IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO.

2, APRIL 2002 297

Passive Protection Strategy for a Drive System With

a Matrix Converter and an Induction Machine
Jochen Mahlein, Manfred Bruckmann, and Michael Braun

Abstract—In this paper, the design and testing of a new

protection strategy for a matrix converter feeding an induction
motor with a squirrel cage rotor is described. The new protection
strategy with excellent overvoltage protection allows the removal
of the large and expensive diode clamp.
Index Terms—Clamp circuit, insulated gate bipolar transistor
protection, matrix converter, over-voltage.

I. INTRODUCTION

T HE matrix converter is a simple —to- phase converter.

For common usage, and are normally set to 3. By
using a matrix of nine switches, the matrix converter is able to
generate a variable output voltage system of a desired frequency
Fig. 1. Matrix converter with a diode clamp circuit.
and magnitude [1]–[12]. The switches must be able to block
voltage and conduct current in both of its directions. They are
named bidirectional switches (BDSs). The matrix converter has
a maximum sinusoidal voltage transfer ratio of 0.86 and offers
power regeneration. If an ordinary LC filter is added, the grid
is stressed by sinusoidal current only. As mentioned above, the
matrix converter has some advantages and disadvantages versus
the dc-link voltage converter with a diode bridge as well as the
dc-link voltage converter with power regeneration and has to
compete with these two converter types.
A large problem of this converter is the absence of any pas-
sive free-wheeling path for the load current. This problem has
often been overcome by adding a simple rectifier bridge at the
input and output lines of the converter as shown in Fig. 1 [1],
[4], [6]. This diode bridge increases the number of the required
semiconductors and the costs of the converter. In [11], the diode
bridge is realized by adding extra diodes to the diodes contained
in the BDS, building a rectifier used as a free-wheeling path for
the load current. In all applications, the dc-link capacitor has to Fig. 2. BDS combinations.
be charged up by an extra circuit before the converter starts its
operation. This circuit causes additional costs. It can be shown low current ripple, it is necessary to choose a high switching
that the diode bridge is not necessary to operate the converter frequency for the converter. Insulated gate bipolar transistors
with an induction motor (IM). To design a matrix converter, a (IGBTs) offer these demands from low to high power levels
BDS has to be created by using standard power semiconductor making them a perfect device for this application.
devices because a BDS is not yet available as a single device.
For a proper commutation, it will be necessary to have a bidi- II. IGBT SWITCH COMBINATIONS
rectional device which can be controlled independently in both
current flow directions. To achieve a small filter design and a A BDS can be realized by connecting IGBTs in several ways
(Fig. 2). Each combination has different advantages as described
in [10]. The most used combinations are b) and c) because of
Manuscript received April 12, 2001; revised August 30, 2001. Abstract pub-
lished on the Internet January 9, 2002. their low conducting losses and the number of galvanic insulated
J. Mahlein and M. Braun are with the Elektrotechnisches Institut, University power supplies. These combinations fulfill the requirements of
of Karlsruhe, D-76128 Karlsruhe, Germany (e-mail: mahlein@eti.etec. a BDS as mentioned above. Combination b) offers a very easy
uni-karlsruhe.de).
M. Bruckmann is with Siemens AG, D-91056 Erlangen, Germany. way to build the converter in a modular structure as required
Publisher Item Identifier S 0278-0046(02)02893-9. for high power. The most favored combination for low power in
0278-0046/02$17.00 © 2002 IEEE
298 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL 2002

Fig. 3. Single output phase.

industrial usage is combination c). It uses a minimum of six in-

sulated power supplies to build up the converter. A new device, Fig. 4. Legal commutation sequences.
the reverse blocking IGBT (RBIGBT) [17] offers a simple an-
tiparallel structure for a BDS with excellent conducting losses
and all positive features of the other switches described. Unfor-
tunately, the switching losses of these devices are rather high
which will decrease the switching frequency of the converter.
This will lead in a larger filter design.

III. COMMUTATION SEQUENCES

A basic rule of matrix converter commutation is a switching
sequence which never interrupts the output currents to the
inductive load or short-circuit the input voltage sources. This
makes clear the demand for a BDS that can control the current
in both directions independently. Because of its extra power
losses and the extra costs, a clamp circuit is not desired. A
so-called “step-by-step” commutation strategy has to be chosen.
Existing commutation strategies can be voltage controlled as
described in [1] or current controlled as given in [7]. A mixture
of both is conceivable but not necessary. All possible switching
sequences can be investigated on a circuit as shown in Fig. 3.
The starting situation is a load current with transistors 1 and
2 switched on. For example, a valid switching sequence is 2314
which means that in the first step transistor 2 is switched off.
In the second step, transistor 3 is switched on, transistor 1 is
switched off in the third step, transistor 4 is turned on in the
fourth step. After the sequence has passed, the load current
has commutated on transistor 3. In order to deal with the delay
times of the IGBTs, a dead time between each step has to be
kept.
The results of all valid switching combinations can be seen
in Fig. 4. Combinations which are marked with ellipses need
the knowledge of the sign of the load current . The combi-
nations marked with rectangles need the sign of the voltage
. Nonmarked sequences will lead to a proper commutation
also, but the knowledge of voltage polarity and current direc-
tion is needed, making the signal processing more complicated
than required. The shown sequences are called “four-step” com-
mutations. As mentioned before, the matrix converter has no
passive free-wheeling path for the load current. In any event,
the topology of the converter offers the possibility to switch
free-wheeling paths as shown in [6] and [8]. In both approaches,
a so-called “two-step” commutation is suggested. The two-step
commutation obtains a free wheeling path by switching extra
switches of the matrix converter which will lead the load current
only in case of error. This strategy offers a good and easy way to
Fig. 5. Single-phase equivalent circuit diagram of the induction machine with
a squirrel cage in addition to protection circuits.
protect the IGBTs used for the BDS against over-voltage as long
as the signal processing is working. A shut down of the grid,
error in the signal processing, or emergency break will cause
problems which may lead to the destruction of the converter.
IV. OVERVOLTAGE PROTECTION
The diode bridge as described in Fig. 1 was the only solu-
tion for the four-step commutation. Due to the “step-by-step”
switching sequence as shown above, no extra free-wheeling path
for commutation is needed. The problem of breaking the load
current exists only in the case of a hard pulse-off shut down of
the converter. This will appear at the emergency OFF or con-
verter error. A shut down of the converter under normal opera-
tion mode can be done by reducing the power to the machine,
causing no interruption of the motor current. A free-wheeling
path for the load current at a pulse-off shut down has to be found.
V. VARISTOR PROTECTION
Assuming that the matrix converter should be a low-cost so-
lution in a low-power range, it should be used in teamwork with
an induction motor with a squirrel cage rotor. If the induction
motor is shut down by a pulse-off, the recovered energy is ex-
pected to be the stored energy in the motor leakage inductances.
A worst-case estimation for a whole induction motor protected
by a diode clamp circuit is given in [4]
(1)
is the maximum amount of the recovered energy
transferred to the clamp capacitor, is the stator leakage
inductance, and the rotor leakage inductance calculated
to the stator side. is the magnitude of the stator current. A
MAHLEIN et al.: PASSIVE PROTECTION STRATEGY FOR A DRIVE SYSTEM
Fig. 6. Estimated maximum energy of the motor leakage inductances from the IM equivalent circuit diagram.
worst-case assumption is that the magnetizing current of the ma-
chine equals zero. Based on this formula, the maximum energy
of a single phase of the machine can be assumed to be
(2)
The estimations given here are taken from the single-phase
equivalent circuit diagram of the induction machine with a
capacitor connected by a rectifier bridge to the motor lines
(Fig. 5). The values of a two-pole and four-pole induction
motor and the amount of the recovered energy for single-phase
equivalent circuit diagram can be seen in Fig. 6. It is shown that
the recovered energy is rather small. If a varistor is connected
between the motor lines, the stator current will go down faster
with a “time constant”
(3)
which is not fixed during the whole pulse-off due to the expo-
nential resistant characteristic of the device. The value of
depends on the voltage of the magnetizing inductance in the
stator loop during the transient process but is rather high com-
pared to the stator resistance of the machine. Simulations on
the circuit in Fig. 5 have shown that the stator current reaches
a value of zero faster by connecting a varistor as shown in this
figure, and, as a result, some of the stored energy as calculated
in (1) is transferred to the rotor side where it is absorbed by the
rotor resistance. A closed formula for the recovered energy on
the stator side can hardly be given in this paper. As a maximum
value for one phase, the results from (2) can be used. Having a
look at varistor data sheets [15], [16], the amount of energy is
not given but the duration and the magnitude of the pulse current
to be surged. Simulation results and measurements have shown
that a duration on standard machines [14] will not last longer
299
Fig. 7. Matrix converter with varistor protection.
than 200 s. A carefully selected varistor is able to stand this
stress for more than 1 million pulse-off cycles which is enough
compared to the lifetime of the converter. Higher stress of the
varistors such as short circuit or ground failure will decrease the
number of pulse-off cycles. Another point is the admission of
higher motor currents at speed changes to improve the dynamic
of the drive system. This is a critical aspect because a doubled
nominal current will lead to a 16 times greater value in the stored
energy of the machine, according to (1). As a result, a larger suit-
able varistor device has to be selected from the data sheets. The
selected varistors can be added in triangle configuration at the
input and output lines of the converter (Fig. 7). The varistors at
the input lines protect the converter against voltage transients
from the grid which is needed for electromagnetic immunity
(e.g., IEC 801 standards). The varistors at the motor lines take
the pulse-off energy of the induction machine. Measurements on
the varistor behavior have been performe on a 30-kW, four-pole
induction motor. Fig. 8 shows a pulse-off at 50-A stator current.
300
Fig. 8. Varistor voltage/current at pulse-off.
Fig. 9. Equivalent circuit diagram of a varistor.
Fig. 10. Varistor current during normal operation.
The energy absorbed by the varistor in Fig. 8 is about 3 Ws. This
was expected due to (2).
During normal operation of the converter, the varistor be-
havior has been examined. In Fig. 9, the equivalent circuit dia-
gram of the real varistor is shown. The varistor behavior can be
explained by an ideal varistor having a parasitic capacitor in par-
allel. The parasitic capacity is caused by the connection flange
and the grain boundaries of the metal oxide material. The typical
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL 2002
Fig. 11. Unbalanced blocking voltages at pulse-off.
value reaches from a few up to thousands of picofarads at 1 kHz
depending on the size and electrical values of the device. A small
stray inductivity is in serial with this configuration. Its amount
is rather small and can be neglected in this application. Fig. 10
shows the varistor behavior during normal operation mode. The
current flowing through the device is the reload current of the
parasitic capacity at polarity changes of the output voltages of
the converter. This current is small and brief. The losses caused
are not worth mentioning. The calculated parasitic capacity out
of this measurements is 500 pF which was expected due to the
data sheets of the device [16].
To select a suitable varistor device, the problem of unbal-
anced blocking voltages has to be solved. Examining Fig. 11,
the varistors are connected between the load lines. The tran-
sient voltage/current characteristic curve of a typical varistor
will have its clamp voltage over the maximum blocking voltage
of a single IGBT at maximum load current. If a shutdown of the
MAHLEIN et al.: PASSIVE PROTECTION STRATEGY FOR A DRIVE SYSTEM 301

Fig. 12. Gate driver with suppressor diode protection.

(a)

converter is initiated, all BDSs are gated off. A problem occurs if

a BDS (e.g., BDS32) is delayed in its blocking capability before
the rest of the BDSs. The neighbored IGBT (BDS31) having its
full blocking capability may get the maximum clamping voltage
of the varistor causing damage to this device. An extra circuit
has to be added to protect the single IGBT against this transient
over-voltage at a pulse-off.

VI. VARISTOR/SUPPRESSOR DIODE PROTECTION

To protect a single IGBT against over-voltage, a simple cir-
cuit can be found. The IGBTs are protected by a suppressor
diode. This diode has the characteristic of a Z-diode with a high
breakdown voltage. The value of the breakdown voltage has to (b)
be lower than the maximum blocking voltage of the IGBT. The Fig. 13. Varistor/suppressor diode protection. (a) Voltages. (b) Currents.
circuit for a single IGBT can be seen in Fig. 12. If the collector
emitter voltage of the IGBT increases to a value higher than that
of the breakdown voltage of the suppressor diode, this diode
becomes conductive. The gate of the IGBT is charged again.
The IGBT becomes conductive in its nonsaturated character-
istic. This operation causes high losses in the transistor. For-
tunately, this happens only until all IGBTs are off. The losses
will not do any harm to the IGBT chip. Each IGBT can be pro-
tected this way. To ensure a good performance and lifetime of
the converter a combination of both protection strategies is sug-
gested. In Fig. 13, the measurements on a varistor/suppressor
diode protected matrix converter with 7.5-kW induction motor
can be seen. In the voltage diagram the collector emitter volt-
ages of two neighbored IGBTs are shown. If a pulse-off is initi-
ated, the collector emitter voltages on both transistors increase
rapidly over 800 V. The suppressor diodes are getting active and
clamping the IGBTs to a maximum of 800 V. This process is Fig. 14. Suppressor diode behavior.
rather short, which can be seen in Fig. 14. The suppressor diode
gives a short but high current pulse to the gate, cutting the sharp
VII. SUPPRESSOR DIODE PROTECTION
voltage peak of the IGBT at a turn off. The switch off time of
IGBT 2 is delayed by this process, giving time to IGBT 1 to take Looking at Fig. 15, the energy absorption capability of
its blocking voltage. The varistor becomes active in its low resis-
IGBTs is shown. A 100- s pulse is assumed due to the re-
tance characteristic. As the current diagram in Fig. 13 shows, the
is fully taken by the varistor for the duration of sults of the measurements as shown in Fig. 8. The question if
load current
the process. These measurements show that the IGBT is hardly the matrix converter can be protected by its own IGBTs using
stressed by this protection method. The main part of the energy suppressor diodes only without varistors must be discussed.
is moved to the varistors which are able to take this energy quite Compared to the expected maximum energy of the induction
easily. The problem of unbalanced voltage at pulse-off has been motor (Fig. 6), it is not easy to decide if an IGBT can stand the
overcome. stress of the pulse-off energy without a varistor device. For a
302 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 49, NO. 2, APRIL 2002

TABLE I
SECURITY FACTOR OF CHOSEN IGBTs

Fig. 15. Energy absorption capability of IGBTs.

certain drive, the capability of intrinsic safety can be classified

by defining a safety factor

Energy absorption capability IGBT

Safety factor (4)
Recovered energy motor

For example, a 7.5-kW two-pole induction motor at a

230/400-V grid has a nominal current of 14 A [14]. Assuming
that the chosen nominal current of IGBTs should be about
double the nominal motor current, a 35-A transistor type is
selected. According to Figs. 6 and 15, it will stand the stress
of pulse-off without problems. The resulting safety factor is
1.05 (see Table I). In this case, no varistor at the load side
is needed. The matrix converter can be protected only by
suppressor diodes. The calculations for larger motors show
smaller results. For a 15-kW motor, the safety factor is 0.54
which is to small. An IGBT with higher nominal current must
be chosen.
To avoid the function of the varistor in an experimental setup,
the suppressor breakdown voltage has been reduced to 600 V.
At this voltage, the varistor resistance stays rather high and Fig. 16. IGBT behavior at pulse-off.
does not take much current. Measurements on these suppressor
diode-protected IGBTs have been done on a smaller 7.5-kW at a pulse-off during this operation has a much higher value,
induction motor (Fig. 16). A large amount of the load current causing additional stress to the IGBTs that will be damaged. As
is flowing through the IGBT during the whole pulse-off. a result, a larger IGBT has to be chosen leading to extra costs.
The varistor is nearly inactive. This measurement shows that the An advanced packaging technology or the usage of robust semi-
varistors can be removed in this application. The matrix con- conductor material such as silicon carbite may overcome this
verter can be protected by its own IGBTs. problem.
The question of whether an IGBT can stand this stress in the If no additional stress is desired to the IGBTs, the combi-
long run cannot be answered by this study. Further investiga- nation of varistor and suppressor diode protection is an excel-
tions on IGBT lifetime behavior must be performed. In addition, lent combination to protect the matrix converter in all operation
a few other facts must be taken into account for a matrix con- modes. No precharge circuit is needed. The converter can be set
verter drive system. In the converter design, the case of an over to operation without any delay times. Because of the small peak
current shut down has to be considered. Another point is the ad- power of the suppressor diodes taken at pulse-off, the device can
mission of a higher motor current (e.g., 150–200%) at typical be integrated with the gate unit to the power modules. This in-
speed changes in standard drive systems. The recovered energy telligent module would be a step closer to system integration.
MAHLEIN et al.: PASSIVE PROTECTION STRATEGY FOR A DRIVE SYSTEM
VIII. CONCLUSION
In this paper, a new strategy has been presented to protect
a matrix converter system with an induction motor against
over-voltage in all operation modes. The suggested strategy
with suppressor diodes and varistors works very well. Mea-
surements and results on existing drives have been presented
and discussed. The strategy offers an additional possibility
to remove the needed varistors. In this low-cost strategy, the
IGBTs are used to protect themselves against over-voltages
from the load. This operation mode is practicable but causes
hard working conditions to the semiconductors, which must be
considered.
REFERENCES
[1] A. Alesina and M. Venturini, “Analysis of optimum-amplitude nine-
switch direct AC–AC converters,” IEEE Trans. Power Electron., vol. 4,
pp. 101–112, Jan. 1989.
[2] M. Braun, “Ein dreiphasiger Direktumrichter mit Pulsweitenmodulation
zur getrennten Steuerung der Ausgangsspannung und der Eingangs-
blindleistung,” Dr.-Ing. dissertation, Tech. Hochschule Darmstadt,
Darmstadt, Germany, 1983.
[3] W. Söhner, “Der selbstgeführte Direktumrichter und seine Anwendung
als Matrix-Converter zur Speisung von drehzahlgeregelten Antrieben
mit Asynchronmaschine,” Ph.D. dissertation, Univ. Fridericiana, Karl-
sruhe, Germany, 1993.
[4] P. Nielsen, “The matrixconverter for an induction motor drive,” Ph.D.
dissertation, Aalborg Univ., Aaalborg East, Denmark, 1996.
[5] M. Marcks, “Steuerung, Regelung und Konstruktion eines Matrixum-
richters sowie seine Möglichkeiten zum stromlosen Schalten,” Dr.-Ing.
dissertation, Tech. Univ. Darmstadt, Darmstadt, Germany, 1998.
[6] M. Ziegler, “Performance of a two steps commutated matrix converter
for AC-variable-speed drives,” in Proc. EPE’99, Lausanne, Switzerland,
1999, CD-ROM Paper 258.
[7] P. Wheeler and D. Grant, “A low loss matrix converter for AC variable-
speed drives,” in Proc. EPE, Brighton, U.K., 1993, pp. 27–32.
[8] L. Empringham, P. Wheeler, and J. Clare, “Bi-directional switch com-
mutation for matrix converters,” in Proc. EPE, Lausanne, Switzerland,
1999, CD-ROM Paper 409.
[9] L. Huber and D. Borojevic, “Space vector modulated three-phase to
three-phase matrix converter with power factor correction,” IEEE Trans.
Ind. Applicat., vol. 31, pp. 1234–1246, Dec. 1995.
[10] J. Mahlein, O. Simon, and M. Braun, “A matrix converter with space
vector modulation enabling overmodulation,” in Proc. EPE, Lausanne,
Switzerland, 1999, CD-ROM Paper 394.
[11] P. Nielsen, F. Blaabjerg, and J. K. Pedersen, “Novel solutions for pro-
tection of matrix converter to three phase induction machine,” in Conf.
Rec. IEEE-IAS Annu. Meeting, vol. 2, 1997, pp. 1447–1454.
[12] O. Simon and M. Braun, “Theory of vector modulation for matrix con-
verter,” in Proc. EPE, Graz, Austria, 2001, CD-ROM Paper 19.
303
[13] Power Semiconductors & IGBT Modules—Siemens Data Book 05.96,
Siemens AG, Munich, Germany, 1996.
[14] Drehstrom-Niederspannungsmotoren—Siemens Data Book M11,
Siemens AG, Munich, Germany, 1998.
[15] Data Book “SIOV Metal Oxide Varistors”, Siemens Matsushita Com-
ponents, Munich, Germany, 1999.
[16] “Data sheet 4366.2; ‘Ultramov™ series—High energy radial lead metal
oxide varistors’,” Harris Semiconductor, Feb. 1998.
[17] Data sheet “IXRH 50R80/60 IGBT with reverse blocking capability”.
IXYS Semiconductors Advanced Technical Information. [Online].
Available: www.ixys.net/l400.pdf.
Jochen Mahlein received the diploma engineer
title from the University of Karlsruhe, Karlsruhe,
Germany, in 1996.
Since 1996, he has been an Assistant Professor
in the Electrotechnical Institute, University of
Karlsruhe. He teaches courses in “Control of Elec-
trical Drives” and “Practice of Electrical Drives.” His
research areas are direct converters, soft-switching
converters, and control of electrical drives.
Manfred Bruckmann received the Dipl.-Ing.
degree in electrical engineering from Friedrich
Alexander University, Erlangen, Germany, in 1991.
Since 1991, he has been with the Department for
Innovation and Technology, Siemens AG, Erlangen,
Germany. His research interests are new power semi-
conductor devices, high-voltage IGBT applications,
and new converter topologies.
Michael Braun received the Dipl.-Ing. and Dr.-Ing.
degrees from the Technische Hochschule Darmstadt,
Darmstadt, Germany, in 1978 and 1983, respectively.
From 1983 to 1994, he was with Siemens AG,
Erlangen, Germany, where he was engaged in the de-
velopment of power electronics and electrical drives.
In 1994, he joined the University of Karlsruhe,
Karlsruhe, Germany, where, since 1994, he has
been a Professor and Head of the Electrotechnical
Institute. His research areas are power electronics,
control of electrical drives, and mechatronics.