters
(3) of the model are derived using the frequency responses of
the phase-to-neutral impedance ( ) and phase-to-ground
impedance ( ), as suggested in [10] and [11].
(4)
Figs. 6 and 7 show schematics of the experimental setup to
measure the frequency responses of the phase-to-neutral and
(5) phase-to-ground impedances. The machines tested in laboratory
are dual-voltage ac induction motors. Dual-voltage machines
can commonly be connected in “delta” or “star” configuration,
(6) for a lower or a higher voltage, respectively. Usually, some of
the terminals of these machines are inaccessible, such as the ter-
minals numbered 10, 11, and 12 in Figs. 6 and 7. Therefore, the
(7)
following procedures are taken.
1300 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 38, NO. 5, SEPTEMBER/OCTOBER 2002

Fig. 7. Measurement of the phase-to-ground impedance Z .

Fig. 10. Phase-to-ground impedance frequency response. Solid line:

simulation results; stars: experimental results.

Fig. 8. Per-phase representation of the ac motor including the high-frequency

model and the dynamic dq model.

Fig. 11. Motor input impedance frequency response. Solid line: simulation
results; stars: experimental results.

TABLE II
PARAMETERS OF THE HIGH-FREQUENCY MOTOR MODEL FOR VARIOUS RATING

Fig. 9. Phase-to-neutral impedance frequency response. Solid line: simulation

results; stars: experimental results.

• Since terminals 10–12 are not accessible, the phase-to-

neutral and phase-to-ground impedances are measured for
the machine connected in delta configuration.
• Since the internal neutral (terminals 10–12) is not ac-
cessible, the phase-to-neutral impedance is measured
between terminals 1–2–3 (phase) and terminals 4–5–6 dynamic – model is partly responsible for capturing the
(external neutral). The measured values are, thus, divided low-frequency transients, while the remaining – – network
by 2, in order to account for the other parallel set of is responsible to represent the high-frequency phenomena.
windings (7–10, 8–11, and 9–12). Winding-to-ground capacitance and winding turn-to-turn ca-
Fig. 8 shows the proposed per-phase high-frequency motor pacitance play the major role in the high-frequency phenomena
model that is used in the calculation of the overvoltage analysis. [10], [11]. Their relation with the leakage inductance forms the
The suggested model is a lumped-parameter representation dominant poles in the frequency response.
of the motor input impedance for which the calculated fre- The parameter represents the winding-to-ground capac-
quency response matches the experimental response. The itance. The parameter is added in the circuit to represent
MOREIRA et al.: CABLE AND INDUCTION MOTOR OVERVOLTAGE STUDIES IN LONG CABLE DRIVES 1301

(a) (b)

Fig. 12. Overvoltage waveforms (3-hp induction motor; 20-m cable length). (a) Simulation. (b) Experimental.

(a) (b)

Fig. 13. Overvoltage waveforms (3-hp induction motor; 40-m cable length). (a) Simulation. (b) Experimental.

the dissipative effects that are present in the motor frame resis- The calculated frequency response using the proposed high-fre-
tance. The circuit formed by the parameters , and is quency model has been shown to be in good agreement with
the part of the network responsible to capture the second reso- the experimental results, and the proposed approach may be
nance in the frequency response, which is related to the winding used to represent the motor input impedance in the overvoltage
turn-to-turn capacitance. The parameter is responsible to ac- simulation analysis. Table II summarizes the parameters of the
count for the losses introduced by eddy current inside the mag- high-frequency model for various hp ratings of induction motor
netic core.
In order to estimate the parameters of the high-frequency
part of the model, it is suggested to replace the – model by (8)
a lumped-inductance , which represents the leakage induc-
tance of the machine winding. The high-frequency model pro- (9)
posed in Fig. 8 has been evaluated for various induction motor
power ratings. The behavior of the phase-to-neutral, phase-to- (10)
ground, and motor input impedances as a function of frequency
(11)
for a typical random-wound 3-hp induction motor are shown in
Figs. 9–11, respectively. From the frequency responses obtained (12)
experimentally, the parameters of the high-frequency model can
be calculated using the expressions (8)–(14) [10], [11], in which (13)
and are the lowest (1 kHz) and the highest (2 MHz)
test frequencies in the impedance measurements, respectively. (14)
1302 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 38, NO. 5, SEPTEMBER/OCTOBER 2002

(a) (b)

Fig. 14. Overvoltage waveforms (3-hp induction motor; 70-m cable length). (a) Simulation. (b) Experimental.

IV. SIMULATION AND EXPERIMENTAL ANALYSIS

OF THE OVERVOLTAGE
A flexible program was built in MATLAB that allows one to
evaluate the overvoltage phenomenon accurately for different
cable lengths and for different voltage pulse rise times. In the
program, one chooses the number of lumped sections as well
as the simulation time step. The whole system can be simulated
in a two-axes ( – ) approach, which reduces the number of
differential equations. The numerical integration is realized
using Runge–Kutta Method order 4 [4]. Simulation results
were obtained for a system with a 3-hp motor and #6 AWG
driving cable with different lengths. Experimental results were
also obtained to validate the simulation study. Figs. 12–14 show
the overvoltage waveforms for different power cable lengths.
In each figure, the left-side plots are the simulation realized in
MATLAB using the aforementioned high-frequency models, Fig. 15. Line-to-line voltage peak in pu versus rise time (3-hp induction motor;
20-m cable length). Solid line: simulation; stars: experimental.
and the right-side plots are the experimental waveforms. The
accuracy of the simulation models in predicting the overvoltage
is readily evident from the plots. Additional simulation and
experimental results are shown in Figs. 15 and 16, where the
behavior of the line-to-line voltage in the machine terminals
can be examined as a function of the voltage pulse rise times
for 20- and 70-m cable length, respectively. It can be concluded
from these results that the proposed overvoltage simulation
approach is a useful tool to predict, with reasonable accuracy,
the magnitude of the overvoltage in the machine winding.

V. MODELING AND DESIGN OF FILTER NETWORKS

Several filtering techniques have been proposed to mitigate
the overvoltage problem in long cable PWM drives [6]–[8], [12].
In general, these techniques include matching the cable charac-
teristic impedance to the motor input impedance and increasing
the voltage pulse rise time initiated at the inverter. The most rel- Fig. 16. Line-to-line voltage peak in pu versus rise time (3-hp induction motor;
evant filter networks are found to be: and filters at the 70-m cable length). Solid line: simulation; stars: experimental.
motor terminals [6], [7], [12], and clamping filters
at the inverter output [7], [8]. Table III presents a summary of the peak voltage at the motor terminals. In reality, a closed-form ex-
configuration and design equations of these approaches. These pression that gives the voltage peak in the motor terminal as a
equations are not sufficient to provide a suitable design for the function of the filter parameters is too complex to be useful for
filter topologies if one is looking for lower filter losses and lower design purposes.
MOREIRA et al.: CABLE AND INDUCTION MOTOR OVERVOLTAGE STUDIES IN LONG CABLE DRIVES 1303

TABLE III
DESIGN EQUATIONS OF THE MOST SUITABLE FILTER NETWORKS

Fig. 18. Voltage reflection coefficient at the motor terminals as a function of

Fig. 17. Matching characteristics between cable characteristic and motor input frequency for various filter resistors (cable: #6 AWG; motor: 3 hp).
impedances as a function of frequency (cable: #6 AWG; motor: 3 hp).

An alternative approach for the filter design could be based on

simulation analysis rather than a closed-form expression deriva-
tion. It is suggested here to utilize the overvoltage simulation
program developed in MATLAB to find more suitable values
for the filter topologies parameters. Simulation charts can be
derived for each filter approach showing motor terminal voltage
and filter losses as a function of the filter parameters. When per-
tinent, the concepts behind the filtering techniques may be used
together with the simulation analysis in order to derive the most
suitable design. Such design procedure for the filter is pre-
sented further for a 3-hp motor drive and #6 AWG power cable.
The first step is to verify the matching characteristics
between cable characteristic impedance and motor input
impedance, since this is the starting point of the doubling Fig. 19. Overvoltage versus filter losses (at 5 kHz) for various numbers of
effects occurrence. Fig. 17 shows the frequency dependence filter capacitors (in star connection) and for various cable lengths (motor: 3 hp;
of these impedances. It can be observed for frequencies above R = 
42
; voltage pulse rise time 100 ns).
tens of kilohertz that the impedances do match poorly. It is
not adequate to analyze the impedance matching for unique analysis must be carried out for a broad range of frequencies,
value of impedances for a particular frequency. Rather, the which would include the frequency responses of the cable
1304 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 38, NO. 5, SEPTEMBER/OCTOBER 2002

(a) (b)
Fig. 20. (a) Simulation and (b) experimental results of the RC filter at the motor terminals (motor: 3 hp; cable length: 70 m; RC filter: R = 126
and
C 
= 10 nF (delta connection); voltage pulse rise time 100 ns).

characteristic and the input motor impedances. The suggested

design practice is to make the filter resistance equal to the cable
characteristic impedance since the motor input impedance is
much greater than the cable characteristic impedance. The key
issue is the choice of the frequency at which the filter resistance
should be made equal to the cable characteristic impedance.
In the technical literature, the impedance matching is usually
analyzed at low frequencies around 100–200 Hz. This is not a
good choice because the overvoltage problem appears for high
frequencies above tens of kilohertz. Therefore, it is proposed to
match the filter resistance to the value of the cable characteristic
impedance for very high frequencies as shown in (15) for a star
connection of the filter, where and are parameters of
the per phase high-frequency cable model per-unit length

- (15)
Fig. 21. Experimental results of the line-to-line voltage at low motor current
level (motor: 3 hp; cable length: 70 m; RC filter: R = 126
and C =
Fig. 18 shows the simulation results of the voltage reflection 
10 nF (delta connection); voltage pulse rise time 100 ns.
coefficient at motor terminals for three values of filter resis-
tance, while the capacitance was fixed at F. the literature for the filter networks placed at the motor termi-
It can be observed that the best option is the value established in nals. In this case, the total inductance of the line must be consid-
(15), which gives the closest to a zero magnitude for the voltage ered during the filter capacitor design. Thus, it can be concluded
reflection coefficient and, thus, the lowest peak overvoltage at that the chart proposed by Fig. 19 is indeed very useful on the
the motor terminals. filter network design.
In the case of the filter capacitor, the objective is, through the Several simulation and experimental analyses have been real-
results of several simulation runs, to choose a capacitor value ized using the filter at the motor terminals. Selected results
for which the filter has lower losses and the motor experiences are presented here. Fig. 20 shows simulation and experimental
the lower line-to-line peak voltage. Fig. 19 shows the simula- results for the 70 m cable length case using
tion results for several values of filter capacitance. The choice and nF (filter in a delta connection). These plots
of the filter capacitor becomes a tradeoff between filter losses show the characteristics of the line-to-line voltage and the
and peak voltage at the machine terminals. As the capacitance current through one of the filter branches. The correlation
becomes large, the line-to-line peak voltage becomes lower and between simulation and experimental results can be readily
the filter losses become higher. Therefore, one must decide on verified.
the maximum allowable filter losses and then choose the filter One important missing detail that should be mentioned here
capacitance to obtain the desired peak voltage at the motor ter- is related to the initial charge conditions across the various en-
minals. It can be also verified from Fig. 19 that the filter capac- ergy storage elements of the cable model during the simulation.
itor choice to obtain the same level of line-to-line voltage is not At this point, it has not been possible to initialize the charge
unique, independent of cable length, as commonly expected in conditions of the cable in the proposed simulation program. It
MOREIRA et al.: CABLE AND INDUCTION MOTOR OVERVOLTAGE STUDIES IN LONG CABLE DRIVES 1305

Design equations of the most relevant filter networks have

been presented. Using the simulation program, a more thorough
design approach has been proposed examining the filter at
the motor terminals. Using this approach, the filter resistance
can be chosen according to the matching between cable charac-
teristic and motor input impedances in the frequency domain. In
addition, the filter capacitor is selected according to the desired
voltage stress and filter losses. The proposed simulation model
can be used in the design of the other filter networks as well.

REFERENCES
[1] E. Persson, “Transient effects in applications of PWM inverters to in-
duction motors,” IEEE Trans. Ind. Applicat., vol. 28, pp. 1095–1101,
Sept./Oct. 1992.
[2] A. H. Bonnett, “Analysis of the impact of pulse-width modulated in-
verter voltage waveforms on AC induction motors,” IEEE Trans. Ind.
Applicat., vol. 32, pp. 386–392, Mar./Apr. 1996.
Fig. 22. Experimental results of the line-to-line voltage at high motor current
level (motor: 3 hp; cable length: 70 m; RC filter: R = 126
and C =
[3] R. Kerkman, D. Leggate, and G. Skibinski, “Interaction of drive mod-

10 nF (delta connection); voltage pulse rise time 100 ns. ulation and cable parameters on AC motor transients,” in Conf. Rec.
IEEE-IAS Annu. Meeting, vol. 1, San Diego, CA, 1996, pp. 143–152.
[4] A. F. Moreira, T. A. Lipo, G. Venkataramanan, and S. Bernet, “Mod-
eling and evaluation of dv=dt filters for AC drives with high switching
is however true that the current level in the machine for which speed,” in Proc. 9th European Conf. Power Electronics and Applications
the switching transient occurs will produce different amplitudes (EPE’01), Graz, Austria, Aug. 27–29, 2001, CD-ROM.
of overvoltage in the machine terminals. Figs. 21 and 22 show [5] P. C. Krause and K. Carlsen, “Analysis and hybrid computer simulation
of multiconductor transmission systems,” IEEE Trans. Power App. Syst.,
experimental results demonstrating this fact. It can be observed vol. PAS-91, pp. 465–477, Mar./Apr. 1972.
that for higher current amplitudes, the line-to-line voltage is also [6] A. von Jouanne, D. Rendusara, and P. N. Enjeti, “Filtering techniques
higher. Incorporation of these issues is the subject of ongoing in- to minimize the effect of long motor leads on PWM inverter-fed AC
vestigations and the results will be published in the future. motor drive systems,” IEEE Trans. Ind. Applicat., vol. 32, pp. 919–926,
July/Aug. 1996.
[7] A. von Jouanne and P. N. Enjeti, “Design considerations for an inverter
output filter to mitigate the effects of long motor leads in ASD applica-
VI. CONCLUSIONS tions,” IEEE Trans. Ind. Applicat., vol. 33, pp. 1138–1145, Sept./Oct.
1997.
This paper has proposed a MATLAB-based program to be [8] S. J. Kim and S. K. Sul, “A novel filter design for suppression of high
voltage gradient in voltage-fed PWM inverter,” in Proc. IEEE APEC’97,
used in the analysis of the overvoltage phenomena in long cable vol. 1, Atlanta, GA, Feb. 23–27, 1997, pp. 122–127.
PWM drives. The following are the main characteristics of the [9] G. Skibinski, R. Kerkman, D. Leggate, J. Pankau, and D. Schlegel, “Re-
simulation program. flected wave modeling techniques for PWM AC motor drives,” in Proc.
IEEE APEC’98, vol. 2, Anaheim, CA, Feb. 15–19, 1998, pp. 1021–1029.
• The voltage pulse initiated at the inverter can be easily ad- [10] G. Grandi, D. Casadei, and A. Massarini, “High frequency lumped-pa-
justed to better represent the reality. This is an important fea- rameter Model for AC motor windings,” in Proc. EPE’97, vol. 2, Trond-
ture, since the characteristics of the overvoltage are strongly heim, Norway, 1997, pp. 578–583.
[11] A. Boglietti and E. Carpaneto, “Induction motor high frequency model,”
dependent on the harmonic contents of the voltage pulse. in Conf. Rec. IEEE-IAS Annu. Meeting, Phoenix, AZ, Oct. 3–7, 1999, pp.
• The power cable is represented as a transmission line with 1551–1558.
lossy characteristics and modeled using several lumped- [12] G. Skibinski, “Design methodology of a cable terminator to reduce re-
flected voltage on AC motors,” in Conf. Rec. IEEE-IAS Annu. Meeting,
parameter segments. The inclusion of distortion is very San Diego, CA, 1996, pp. 153–161.
important to obtain a close representation of the overvoltage. [13] P. C. Magnusson, G. C. Alexander, and V. K. Tripathi, Transmission
• A new model for the motor input impedance has been also Lines and Wave Propagation, 3rd ed. Boca Raton, FL: CRC Press,
suggested for the high-frequency analysis. Such a motor 1992, pp. 136–140.
model is very important to capture the high frequency
transients.
The parameters of the cable and motor models have been
derived from frequency response analysis of the cable charac-
teristic and motor input impedances, respectively, and a better Alessandro F. Moreira (M’94) was born in Coronel
Fabriciano, Brazil. He received the Electrical En-
representation of the cable–motor system is achieved for the gineering and the Master of Electrical Engineering
overvoltage analysis. degrees from the Universidade Federal de Minas
Several simulation and experimental results have been pre- Gerais, Belo Horizonte, Brazil, in 1991 and 1993, re-
spectively. He is currently working toward the Ph.D.
sented showing the usefulness of the simulation program on the degree at the University of Wisconsin, Madison.
analysis of the overvoltage. Overvoltage charts as a function of Since 1993, he has been with the Department
the voltage pulse rise time and cable length were generated that of Electrical Engineering, Universidade Federal de
Minas Gerais, as an Assistant Professor. His areas
allow one to accurately predict the voltage stress for various of interest include electric drive systems, power
situations. devices, and electromagnetic compatibility.
1306 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 38, NO. 5, SEPTEMBER/OCTOBER 2002

Thomas A. Lipo (M’64–SM’71–F’87) is a native of Steffen Bernet (M’97) was born in Ilmenau, Ger-
Milwaukee, WI. He received the B.E.E. and M.S.E.E. many, in 1963. He received the M.S. degree from the
degrees from Marquette University, Milwaukee, WI, Technical University of Dresden, Dresden, Germany,
in 1962 and 1964, respectively, and the Ph.D. degree in 1990 and the Ph.D. degree from the Technical
in electrical engineering from the University of Wis- University of Ilmenau, Ilmenau, Germany, in 1995,
consin, Madison, in 1968. both in electrical engineering. His doctoral thesis
From 1969 to 1979, he was an Electrical Engineer contained the investigation of power semiconductors
in the Power Electronics Laboratory of Corporate in soft switching converters.
Research and Development, General Electric He was a Development Engineer with the Depart-
Company, Schenectady, NY. He became a Professor ment of Private Communication Systems, Siemens,
of Electrical Engineering at Purdue University, West from 1994 to 1995. During 1995 and 1996, he
Lafayette, IN, in 1979 and, in 1981, he joined the University of Wisconsin, was a Post-Doctoral Researcher in the Electrical and Computer Engineering
Madison, in the same capacity. He is presently the W.W. Grainger Professor Department, University of Wisconsin, Madison. In 1996, he joined ABB
for Power Electronics and Electrical Machines, Co-Director of the Wisconsin Corporate Research, Heidelberg, Germany, where he led several power
Electric Machines and Power Electronics Consortium, and Director of the electronics research projects and developed various novel high-power con-
Wisconsin Power Electronics Research Center. verters including resonant snubber based current source converters, matrix
Dr. Lipo has received the Outstanding Achievement Award from the IEEE In- converters, and three-level voltage-source inverters. He led the Electrical
dustry Applications Society, the William E. Newell Field Award from the IEEE Drive Systems Group, ABB Corporate Research, Heidelberg, from 1998
Power Electronics Society, and the 1995 Nicola Tesla IEEE Field Award from to 2001. In 1999, he was appointed Subprogram Manager for Power Elec-
the IEEE Power Engineering Society for his work. Over the past 30 years, he tronics Systems, ABB Corporate Research Worldwide. In 2001, he joined
has served the IEEE in numerous capacities, including President of the IEEE the Technical University of Berlin, Berlin, Germany, as a Professor of Power
Industry Applications Society. Electronics. His main research areas are high-power converter topologies,
power semiconductors, and motor drives.

Giri Venkataramanan studied electrical engi-

neering at the Government College of Technology,
Coimbatore, India, California Institute of Tech-
nology, Pasadena, and the University of Wisconsin,
Madison.
After teaching electrical engineering at Montana
State University, Bozeman, he returned to the Uni-
versity of Wisconsin, Madison, as a faculty member
in 1999, where he continues to direct research in
various areas of electronic power conversion as
an Associate Director of the Wisconsin Electric
Machines and Power Electronics Consortium (WEMPEC). He is the holder
of four U.S. patents and has authored a number of published technical
papers.