Julkaisu 686 Publication 686

Matti Jussila
Comparison of Space-Vector-Modulated Direct and
Indirect Matrix Converters in Low-Power Applications
Tampere 2007


Tampereen teknillinen yliopisto. J ulkaisu 686
Tampere University of Technology. Publication 686
















Matti J ussila

Comparison of Space-Vector-Modulated Direct and
Indirect Matrix Converters in Low-Power Applications

Thesis for the degree of Doctor of Technology to be presented with due permission for
public examination and criticism in Rakennustalo Building, Auditorium RG202, at
Tampere University of Technology, on the 23rd of November 2007, at 12 noon.
























Tampereen teknillinen yliopisto - Tampere University of Technology
Tampere 2007




















































ISBN 978-952-15-1862-1 (printed)
ISBN 978-952-15-1906-2 (PDF)
ISSN 1459-2045


iii
Abstract
Matrix converters are frequency converters which do not contain a direct current link circuit
with passive components, unlike conventional frequency converters. Thus, matrix converters
may provide a solution for applications where large passive components are not allowed or a
purely semiconductor-based solution provides an economically more efficient result than
conventional frequency converters. However, the lack of a link circuit may also be a drawback
in non-ideal operation conditions and the direct structure also places restrictions on converter
capability.
This thesis concerns two space-vector-modulated matrix converter topologies. The topologies
are a direct and indirect matrix converter, which are compared in low-power applications. The
comparison is based on the space vector analysis and modelling of the matrix converter
topologies concerned. The comparison is confirmed both in the simulations and in the
experimental tests with the prototypes built. In addition to the comparison, the thesis also
contains the space vector form analysis of the matrix converter operation under distorted
supply and load conditions.
The space vector theory is used for the explicit presentation of matrix converter modulation. In
addition, some space vector modulation methods are compared from the point of view of the
power losses and the output common-mode voltages produced. The space vector approach is
used to analyse the migration of supply voltage distortion to output and its reflection back to
the supply side. Three computationally simple methods for mitigation of the distortion
migration are also compared. These new analytical results are confirmed by simulations and
measurements: the distortion migration can be mitigated but not removed totally and the
mitigation also increases the complexity of control and decreases supply current quality.
The comparison of direct and indirect matrix converter topologies is based on the analyses of
their non-ideal characteristics. Their voltage transfer characteristics, semiconductor power
losses and supply current qualities are studied. The analyses are again tested either
computationally or in the simulations and experiments. In addition to the passive RL load, the
applications also contain a space-vector-controlled cage induction machine drive and a
permanent magnet synchronous machine drive. The research shows that the two topologies are
similar in the ideal case but after non-ideal characteristics are introduced, the direct topology
shows better characteristics in nearly all loading situations tested, i.e. the direct matrix
converter has more linear voltage transfer characteristics, lower power losses and higher
supply current quality than the indirect matrix converter.
Index terms: Frequency converter, matrix converter, space vector modulation.
iv

v
Preface
This work was carried out in the Institute of Power Electronics at Tampere University of
Technology during the years 20022007. The research was funded by Tampere University of
Technology, Tekes (Finnish Funding Agency for Technology and Innovations), Fortum
Foundation and industrial partners (ABB Oy Drives, Eaton Power Quality, Hyvinkn
Techvilla, Kalmar Industries, KCI Konecranes Oyj, Kemppi Oy, Kone Oyj, Vacon Oyj).
First of all, I thank my supervisor, Professor Heikki Tuusa, who made my research possible
and guided me through the difficult regions of power electronic converters. In addition, I thank
Dr. Mika Salo for his guidance at the beginning of the research work, Mr. Pentti Kivinen for
all the help with the prototypes and Jarno Alahuhtala, M.Sc., for research cooperation. I owe a
debt of gratitude to all colleagues at the Institute of Power Electronics during the years 2000
2007, who shall never be forgotten. Special thanks go to my former roommate Mikko
Routimo, Lic. Tech., to Dr. Tero Viitanen and to Juha Turunen, Lic. Tech., for their friendship
and company at sauna parties. Similar gratitude is owed to Dr. Matti Eskola, whose
cooperation in research and contribution to history-related discussions deserve much praise.
I am also very thankful to Danny Donoghue, who proofread the manuscript and improved my
language significantly. In addition, I thank Professor Jorma Luomi and Dr. Pat Wheeler for the
pre-examination of this thesis.
I express my grateful appreciation for the personal grants from Emil Aaltonen Foundation,
Fortum Foundation and Tekniikan edistmissti (Foundation of Technology). The financial
assistance and mental support provided by their grants has been of crucial importance to me.
Finally, I thank my parents for their sympathy.






Vantaa, October 2007



Matti Jussila




vi
Contents
Abstract................................................................................................................................... iii
Preface....................................................................................................................................... v
Contents .................................................................................................................................. vi
List of Publications .............................................................................................................. viii
Related Publications.......................................................................................................... ix
Symbols and Abbreviations .................................................................................................... x
1 Introduction........................................................................................................................ 1
1.1 Overview of PWM Frequency Converters with dc Link............................................. 2
1.2 History of Direct Frequency Converters...................................................................... 4
1.3 Objectives and Outline of the Thesis........................................................................... 7
1.4 Contribution of the Thesis ........................................................................................... 8
2 Matrix Converter Systems ................................................................................................ 9
2.1 Topologies ................................................................................................................... 9
2.1.1 Bidirectional Switches ................................................................................................ 9
2.1.2 Single-Stage Topologies ........................................................................................... 10
2.1.3 Two-Stage Topologies .............................................................................................. 12
2.2 Ensuring Safe Operation............................................................................................ 15
2.2.1 Switch Commutations............................................................................................... 15
2.2.2 Overvoltage Protection ............................................................................................. 21
2.3 Conclusion ..................................................................................................................24
3 Matrix Converter Control .............................................................................................. 25
3.1 Space Vector Theory ................................................................................................ 25
3.2 Overview of Matrix Converter Modulation............................................................... 28
3.2.1 Review of Modulation Methods ............................................................................... 30
3.2.2 Compound Control and Modulation Strategies......................................................... 32
3.3 Space Vector Modulation .......................................................................................... 33
3.3.1 Derivation of Duty Cycles ........................................................................................ 33
3.3.2 Switching Pattern Optimisation ................................................................................ 38
3.3.3 Angle Detection of Input Reference Space Vector ................................................... 44
3.4 Conclusion ..................................................................................................................46
4 Matrix Converter Modelling........................................................................................... 47
4.1 Main Circuit Models.................................................................................................. 47
4.2 Converter Models ...................................................................................................... 51
4.2.1 Models of Direct Matrix Converter .......................................................................... 51
4.2.2 Models of Indirect Matrix Converter ........................................................................ 53
4.2.3 Comparison of Simulation Models ........................................................................... 54
Contents vii
4.3 Induction Machine Modelling and Control System.................................................. 55
4.3.1 Cage Induction Machine Model ................................................................................ 55
4.3.2 Cage Induction Machine Control System...................................................................57
4.3.3 Induction Motor Drive Simulations............................................................................58
4.4 Conclusion.................................................................................................................. 61
5 Experimental Setup ........................................................................................................ 62
5.1 Prototype Implementations ....................................................................................... 62
5.1.1 Modulator Implementations....................................................................................... 66
5.1.2 Induction Motor Drive............................................................................................... 68
5.2 Measurement Equipment........................................................................................... 69
5.3 Experimental Results................................................................................................. 69
5.3.1 Passive Load .............................................................................................................. 69
5.3.2 Induction Motor Load................................................................................................ 72
5.4 Conclusion.................................................................................................................. 74
6 Matrix Converter under Distorted Conditions............................................................. 75
6.1 Space Vector Analysis of Distortion Migration in Matrix Converters ..................... 75
6.1.1 Input Voltage Distortion Migration ........................................................................... 76
6.1.2 Output Current Distortion Migration......................................................................... 78
6.1.3 Confirmation of Migration Analyses......................................................................... 79
6.2 Mitigation of Voltage Distortion Migration.............................................................. 80
6.2.1 Control Methods ........................................................................................................ 81
6.2.2 Comparison of Control Methods ............................................................................... 83
7 Comparison of Non-Ideal Characteristics of Direct and Indirect Matrix Converter 85
7.1 Supply Current Quality ............................................................................................. 85
7.1.1 Comparison of DMC and IMC Current Distortions .................................................. 85
7.2 Voltage Transfer Characteristics............................................................................... 87
7.2.1 IMC Voltage Transfer Characteristics....................................................................... 88
7.2.2 DMC Voltage Transfer Characteristics ..................................................................... 90
7.2.3 Comparison of DMC and IMC Voltage Transfer Characteristics ............................. 91
7.3 Semiconductor Power Losses.................................................................................... 94
7.3.1 On-State Losses ......................................................................................................... 94
7.3.2 Switching Losses ....................................................................................................... 96
7.3.3 Comparison of DMC and IMC Power Losses ........................................................... 97
7.4 Summarised Comparison of Direct and Indirect Matrix Converter ........................ 100
8 Summaries and Contribution of the Publications...................................................... 101
9 Conclusions .................................................................................................................... 104
References ............................................................................................................................ 106
Appendix A........................................................................................................................... 116
Appendix B........................................................................................................................... 118
Publications


viii
List of Publications
This thesis consists of an overview and the following publications:
[P1] Jussila, M., Salo, M. and Tuusa, H. (2004). Induction motor drive fed by a vector
modulated indirect matrix converter. Proceedings of the 2004 IEEE Power Electronics
Specialists Conference, PESC04, Aachen, Germany, June 2025, vol. 4, pp. 2862
2868.
[P2] Jussila, M., Alahuhtala, J. and Tuusa, H. (2006). Common-mode voltages of space-
vector modulated matrix converters compared to three-level voltage source inverter.
Proceedings of the 2006 IEEE Power Electronics Specialists Conference, PESC06,
Jeju, Korea, June 1822, pp. 923929.
[P3] Jussila, M. and Tuusa, H. (2007). Comparison of simple control strategies of space-
vector modulated indirect matrix converter under distorted supply voltage. IEEE
Transactions on Power Electronics, vol. 22, January 2007, pp. 139148.
[P4] Jussila, M. and Tuusa, H. (2007). Illustration of relation between load and supply
current distortions in direct and indirect matrix converters. Proceedings of the 2007
IEEE Power Electronics Specialists Conference, PESC07, Orlando, USA, June 17
21, pp. 25222528.
[P5] Jussila, M., Eskola, M. and Tuusa, H. (2006). Characteristics and comparison of output
voltage non-idealities of direct and indirect matrix converters. International Review of
Electrical Engineering by Praise Worthy Prize, IREE, vol. 0, no. 0, February 2006, pp.
7482.
[P6] Jussila, M. and Tuusa, H. (2007). Semiconductor power loss comparison of space-
vector modulated direct and indirect matrix converter. Proceedings of the 2007 IEEE
Power Conversion Conference, PCC 2007, Nagoya, Japan, April 25, pp. 831838.
[P7] Jussila, M. and Tuusa, H. (2006). Comparison of direct and indirect matrix converters
in induction motor drive. Proceedings of the 2006 IEEE Industrial Electronics
Conference, IECON06, Paris, France, November 610, pp. 16211626.
The author has written Publications [P1][P7] with help and guidance from Professor Heikki
Tuusa. In Publication [P1], help and guidance were also provided by Dr. Mika Salo. The
author was responsible for all experiments and simulations. The author made all the prototypes
and the software used in the experimental tests and the simulation models with two
exceptions: 1) In Publication [P2], Jarno Alahuhtala, M.Sc. provided guidance with the text
relating to modulation of a three-level voltage source inverter and implemented the simulation
List of Publications ix
model of the three-level voltage source inverter. 2) In Publication [P5], the control system of
the permanent magnet motor used in the measurements was designed and implemented by Dr.
Matti Eskola, who also provided guidance with the text concerning permanent magnet motor
control.
Related Publications
The following publications relate to the topic of this thesis:
[RP1] Jussila, M., Salo, M. and Tuusa, H. (2002). Comparison of modulation methods for
matrix converters. Proceedings of the 2002 Nordic Workshop on Power and Industrial
Electronics, NORPIE 2002, Stockholm, Sweden, August 1214, 6 p.
[RP2] Jussila, M., Salo, M. and Tuusa, H. (2003). Implementation of a three-phase indirect
matrix converter with an indirect vector modulation method applying a
microcontroller. Proceedings of the 2003 International Exhibition and Conference for
Power Electronics, Intelligent Motion, Power Quality, PCIM Europe 2003,
Nuremberg, Germany, May 2022, vol. Intelligent Motion, pp. 5357.
[RP3] Jussila, M., Salo, M. and Tuusa, H. (2003). Realization of a three-phase indirect matrix
converter with an indirect vector modulation method. Proceedings of the 2003 IEEE
Power Electronics Specialists Conference, PESC03, Acapulco, Mexico, June 1519,
vol. 3, pp. 689694.
[RP4] Jussila, M., Salo, M. and Tuusa, H. (2003). Comparison of two vector modulated
matrix converter topologies. Proceedings of the 2003 European Conference on Power
Electronics and Applications, EPE 2003, Toulouse, France, September 24, 9 p.
[RP5] Jussila, M., Salo, M., Khknen, L. and Tuusa, H. (2004). A vector modulated three-
phase four-quadrant rectifier application to a dc motor drive. Proceedings of the
2004 Nordic Workshop on Power and Industrial Electronics, NORPIE 2004,
Trondheim, Norway, June 1416, 5 p.
[RP6] Eskola, M., Jussila, M. and Tuusa, H. (2004). Indirect matrix converter fed PMSM-
sensorless control using carrier injection. Proceedings of the 2004 IEEE Power
Electronics Specialists Conference, PESC04, Aachen, Germany, June 2025, vol. 5,
pp. 40104016.
[RP7] Jussila, M. and Tuusa, H. (2005). Space-vector modulated indirect matrix converter
under distorted supply voltage Effect on load current. Proceedings of the 2005 IEEE
Power Electronics Specialists Conference, PESC05, Recife, Brazil, June 1216, pp.
23962402.
[RP8] Jussila, M., Eskola, M. and Tuusa, H. (2005). Analysis of non-idealities in direct and
indirect matrix converters. Proceedings of the 2005 European Conference on Power
Electronics and Applications, EPE 2005, Dresden, Germany, September 1114, 10 p.
x
Symbols and Abbreviations
Some of Publications [P1][P7] contain notations different from those in the list below.
Symbols follow [IEC95] and [IEC04], with some exceptions made to avoid ambiguity.
List of Abbreviations
A/D analogue to digital
ac alternating current
BBVSC back-to-back converter
CSC current source converter
CSVM conventional space vector modulation method
dc direct current
DMC direct matrix converter
ECSVM easy commutation space vector modulation method
GAL gate array logic
IGBT insulated gate bipolar transistor
ILB load-side bridge of indirect matrix converter
IM induction machine
IMC indirect matrix converter
ISB supply-side bridge of indirect matrix converter
ISVM improved space vector modulation method
MC matrix converter
Meas. measured
Meth. method
MOSFET metal-oxide-silicon field effect transistor
NZSVM non-zero space vector modulation method
OECD Organisation for Economic Co-operation and Development
PCC point of common coupling
PI proportional-integral
PLL phase-locked loop
PMSM permanent magnet synchronous motor
PWM pulse width modulation
RBIGBT reverse blocking IGBT
rms root-mean-square
Sim. simulated
SVM space vector modulation
THD total harmonic distortion
Symbols and Abbreviations xi

TPU time processor unit
VSC voltage source converter (two-level)
VSI voltage source inverter (two-level)
3LBBVSC three-level back-to-back voltage source converter
3LVSC three-level voltage source converter
List of Symbols
A, B, C output or load phases
Ax magnitude of x times frequency resolution component in spectral plots
A1 magnitude of fundamental wave in spectral plots when the frequency
resolution is the fundamental frequency
a e
j2/3

a, b, c input or supply phases
B friction constant
C capacitance, capacitor
d duty cycle, relative time
E energy
e controller error variable
f frequency
G transfer function
I, i current: rms value, instantaneous value
J moment of inertia
j imaginary unity
k present sample in time-discrete system
K
P
proportional action coefficient of PI controller
L, L
m
inductance or inductor, magnetising inductance
m, m
ideal
, m
real
modulation index, modulation index when ideal conditions are assumed,
modulation index when real conditions are assumed
N neutral
n rotation speed (round per minute), positive integer
n negative bar of dc link
P, p active power: average value, instantaneous value
p positive bar of dc link
p
p
number of pairs of poles
R resistance, resistor
S switch
sw switching function
T torque, time period
T
C
, T
I
calculation/sample time, integral action time of PI controller
t time
U, u voltage: rms value, instantaneous value
xii
y controller output variable
x arbitrary variable
Z impedance
increment or error
angle, alignment angle of a space vector, phase angle
input-to-output voltage transfer ratio
phase displacement angle
flux linkage
angular frequency
List of Subscripts
aav average of absolute value
av average value
bl blanking quantity
CE collector-emitter quantity
cm commutation quantity
cms second commutation quantity
d diode parameter or value
dc dc link quantity
E emitter quantity
e electromagnetic quantity
env+ positive envelope
f filter parameter or quantity
FM flux model quantity
i input- or current-related quantity
k quantity of an arbitrary reference frame
kd, kq real and imaginary components of a space vector in an arbitrary reference
frame
LL line-to-line quantity
load load quantity
m modulation quantity
max the maximum
mech mechanical quantity
med the middlemost
med+ the nearest under the positive envelope
min the minimum
mr quantity in rotor-flux-oriented reference frame, magnetisation quantity
o output or load quantity
on on-state quantity
P parallel-connected
Symbols and Abbreviations xiii

r rotor quantity
ref reference value
res resolution
S series-connected, switch quantity or parameter
S75, S40 semiconductor component types SKM75 and SKM40, respectively
s stator or sampling quantity
slip slip-related quantity
sup supply quantity
sw switching quantity
toff turn-off parameter or quantity
ton turn-on parameter or quantity
th threshold quantity
u voltage-related quantity
x arbitrary quantity in stationary reference frame
x quantity of arbitrary distortion component
x, y real and imaginary components of a space vector in a rotor-flux-oriented
reference frame
0, 0 zero sequence component or zero state, zero vectors
1 fundamental component
16 serial numbers of switching states
, real and imaginary components of a space vector in stationary reference
frame
, space vectors adjacent to supply current reference vector: larger and smaller
angle, respectively
, space vectors adjacent to load voltage reference vector: smaller and larger
angle, respectively
leakage component
List of Other Notations
X rms value of ac quantity, average of dc quantity, constant quantity
X matrix
x peak value or magnitude of x
x space vector
|x| magnitude of the space vector x
x
*
complex conjugate of the space vector x
Re{x} real part of the space vector x
Im{x} imaginary part of the space vector x
x
(x,y,z)
x
x
, x
y
and x
z




xiv

1
1 Introduction
In the modern world, the demand for controllability and efficiency has been and is still
constantly increasing in technology. This also concerns electric drives. In the past, controllable
electric drives were mostly constructed with rotating converters, i.e. using a combination of
several electric machines, like the Ward Leonard drive [Leo01]. Stationary converters have
also been studied since the early 20
th
century [Wyk94], [Gyu70] but the number of
applications remained low until power semiconductors were developed in the 1950s, when
controlled switches, i.e. power transistors and thyristors were developed [Hal52], [Mol56],
[Mau57], [Adl84]. After that, most thyristor-based converter systems were developed in about
ten years [Gyu70], [Pel71], [Maz73]. Thus, the invention of the thyristor can be considered as
the birth of modern power electronics, which aims at the effective control of electric drives
and systems.
Power electronics also has other objectives apart from effective control. One of them is the
enhancement of electric power quality, including e.g. current and voltage distortion, voltage
dips and supply voltage interruptions. Primary power electronic systems such as diode and
thyristor converters have usually been considered sources of quality problems because their
supply current may contain high amounts of low-order harmonic components, which have
been considered a power quality problem for nearly as long as alternating current (ac)
networks have been in use [Owe98]. However, modern power electronics can be applied to
solve or mitigate such power quality problems. Modern fully controllable power
semiconductor switches such as the metal-oxide-silicon field effect transistor (MOSFET) and
insulated gate bipolar transistor (IGBT) were introduced in the late 1970s and the early 1980s,
respectively [Tem78], [Bal82], [Adl84]. These voltage-controlled switching devices have
made it possible to apply switching frequencies of dozens of kilohertz with pulse width
modulation (PWM) techniques. This enables sinusoidal supply current waveforms in the
power electronic equipment, which can also control reactive power and compensate harmonics
created by distortive loads.
Power electronic converters are used for a wide power range and in various applications, i.e.
wherever electric power has to be converted. The number of such systems is large and still
growing. Thus, there are no reasons to assume that the need for power converters will decrease
in the future: the electric power will still be converted from ac (alternating current) to direct
current (dc), from dc to ac, from dc to dc and from ac to ac. The last is used especially in
industrial adjustable-speed motor drives. As an example of the significance of motor drives,
industry and construction accounted for 54% of electricity consumption in 2006 in Finland
[Ene07]. Since the proportion of motor drives in industrial electricity consumption is
2 Chapter 1

approximated to be roughly 70%, industrial electric motors consume about 38% of the total
electric energy. When agriculture, the public and service sectors and households are also taken
into account, electric motor drives consume approximately 50% of the total electricity in
Finland. In the European Union, industry consumed about 41% of the total electricity in 2004
[Eur05]. In 2003, industry consumed about 39% of the total electricity in the OECD
(Organisation for Economic Co-operation and Development) countries and about 61% in the
non-OECD countries [EIA06]. Thus, the improvements in electric drive efficiency and control
are key issues in focusing on energy savings for sustainable development. This is already
reality because most new industrial motor drives are ac drives and the proportion of
adjustable-speed ac drives is increasing.
The state-of-the-art ac to ac converter has a dc link. Thus, conventional PWM frequency
converters incorporating passive dc link components are introduced briefly. After that, the
development of frequency converters without a dc link is presented in brief. Finally, the
objectives and the outline of the thesis are described. The contribution of the thesis is
summarised at the end of the chapter.
1.1 Overview of PWM Frequency Converters with dc Link
In most cases, the term frequency converter refers to a two-level voltage source inverter
(VSI) whose dc link is supplied by a three-phase diode bridge, as presented in Figure 1.1a,
where the inverted bridge consists of IGBTs. The frequency converter presented in Figure 1.1a
is called a two-level voltage source converter (VSC). For the load side (A, B and C) currents,
it is a quite optimal solution, but its supply side (a, b and c) currents are highly distorted,
containing high amounts of low-order harmonics [Kaz02]. Via the impedance of the mains, the
low-order current harmonics may distort the voltage of the point of common coupling (PCC),
which may further interfere with the other electric systems in the network [Arr03].
A conventional solution for the problem caused by VSC diode bridges is to use a similar IGBT
bridge as a supply bridge, too, i.e. back-to-back converter (BBVSC), which is presented in
Figure 1.1b. The BBVSC was introduced in the late 1970s [Wil78]. A PWM BBVSC produces
sinusoidal supply current waveforms. It is a boost-type converter, i.e. its dc link side voltage
has to be higher than the peak value of the supply line-to-line voltage. The BBVSC contains a
dc link capacitor which separates the supply- and load-side bridges, making their separate
control possible. Thus, the BBVSC can control the dc link voltage, compensate reactive power
in the supply and supply the load at the same time [Lip88], [Moh95].
The dc link capacitor of the VSCs and the BBVSCs is a bulky component with a limited
lifetime. In addition, conventional capacitors cannot be used in some special applications, such
as in aeronautics and deep-sea or space systems [Whe03], [Bha05]. In the BBVSCs, supply
filter inductors are also required, which can be considered a severe problem because the
inductors are bulkier and heavier than the dc link capacitor in low and medium power
Introduction 3

converters. However, VSI-based converters are well-known and widely used, so that the
commercial advantages due to mass production remain undeniable.
A theoretical option to voltage source converters is the PWM current source converter (CSC),
presented in Figure 1.2 [Enj91], [Tuu93], [Sal02]. According to [Lip88], the CSC was first
suggested in the 1980s. It produces sinusoidal supply current waveforms like the BBVSC and
can also compensate the reactive power of the supply.
Instead of a dc link capacitor, the CSC contains a dc link inductor, which is generally bulkier
and heavier than the link capacitor in voltage source converters. The supply filter of the CSC
consists of capacitors and inductors, i.e. an LC-type filter. Supply filter capacitors provide a
current path for high frequency components. Thus, the inductance value and the physical size
of CSC supply filter inductors can be smaller than those of BBVSC filter inductors, which
have to filter high frequency components alone. The capacitance value and physical size of
CSC ac capacitors are also small compared to the dc link capacitor of the BBVSC.
In addition, the CSC usually requires series-connected diodes with every IGBT. This increases
semiconductor on-state losses and the complexity of the main circuit. Both losses and
additional components increase the price of the system, too. A reverse blocking IGBT
(RBIGBT), introduced in the early 2000s, makes serial diodes unnecessary [Lin01], but the
RBIGBT is still quite a new device so that its development is still not complete and its
availability is limited, too.
(a)
A
B
C
a
b
c
(b)
a
b
c
A
B
C

Figure 1.1 Two-level voltage source frequency converter: (a) conventional topology with diode bridge on
supply side (VSC), (b) back-to-back converter (BBVSC).
A
B
C
a
b
c

Figure 1.2 Two-level current source frequency converter (CSC).
The PWM-controlled VSI produces phase voltages whose instantaneous sum is not zero, i.e.
common-mode voltage is generated. This may cause undesired high frequency currents in the
load system, e.g. bearing currents in electric machines and load mechanics [Che96], [Bus97],
4 Chapter 1

[Oll97]. A conventional mitigation strategy for common-mode voltages is to increase the
number of voltage levels [Hol03]. A three-level voltage source inverter supplied by a diode
bridge, i.e. a three-level voltage source converter (3LVSC), is presented in Figure 1.3. The
common-mode voltage produced by the 3-level VSI has lower voltage steps and a lower peak
value than the VSI [P2]. The 3LVSC is also commercially available. However, lower voltage
stresses of the semiconductors compared to the VSC is a more conventional reason to apply
the 3LVSC, so that it has conventionally been used in applications where voltages exceed the
limits of the semiconductors [Hol03]. Compared to the conventional VSI, the number of active
switches is double in the 3-level VSI. The total number of switches reaches 18 when the 2-
level rectifier of the BBVSC is applied as the supply bridge. If the supply-side bridge is based
on the 3-level VSI, i.e. 3-level BBVSC (3LBBVSC), the system requires 24 switches. The
number of voltage levels can be increased beyond three by increasing the number of series-
connected switches and capacitors [Hol03]. This leads to smoother output voltages and
decreases the voltage stress of the switches. However, the complexity of the system increases.
a
b
c
A
B
C

Figure 1.3 Three-level voltage source frequency converter with diode bridge on the supply side (3LVSC).
1.2 History of Direct Frequency Converters
Section 1.1 introduced converters containing either dc link capacitor(s) or inductor(s). They
both are physically large and/or heavy in most cases. In addition electrolytic capacitors are
subject to ageing, for example, and cannot be used in all applications as presented. The passive
components also cause power losses. Moreover, the converters presented do not fulfil the idea
of a purely semiconductor-based converter. Thus, direct frequency converters, converting ac
power to ac power directly without dc link passive components, have been studied, too.
According to [Gyu70] and [Wyk94], the idea of direct frequency conversion was originally
presented in the 1920s and even applied in the 1930s. The first semiconductor-based direct
frequency converters were developed in the 1960s after the invention of the thyristor [Pel71],
[Maz73], [Gyu76]. A possible main circuit of a phase-controlled thyristor-based three-to-three
phase cycloconverter is presented in Figure 1.4. The circuit presented is a six-pulse
cycloconverter, which assumes isolated loads so that a supply transformer is not necessary.
For simplicity, the cycloconverter in Figure 1.4 does not include circulating current reactors
either, which are sometimes used to enhance load power quality with discontinuous load
current [Pel71]. Although reactors and supply transformers are sometimes avoided, they are
necessary in many cases to make the cycloconverter system possible in practice, e.g. the
Introduction 5

supply transformer is not avoided with non-isolated load because in this case all the supplies
of all three bridges are required to be isolated from each other. In addition, the six-pulse three-
to-three phase cycloconverter requires 36 and a twelve-pulse version requires 72 thyristors.
The load voltage and input current waveforms of the cycloconverter are heavily distorted and
the fundamental power factor of the input is quite poor, too, irrespective of the fundamental
power factor of the load [She04]. In practice, its load frequency is also usually limited to half
of the supply frequency because normal loads cannot tolerate the voltage distortion produced
with higher input-to-output frequency ratios [Pel71]. Thus, the advantages remaining are
robustness of the thyristor technology and low losses. Considering the drawbacks and
limitations mentioned, the cycloconverter cannot be seen as an optimal solution for medium
power converters, but it is nowadays applied mostly with higher power levels where fully
controllable semiconductor devices, like IGBTs or MOSFETs, cannot be used yet.
Load
A
Load
B
Load
C
a
b
c

Figure 1.4 Six-pulse cycloconverter with isolated load phases without circulating current reactors.
The idea of the silicon-based forced commutated cycloconverter was presented and analysed
first in [Gyu70] and later in [Gyu76], the latter being known more widely due to its easier
availability. Over the decades, matrix converter (MC) has become established as the name of
this kind of direct frequency converter. The principle of a three-to-three phase MC is presented
in Figure 1.5, where each ideal switch describes a bidirectional switch which can conduct
current and block voltage in both directions depending only on the control signal of the switch.
a b c
A B C

Figure 1.5 Principle of matrix converter (MC).
Compared to Figures 1.11.4, the circuit in Figure 1.5 may appear simple. However, it
presents only the principle. In practice, the ideal switches assumed are not available and a
supply-side filter is also necessary. Thus, a more realistic circuit of a direct matrix converter
(DMC) is presented in Figure 1.6a, where the ideal switches are replaced by IGBTs and
diodes. The circuit has no dc link, but it still requires passive ac filter components, so that the
MC cannot be based on semiconductors purely in practice.
6 Chapter 1

A theoretically identical choice for the DMC is the indirect matrix converter (IMC), presented
in Figure 1.6b, where p and n are the dc link bars. The idea of the IMC for MC analyses was
introduced in the 1980s [Zio85] and an actual IMC circuit was proposed in the 1990s [Min93].
Below, the terms direct and indirect matrix converters, i.e. the DMC and the IMC,
respectively, are restricted to mean a three-to-three phase version with the same power transfer
capability as the circuits in Figure 1.6 have.
A rough comparison between the PWM converters discussed above is presented in Table 1.1.
A more extensive comparison between VSC and MC can be found e.g. in [Ber02].
(a)
B
a
b
c
A
C

(b)
a
b
c
n
p
A
B
C

Figure 1.6 (a) A direct matrix converter (DMC). (b) An indirect matrix converter (IMC).
Table 1.1 A rough comparison of the PWM frequency converters reviewed in Sections 1.1 and 1.2.
2-level converters 3-level converters Direct converters
VSC BBVSC CSC 3LVSC 3LBBVSC Cycloconv. DMC/IMC
Number of
diodes
12 12 12 24 36 None 18
Number of
active devices
6 12 12 12 24
36
thyristors
18
Dc link
components
Capacitor(s) Capacitor(s) Coil Capacitors Capacitors None None
Supply current Distorted Sinusoidal Sinusoidal Distorted Sinusoidal Distorted Sinusoidal
Input-to-output
voltage ratio
01
0 3 / 2
(or more)
01
(or more)
01
0 3 / 2
(or more)
0 / 3 3 0 2 / 3
Supply filter
inductor
Usually Necessary
Depends on
application
Usually Necessary
Usually
transformer
Depends on
application
Supply filter
capacitor
None
Depends on
application
Necessary None
Depends on
application
None Necessary

Introduction 7

The MC has no natural freewheeling paths for inductive current as has a VSC. On the other
hand, the MC may short-circuit the supply, unlike the CSC. Thus, the commutation of
semiconductor switches was a problem, too, until late the 1980s when the first safe
commutation method was introduced [Bur89]. On the other hand, modern active
semiconductors with fast switching capabilities were introduced just couple of years before as
presented at the beginning of this chapter. First modulation methods for MCs were also
presented in 1980s [Ven80a], [Ven80b], [Zio85] [Ale89], [Hub89], [Oya89], [Wie90],
[Whe02]. Thus, the main problems with MCs were mainly overcome until 1990. In addition,
an MC system recovers faster after a power grid failure than conventional BBVSC systems
containing dc link capacitor requiring charging and causing inrush currents [Kan02]. The
RBIGBTs may also provide some improvements for the MCs in the future as with the CSC.
Thus, it is possible that MCs will become common in practical applications.
1.3 Objectives and Outline of the Thesis
The general motivation for this thesis was provided by the desire to determine the suitability of
matrix converters for different electric drives and to compare direct and indirect matrix
converters, i.e. the DMC and the IMC. Thus, the objectives of the thesis are
1. to adapt the space-vector-based modelling of matrix converters (MCs) to match the
modelling of conventional voltage and current source converters
2. to examine the most significant sources of non-ideal behaviours in MCs and their
effects on current distortion, on control accuracy of output voltage and on power loss
3. to compare the characteristics of the DMC and the IMC in ideal and in non-ideal
conditions
4. to compare the effects of non-ideal MCs on the operation of motor drives
5. to examine the effects of modulation methods on the common-mode voltages in the
MC output
6. to examine the effects of supply voltage distortion and non-ideal load on output voltage
and output and supply currents
7. to compare the capability of different control methods to prevent the migration of
distortion through a converter.
The study has been performed mainly using the same procedure: 1) the problem is analysed
and equations predicting behaviour are derived; 2) the system is studied applying numerical
solving of the derived equations or simulations, or of both in some cases; 3) the analysis and
the calculation or simulation results are confirmed by experimental tests applying the DMC
and IMC prototypes built in the course of the research.
This thesis consists of an overview and seven Publications [P1][P7] which are arranged to
form a progressive presentation of the scientific contribution. The introduction in Chapter 1 is
followed in Chapter 2 by a presentation of the basic matrix converter circuits and technologies.
8 Chapter 1

Chapter 3 presents the modulation of the matrix converters and a brief review of different
methods. That is followed by the presentation of space vector modulation and the comparison
of the common-mode voltages produced by different space vector modulation methods [P1]
[P2]. Chapter 4 introduces the models of the MC circuits and machine drive control systems
with their simulation models. Prototypes and experimental setups are described in Chapter 5.
Chapter 6 presents the modelling of the distortion migration in MCs and simple distortion
mitigation methods together with the simulations and the experiments [P3][P4]. Chapter 7
summarises the comparison between the topologies and presents the analyses of the non-ideal
characteristics together with their verification [P4][P7]. Chapter 8 summarises Publications
[P1][P7] and their scientific contribution. Chapter 9 concludes the overview part of the thesis.
1.4 Contribution of the Thesis
The main contributions of the thesis are
The space vector modulation method, which was introduced in [Hub95] and developed
in [Nie96b], is derived purely in space vector form and is applied in the real indirect
matrix converter [P1].
The common-mode voltages produced by different matrix converter space vector
modulation methods are analysed and compared. In addition, the common-mode
voltages produced by the matrix converter modulation methods are compared to the
common-mode voltages produced by a space-vector-modulated three-level voltage
source converter [P2].
The true characteristics of matrix converters with supply voltage distortion and load
current distortion, which are not commonly considered in the literature, are presented.
The space-vector-based analysis and illustration of the effects of supply voltage
harmonic distortion on the load current of matrix converters and related migration of
the load current distortion back to supply are presented [P3][P4]. Compared to the
previous work presented in [Nie96b], the analysis of [P3] includes all harmonics in
addition to the negative sequence fundamental component considered in [Nie96b].
Also, the main concern in [Nie96a], [Cas98a] and [Cas98b] has been the supply current
quality, which cannot be the main concern in practice. The migration of load current
disturbance has not been presented before [P4].
The non-ideal characteristics of direct and indirect matrix converters are analysed and
compared. The non-ideal characteristics considered are power losses, supply current
distortion and input-to-output voltage transfer characteristics [P4][P7]. The analytical
DMC and the IMC power loss models have not been presented before. The analytical
model of input-to-output voltage transfer characteristics of the IMC has not been
presented before and the input-to-output voltage transfer characteristics model of the
DMC is simpler than the previously presented model in [Lee04] and [Lee05a].
The contributions of Publications are presented separately and more precisely in Chapter 8.
9
2 Matrix Converter Systems
Gate-controlled semiconductor-based matrix converters (MCs) have been found an interesting
research topic ever since their invention. However, its commercial applications are still few.
To the best of the authors knowledge, only a single Japanese manufacturer is marketing MCs
at the moment.
This chapter is a literature review which presents the basic MC technology and MC protection
issues as a background for [P1][P7]. First, the basic switch configurations and optional
single- and two-stage MC topologies are presented. After that, the methods and the systems
providing safe operation of the MCs are considered briefly.
2.1 Topologies
The ideal three-to-three phase MC circuit presented in Figure 1.5 can be implemented in many
ways. The basic two MCs are direct MC (DMC) and indirect MC (IMC), as shown in Figure
1.6. Their other bidirectional switch configurations are dealt with in this section, which also
presents optional single- and two-stage MC topologies as a background for [P1][P7].
2.1.1 Bidirectional Switches
The possibilities to implement the bidirectional switch with IGBTs are presented in Figure 2.1.
Any other fully controllable semiconductor switches could also be used instead of the IGBT.
The configuration in Figure 2.1a contains only one active switch and is the simplest choice.
However, it cannot be used in MCs with most safe commutation methods, as will be presented
in Section 2.2. In addition, it has more conducting components in a supply-to-load current path
than the configurations in Figures 2.1bf. However, it could be used with some two-stage MC
topologies presented in Section 2.1.3 and it was applied with MOSFET in one of the first
experimental verifications of MC drives in 1988 [Nef88].
Conventionally, the most popular MC switches have been common-emitter and -collector
configurations (Figures 2.1bc) and their combinations [Nie99], [Whe02], [Jus05]. They both
always have a single active switch and a single diode conducting per output phase. In practice,
their only difference is the number of required isolated emitter potentials. For example, a
common-collector switch-based DMC requires only six isolated gate control units, whereas
the common-emitter-based DMC, shown in Figure 1.6a, requires nine. The common-collector-
based IGBT modules containing all switches of the DMC have been available for low power
range [Bru01] and IGBT modules with a common-emitter-connected bidirectional switch are
also available [Sem04].
10 Chapter 2

(a)
(b)
(c)
(d)
(e)
(f)

Figure 2.1 Bidirectional switch configurations in practice: (a) switch and diode bridge, (b) common-
emitter, (c) common-collector, (d) diode and switch in series, (e) diode and switch in series
and (f) antiparallel RBIGBT configuration.
The separated serial combinations (Figures 2.1de) are not as commonly used as the previous
two and they have similarities with the switches required in the current source converter
(CSC). However, differences arise with the number of isolated gate drivers: with the DMC, the
configuration in Figure 2.1d requires eighteen isolated gate drivers, whereas the configuration
in Figure 2.1e requires only six, as in the common-collector configuration.
The similarity between the MCs and the CSCs continues with the benefits provided by the
RBIGBTs (Figure 2.1f) [Ber96], [Lin01]: there is only a single RBIGBT conducting per
output phase. However, they are still quite new and their switching characteristic is not
necessarily comparable to that of conventional IGBTs yet [IXY05], [Fri06], [Klu06a],
[Zho07]. On the other hand, contrary results have also been presented in [Tak04], [Nai04],
[Jia05], [Lut05] and [Bla04], but all of these sources consider only a simplified circuit, a
single switch, or a pair of switches. The same holds for [Mot04], even though it introduces an
RBIGBT module containing DMC switches. Most experimental results for a full DMC system
with eighteen RBIGBTs are presented with voltage levels lower than European 400 V, e.g. in
[Ito05b], [Bal06], [Sun06] and [Zho07]. In addition, there seem to be differences which
depend on the manufacturer whose RBIGBT is applied in each case.
2.1.2 Single-Stage Topologies
The DMC, i.e. a single-stage three-to-three MC, presented in Figures 1.5 and 1.6a is not the
only possible single-stage MC; rather, the number of phases can be arbitrary, as presented e.g.
in [Ten92]. The DMC in Figure 1.6a consists of three output-phase-specific switch groups (a
single switch group is presented in Figure 2.2b). Compared to the DMC, the principles of the
modification are: 1) adding/removing switches to/from the switch group when the number of
input phases varies, 2) adding/removing switch groups when the number of output phases
varies, 3) combining 1) and 2) when the number of both input and output phases varies.
A single-to-single or single-to-two phase MC is presented in Figure 2.2a and studied e.g. in
[Kho88], [Zuc97] and [Idr06]. As can be seen, it is a H-bridge with bidirectional switches. The
simplest three-to-single phase MC is presented in Figure 2.2b, where the supply and load have
a common neutral [You99]. Another three-to-single phase MC is presented in Figure 2.2c,
Matrix Converter Systems 11

where two switch groups are used [Ten92], [Pac03], i.e. the circuit has two output phases so
that it can be called a three-to-two phase MC, too. The circuit in Figure 2.2c can also serve as
a four-quadrant buck-type rectifier, as suggested in [Hol92]. Due to its four-quadrant
operation, it can be applied to a dc motor drive, as presented in [RP5]. Functionally, the four-
quadrant rectifier in Figure 2.2c equals the combination of two antiparallel-connected rectifier-
side bridges of the CSC (Figure 1.2), which also holds for the control system, as shown in
[RP5].
(a) (b)
Switch
group

(c)
Figure 2.2 Some single-stage MC circuits: (a) single-to-single or single-to-two phase MC, (b) three-to-
single phase MC with common neutral, i.e. single switch group of output phase, and (c) three-
to-single or three-to-two phase MC or four-quadrant buck-type rectifier.
In the case of single-stage MCs with more than three output phases, the only one studied
experimentally is a three-to-four phase or a four-wire single-stage MC, presented in Figure
2.3, suggested in [Whe05] and applied in [Kat05]. In both [Whe05] and [Kat05], the fourth
output phase N is neutral, i.e. the system connects a four-wire system to a three-wire system.
The circuit requires 24 IGBTs and diodes.
A B C
a b c
N

Figure 2.3 Three-to-four phase single-stage MC.
12 Chapter 2

2.1.3 Two-Stage Topologies
In the single-stage MC circuits presented, every output phase can be connected to any input
phase without restriction. However, the same performance can be achieved in practice even
when the possibility to connect input and output phases is restricted. An ideal two-stage MC is
presented in Figure 2.4. The two-stage or indirect approach to MC analysis and modulation
was introduced already in the 1980s [Zio85], [Oya89], [Hub89], [Whe02], whereas two-stage
MC topologies as real converters came to be studied slightly later [Hol89], [Min93].
a
b
c
A
B
C
u
dc
p
n
i
dc

Figure 2.4 Ideal two-stage matrix converter, i.e. ideal indirect MC.
The two-stage MC topology proposed in [Hol89] consists of a conventional VSI bridge
(Figure 1.1) and a rectifier stage, where each bidirectional switch consists of a pair of
antiparallel thyristors. [Hol89] proposes a control system where VSI is controlled by such a
PWM method that causes dc link current i
dc
to become zero once in a modulation period.
Thus, the thyristors of the rectification stage require no auxiliary commutation circuits and can
still be controlled by PWM control. However, the system has not been verified in experimental
tests to the best of the authors knowledge. A more primitive system, presented in Figure 2.5,
is proposed in [Kim00], where a conventional back-to-back converter (BBVSC, Figure 1.1b)
without a dc link capacitor and with an LC-type supply filter is introduced. The supply-side
bridge, i.e. the rectifier stage of this simplified two-stage MC, is controlled so that its IGBTs
are turned on at the same time as its diodes turn on [Kim00]. Thus, the rectifier stage allows
the dc link current i
dc
to change direction without restriction. Due to the rectifier stage control,
the dc link voltage u
dc
follows the envelope of the supply line-to-line voltages and the supply
current waveform is roughly a four-step-square wave if the link current i
dc
is assumed
constant. The inverter stage may again be controlled like the VSI.
A
B
C
u
dc
p
n
i
dc
a
b
c

Figure 2.5 Simplified two-stage matrix converter.
Matrix Converter Systems 13

As presented in Chapter 1, one of the main motivations in MC research has been the
possibility to attain sinusoidal supply current waveforms. That is not possible with the
simplified two-stage MC. Thus, the most studied three-to-three phase two-stage MC is the
indirect matrix converter (IMC), presented in Figure 2.6, where common-emitter-configured
switches are applied in the IMC supply bridge (ISB) and PWM control is possible. The IMC
load bridge (ILB) is again the conventional VSI. Due to the ISB configuration, the IMC can
produce sinusoidal supply current waveforms. Generally, the IMC in Figure 2.6 should
provide performance identical to that of the DMC when modulated with the same strategy.
This is discussed more extensively in Chapters 37 and in [P2] and [P4][P7].
a
b
c
n
p
A
B
C
u
dc
i
dc

Figure 2.6 Indirect matrix converter (IMC).
To the best of the authors knowledge, the IMC was first proposed and also confirmed
experimentally in [Min93] and then researched further e.g. in [Iim97], [Wei01], [Zwi01] and
[Iim04]. As with the DMC, the ISB switches may also be implemented with all configurations
presented in Figure 2.1, e.g. with RBIGBTs as in [Fri06]. In the ISB, the poor switching
characteristics of the RBIGBTs are not necessarily a problem because they can be switched
with zero current, as suggested in [Hol89], when a suitable modulation method is applied for
the IMC [Fri06]. That kind of modulation also allows the configuration in Figure 2.1a to be
used in the ISB [Wei01], [Kol02]. This would decrease the number of active devices to twelve
and increase the number of diodes to thirty, whereas the IMC in Figure 2.6 has eighteen
IGBTs and diodes. However, that kind of converter would have four conducting devices in the
current path, whereas the IMC in Figure 2.6 has only three conducting devices in the current
path.
The total number of semiconductor devices can also be reduced using an approach where the
modulation is not restricted [Kol02], [Wei02]. Figure 2.7 presents a three-to-three two-stage
MC with the same operation as the IMC, but the ISB contains only nine active devices and
thus it is only a reduced IMC. The reduced IMC in Figure 2.7 contains fifteen IGBTs and
eighteen diodes and has four devices in the current path. If the rectifier bridge is reduced
further, as presented in the three-to-three phase two-stage MC in Figure 2.8, the dc link current
i
dc
can no longer flow in the negative direction. Thus, the fundamental output power factor
may range only between unity and 2 / 3 and its maximum output fundamental displacement
angle is 30. However, the two-stage MC in Figure 2.8 contains only nine IGBTs and 18
diodes.
14 Chapter 2

As can be seen in comparing the two-stage circuits in Figures 2.62.8 to the DMC circuit in
Figure 1.6a, the DMC has always less semiconductor devices on the current path from supply
to load. However, the two-stage MC offers a possibility to decrease the number of active
semiconductor devices, as presented above. The decrease is greater when the number of load
phases is increased, i.e. with a three-to-four phase two-stage MC, which was suggested in
[Yue06] and is presented in Figure 2.9. As can be seen, the three-to-four phase two-stage MC
in Figure 2.9 requires only two additional active devices and diodes, i.e. twenty IGBTs and
diodes. That is four less compared to the respective single-stage MC in Figure 2.3. In addition,
the reduced ISB presented in Figure 2.7 could be used instead of the full ISB without
limitations on operation, which decreased the number of active devices to seventeen.
A
B
C
a
b
c
n
p
u
dc
i
dc

Figure 2.7 Two-stage MC with 15 active devices, i.e. reduced IMC.
A
B
C
a
b
c
n
p
u
dc
i
dc

Figure 2.8 Two-stage MC with 9 active devices and reduced displacement power factor of the output.
a
b
c
n
p
A
B
C
u
dc
i
dc
N

Figure 2.9 Three-to-four phase two-stage MC.
The two-stage MC contains also a possibility to attain three-level operation with an auxiliary
circuit, which is impossible in the single-stage MCs. A three-level two-stage MC, introduced
in [Klu06b], is presented in Figure 2.10. In the three-level two-stage MC, both link bars can be
Matrix Converter Systems 15

connected to the supply neutral point. Thus, three voltage levels are available for the load,
which can be used to decrease common-mode voltage steps. As discussed in Chapter 1 and in
[P2], a three-level inverter system can decrease common-mode load voltage steps, which can
be seen as a possible benefit. The circuits presented in Figures 2.2, 2.3, 2.5 and 2.72.10 are
examples only. Their deeper analysis and specific features are beyond the scope of this thesis,
but they are presented to show the MC circuit variety.
a
b
c
n
p
A
B
C
u
dc
i
dc

Figure 2.10 Three-level two-stage MC.
2.2 Ensuring Safe Operation
In all MC topologies, some basic operational limitations exist. Due to the voltage source type
input, i.e. supply filter capacitor, load must be resistive or inductive. If the path of the
inductive load current is cut abruptly, high load voltage spikes may arise. Thus, the load
current path must always be ensured because purely resistive circuits do not exist in practice.
On the other hand, input phases should never be short-circuited to avoid input current spikes.
In the MCs, every output phase must always be connected to a single input phase.
This section is mainly a literature review of commutation methods and auxiliary circuits
applied in the MCs to ensure safe operation as a background for [P1][P7]. First, safe
commutation methods are reviewed, after which overvoltage protection is discussed. Neither
of the issues is presented in [P1][P7]; they are only used there.
2.2.1 Switch Commutations
Problems with short-circuiting of the input phases or cutting of the output current path can be
avoided in normal conditions by controlling the turn-on and turn-off switchings of every
device separately. These methods are called commutation methods. Their basic idea is to
perform the commutations in steps, avoiding short- and open-circuit situations, and they are
always applied separately for each output phase switch group, as presented in Figure 2.11a. In
addition, they require such bidirectional switch configuration that the direction of current flow
through the switch can be controlled. Thus, the configuration presented in Figure 2.1a
containing a single active device cannot provide safe commutation methods.
Both basic commutation methods of the DMC were suggested in the late 1980s [Bur89],
[Oya89]. One is based on output current direction information and the other on input voltage
16 Chapter 2

polarities. They and their variations are presented next. All of them were originally developed
for the DMC, but they all can be used in the IMC, too. The commutation of the IMC is
presented at the end of this section.
Output-Current-Direction-Based Commutation
The method suggested in [Bur89] is based on output current direction and its principle is
presented in Figure 2.11. Figure 2.11a presents the switch group S
A(a,b,c)(1,2)
of output phase A,
whose current is i
A
, and Figure 2.11b presents the gate control signals u
SA(a,b)(1,2)
of the
switches S
A(a,b)(1,2)
, respectively, when positive output current commutes from input phase a to
input phase b. Before the commutation procedure, both devices S
Aa1
and S
Aa2
, forming switch
S
Aa
, are on and S
Aa1
conducts. The procedure starts with the first step at instant t
1
when the
non-conducting device of the conducting switch (S
Aa2
) is turned off. The second step is
performed at instant t
2
when the device S
Ab1
, which conducts in the direction of the current i
A
,
is turned on. The third step is to turn off the originally conducting device S
Aa1
at instant t
3
. The
fourth step is to turn on the non-conducting device S
Ab2
of the new switch, in which the device
S
Ab1
is conducting, at instant t
4
. The sequences with both current directions are presented in
Figure 2.12 in part 4-step.
(a)
A
a b c
S
Aa1
S
Aa2
S
Ab1
S
Ab2
S
Ac2
S
Ac1
i
A
u
iab
(b)
u
SAa1
u
SAa2
u
SAb1
u
SAb2
t
1
t
2
t
3
t
4
t

Figure 2.11 (a) Switch group of output phase A. (b) Control signals u
SAa(1,2)
and u
SAb(1,2)
of devices S
Aa(1,2)

and S
Ab(1,2)
, respectively, in current-sign-based four-step commutation, when output current i
A

commutes from input phase a to input phase b and i
A
> 0, i.e. flows in direction 1.
In the situation shown in Figure 2.11, the commutation occurs at t
2
if u
iab
< 0 and at t
3
if
u
iab
> 0, where u
iab
is the input line-to-line voltage. When the commutation occurs at t
2
, the
turn-off of S
Aa1
is soft, and both the turn-on of S
Ab1
and the turn-off of the antiparallel diode of
S
Aa2
are hard because of the reverse recovery current in turn-off of power diodes [Moh95].
When the commutation occurs at t
3
, only the turn-off of S
Aa1
is hard and other switchings are
soft because the positive u
iab
is blocked by the non-conducting antiparallel diode of S
Aa1
,
whose currentless turn-off does not cause reverse recovery current [Bla03]. In the modulation
method applied in [P1][P7], both commutation types occur equally in a modulation period.
Thus, it can be said for simplicity that one type of commutation is totally hard and the other is
totally soft. The total number and energy losses of the commutations match even though their
parts are distributed in reality as shown above.
The four-step commutation method allows the change in output current direction automatically
because after the commutation procedure both devices of the switch are on as presented above.
However, the turn-on and turn-off operations of the non-conductive devices are mostly
Matrix Converter Systems 17

useless. Thus, the second output-current-direction-based strategy is a commutation with two-
steps, where only the conducting device is turned on and all switchings of the devices are
unnecessary [Emp98]. The output-current-direction-based two-step commutation contains
steps 2 and 3 of the four-step method, i.e. the instants t
2
and t
3
in Figure 2.11b [Emp98]. Both
possible sequences are presented in Figure 2.12 in part 2-step.
In the two-step commutation method, the change in the current direction has to be taken into
account separately. The detection of output current zero crossing, which contains PWM ripple,
is not a fault-free task. Start-up is also a problem because the current direction is then
undefined. In practice, two-step commutation requires additional logic to take care of the zero
current and start-up situations. The solution is based on threshold levels on both sides near
zero value [Emp98]. When the current crosses the threshold level, the conducting device is
turned off and the device conducting in opposite direction is turned on after a blanking time.
The same logic takes care automatically of the start-up, changing the device turned on after the
blanking time if the current does not start to flow via the first device turned on. Thus, the
current-based two-step commutation requires more complex logic than the four-step method.
In addition, it requires information on current magnitude, unlike the four-step method.
S
Aa1
, S
Aa2
S
Aa1
S
Aa1
, S
Ab1
S
Ab1
S
Ab1
, S
Ab2
S
Aa1
, S
Aa2
S
Aa2
S
Aa2
, S
Ab2
S
Ab2
S
Ab1
, S
Ab2
S
Ab1
, S
Ab2
i
A
< 0
i
A
> 0
4-step
S
Aa1
, S
Aa2
S
Ab1
S
Ab1
, S
Ab2
S
Aa1
, S
Aa2
S
Ab2
S
Ab1
, S
Ab2
S
Ab1
, S
Ab2
i
A
< 0
i
A
> 0
3-step
off
S
Aa2
on
S
Ab1
on
S
Ab2
off
S
Aa1
off
S
Aa2
on
S
Ab1
on
S
Ab2
off
S
Aa1
off S
Aa1
on S
Ab1
on
S
Ab2
S
Aa1
S
Aa2
off
S
Aa2
off
S
Aa1
on
S
Ab1
off S
Aa2
on S
Ab2
S
Aa1
S
Ab1
S
Aa2
S
Ab2
i
A
< 0
i
A
> 0
2-step
off S
Aa1
on S
Ab1
on S
Ab2
S
Aa1
, S
Ab1
S
Aa2
, S
Ab2
off S
Aa2
S
Aa1
S
Ab1
S
Aa2
S
Ab2
i
A
< 0
i
A
> 0
1-step
off S
Aa1
on S
Ab1
on S
Ab2
off S
Aa2

Figure 2.12 Block diagram of the devices turned on with both current directions in current-sign-based
commutation methods when output current i
A
commutes from input phase a to input phase b.
The leftmost blocks present the initial states and the rightmost blocks present the target state.
The current-direction-based commutation methods can be reduced to a three-step or single-
step commutation method, both presented in Figure 2.12 in parts 3-step and 1-step,
18 Chapter 2

respectively [Whe04]. In the three-step commutation method, the target device and the
originally conducting device are turned on and off, respectively, at the same instant, i.e. t
2
= t
3

in Figure 2.11b. This can be done with the semiconductor devices that have longer turn-off
than turn-on delay in all possible conditions. Then, the overlapping caused by the difference in
delays is long enough to secure the current path, i.e. the originally conducting device turns off
after turn-off delay. The same can be applied in the case of the two-step current-direction-
based commutation method where this means that only a single step occurs, thus leading to the
single-step commutation method. The reduced commutation methods do not require as long a
time as even-numbered commutation methods [Whe04]. Thus, they could lead to more ideal
voltage transfer characteristics than the characteristics analysed in [P5] using the four-step
method.
Input-Voltage-Polarity-Based Commutation
The four-step commutation method suggested in [Oya89] is based on the information on the
input line-to-line voltage polarities, which is the other method that can be used to ensure safe
commutation. In the four-step voltage-polarity-based method, the devices are turned on and
off in different order compared to the four-step output current direction based method. The
sequences with both possible voltage polarities are presented in Figure 2.13 when the current
commutes from S
Aa
to S
Ab
in the circuit presented in Figure 2.11a.

S
Aa1
, S
Aa2
S
Aa1
, S
Aa2
,
S
Ab1
S
Aa2
, S
Ab1
S
Ab1
, S
Ab2
u
iab
< 0
u
iab
> 0
4-step
on
S
Ab1
off
S
Aa1
off
S
Aa2
on
S
Ab2
u
ia
> u
ib
> u
ic
2-step
on S
Ab1
off S
Aa1
S
Aa2
, S
Ac1
S
Aa1
, S
Aa2
S
Aa1
, S
Aa2
,
S
Ab2
S
Aa1
, S
Ab2
S
Ab1
, S
Ab2
on
S
Ab2
off
S
Aa2
off
S
Aa1
on
S
Ab1
S
Ab1
, S
Ab2
,
S
Aa2
S
Ab1
, S
Ab2
,
S
Aa1
u
ia
> u
ic
> u
ib
S
Aa1
, S
Aa2
,
S
Ac1
on S
Ab2
S
Ab1
, S
Ab2
,
S
Ac1
, S
Aa2
on S
Ab2
off S
Aa1
S
Aa2
, S
Ab1
S
Aa1
, S
Aa2
,
S
Ab1
S
Ab1
, S
Ab2
,
S
Aa2
u
ic
> u
ia
> u
ib
on S
Ab2
off S
Aa1
S
Ab1
, S
Ac2
S
Aa1
, S
Aa2
,
S
Ab1
, S
Ac2
S
Ab1
, S
Ab2
,
S
Ac1
off S
Aa2
u
ib
> u
ia
> u
ic
on S
Ab1
off S
Aa1
S
Ab2
, S
Ac1
u
ib
> u
ic
> u
ia
S
Aa1
, S
Aa2
,
S
Ac1
, S
Ab2
S
Ab1
, S
Ab2
,
S
Ac1
on S
Ab1
off S
Aa2
S
Aa1
, S
Ab2
S
Aa1
, S
Aa2
,
S
Ab2
S
Ab1
, S
Ab2
,
S
Aa1
u
ic
> u
ib
> u
ia
off S
Aa2
S
Aa1
, S
Ac2
S
Aa1
, S
Aa2
,
S
Ac2
S
Ab1
, S
Ab2
,
S
Aa1
, S
Ac2
off S
Aa2
on S
Ab1
on S
Ab2

Figure 2.13 Block diagram of the devices turned on in all situations with voltage-polarity-based
commutation methods when output current i
A
commutes from input phase a to input phase b.
The leftmost blocks present the initial states and the rightmost blocks present the target state.
As with the current-direction-based four-step method, the commutation of current occurs after
step 2 or after step 3 with the voltage-polarity-based method because at step 1 the device
Matrix Converter Systems 19

turned on is always reverse-biased and the device turned on at step 3 is always forward-biased.
Thus, the commutation is soft at step 2, i.e. when i
A
> 0, and hard at step 3, i.e. when i
A
< 0,
when the definitions presented above for current-based methods are used.
The input-voltage-based two-step commutation method presented in [Zie98] and [You99] does
not have the same analogy with the input voltage polarity based four-step voltage
commutation as the current-direction-based four- and two-step methods. The basic solution of
[Zie98] and [You99] is that 1) the negative direction conducting device of the input phase,
which forms the positive envelope of the input phase voltages instantaneously, is always
turned on and 2) the positive direction conducting device of the input phase, which forms the
negative envelope of the input phase voltages instantaneously, is always turned on, too. Thus,
the current path always exists without short-circuit so that safe commutation with two steps is
possible. Possible combinations when output current i
A
commutes from input phase a to b are
presented in Figure 2.13 in part 2-step, which is based on the method presented in [You99].
In certain situations, the voltage-based two-step commutation leads to an intermediate state
where current commutates to the input phase which is neither the initial nor the target phase.
That causes an additional commutation, which is hard with the other output current direction.
This is a drawback compared to the other methods presented above.
In both voltage-based commutation methods, the commutation occurs either after the turn-off
of the initially conducting device or after the turn-on of the target device. The simultaneous
turn-off and turn-on of those devices would short-circuit input phases because the operation
would create two bidirectional switches turned on simultaneously. Thus, the input-voltage-
based commutation methods are safe as four- and two-step versions only. However, single-
step commutation can be achieved when current direction detection is combined with voltage-
based two-step commutation [Zie01]. The advantages of that kind of system are quite limited
because safe operation can be achieved with the current direction only as presented above.
Comparison of Current- and Voltage-Based Commutation Methods
Unlike the current-direction-based commutation methods, the bidirectional current path
always exists in all the voltage-based commutation. However, the input is short-circuited if the
input voltage signs or order is detected wrongly. This can be fatal due to the discharge of the
supply filter capacitors, which can be detected by neither output nor supply current sensors. In
the current-direction-based commutation methods, an incorrectly detected current sign opens a
current path, creating an overvoltage condition which can be handled with an overvoltage
clamp circuit. In addition, a current sign detection error is likely to occur near zero and the
energy to be dissipated is low. Voltage detection errors occur near zero, too, but the short-
circuiting of the input capacitors may still be fatal.
In all cases, problems may arise if either current sign or voltage polarity information changes
during the commutation. In the two-step current-based method, the additional logic for zero
crossing solves the problem automatically. The latching of current sign information until the
on-going commutation is fulfilled can also be applied in the four-step current-based
20 Chapter 2

commutation. This causes only a short interruption of the current path with zero current,
causing no problem. In the voltage-based methods, the latching of the polarity information
short-circuits the input. Thus, a more secure method could be to change the voltage polarity
information immediately when the polarity changes. This results in cutting off the current path
and causes an overvoltage situation, which is easier to handle than overcurrent, as discussed
above.
The delays in all the commutation methods do not need to be equal. Their values depend on
the semiconductors used and the three- and single-step methods can be seen as ultimate
situations. The delays can also be adaptive, depending on the output current level [Kaz02].
Input voltage measurement in voltage-based methods can be considered easier and more cost-
effective than output current measurement with current sensors. However, when the output
current direction is the only information required, it can be defined by measuring voltages
across the semiconductor devices of the switches [Emp98]. That leads to several voltage
measurements which should detect zero crossing accurately and have the ability to handle the
peak-value of input line-to-line voltage, too. Another method is to add an antiparallel-
connected diode pair to each output phase and measure voltage across the diodes, which
requires only a single narrow voltage band measurement per output phase [Ker00]. However,
the diode pair causes power losses, decreases voltage transfer ratio and is an additional cost.
Safe Commutation in IMC
In the IMC, the commutation methods of the bridges separated are different. As presented, the
commutation methods of the MCs are used separately for each output phase switch group.
Thus, all methods presented above are directly suitable for the ISB, which consists of two such
groups. The only exception is that dc link bars p and n define the switch group instead of
output phases, as presented in Figure 2.6. All methods can also be applied in the rectifier
bridge of the reduced IMC in Figure 2.7. However, the IMC structure also offers a benefit
which is not provided by DMC [Hol89], [Wei01]: When the IMC is modulated so that the ISB
always changes its state during the zero output voltage state of the ILB, i.e. all output phases
are connected either to p or to n bar of the dc link (Figure 2.6), there is no need for the special
commutation method in the ISB because its devices are switched with zero current. Thus, a
conventional blanking time between the turn-off and -on is enough to avoid the short-
circuiting of the input phases. The method also decreases the ISB switching losses to zero.
Naturally, the method can be applied only when the sum of the output currents is zero. In
addition, this is the only way to make it safe to use the type of switches in Figure 2.1a in the
ISB.
The ILB is identical to the VSI and thus their safe commutation is also identical, i.e. the ILB
requires dead or blanking time between turn-off and turn-on in a leg [P5]. Simultaneous
turning on of both the devices in the same leg would short-circuit the input phases. In the
experiments in [P1][P7], the ILB is commutated by applying blanking times with fixed
values.
Matrix Converter Systems 21

The four-step current-sign-based method with fixed times between the steps is applied both in
the DMC and the ISB in [P1][P7]. For the DMC commutation, all three output current
directions are required, whereas only the dc link current direction is required for the ISB.
2.2.2 Overvoltage Protection
The commutation methods presented above lead to safe operation under normal conditions.
However, commutation failures are still possible as a result of electromagnetic interference,
power failure, etc. In addition, the commutation methods may cut the output current path in
some situations, leading to overvoltage due to leakage inductances, as presented above.
Overvoltage from a supply grid should not be ignored either [Nie99]. As a result, the lack of a
natural freewheeling path leads to the need for an additional circuit providing energy storage
and preventing overvoltage in the error cases [Nef88].
The danger of overvoltage situations can also be decreased when the modulation method and
physical device placement are combined so that every commutation always occurs between
physically adjacent switches [Sch04]. This minimises the effects of leakage inductance in
bonding wires so that the magnitude of commutation voltage spikes diminishes.
Overvoltage Protection in DMC
In terms of current path and overvoltage, the MC is similar to the CSC. Thus, the overvoltage
clamp circuit of the DMC, suggested in [Ale89] and presented in Figure 2.14, is quite the same
as that used with the CSCs [Put01]. The clamp circuit contains a varistor, which limits the
maximum voltage without a large capacitor. The configuration presented in Figure 2.14 is also
that applied in [P2] and [P4][P7]. Another solution is to use the clamp circuit capacitor
without the parallel varistor, but then the capacitor has to be capable of storing the maximum
possible energy alone without overvoltage, as suggested e.g. in [Nef88] and [Nie99]. The latter
also suggests a reduction in the number of separate clamp circuit diodes from twelve to six, as
shown in Figure 2.15. This reduction is possible only if all switch groups contain both the
common-emitter and common-collector connected switches.
Instead of using a varistor, the voltage of the clamp circuit capacitor can also be controlled
with an active semiconductor switch and resistor, as suggested in [Tuu92], or with a
semiconductor switch gated so that the overvoltage drives it to the linear operation region,
where it dissipates the energy, as suggested in [Ito05b]. In addition, the capacitor can supply
power to auxiliary converter control circuits, as presented in [Nef88], where the auxiliary
power supply consumes the capacitor energy and controls the clamp capacitor voltage.
The overvoltage protectors presented above are based on a common circuit structure. Another
approach is to use a separate protection circuit or device for every switch, as proposed in
[Nef88] and [Mah02]. The latter reference suggests delta-connected varistors both for the
input and the output side of the DMC, mounted directly on the terminals. In the case of
overvoltage, the varistors turn on and dissipate the energy of the voltage spike. According to
[Mah02], single devices may also require separate protection by a gate control system, which
22 Chapter 2

drives the device to the linear operation region to dissipate voltage spike energy when
required. Due to the limited energy dissipation capability of varistors, they can be used alone
in low-power applications only [Mah02]. With higher power, other solutions must be used.
When a varistor is used to protect the converter in parallel with a capacitor, as e.g. in [Put01],
[Sal02] and [P1][P7], some design issues should be noted. The varistor voltage should be
chosen so that the semiconductors are not damaged. On the other hand, the varistor voltage
should not be too low because the clamp capacitor should store the amount of energy that the
varistor voltage is exceeded only when the malfunction of the converter is serious. In any case,
the clamp circuit should be capable to dissipate the energy stored in the leakage inductances of
the load together with the resistive components of the load [Nie99], [Mah02]. The maximum
ratings for the pulse magnitude and duration of the varistor current should also be taken into
account and they both are affected by the circuit time constant parameters [Mah02]. Thus, the
use of varistors requires careful design and their usability depends on the application.
A
B C
a b c

Figure 2.14 DMC with overvoltage clamp circuit.
A
B C
a b c

Figure 2.15 DMC with reduced overvoltage clamp circuit.
Overvoltage Protection in IMC
Everything said about the overvoltage protection of the DMCs above is also valid for the IMC.
Natural freewheeling current paths do not exist in IMCs, so that overvoltage clamp circuits are
required. The clamp circuit configuration applied in [P1][P7] is presented in Figure 2.16. In
theory, the supply-side clamp circuit in Figure 2.16 is adequate to secure the current paths
alone. In practice, the dc link bars contain leakage inductance, which separates the supply-side
and load-side ends of the dc link bars. Thus, there is also a clamp circuit on the load side for
the load-side ends of the bars. For the leakage inductance, the IMC circuit in Figure 2.16 also
Matrix Converter Systems 23

contains the small dc link capacitor C
dc
, which minimises the effects of leakage inductance in
dc link bars and semiconductor module bounding wires [Kim00], [P4].
[Klu05b] suggests that the IMC overvoltage clamp circuit can be considerably reduced
because the freewheeling path of the ILB can be provided by series connection of a diode and
a capacitor in the dc link, as presented in Figure 2.17. However, that solution does not take
into account the effects of the supply grid and does not protect the ISB, so that protection has
to be provided in some other way. A solution based on reduced clamp circuit of the DMC and
simultaneous protection of both bridges is presented in Figure 2.18 [Klu05b]. However, the
clamp circuit configuration presented in Figure 2.18 ignores the possible effects of the leakage
inductance in dc link bars, unlike the full clamp circuit in Figure 2.16.
a
b
c
n
p
A
B
C
C
dc

Figure 2.16 IMC with two overvoltage clamp circuits.
a
b
c
n
p
A
B
C
C
dc

Figure 2.17 IMC with reduced overvoltage clamp circuit of ILB and unprotected ISB.
a
b
c
n
p
A
B
C
C
dc

Figure 2.18 IMC with reduced overvoltage clamp circuit.
24 Chapter 2

2.3 Conclusion
The basic MC technology and MC protection issues have been presented in this chapter. First,
the basic switch configurations and optional single- and two-stage MC topologies have been
presented. After that, the commutation methods and the systems providing safe operation of
the MCs have been considered briefly.
It has been found that the switches of the MCs can be implemented with different
configurations. RBIGBT technology would provide benefits in MCs. However, they are not
mature technology yet. Different single-stage and two-stage MC topologies with different
number of input and output phases have been suggested in the literature. In theory, the number
of combinations of input and output phases is unlimited. In this thesis, three-to-three phase
DMC and IMC are the only topologies researched further. Other topologies have been
introduced as a background for [P1][P7].
The bidirectional switch configurations of the MCs do not provide safe operation
automatically. Thus, several multi-step commutation methods have been suggested in the
literature. Some of them are based on output current direction information and the others on
input voltage polarities. After introducing the basic commutation methods they have been
compared. It has been found that the methods based on output current direction are safer than
the commutation methods based on input voltage polarities because the misdetection of the
output current direction does not have as serious consequences as the misdetection of the input
voltage polarity. In addition, the detection of current direction is more robust than the
detection of voltage polarity. Thus, the four-step output current detection commutation has
been chosen to be used in this thesis for the DMC and the ISB [P1][P7]. The commutation of
the ILB has been found to be identical to the commutation of the VSI, i.e. blanking times are
used between switchings.
In addition to the safe commutation methods, the MCs require overvoltage protection because
the commutation failures are still possible. It has been found that several different methods for
the overvoltage protection have been suggested. In most cases, a separate circuit to store or
dissipate the energy have been used. For simplicity, the diode-bridge-based circuit the dc link
of which contains capacitor and varistor in parallel has been chosen to be used in the
experimental setups of this thesis because only the low-power applications are considered
[P1][P7].
25
3 Matrix Converter Control
There are several different open- and closed-loop control methods suitable for matrix
converters (MCs). The open-loop methods can be considered pure modulation strategies if
load control systems are ignored. In addition, the closed-loop control systems do not usually
control the converter itself, but rather the load, so that they also contain a modulator. Thus,
modulation is one of the key issues with MCs.
This chapter seeks to present an extensive coherent of the modulation method applied in [P1]
[P7]. First, the space vector theory is introduced briefly as background. After that, the basics
of MC modulation are presented. That is followed by a brief review of MC modulation
schemes presented in the literature. After that, the derivation of the space vector modulation
(SVM) method for the MCs applied in [P1][P7] is presented, based mostly on [P1], but in
more coherent and extensive form. Then, four SVM methods are compared, concentrating on
the common-mode voltages based on [P2]. At the end of the chapter, also the synchronisation
of the MC operation to the supply applied in [P1][P7] is described, with a brief comparison
with other methods suggested in the literature.
3.1 Space Vector Theory
The presentation of modulation and control methods is based on space vectors. Space vectors
present a three-phase system of time-variant quantities such as voltage, current or flux in a
complex plane and were originally used to model the dynamic behaviour of ac machines
[Kov62]. However, they can be used to describe any kind of three-phase system, e.g. three-
phase power electronic converters and their modulation, as presented in power electronic
literature, e.g. in [Aka84], [Bro88], [Hol89], [Hol03] and [Nov96]. In addition, space vector
theory is used in the control of ac machines, as presented e.g. in [Nov96], [Vas90], [Vas92]
and [Kaz02]. Next, a definition and some transformation equations of the space vector applied
in [P1][P7] and below are given.
Let x
a
, x
b
and x
c
be the instantaneous values of arbitrary three-phase variables forming the
space vector x [Kov62]:
( )
c
2
b a
3
2
x a x a x x + + = , (3.1-1a)
where a = e
j2/3
. When quantities are purely sinusoidal and symmetric, and they have constant
angular frequency, the space vector x can be presented as
26 Chapter 3

t t
x x x
x x
j j
e e

= = , (3.1-1b)
where |x| is the magnitude of the space vector x, which is equal to the amplitude of the phase
magnitudes: x x x x x | |
c b a
= = = = ,
x
is the angular frequency of x
(a,b,c)
and t is time. For the
sake of simplicity, time is not presented in parentheses after variables x even though they are
time-variant. When the three-phase system considered contains a zero sequence component or
an average x
0
, it has to be taken into account separately:
x
0
= (x
a
+ x
b
+ x
c
)/3 (3.1-2)
It is often useful to present x using two-axis components, i.e. to separate it into the projections
of the real- and imaginary-axis, Re and Im, respectively:
x = Re{x}+jIm{x} = x

+ j

x

. (3.1-3)
In a stationary reference frame, the transformation from a three-phase to the two-axis and zero
sequence components can be performed as follows [Kov62]:

c
b
a
0

2 / 1 2 / 1 2 / 1
2 / 3 2 / 3 0
2 / 1 2 / 1 1
3
2
x
x
x
x
x
x
. (3.1-4)
Thus, the inverse transformation is

c
b
a
1 2 / 3 2 / 1
1 2 / 3 2 / 1
1 0 1
x
x
x
x
x
x
. (3.1-5)
Sometimes it is advantageous to use some other reference frame instead of the stationary one.
Let
k
be the instantaneous angle of an arbitrary reference frame, let
x
be the angle of the x in
the stationary reference frame at the same instant as
k
, i.e.
x
=
x
t, and let x
k
be the space
vector x transformed from the stationary to the arbitrary reference frame:
k k x k x
j j j ) j(
k
e e e e

= = = x x x x
t
(3.1-6)
The transformations are then in matrix form
( ) ( )
( ) ( )

c
b
a
k k k
k k k
0

k k
k k
0
kq
kd
2 / 1 2 / 1 2 / 1
3 / 4 sin 3 / 2 sin sin
3 / 4 cos 3 / 2 cos cos
3
2
1 0 0
0 cos sin
0 sin cos
x
x
x
x
x
x
x
x
x




(3.1-7)
and the inverse transformations are, respectively,
Matrix Converter Control 27
( ) ( )
( ) ( )

0
kq
kd
k k
k k
k k
c
b
a
0
kq
kd
k k
k k
0

1 3 / 4 sin 3 / 4 cos
1 3 / 2 sin 3 / 2 cos
1 sin cos
1 0 0
0 cos sin
0 sin cos
x
x
x
x
x
x
x
x
x
x
x
x





. (3.1-8)
It is possible to express the instantaneous active power p of a three-phase three-wire system
having phase voltages u
(a,b,c)
and currents i
(a,b,c)
using the space vector theory [Kov62],
[Aka84], [Vas92]:
{ } ( ) ( )
kq kq kd kd c c b b a a
2
3
2
3
Re
2
3
i u i u i u i u i u i u i u i u p + = + = = + + =

, (3.1-9)
where i
*
denotes the complex conjugate of i, the space vectors u and i are formed by u
(a,b,c)
and
i
(a,b,c)
, respectively, and u
(,)
, i
(,),
u
(kd,kq)
and i
(kd,kq)
are the two-axis components of u and i,
following (3.1-4) and (3.1-7).
Sometimes the relations between sine and cosine and their exponent forms are useful in the
analysis with space vectors:
j 2
e e
sin
j j x x
x

= , x (3.1-10a)
2
e e
cos
j j x x
x

+
= , x (3.1-10b)
Finally, the input phase voltage and current and the output voltage and current space vectors of
an MC are denoted by u
i
, i
i
, u
o
and i
o
, respectively:
( )
ic
2
ib ia i
3
2
u a u a u u + + = , (3.1-11) ( )
ic
2
ib ia i
3
2
i a i a i i + + = , (3.1-12)
( )
C
2
B A o
3
2
u a u a u u + + = , (3.1-13) ( )
C
2
B A o
3
2
i a i a i i + + = , (3.1-14)
where u
i(a,b,c)
refer to the input phase voltages, i
i(a,b,c)
to the input currents, u
(A,B,C)
to the output
phase voltages and i
(A,B,C)
to the output currents. When phase quantities are purely sinusoidal
and symmetric, and they have constant angular frequency, the space vectors in (3.1-11)(3.1-
14) can be presented as
t t
u u u
i i
j
i
j
i i
e e

= = , (3.1-15)
( ) ( )
1 , i i 1 , i i
j
i
j
i i
e

e

= =
t t
i i i , (3.1-16)
t t
u u u
o o
j
o
j
o o
e e

= = , (3.1-17)
( ) ( )
1 , o o 1 , o o
j
o
j
o o
e

e

= =
t t
i i i , (3.1-18)
where
i
refers to the input angular frequency,
o
to the output angular frequency,
(i,o),1
are
the fundamental displacement angles between voltage and current of the input and the output,
respectively, and
i
,
i
,
o
and
o
are the magnitudes of the respective quantities.
28 Chapter 3

3.2 Overview of Matrix Converter Modulation
This section deals with the modulation of MCs in a general manner and also presents a brief
review of different modulation and control methods suggested for MCs in the literature. Thus,
this section serves as an introduction for the derivation of the space vector modulation (SVM)
method presented in Section 3.3 and used in [P1][P7].
In theory, an ideal direct three-to-three phase MC (DMC) as in Figure 3.1 has 512, i.e. 2
9
,
different switching states. However, short-circuiting of the input phases (a, b, c) is not allowed
and the output current path cannot be opened. Thus, every output phase (A, B, C) must always
be connected to a single input phase. This reduces the number of switching states to 27, as
presented in Table 3.1. Thus, when the switching functions sw
(A,B,C)(a,b,c)
are used to present the
state of the switches S
(A,B,C)(a,b,c)
(Figure 3.1), respectively, the system always has to follow
sw
Aa
+ sw
Ab
+ sw
Ac
= 1, sw
Ba
+ sw
Bb
+ sw
Bc
= 1, sw
Ca
+ sw
Cb
+ sw
Cc
= 1, (3.2-1)
where sw
(A,B,C)(a,b,c)
has values 0 or 1 describing on-state and off-state, respectively. Thus, the
output phase voltages u
(A,B,C)
are instantaneously

ic
ib
ia
Cc Cb Ca
Bc Bb Ba
Ac Ab Aa
C
B
A
u
u
u
sw sw sw
sw sw sw
sw sw sw
u
u
u
4 4 4 3 4 4 4 2 1
SW
, (3.2-2)
where u
i(a,b,c)
are the input phase voltages. Respectively, the input currents i
i(a,b,c)
are

C
B
A
Cc Bc Ac
Cb Bb Ab
Ca Ba Aa
ic
ib
ia
i
i
i
sw sw sw
sw sw sw
sw sw sw
i
i
i
, (3.2-3)
where i
(A,B,C)
are the output currents. As a result, equations (3.2-1)(3.2-3) describe the state of
the DMC in all 27 states permitted.
a
b
c
A
B
C
S
Aa
S
Ab
S
Ac
S
Ba
S
Bb
S
Bc
S
Ca
S
Cb
S
Cc
i
A
i
B
i
C
i
ic
i
ib
i
ia
u
ic
u
ia
u
ib
u
A
u
B
u
C

Figure 3.1 Ideal DMC with notations.
Matrix Converter Control 29
In Table 3.1, the 27 switching states permitted are divided into three groups, each of which has
different characteristics. Group I contains switching state connecting every output phase to a
different input phase, which produces voltage space vectors rotating with the input angular
frequency
i
. Below, it is assumed that the input voltages are sinusoidal and symmetric with
constant magnitude
i
and angular frequency
i
. For example, the value of the switching state
matrix SW with State I-iii is

=
1 0 0
0 0 1
0 1 0
SW ( )
( ) 6 / 7 j
i ic
2
ia ib o
i
e
3
2
3
2

= + + =
t
u u a u a u u

, (3.2-4)
whose angle varies with time and whose magnitude is constant. In Table 3.1, the switching
states of Group II use only two input phase voltages at a time. Thus, the switching states of
Group II produce voltage space vectors with constant angle and varying magnitude with the
assumption of ideal input voltages made above. For example, State IIB-i means
Table 3.1 Permitted switching states of DMC.
Phase
Output voltage Input current Switching function values
Group
State
A B C u
AB
u
BC
u
CA
i
ia
i
ib
i
ic
sw
Aa
sw
Ab
sw
Ac
sw
Ba
sw
Bb
sw
Bc
sw
Ca
sw
Cb
sw
Cc
i
a b c u
ab
u
bc
u
ca
i
A
i
B
i
C
1 0 0 0 1 0 0 0 1
ii
a c b u
ca
u
bc
u
ab
i
A
i
C
i
B
1 0 0 0 0 1 0 1 0
iii
b a c u
ab
u
ca
u
bc
i
B
i
A
i
C
0 1 0 1 0 0 0 0 1
iv
b c a u
bc
u
ca
u
ab
i
C
i
A
i
B
0 1 0 0 0 1 1 0 0
v
c a b u
ca
u
ab
u
bc
i
B
i
C
i
A
0 0 1 1 0 0 0 1 0
I
vi
c b a u
bc
u
ab
u
ca
i
C
i
B
i
A
0 0 1 0 1 0 1 0 0
i
a c c u
ca
0 u
ca
i
A
0 i
A
1 0 0 0 0 1 0 0 1
ii
b c c u
bc
0 u
bc
0 i
A
i
A
0 1 0 0 0 1 0 0 1
iii
b a a u
ab
0 u
ab
i
A
i
A
0 0 1 0 1 0 0 1 0 0
iv
c a a u
ca
0 u
ca
i
A
0 i
A
0 0 1 1 0 0 1 0 0
v
c b b u
bc
0 u
bc
0 -i
A
i
A
0 0 1 0 1 0 0 1 0
IIA
vi
a b b u
ab
0 u
ab
i
A
-i
A
0 1 0 0 0 1 0 0 1 0
i
c a c u
ca
u
ca
0 i
B
0 i
B
0 0 1 1 0 0 0 0 1
ii
c b c u
bc
u
bc
0 0 i
B
i
B
0 0 1 0 1 0 0 0 1
iii
a b a u
ab
u
ab
0 i
B
i
B
0 1 0 0 0 1 0 1 0 0
iv
a c a u
ca
u
ca
0 i
B
0 i
B
1 0 0 0 0 1 1 0 0
v
b c b u
bc
u
bc
0 0 i
B
i
B
0 1 0 0 0 1 0 1 0
IIB
vi
b a b u
ab
u
ab
0 i
B
i
B
0 0 1 0 1 0 0 0 1 0
i
c c a 0 u
ca
u
ca
i
C
0 i
C
0 0 1 0 0 1 1 0 0
ii
c c b 0 u
bc
u
bc
0 i
C
i
C
0 0 1 0 0 1 0 1 0
iii
a a b 0 u
ab
u
ab
i
C
i
C
0 1 0 0 1 0 0 0 1 0
iv
a a c 0 u
ca
u
ca
i
C
0 i
C
1 0 0 1 0 0 0 0 1
v
b b c 0 u
bc
u
bc
0 i
C
i
C
0 1 0 0 1 0 0 0 1
IIC
vi
b b a 0 u
ab
u
ab
i
C
i
C
0 0 1 0 0 1 0 1 0 0
i
a a a 0 0 0 0 0 0 1 0 0 1 0 0 1 0 0
ii
b b b 0 0 0 0 0 0 0 1 0 0 1 0 0 1 0 III
iii
c c c 0 0 0 0 0 0 0 0 1 0 0 1 0 0 1
30 Chapter 3

=
1 0 0
0 0 1
1 0 0
SW
( )
( )
3 / 5 j i i 3 / 5 j ica
ic
2
ia ic o
e
3
6 / 7 sin 2
e
3
2
3
2
= = + + =
t u u
u a u a u u

, (3.2-5)
whose angle is 5/3 and whose magnitude varies. The last group in Table 3.1 is Group III,
where all output phases are connected to the same input phase, producing zero output voltages.
For example, State III-iii means

=
1 0 0
1 0 0
1 0 0
SW ( ) 0
3
2
ic
2
ic ic o
= + + = u a u a u u . (3.2-6)
3.2.1 Review of Modulation Methods
Next, different ways to define the values of switching function matrix SW are reviewed
briefly. Below, it is again assumed that input voltages are sinusoidal and symmetric with
constant magnitude and angular frequency.
The first modulation method presented for the DMC is the Venturini method suggested in
1980 [Ven80a], [Ven80b]. In this original Venturini method, every output phase voltage is
formed separately and fundamental wave voltage references are always between the negative
and positive envelope of the input voltage. Thus, the maximum attained input-to-output
voltage transfer ratio () is . The method was further developed in the late 1980s to attain the
full voltage transfer ratio of MCs, i.e. 2 / 3 [Ale88], [Ale89]. The increase in the voltage
transfer ratio was achieved by adding two frequency components to the fundamental output
phase voltage reference. The frequencies of the additional components are the third-order
harmonic of the input frequency and the third-order harmonic of the output reference
frequency. When the magnitudes of those third-order components are chosen correctly, the full
voltage transfer ratio is achieved. The approach in the Venturini method is direct and thus the
totally independent control of each output phase leads to the unrestricted use of the switching
states, i.e. all states presented in Table 3.1 are possible. This can be an advantage when the
number of output phases is increased or the separate control of output phase voltages is
preferable otherwise. However, it is not easy to avoid the combinations producing high
common-mode voltages, i.e. zero sequence as in (3.1-2). To do so would require a much more
complex modulation system.
MCs were analysed applying the ideal indirect MC (IMC) circuit in Figure 3.2 already in the
1980s [Zio85], [Hol89], [Hub89], [Oya89], [Whe02]. The indirect modulation approach
always leads to the limitation of possible switching states because the states of Group I cannot
be formed with the IMC. Despite that, the full input-to-output voltage ratio of 2 / 3 is still
always achieved.
Matrix Converter Control 31
The indirect approach also made it possible to apply the space vector modulation (SVM) with
MCs as introduced in [Hub89]. With this first SVM method, the ISB is controlled so that the
fictitious dc link voltage u
dc
is the positive envelope u
i,LL,env+
of the input line-to-line voltages:
u
dc
= u
i,LL,env+
, whose average / 3
LL i, dc
u U = , (3.2-7)
where
i,LL
is the magnitude of the ideal input line-to-line voltage. The method uses only two
input phases during the modulation period T
m
. Thus, it could be implemented with the
simplified two-stage MC presented in Figure 2.5 if a two-stage topology were used. However,
a full DMC circuit is required if a single-stage MC topology is used.
The SVM was extended to produce sinusoidal supply currents in the early 1990s [Hub91],
[Wie90], [Cas93], [Hub95]. With the more developed SVM introduced in [Hub91], the
fictitious IMC supply bridge (ISB) is modulated using all three input phases during T
m
. That
produces sinusoidal supply currents and a PWM-shaped link voltage u
dc
, whose average U
dc
is
U
dc
= d

u
i,LL,env+
+ d

u
i,LL,med+
, (3.2-8)
where d

and d

are the duty ratios of two ISB switching states used in the modulation period
and u
i,LL,med+
is the line-to-line voltage nearest below the positive envelope, as shown in Figure
3.3. An extensive comparison of the SVM and Venturini methods can be found in [Zha98].
A
B
C
S
pa
S
pb
S
pc
S
na
S
nb
S
nc
S
Ap
S
Bp
S
Cp
S
An
S
Bn
S
Cn
u
dc
p
n
i
dc
i
ic
i
ib
i
ia
i
C
i
B
i
A
a
b
c
u
ic
u
ia
u
ib
u
dc
2
u
dc
2
u
A
u
B
u
C

Figure 3.2 Ideal IMC with notations.
u
icb
u
iab
u
iba

i
t
u
i,LL,med+
u
i,LL,env+
u
ibc
u
icb
u
ica
u
iac
voltage range
forming u
dc

Figure 3.3 Input line-to-line voltages and the forming of the dc link voltage u
dc
with notations.
The indirect approach also makes it possible to apply the carrier-wave-based modulation
strategies of conventional dc link converters in the modulation of the MCs. [Oya89] introduces
a carrier wave modulation method that cannot produce a sinusoidal supply current waveform,
i.e. the result is equal to the first SVM of the MCs. If the method suggested in [Oya89] were
applied to the IMC, its fictitious dc link voltage u
dc
would follow (3.2-7). [Oya93] and
32 Chapter 3

[Oya96] present developed versions with sinusoidal supply currents like the developed SVM.
The same holds for the carrier-based method suggested in [Ito05a]. The modulation method
for the real IMC suggested in [Min93] leads to basically similar waveforms as the developed
SVM and carrier wave modulation methods [Hub91], [Oya93].
A totally opposite approach for the modulation of the MCs is presented in [Mil00], where a
direct switching matrix approach is used to derive a modulation method, applying only the
switching states of Groups I and III presented in Table 3.1. However, that kind of modulation
does not have any benefits compared to the methods considered above.
3.2.2 Compound Control and Modulation Strategies
All the modulation methods described above were based on a conventional scheme where the
modulator is a separate part getting reference values from a higher level system, e.g. a motor
controller. However, other kinds of methods to define the switching state matrix have also
been presented in the literature and a brief review is given here as background information.
[Ml03] and [Ml05] suggest a compound method which uses all switching states presented in
Table 3.1. The state is chosen so that it generates the optimum operation for both the load and
supply sides. The calculation is based on the space vector model containing the whole system,
whose condition is predicted for every switching state and then the state generating the
optimum conditions for the next calculation period is chosen. Compared to the methods above,
the system requires fast measurements from the supply and output currents and the input
voltages, none of which are required in pure modulators, which require synchronisation with
the input only in the simplest case [P3]. In addition, the calculation of the whole system model
for all 27 switching states is computationally heavy with a high-performance drive.
A supply power control approach has been suggested in [Nog06], where the system is based
on the indirect approach and the separate control of the fictitious IMC load bridge (ILB) and
the ISB. The ILB control is based on the load current feedback control giving output voltage
references for a carrier-based modulator. Active and reactive supply powers are defined by
measured supply voltages and currents. These actual power values are then subtracted from
their references, which are the load active power based on measured load quantities and the
reactive supply power reference given by a higher-level system. The error components of the
supply power are then used together with the measured supply voltage to define the optimum
state for the fictitious ISB. The system measures all supply and output currents and voltages.
Both [Ham06] and [Pin06] propose the use of sliding mode control to define the optimum
switching state in MCs. In [Ham06], the sliding mode control is applied in the IMC to control
the ISB, whose switching states are then set in a synthesis with the ILB, using the SVM. In
[Pin06], the DMC is controlled with a sliding mode controller which defines the optimum
switching state by applying space vectors. Both sliding mode controllers contain a system
model and thus they can be considered computationally heavy compared to modulators. On
the other hand, the controller proposed in [Pin06] automatically contains the load control
Matrix Converter Control 33
system. That reduces the difference compared to modulator systems, which in practice often
require an additional control system for the load.
The separate control system can be applied purely to the load control, too, as suggested in
[Cas98c] and verified in [Cas01b]. In these, the direct torque control of a load machine is
applied to define the optimum state of the fictitious ILB, whereas the switching state of the
fictitious ISB is chosen so that the input power factor becomes unity. The states of both
bridges are constant for a calculation period. Thus, the switching frequency is variable, which
causes resonance in the supply-side LC filter. A modified version of the direct torque control
in MCs is presented in [Cas01a] and [Ort05], where two ISB states are taken in every
calculation period. Thus, the switching frequency becomes constant and can be taken into
account in the design of the supply LC filter. Another direct control method without separate
modulator is presented in [Mut02]. The method is based on the measurements of both the
supply and load currents. A closed-loop system controls the load currents, whereas the supply
currents are used to define an optimum switching state.
3.3 Space Vector Modulation
This section presents the space vector modulation (SVM) method for the MCs applied in [P1]
[P7]. The section deals with the derivation of the duty cycles and the optimum pulse patterns,
both based on [P1][P2]. In addition, the section deals with the synchronisation of the MC to
the supply voltage, extending the presentation given in [P3].
3.3.1 Derivation of Duty Cycles
The SVM method applied in [P1][P7] has been suggested in [Hub95]. The derivation of the
SVM has usually been presented in matrix form, as e.g. in [Hub95] and [Zwi01]. In addition,
the derivation has been presented conventionally using the space vector of the output line-to-
line voltages, as e.g. in [Cas93], [Hub95], [Nie96b] and [RP3]. This is a natural approach with
the DMC having no dc link midpoint, which is a common reference point in the SVM of the
VSI, where the space vector of the output phase voltage is commonly used. To show the
similarity of conventional VSCs and CSCs to MCs, the phase voltage space-vector-based
approach has been used in [P1]. That is also the approach used in this section, which is mostly
the extended presentation of [P1]. Below, ideal sinusoidal and symmetric input and output
quantities with constant frequencies and magnitudes are assumed. It is also assumed that the
input and output quantities remain constant during the calculation period. The notations are
based on the ideal IMC circuit presented in Figure 3.2.
Supply Bridge (Fictitious ISB) [P1]
It is possible to describe the switching states of the input phases a, b and c shown in Figure 3.2
by switching functions sw
a
, sw
b
and sw
c
, respectively. As presented in (3.2-1), the short-
circuiting of the input phases is not allowed. Thus, the phase switching functions can have
three values: 1, 0 and 1, denoting a connection to the dc link bar p (the upper switch on), the
34 Chapter 3

unconnected state (both switches off) and a connection to the dc link bar n (the lower switch
on), respectively. As a result, the input switching functions always fulfil
sw
a
+ sw
b
+ sw
c
= 0. (3.3-1)
Thus, the input currents are: i
ia
=

sw
a
i
dc
,

i
ib
= sw
b
i
dc
and i
ic
= sw
c
i
dc
, which can be substituted
into (3.1-12) and (3.1-16) due to the ideal sinusoidal currents assumed:
( )
( )
1 , i i
j
i dc dc i dc c
2
b a i
e
3
2

= = + + =
t
sw i i sw i sw a sw a sw i , (3.3-2)
where sw
i
is the space vector of the input-side switching functions and i
dc
is the dc link current.
The substitution of the allowed switching function values into (3.3-2) produces the current
vectors presented in Figure 3.4a when i
dc
> 0. When i
dc
< 0, the directions of the current
vectors are reversed, as presented in Figure 3.4b. In Figure 3.4, i
x
(sw
a
,sw
b
,sw
c
) denotes current
vectors when x and x [1,6]. For example, vector i
1
is the switching state where sw
a
= 1,
sw
b
= 0 and sw
c
= 1; thus i
ia
= i
dc
, i
ib
= 0 and i
ic
= i
dc
, i.e. the phase a is connected to the dc
link bar p, the phase b is unconnected and the phase c is connected to the dc link bar n.

i
d

i
i,ref
I
III II
IV
V VI
i
2
(0,1,-1)
i
6
(1,-1,0)
i
3
(-1,1,0)
i
5
(0,-1,1)
i
1
(1,0,-1)
i
4
(-1,0,1)
IV
VI V
I
II III
i
x
(sw
a
,sw
b
,sw
c
)
i
5
(0,-1,1)
i
3
(-1,1,0)
i
6
(1,-1,0)
i
2
(0,1,-1)
i
4
(-1,0,1)
i
1
(1,0,-1)
Im
Re Re
Im
i
dc
> 0 i
dc
< 0

(a) (b) (c)
Figure 3.4 Input current space vectors produced by allowed switching function values when (a) i
dc
> 0
and (b) i
dc
< 0. (c) Synthesis of the input current reference vector i
i,ref
.
Figure 3.4c presents the synthesis of the input current reference vector i
i,ref
in the modulation
period as an average of the two adjacent current vectors:
i
i,ref
= d

+ d

, (3.3-3)
where the duty ratios of the current vectors i

and i

are d

and d

, respectively. The angle of

and d

are

=
i
dc
ref i,

sin
i
i
d , (3.3-4a) ( )
i
dc
ref i,

sin
i
i
d = . (3.3-4b)
When the fictitious ISB is assumed lossless, the instantaneous dc link power u
dc
i
dc
is equal to
the instantaneous input active power p
i
following (3.1-9):
Matrix Converter Control 35
{ }
dc dc
*
i i i
Re
2
3
i u i u p = = . (3.3-5)
After substituting (3.3-2) into (3.3-5) and then dividing the result by i
dc
, we get u
dc
:
{ }
*
i i dc
Re
2
3
sw u u = . (3.3-6)
As presented in (3.3-3), two switching states are used in every modulation period T
m
to
synthesise the reference current vector. The switching function vector sw
i
gets the two values
sw

and sw

respective to the current vectors i

and i

, and their duty cycles d

and d

. Thus, the
dc link voltage u
dc
gets two values in the modulation period T
m
and its average value u
dc,av
in

T
m
is
{ } { } ( )
dc, dc,
*
i
*
i av dc,
Re Re
2
3
u d u d sw u d sw u d u + = + = , (3.3-7)
where u
dc,
and

u
dc,
are the two dc link voltage values during T
m
. According to (3.3-7), the dc
link voltage u
dc,av
gets the maximum value when the current ratio |i
i,ref
|

/

|i
dc
| is the maximum,
which is unity, when sinusoidal currents are aimed at. This means that the fictitious ISB does
not restrict the input current; that depends only on the load and required input fundamental
phase angle
i,1
. Thus, equation (3.3-4) reduces to

=
i
3

sin d , (3.3-8a) ( )
i
sin = d . (3.3-8b)
Load Bridge (Fictitious ILB) [P1]
It is possible to describe the switching states of the output phases A, B and C, shown in Figure
3.2, by switching functions sw
A
, sw
B
and sw
C
, respectively. As presented in (3.2-1), the open
circuit of the load is not allowed. Thus, the phase switching functions can have two values: 1
and 1, denoting a connection to the dc link bar p (the upper switch on) and a connection to
the dc link bar n (the lower on), respectively. As a result, they always fulfil
sw
A
+ sw
B
+ sw
C
0. (3.3-9)
Thus, the output phase voltages with respect to the midpoint of the fictitious dc link are:
u
A
=

sw
A
U
dc
/2,

u
B
= sw
B
U
dc
/2 and u
C
= sw
C
U
dc
/2, which can be substituted into (3.1-13):
( )
( ) t
sw
U U
sw
U
sw a sw a sw u
o
j
o
dc dc
o
dc
C
2
B A o
e
2 2 2 3
2

= = + + = , (3.3-10)
where sw
o
is the space vector of the output-side switching functions and the ideal output
voltages are assumed, i.e. (3.1-17) holds.
The substitution of the allowed switching function values into (3.3-10) produces the voltage
vectors presented in Figure 3.5a, where u
x
(sw
A
,sw
B
,sw
C
) denotes voltage vectors when x
and x [1,6]. For example, vector u
1
refers to the state where sw
A
= 1, sw
C
= 1 and sw
C
= 1,
i.e. the phase A is connected to the dc link bar p and both the phase B and the phase C are
36 Chapter 3

connected to the dc link bar n. Thus, the instantaneous line-to-line voltages are u
AB
= u
dc
,
u
BC
= 0 and u
CA
= u
dc
. The zero vectors u
0+,
u
0
are produced by connecting all output phases
to the bar p and to the bar n, respectively.
Figure 3.5b presents the synthesis of the output phase voltage reference vector u
o,ref
in the
modulation period as an average of the two adjacent voltage vectors:
u
o,ref
= d

+ d

, (3.3-11)
where the duty ratios of the voltage vectors u

and u

are d

and d

, respectively. The angle of

=
o o
dc
ref o,

sin
3

sin
3
m
U
u
d , (3.3-12a)
( ) ( )
o o
dc
ref o,

sin sin
3
m
U
u
d = = , (3.3-12b)
because |u

| = |u

| = 2u
dc,av
/3. In (3.3-12), m denotes the modulation index of the ILB, ranging
from zero to unity.
I III
II
IV
V
VI
u
0+
,u
0-
u
x
(sw
A
,sw
B
,sw
C
)
u
o,ref
u

o
u

u
1
(1,-1,-1)
u
2
(1,1,-1) u
3
(-1,1,-1)
u
6
(1,-1,1)
u
4
(-1,1,1)
u
5
(-1,-1,1)
Im
Re
u
o,ref
u

(a) (b) (c)
Figure 3.5 (a) Output phase voltage space vectors produced by allowed switching function values, where
u
o+
is state (1,1,1) and u
o
is state (1,1,1). (b) Synthesis of the output voltage reference
vector u
o,ref
. (c) Synthesis of u
o,ref
with the resulting active voltage vectors.
Synthesis of Fictitious Supply and Load Bridges [P1]
As presented in Figure 3.2, the fictitious ISB and ILB are connected to each other directly in
the ideal IMC. Thus, equation (3.3-7) can be substituted into (3.3-10), from which we get the
average u
o,av
of the output phase voltage vector in the modulation period T
m
:
( )
dc, dc, av o, av o,
2
1
u d u d sw u + = , (3.3-13)
which takes into account that the average output switching function vector sw
o,av
in T
m
is
sw
o,av
= d

sw

+ d

sw

. (3.3-14)
Thus, we get for the average output phase voltage vector u
o,av
in the modulation period T
m
:
Matrix Converter Control 37
( )( )
{ } { } { } (
{ })

*
i
*
i
*
i
*
i
dc, dc, av o,
Re
Re Re Re
4
3
2
1
u d u d u d u d u d d u d d u d d u d d
sw u sw d d
sw u sw d d sw u sw d d sw u sw d d
u d u d sw d sw d u
+ + + = + + + =
+
+ + =
+ + =
(3.3-15)
Equation (3.3-15) illustrates that in the modulation period T
m
the output voltage is synthesised
by applying both ILB states during both ISB states, i.e. four active voltage vectors u

, u

, u

and u

are used, as shown in Figure 3.5c. The duty cycles of the active voltage vectors are got
with (3.3-8) and (3.3-12):
)
3

sin( )
3

sin(
i o
= = m d d d , (3.3-16a)
) sin( )
3

sin(
i o
= = m d d d , (3.3-16b)
)
3

sin( ) sin(
i o
= = m d d d , (3.3-16c)
) sin( ) sin(
i o
m d d d = = . (3.3-16d)
The rest of the modulation period T
m
, not used for the active duty cycles, is used for the zero
state or the zero vectors u
0(+,)
, whose duty cycle is d
0
:
d
0
= 1 (d

+d

+ d

+ d

). (3.3-16e)
Non-relative duty cycle times T

, T

, T

, T

and T
0
are got by multiplying the duty cycles in
(3.3-16) by the modulation period T
m
. The modulation index m in (3.3-16) is the same as in
(3.3-12), where it is in the link-voltage-based form. However, it may also be expressed as a
direct relation between the output and input voltages as presented next.
As presented above, it is assumed that the input voltages are sinusoidal and symmetric, so that
the input voltage does not contain any components other than the fundamental wave, i.e.
u
i
= u
i,1
. That is also assumed for the input current reference i
i,ref
. It is also assumed that
i
i
= i
i,ref
. As presented in (3.1-15) and (3.1-16), the fundamental displacement angle between
the input current and phase voltage is
i,1
. Thus, the switching function vector sw
i
follows i
i,ref
,
as presented in (3.3-2). In addition, the magnitude of the instantaneous switching function
vector sw
i
has to be equal to the magnitude of the average input switching function vector
sw
i,av
, which contains only the fundamental component. As a result, it is possible to write (3.3-
7) for the average dc link voltage U
dc
in polar coordinates:
( )
{ }
( )
{ }
1 , i i i 1 , i i i
j
av i, i,1
j
av i,
j
i,1 dc
e Re
2
3
e e Re
2
3
+
= =
t t t t
sw u sw u U . (3.3-17)
As presented above (3.3-8), |i
i,ref
|

/

|i
dc
| = 1, leading to

|sw
i
| = 1 by (3.3-2). Thus,
1 , i i,1 dc
cos
2
3
u U = , (3.3-18)
38 Chapter 3

which can be substituted into the equation of the modulation index m obtained from (3.3-12):
1 , i
ref
1 , i i,1
ref o,
cos 3
2
cos 3
2

= =
u
u
m , (3.3-19)
where
ref
is the reference of the input-to-output voltage transfer ratio and it is assumed that
output voltage reference u
o,ref
contains only the fundamental component. In the linear
modulation range, m [0,1]. The maximum value of the voltage transfer ratio reference
ref

for the ideal MC is 2 / 3 0.866, which is attained when cos
i,1
is unity, i.e.
i,1
= 0.
The results obtained above are the same as the results obtained e.g. in [Hub95]. However, the
derivation is based here purely on the space vectors, which offers a possibility for easy
analysis of MC characteristics, e.g. when distortion migration is analysed as in [P3] and [P4].
Originally, the SVM was developed for the DMC by applying the concept of the fictitious
IMC [Cas93], [Hub95], [Nie96b]. However, the SVM presented is also directly suitable for the
real IMC presented in Figure 2.6 because (3.3-19) together with m [0,1] limits the input
fundamental displacement angle:
i,1
[/2, /2], and the dc link voltage u
dc
produced by the
permitted ISB switching states, presented in Figure 3.4, is always positive. Thus, the diodes of
the ILB are always reverse-biased and the dc link is not short-circuited.
3.3.2 Switching Pattern Optimisation
Above, the voltage transfer ratio and the SVM duty ratios for the MCs were derived, but final
switching patterns were not considered. There are several possibilities for arranging the
switching pattern of the duty cycles presented in (3.3-16), as considered e.g. in [Hub95],
[Nie96b], [Wei02], [Apa03], [Cha03], [Lee05b].
Conventional SVM Method (CSVM)
The modulation method applied in [P1][P7] is a conventional SVM (CSVM), which is based
on the optimisation rules presented in [Nie96b]. In the CSMV, the switching patterns are
arranged to minimise the number of commutations. As a result, only one switching is
performed for every duty cycle change.
For example, when both the output voltage and input reference vectors u
o,ref
and i
i,ref
lie in
sector I, the correlations between active output voltage vectors and switching states, according
to Figures 3.4 and 3.5 and Table 3.1, are
u

IIA-vi: abb, u

IIC-iii: aab, u

IIA-i: acc, u

IIC-iv: aac (3.3-20)
An order minimising the state transitions in (3.3-20) is
abb

aab aac

acc, i.e. u

u

u

. (3.3-21)
To achieve symmetric switching patterns in the modulation period T
m
and operation without
switchings when the modulation period changes without simultaneous sector change, a
Matrix Converter Control 39
double-sided symmetrical pattern with zero vector u
0
, i.e. u
0(+,)
, in the middle of T
m
is used
leading to
u

u

u

u
0
u

, (3.3-22a)
whose duty cycles fulfil, respectively,
T
m
= (d

+ d

+ d

+ d

+ 2d
0
+ d

+ d

+ d

+ d

)

T
m
/

2. (3.3-22b)
According to (3.3-21)(3.3-22) and Table 3.1, the suitable zero vector is the state III-iii: ccc.
The optimum switching patterns can be defined for all 36 combinations of input and output
sectors, as presented in Appendix B. As a result, the rules for optimum pattern generation
presented in [Nie96b] combined with the approach in Section 3.3.1 lead to the following rules,
where the notations of the subscripts are used for clarity:
The pattern on the input side is always --.
The pattern of the output side is --0-- if the sum of the numbers of the sectors in
which the input and output references lie, is even.
The pattern of the output side is --0-- if the sum of the numbers of the sectors in
which the input and output references lie, is odd.
If the input reference lies in an even-numbered sector, i.e. II, IV or VI, the zero vector
u
o+
, i.e. (1,1,1) as presented in Figure 3.5, is used.
If the input reference lies in an odd-numbered sector, i.e. I, III or V, the zero vector u
o
,
i.e. (1,1,1) as presented in Figure 3.5, is used.
The duty cycle patterns of the CSVM are shown in Table 3.2 and some examples are shown in
Table 3.3, both in the row CSVM.
Table 3.2 Duty cycle patterns of the modulation methods considered.
Method
Sum of
sectors
State 1 State 2 State 3 State 4 State 5 State 6 State 7 State 8 State 9
Even d

/2 d

/2 d

/2 d

/2 d
0
d

/2 d

/2 d

/2 d

/2
CSVM
Odd d

/2 d

/2 d

/2 d

/2 d
0
d

/2 d

/2 d

/2 d

/2
Even d

/2 d

/2 d

/2 d

/2 d
0
d

/2 d

/2 d

/2 d

/2

i
< /6
Odd d

/2 d

/2 d

/2 d

/2 d
0
d

/2 d

/2 d

/2 d

/2
Even d
0
/2 d

/2 d

/2 d

/2 d

/2 d

/2 d

/2 d
0
/2
ISVM

i
/6
Odd d
0
/2 d

/2 d

/2 d

/2 d

/2 d

/2 d

/2 d
0
/2
Even d

/2 d

/2 d
0
/2 d

/2 d

/2 d
0
/2 d

/2 d

/2
ECSVM
Odd d

/2 d

/2 d
0
/2 d

/2 d

/2 d
0
/2 d

/2 d

/2
State 1 State 2 State 3 State 4 State 5 State 6 State 7 State 8 State 9 State 10 State 11
Even d
0
/4 d

/2 d

/2 d

/2 d

/2 d
0
/2 d

/2 d

/2 d

/2 d

/2 d
0
/4
NZSVM
Odd d
0
/4 d

/2 d

/2 d

/2 d

/2 d
0
/2 d

/2 d

/2 d

/2 d

/2 d
0
/4
T
m
/2 T
m
/2

Instead of the zero state, any other state could also be taken in the middle if the sequence were
organised differently. However, the minimum duty cycle duration is limited in real converters
due to the limited operation speed of semiconductors. Thus, if the zero state is divided
40 Chapter 3

between some of the States 24 and 68, the limitation takes place with double as long duty
cycle duration as in the middle of the CSVM presented. According to [Apa03], the lack of
zero states during the modulation period increases the modulation frequency output voltage
distortion. If the total number of zero states is increased by dividing the zero duty cycle into
several parts, e.g. adding zeros to the beginning and the end of the CSVM sequence, the
number of switchings increases by two, leading to higher switching losses than in the CSVM.
On the other hand, the modulation methods with only one or two zero states generate more
output voltage distortion at very low modulation depths than the methods with three or more
zero states as presented in [Apa03]. However, that cannot be considered as important as low
power losses under nominal operation point. In addition, the modulation method can be
changed depending on the instant requirements for modulation, which is the state-of-the-art
solution in commercial frequency converters, i.e. more zero states could be used at low
modulation depths only.
Table 3.3 Examples of supply phases connected to output phases with the modulation methods.
Sector of State 1 State 2 State 3 State 4 State 5 State 6 State 7 State 8 State 9
Method
i
i,ref
u
o,ref
ABC ABC ABC ABC ABC ABC ABC ABC ABC
I I abb aab aac acc ccc acc aac aab abb
CSVM
II I aac acc bcc bbc bbb bbc bcc acc aac
I I abb aab aac acc ccc acc aac aab abb

i
< /6
II I aac acc bcc bbc bbb bbc bcc acc aac
I I bbb abb aab aac acc aac aab abb bbb
ISVM

i
/6
II I aaa aac acc bcc bbc bcc acc aac aaa
I I abb aab aaa aac acc aac aaa aab abb
ECSVM
II I aac acc ccc bcc bbc bcc ccc acc aac
State 1 State 2 State 3 State 4 State 5 State 6 State 7 State 8 State 9 State 10 State 11
ABC ABC ABC ABC ABC ABC ABC ABC ABC ABC ABC
I I cbb abb aab aac acc bcc acc aac aab abb cbb
NZSVM
II I aab aac acc bcc bbc bba bbc bcc acc aac aab
T
m
/2 T
m
/2
Comparison of Common-Mode Voltages Produced by Different SVM Methods
In addition to switching losses and distortion, the modulation method also affects the
common-mode voltage of output as discussed in [Cha03], [Lee05b] and [P2]. The common-
mode voltage is the average value of the instantaneous output phase voltages and it is seen
between the load neutral point and the ground. Thus, it is the zero sequence component of the
output voltage and its instantaneous value is defined by (3.1-2).
The correspondence between vector sectors shown in Figures 3.43.5 and input and output
waveforms are presented in Figure 3.6, where it is assumed that the input displacement power
factor is set at unity (
i,1
=

0). In Figure 3.6, u
i,max
, u
i,med
and u
i,min
are the maximum, the
middlemost and the minimum input phase voltage, respectively, and
(i,o)
are the angles of the
input current and the output voltage reference vectors inside the sector, respectively.
With active voltage vectors, two input voltages are connected to output phases, as shown in
Tables 3.1 and 3.3. For example, when both reference vectors lie in sector I, State 2 of the
Matrix Converter Control 41
CSVM is aab according to Table 3.3. When symmetric input voltages are assumed, that
produces common-mode voltage u
0
by (3.1-2)
u
0
= (u
ia
+ u
ia
+

u
ib
)/3 = (2u
ia
u
ia
u
ic
)/3 = u
iac
/3. (3.3-23)
Line-to-line voltage u
iac
has peak value when
i
= /3 in input sector I. Thus, the maximum
possible u
0
in (3.3-23) is
i,LL
/3 190 V with the 400-V line-to-line supply voltage. With the
CSVM, the zero states are also used and during them all output phases are connected to the
same input phase, as shown in Tables 3.1 and 3.3. For example, when both reference vectors
lie in sector I, the zero state, i.e. State 5 in Table 3.3, is ccc, which produces
u
0
= (u
ic
+ u
ic
+

u
ic
)/3 = u
ic
. (3.3-24)
As shown in Figure 3.6a, the maximum absolute value of u
ic
in input sector I is

i
sin(/3) 280 V with the 400-V supply. As can be seen in Figure 3.6a, input phase voltage
u
ic
has the lowest absolute value, i.e. it is u
i,med
, in sector I when
i
< /6. When
i
/6, u
ic
is
u
i,min
in input sector I. Analysing the CSVM with
i,1
= 0 as in Figure 3.6a, it can be found that
the situation is similar for every input sector, i.e. u
i,min
and u
i,max
are used during the zero
vector. The common-mode voltage step occurs at every switching. For example, the transition
from State 1 to State 2 when both reference vectors lie in sector I causes
u
0,1
u
0,2
= (u
ibc
u
iac
)/3 = u
iba
/3. (3.3-25)
Line-to-line voltage u
iba
has peak value when
i
= 0 in input sector I. Thus, the maximum
possible step of the common-mode voltage is
i,LL
/3 190 V with the 400-V supply. As a
result, the maximum common-mode voltage in the CSVM is
i
sin(/3) ( 280 V) and the
maximum common-mode voltage step is
i 3 / 3
( 190 V) [P2].
u
ic
u
ia
u
ib

i
t
I VI V II III IV
V
u
C,ref
u
A,ref
u
B,ref

o
t
I VI V II III IV V Sectors of i
i
when
i
= 0.
IV Sectors
of u
o,ref
u
i,med
u
i,max
u
i,min

i

i

i

i

i

i

o

o

o

o

o

o


(a) (b)
Figure 3.6 Correspondence between sectors and waveforms: (a) input phase voltages u
i(a,b,c)
and the
sectors of Figure 3.4a when cos
i
= 1, (b) output phase voltage references u
(A,B,C),ref
and the
sectors of Figure 3.5a.
In [Cha03], an improved space vector modulation method (ISVM) based on the CSVM is
introduced. The principle of the ISVM is presented in Tables 3.2 and 3.3 in the row ISVM:
the operation is the same as with the CSVM when the angle
i
is below /6. When the angle
i

is greater than /6, the zero state is divided into States 1 and 9 and the principle of single
switching per vector change is retained. Thus, the zero state used in the ISVM is always
42 Chapter 3

formed by u
i,med
, as can be seen in Figure 3.6a. For example, with both reference vectors lying
in sector I, u
ic
is the output voltage used during the zero states only when
i
< /6 and u
ib
is
used instead when
i
/6, as can be seen in Figure 3.6a and Table 3.3. As a result, the
common-mode voltage maximum is no longer attained during the zero states, producing the
maximum of
i
sin(/6) ( 160 V), but rather during the active states, leading to the
maximum of
i 3 / 3
( 190 V). Thus, it is equal to the maximum step, which is the same as
in the CSVM, as can be seen by comparing the state transitions in Table 3.3 and Figure 3.6a.
In addition to common-mode voltage minimisation, the ISVM also decreases switching losses
compared with the CSVM because the average voltage involved in the switchings at the edges
of the zero duty cycle is minimised.
The other possibility to limit the common-mode voltage to the maximum voltage during the
active states is to use the active states only [Lee05b], [P2]. The method is called the non-zero
space vector modulation method (NZSVM) and its principle is described in Tables 3.2 and 3.3
in the row NZSVM. Just like the ISVM, the NZSVM is also based on the CSVM. The zero
duty cycle is now divided into three parts: d
0
/2, d
0
/4 and d
0
/4. As presented in Tables 3.2 and
3.3, the active state taken during both d
0
/4 cycles, i.e. in States 1 an 11, produces opposite
voltage compared to the voltage produced by the state during d
0
/2 in State 6. Thus, these
voltages compensate each other and the total effect is zero when the whole modulation period
is considered. At the same time, the rule is that there is again only one switching allowed for
every normal state transition. For example, when both reference vectors lie in sector I, the
modulation period is started and finished using the state cbb in States 1 and 11, producing the
voltage which is compensated by the voltage produced by state bbc in State 6. This means,
that the sum of the voltages produced by States 1, 6 and 11 in the modulation period is zero.
The main drawback of the NZSVM is the greater number of switchings compared to the
CSVM and the ISVM. Thus, it is also more sensitive to the limitation of duty cycle duration in
the case of the division of the zero state, as presented above.
The division of the zero states into two or three parts as suggested e.g. in [Apa03], generates
non-minimised common-mode voltages: the use of all three zero vectors in a modulation
period evidently leads to the maximum common-mode voltage of
i
, i.e. 326 V.
All methods presented above are directly suitable for both the DMC and the IMC. However,
they do not tap all the possibilities provided by the IMC. As discussed in Chapter 2, the IMC
can be modulated so that the special commutation methods are not required in the ISB. This is
brought about in such a way that the state of a real ISB, i.e. duty cycle change from d

to d

and vice versa, is performed in the middle of the duty cycle d
0
performed by the ILB [Hol89],
[Wei01], [Kol02]. The method is called the easy commutation space vector modulation
method (ECSVM) and its principle is presented in Tables 3.2 and 3.3 in the row ECSVM.
Comparing the ECSVM pulse patterns in Table 3.3 with the input phase voltages in Figure
3.6a and deriving the patterns for all sector combinations, it can be found that the ECSVM
produces the zero states which connect the input voltage with the highest magnitude, i.e. u
i,max

Matrix Converter Control 43
when |u
i,max
| > |u
i,min
| and u
i,min
when |u
i,max
| < |u
i,min
|, to every output phase, resulting always in
the maximum common-mode voltage of
i
, i.e. 326 V. Thus, the ECSVM is not an optimum
solution for common-mode voltages. In addition, the ECSVM does not provide safe
commutation when the duration of the zero duty cycle is not long enough for the commutation
of the ISB. Thus, the ECSVM must always contain a zero state, which limits the attainable
maximum output voltage.
To illustrate the common-mode voltage waveforms, simulated common-mode voltages of all
four methods are presented in Figures 3.73.8 when input frequency f
i
= 50 Hz, supply voltage
is 400 V, output frequency f
o
= 40 Hz and input-to-output voltage ratio = 0.8 [P2]. The
simulation models are ideal Matlab Simulink models, presented more extensively in Chapter
4. Because of the ideal models, the DMC and the IMC are not presented separately [P2]. The
simulation results show that the methods produce the common-mode voltages expected. The
common-mode voltages in the ISVM and in the NZSVM are between 190 V, whereas the
CSVM produces a common-mode voltage between 280 V, as can be seen in Figures 3.7a and
3.8a. As expected, the common-mode voltage of the ECSVM is between 326 V.
(a) (b)
(c) (d)
Figure 3.7 Common-mode voltages of matrix converter space vector modulation methods when
f
o
= 40 Hz and = 0.8: (a) CSVM, (b) ISVM, (c) NZSVM, (d) ECSVM.
(a) (b)
(c) (d)
Figure 3.8 Common-mode voltages of matrix converter space vector modulation methods when
f
o
= 40 Hz and = 0.8: (a) CSVM, (b) ISVM, (c) NZSVM, (d) ECSVM.
44 Chapter 3

Despite the possible advantages of the other methods, the CSVM was chosen to be the
modulation method in the research presented in [P1] and [P3][P7], whereas in [P2] the
different modulation methods were compared. The CSVM can be implemented with simpler
logic than the other methods, thus making the use of simple logic circuits possible.
Table 3.4 shows the switching function space vectors for the DMC and the IMC when they are
modulated with the CSVM, i.e. it presents the rows CSVM of Tables 3.2 and 3.3 in
switching function vector form. If the average dc link voltage were measured from an IMC
circuit modulated with the CSVM in practice, it would not be the same as in (3.3-7) or (3.3-
17). The reason for this is presented in the rules above and it can be seen in Table 3.4, too. To
avoid unnecessary switchings, the state sw

is taken during the zero output voltage state and

in (3.3-7) is replaced by (1

d

)
or by (d

+ d
0
), both signifying the same.
Table 3.4 Switching state vector patterns with the CSVM for the DMC and the IMC.
MC States State 1 State 2 State 3 State 4 State 5 State 6 State 7 State 8 State 9
Even sum
of sectors
sw

sw

sw

sw

sw
0
sw

sw

sw

sw

DMC
Odd sum of
sectors
sw

sw

sw

sw

sw
0
sw

sw

sw

sw

Input state sw

sw

sw

Even sum
of sectors
sw

sw

sw

sw
0
sw

sw

sw

IMC Output
state
Odd sum of
sectors
sw

sw

sw

sw
0
sw

sw

sw

T
m
/2 T
m
/2
3.3.3 Angle Detection of Input Reference Space Vector
The last issue concerning space vector modulation is the method used to define the angle
swi
of the input current reference space vector i
i,ref
or input switching function space vector sw
i

defined in (3.1-12) and (3.3-2), respectively. As discussed after (3.3-19), input fundamental
displacement angle values
i,1
0 lead to the limited voltage transfer ratio. Thus, it is again
assumed that
i,1
= 0, so that the aim is to generate the input reference angle
swi
synchronised
to supply voltage, i.e.
swi
=
i
t.
The synchronisation method used in [P1][P7] is based on the zero crossing detection of
supply phase voltage u
a
fundamental component u
a,1
and a phase-locked loop (PLL). The
system is discussed briefly in [P3] and its basics are presented more extensively e.g. in
[Kau97]. The block diagram of the angle detection system is presented in Figure 3.9.
As presented in Figure 3.9, the input reference vector angle
swi
is based on the phase angle

ua,1
of the fundamental supply phase voltage u
a,1
. The fundamental component u
a,1
is attained
by low-pass filtering the supply phase voltage u
a
. The difference between the zero crossing
Matrix Converter Control 45
times of the fundamental voltage u
a,1
and its estimated phase angle
ua,1
is controlled by a
discrete-time proportional-integral (PI) controller tending to produce zero error. The output of
the PI controller is the angle increment
ua,1
, which is also the reference for the angular
frequency of the input reference vector. At every sample time, the increment
ua,1
is added to
the previous sample time value of angle
ua,1
. A new value of increment
ua,1
is got after
every zero crossing.
Zero crossing
detection
Zero crossing
detection
u
a
+
PI
+

ua,1
Sample
delay
+

ua,1
+
+

swi
Low pass
filter

PLL
Filter delay
compensation
u
a,1
Performed in every calculation period Performed after zero crossings
Time of zero
crossing
Time of
zero crossing

f
Supply
voltage
measurement

Figure 3.9 Detection of supply phase angle
ua,1
using phase-locked loop (PLL) and forming of input
reference angle
swi
.

The angle
ua,1
contains the delay or phase displacement caused by low-pass filtering of u
a
.
However, that delay or phase displacement can be compensated with a constant value
f
, as
shown in Figure 3.9, because the filter parameters and the fundamental supply frequency are
known. The final input reference angle
swi
is got after the subtraction of /2 to compensate
the difference between the phase angles of the fundamental voltage u
a,1
and the supply voltage
fundamental space vector. After that, it is possible to take into account the input displacement
angle
i,1
later in the control system whenever required.
The PLL-based system presented is basically similar to the synchronisation methods suggested
in [Nie96a], [Cas98b], [Bla02], [Liu03] and [Liu04]. In all of them, the input voltages are
measured and used as the basis of the input reference vector angle instead of the supply
voltage as here. This choice has been made to secure the synchronisation because the supply
voltages are evidently less distorted than the input voltage and it is also the method used with
the current source converters in [Sal02]. In addition [Nie96a], [Cas98b], [Bla02], [Liu03] and
[Liu04] deal with the DMC, where the loss of synchronisation is not as fatal as in the IMC, in
which the loss of synchronisation reverses the polarity of the dc link voltage, which short-
circuits the input capacitors via the ILB diodes.
Compared to methods which measure the supply or input voltages and calculate the voltage
space vector, the PLL-based system is simpler and, as shown in [Liu03] and [Liu04], the MC
system is also stable in all operation conditions, unlike a system in which the input vector
angle follows the input voltage angle directly [Cas02], [Cas04], [Cas05]. Thus, the PLL-based
input current reference angle
swi
always tends to produce sinusoidal input currents [Cas98b],
[Bla02]. However, this relates to the angle
swi
only. The input current produced depends on
the load current, as shown in (3.2-3) and in [P4].
46 Chapter 3

3.4 Conclusion
The basics of MC modulation and the derivation of the SVM method for the MCs have been
presented in this chapter. The space vector theory has also been introduced briefly as
background. In addition, different SVM methods have been compared, concentrating on the
common-mode voltages. At the end of the chapter, the synchronisation of the MC operation to
the supply has been described with a brief comparison with some other methods suggested in
the literature.
The basic operation of the MCs can be modelled and analysed with switching matrices. Due to
the requirement of reasonable and safe operation, the number of possible switching states is
reduced from the theoretical maximum of 512 to 27 in the MCs. These allowed switching
states have conventionally been arranged in three main groups, of which two, including 21
switching states, are used in the SVM modulation. It has been found that several different
modulation methods for the MCs have been suggested. The modulation methods can be
divided basically into the carrier wave based methods, the SVM methods and the methods
combining modulation and the control of load. On the other hand, the modulation methods can
be divided into the methods in which the MC is considered to have separate rectification and
inverter stages, i.e. the IMC approach, and the methods in which the dependences between the
input and the output are defined directly, i.e. the DMC approach.
The basic modulation method used in this thesis is the CSVM, the derivation of which has
been presented applying the ideal IMC [P1][P7]. The order of the CSVM pulse pattern has
been set to minimise the number of switchings with as simple modulator logic as possible. The
CSVM has been used as a basic method and three other SVM methods have been presented so
that the pulse patterns of the CSVM are arranged and divided differently in the modulation
period. The comparison of the common-mode voltages produced by different SVM methods
has shown that the CSVM is not the most optimal SVM method but its switching losses should
be the second smallest and it can be implemented using the simplest logic among the
compared methods.
It has been found that the synchronisation of the input current reference vector of an MC to the
input or supply voltage has been studied using different approaches. Those studies have been
presented that the synchronisation methods based on the PLL are robust and they can also be
implemented easily. Thus, the PLL-based synchronisation has also been used in this thesis
[P1][P7].
47
4 Matrix Converter Modelling
This chapter presents the basics of modelling and simulations for the specific issues presented
in Chapters 67 and in [P1][P7]. First, the circuit modelling and simulation systems are
presented. After that, converter models also containing the modulator are introduced with a
brief simulation model comparison. The induction machine control system and its modelling
are discussed briefly at the end of this chapter.
4.1 Main Circuit Models
This section presents the modelling of the matrix converter main circuits and introduces
briefly simulation environments. In [P2][P7], the matrix converters are simulated with
several different models, depending on the results aimed at. The models have been
implemented in Matlab [P6], Matlab Simulink [P2][P5], [P7] or Simplorer software [P3]
[P4]. The combination of Matlab Simulink and Simplorer has also been used [P3][P4], [P6].
All the Simplorer and Matlab Simulink models and their combination models have a basically
similar structure, which is presented in Figure 4.1. In Figure 4.1, u
sup
, u
i
and u
o
are the supply,
input and output voltage vectors, respectively, u
S1
u
S9
are the control signals of the switches,
u
o,ref
is the reference vector of output voltage and i
sup
, i
i
and i
o
are the supply, input and output
current vectors, respectively.
Modulator
Reference
Supply Supply filter Load
Control
MC bridge
u
sup
i
sup
i
i
u
i
i
o
u
o

swi
u
o,ref
u
S1...S9

Figure 4.1 Basic simulation model structure.
As shown in Figure 4.1, the basic simulation model contains the modulator and the DMC or
IMC main circuits presented in Figure 4.2, except that ideal switching devices are used. In
Figure 4.2, PCC denotes the point of common coupling, Z
sup
is the supply impedance, L
f
and
C
f
are the supply filter inductor and capacitor, respectively, and L
o
and R
o
are the load inductor
and resistor, respectively.
The supply was an ideal voltage source having 400 V line-to-line voltage and 50 Hz
frequency. The supply impedance Z
sup
between the voltage source and the PCC was neglected
in [P2] and [P5][P7]. In [P3][P4], Z
sup
was modelled by series-connected resistance and
inductance having values of 0.03 and 0.1 mH, respectively.
48 Chapter 4

Three-phase RL load was used in [P2][P4] and [P6]. In all cases, the value of the load
inductance L
o
was 10 mH. The value of load resistance R
o
depends on the case. With a 5-kVA
load, a 20- R
o
was used.
(a)
L
f
C
f
PCC
Supply Supply filter
u
c
DMC Bridge
Load
i
a
i
b
i
c
L
o
R
o
i
A
i
B
i
C
u
A
u
B
u
C
u
AB
u
ia
S
Ca1
S
Ca2
S
Cb2
S
Cb1
S
Cc2
S
Cc1
S
Bc1
S
Bb1
S
Ba1
S
Bc2
S
Bb2
S
Ba2
S
Ac1
S
Ac2
S
Ab2
S
Ab1
S
Aa2
S
Aa1
u
BC
u
ic
u
a
i
ia
i
ib
i
ic
Z
sup

(b)
n
p
u
dc
i
dc
L
f
C
f
PCC
Supply filter
IMC Supply
Bridge (ISB)
IMC Load
Bridge (ILB)
Dc Link
Supply
i
ia
i
ib
i
ic
Load
L
o
R
o
u
A
u
C
u
AB
i
A
i
B
i
C
u
BC
S
Ap
S
Bp
S
Cp
S
An
S
Bn
S
Cn
S
na1
S
nb1
S
nc1
S
na2
S
nb2
S
nc2
S
pa2
S
pb2
S
pc2
S
pa1
S
pb1
S
pc1
u
c
i
a
i
b
i
c
u
ia
u
ic
u
a
Z
sup

Figure 4.2 Modelled matrix converter circuits with RL-type load: (a) DMC, (b) IMC.
Supply Filter Model
The supply filter design was based on the procedure presented in [Zar94] for the supply filter
of a current source rectifier aiming to fulfil the requirements for modulation frequency
distortion approximately [P3]. The maximum optimisation of component values was not
considered, but the filter was designed to have a reasonable implementation cost. The
resonance frequency was set at 1049 Hz and the rated output apparent power was set at
5 kVA. The filter inductance L
f
is then 2.3%, i.e. 2.3 mH, and the filter capacitance 10%, i.e.
10 F, in Y-connection. A real inductor component has damping in practice. In the models,
that is taken into account with the equivalent circuit presented in Figure 4.3, where u
sup
and i
sup

describe supply phase voltage and current, as e.g. u
a
and i
a
, and u
i
and i
i
describe input phase
voltage and current, as e.g. u
ia
and i
ia
. The circuit contains resistances R
P
and R
S
, which
describe the natural damping of the filter inductor. The parameters used are presented in Table
4.1, where rms denotes root-mean-square value [P1][P7]. The parameters are based on the
components used in the DMC and IMC prototypes described in Chapter 5. The transfer
function G
f,ui
(s) = U
i
(s)/U
sup
(s) describes the dependence of the input voltage u
i
on the supply
Matrix Converter Modelling 49
voltage u
sup
, whose Laplace transformations are U
i
(s) and U
sup
(s), respectively. As a result, the
voltage transfer function G
f,ui
of the LC filter is
( ) ( )
P f f S P
2
f f S P
P f
ui f,
R s L C R R s C L R R
R s L
G
+ + + +
+
= , (4.1-1)
the Bode diagram of which is presented in Figure 4.4. The Bode diagram in Figure 4.4 shows
that the LC filter amplifies the distortion components between 550 and 1400 Hz harmfully.
For the fundamental voltage component, the LC filter causes no noticeable changes, which
confirms the usability of the synchronisation method presented in Section 3.3.3.
C
f
L
f
i
sup
i
i
u
sup
R
P
R
S
u
i

Figure 4.3 Supply LC filter of a single phase including damping resistances.
Table 4.1 Supply filter and system parameters.
Parameter Value
Rated power 5 kVA
Supply line-to-line rms voltage, U
sup,LL
400 V
Fundamental supply frequency, f
i
50 Hz
Filter inductance, L
f
2.3 mH (2.3%)
Filter capacitance, C
f
10 F (10%)
Approximated serial resistance, R
S
55 m
Approximated parallel resistance, R
P
88
Load inductance, L
o
10 mH

Figure 4.4 Bode diagram of LC filter transfer function G
f,ui
in (4.1-1).
Bridge Models
In [P2][P4] and [P7], the characteristic of individual switches was ideal both when the
switches were implemented in the Simplorer, as in [P3][P4], and when they were
50 Chapter 4

implemented in Matlab Simulink, as in [P2] and [P7]. Exception was [P6], in which real IGBT
and diode models with real switching delays and times and on-state losses were used.
With ideal switching characteristics models [P2][P4], [P7], the switches were controlled
directly by the control signals u
S1
u
S9
, as shown in Figure 4.1. In [P6], the real device models
required safe commutation. This was generated in an additional block placed between blocks
Modulator and MC bridge, which is not shown in Figure 4.1. The safe commutation
method used in the DMC and ISB was the four-step commutation and the blanking times were
used in the ILB, as described in Section 2.2.
In theory, the DMC and the IMC behave identically with ideal switches, i.e. their behaviour is
identical only in ideal conditions, which are not present in practice [P4]. When the DMC and
IMC circuits are compared, it is seen that the semiconductor switches are placed differently
and that the IMC has a real dc link. As shown in Figure 4.5, the dc link contains stray
inductance L
dc
, the effects of which are minimised by connecting a capacitor C
dc
to the ILB
terminals, as presented in Section 2.2. The L
dc
value has been estimated to be 1 H and a 0.1-
F capacitor has been used as C
dc
. The real dc link is included in the Simplorer models of the
IMC main circuit [P3][P4], [P6]. Its undamped transfer function G
dc
from the voltage u
dc
to
the voltage over the capacitor C
dc
is
1
1
2
dc dc
dc
+
=
s C L
G , (4.1-2)
thee Bode diagram of which is presented in Figure 4.6. The diagram shows that the dc link has
no effect on the frequency components of the voltage below 150 kHz. Other effects of the dc
link are analysed and illustrated in Section 7.1.
C
dc
L
dc
Supply
bridge
Load
bridge
i
dc,i
i
dc,o
u
dc

Figure 4.5 IMC with non-ideal dc link.

Figure 4.6 Bode diagram of the IMC dc link transfer function G
dc
by (4.1-2).
Matrix Converter Modelling 51
Simulation Environments
With ideal switch models, the simulation step was set at 1 s in Simplorer and was allowed to
range between 0.1 and 1 s in Matlab Simulink simulations. These values are slightly
inaccurate when the PWM waveform of the output voltage is considered, but the difference
between shorter simulation step and the step sizes used was not found to be significant despite
considerably longer simulation durations. Simplorer version 7.0 was used in Simplorer
simulations. In Matlab simulations, Matlab version 6.5 and Simulink version 5.0 were used.
4.2 Converter Models
This section presents the modelling of all matrix converter system. Thus, this section
complements the block diagram shown in Figure 4.1, which also contains a modulator in
addition to the circuit models. Depending on the converter and simulation program, different
modulator models were applied. In all modulator models, the parameters used in the
simulations are the same as in the prototype implementations, which are presented in
Chapter 5, and they are identical in both the DMC models and the IMC models. All modulator
models considered here are based on the CSVM presented in Section 3.3. In all cases, the
modulation period T
m
is 200 s, i.e. the modulation frequency is 5 kHz. The modulator is
updated twice in T
m
, i.e. the duty cycles are updated in the beginning and in the middle of a
modulation period, as presented in Tables 3.23.4. Thus, the calculation time T
C
(= T
m
/

2) is
100 s. Due to the different main circuits, the structures of the modulators are different in the
DMC and IMC models. This also holds for the modulator implementations in the prototypes,
as described in Chapter 5.
4.2.1 Models of Direct Matrix Converter
The block diagram of the basic DMC simulation model in Simplorer environment is presented
in Figure 4.7 [P3]. The system presented in the block diagram represents the implementation
of the CSVM presented in Section 3.3. The same holds for the basic DMC simulation model in
Matlab Simulink presented in Figure 4.8 [P2], [P4], [P6][P7]. Thus, the basics of the block
diagrams in Figures 4.7 and 4.8 are similar. Modulation index m is determined with (3.3-19).
The angle of output voltage reference u
o,ref
inside the sector
o
is determined after the sector of
the output voltage reference is determined using the absolute angle of u
o,ref
. The angle of input
current reference inside the sector
i
is determined after the sector of the input reference angle

swi
has been determined. The exception is that
swi
is not defined in the simulations using the
PLL-based system, as presented in Section 3.3.3, but rather by using directly the supply
voltage fundamental component, which is easily available in the simulation models. Thus, the
resulting
swi
is identical to what the PLL-based system would yield, even with distorted
supply voltages. The angles
(o,i)
and the modulation index m are used in calculating the duty
cycle, i.e. in determining u
((,)(,),0)
, following (3.3-16). Sector numbers of the output and
input reference vectors, i.e. Sector information, are used to define the pulse pattern according
to the CSVM, which is presented in Tables 3.23.4.
52 Chapter 4

Output voltage reference
and supply voltage
Modulation index,
reference vector sectors
and angles determination
Duty cycle calculation
Pulse pattern
determination
Synthesis of duty cycles
and pulse pattern
Output phase switch
group state determination
Control voltage
generation
u
o,ref
u
sup m
d
((,)(,),0)
sw
o(1,2,3,4,5)
phase A
phase B
phase C
u
SA(a,b,c)(1,2)
u
sup

(o,i)
Sector information
u
SB(a,b,c)(1,2)
u
SC(a,b,c)(1,2)
Switching state
number
Main circuit
(E.g. Figure 4.2a)

Figure 4.7 Block diagram of basic DMC Simplorer model.
Output voltage reference
and supply voltage
Modulation index,
reference vector sectors
and angles determination
Duty cycle calculation
Synthesis of duty cycles
and state pattern coding
Output voltage generation
u
o,ref
u
sup m
d
((,)(,),0)
i
o
Supply filter
u
sup

(o,i)
Sector information
sw
(A,B,C)(a,b,c)
Load
Input current generation
State pattern coding
for each output phase
(Appendix B)
phase A
phase B
phase C
i
i
u
o
u
i
Main circuit

Figure 4.8 Block diagram of basic DMC Matlab Simulink model.
In the Simplorer model (Figure 4.7), the pulse pattern determination produces five switching
state vectors sw
o(1,2,3,4,5)
, which are combined with duty cycles according to the sector
information. That determines the switching state number for each instant. This number defines
the switching state of each output phase group. These phase state signals are the basis for the
generation of gate control voltages u
S(A,B,C)(a,b,c)(1,2)
.
The coding system used in the Simulink model (Figure 4.8) and in the DMC prototype
implementation is presented in Appendix B. The synthesis of the duty cycles and phase state
coding system produces the switching state signals sw
(A,B,C)(a,b,c)
, which are then used as
switches to connect the correct input voltage to each output phase and to connect the correct
output current to each input phase, as in [P2] and [P7]. The main circuit part of the Simulink
model (Figure 4.8) can also be implemented in Simplorer using the component-based main
circuit, as in [P4] and [P6]. In that case, the switching state signals sw
(A,B,C)(a,b,c)
in Figure 4.8
are to be transformed to gate control voltages u
S(A,B,C)(a,b,c)(1,2)
, as in Figure 4.7.
Matrix Converter Modelling 53
4.2.2 Models of Indirect Matrix Converter
The block diagram of the basic IMC simulation model in Simplorer environment is presented
in Figure 4.9 [P3]. The block diagram of the basic IMC simulation model in Matlab Simulink
environment is presented in Figure 4.10 [P4], [P6][P7]. Both systems presented represent the
implementation of the CSVM presented in Section 3.3. The similarity between the DMC and
IMC block diagrams in Figures 4.74.10 is obvious. The only differences compared to what
presented in Section 4.2.1 for the DMC are those related to how the gate control signals are
generated. The control signals of the ISB and the ILB are separated in the synthesis of pulse
pattern and duty cycles. Thus, the two-stage structure decreases the number of intermediate
blocks both in the Simplorer and Simulink models compared with the DMC models.
u
o,ref
u
sup
m
phase A,B,C
u
Sp(a,b,c)(1,2)
u
sup

(o,i)
u
Sn(a,b,c)(1,2)
u
S(A,B,C)(p,n)
bar p,n
Output voltage reference
and supply voltage
Modulation index,
reference vector sectors
and angles determination
Duty cycle calculation
Synthesis of duty cycles
and pulse pattern
d
((,)(,),0)
Pulse pattern
determination
sw
o(1,2,3,4,5)
Sector information
Control voltage
generation
Main circuit
(E.g., Figure 4.2b)

Figure 4.9 Block diagram of basic IMC Simplorer model.
u
o,ref
u
sup
m
u
sup

(o,i)
Output voltage reference
and supply voltage
Modulation index,
reference vector sectors
and angles determination
Duty cycle calculation
ISB switching state
determination
d
((,)(,),0)
ILB switching state
determination
u
o
Supply filter
sw
(p,n)(a,b,c)
Load
Output voltage
generation
i
dc
u
i
Main circuit
Voltage sector
sw
(A,B,C)(p,n)
Link current
generation
i
i
i
o
Link voltage
generation
Input current
generation
u
dc
Current sector

Figure 4.10 Block diagram of basic IMC Matlab Simulink model.
As with the DMC, the main circuit of the IMC Simulink model can also be in the Simplorer
model using the component-based main circuit, as in [P4] and [P6]. In that case, the switching
state signals sw
(A,B,C)(p,n)
of the ILB and sw
(p,n)(a,b,c)
of the ISB in Figure 4.10 must be converted
to gate control voltages u
S(A,B,C)(p,n)
and u
S(p,n)(a,b,c)(1,2)
, respectively, as in Figure 4.9.
54 Chapter 4

The basic Simplorer model of the IMC contains the dc link components presented in Figure
4.5, but they are not included in the IMC Matlab Simulink model.
The feedback control in simulations is applied in [P3] with IMC and in [P7] with the ideal MC
model. In both cases, it has been implemented in Matlab Simulink.
4.2.3 Comparison of Simulation Models
Different DMC models were simulated with a three-phase Y-connected RL load consisting of
series-connected 20- resistance and 10-mH inductance. Simulated spectra produced by
different models are presented in Figures 4.114.12, where input-to-output voltage transfer
ratio reference
ref
is 0.8 and output frequency f
o
is 40 Hz. The spectra are plotted up to 2 kHz
with the frequency resolution (f
res
) of 10 Hz and the total harmonic distortion (THD) is
calculated up to 2 kHz, which is the maximum harmonic frequency taken into account in
[IEC00]. In this thesis, the THD for a quantity x is defined as shown in [Moh95]:
1
2
1
2
THD
X
X X
= , (4.2-1)
where X is the rms value of x and X
1
is the rms value of the fundamental component. In the
spectra plots, capital letter A followed by a number denotes the amplitude of the frequency
component multiple for frequency resolution f
res
, e.g. when f
res
= 10 Hz, A4 denotes the
amplitude of 40-Hz component and A5 denotes the amplitude of 50-Hz component.
The Simplorer results with ideal switches are presented in Figure 4.11. The results with the
same ideal Simplorer main circuit but with the modulator in Simulink are presented in Figure
4.12. The difference between the spectra in Figures 4.11 and 4.12 is considerable. It is found
that the Simplorer modulators produce more distortion than the Simulink modulators.
Simplorer models produces e.g. second harmonics to the quantities and an output voltage
fundamental component lower than its reference. With reasonable modulation steps, the
situation was not found to improve significantly. The exact reason for this is unknown, but it
relates to the different numerical solver algorithms because Simplorer results became more
ideal when simulation step was reduced significantly leading to impractical long simulation
times. This should be noted when the results of Simplorer-modulator-based models are
considered below and in [P3]. The same inaccuracy also exists in the IMC models, the results
of which are presented in Figures 4.134.14, attained without dc link model. In other words, in
the ideal case the results should be identical with the respective DMC results.
With ideal switches, the different modulator models should produce identical results because
the IMC dc link in Figure 4.5 is not included in the IMC model. That similarity holds for
Simplorer and Simulink models, as can be seen by comparing the results in Figures 4.11 and
4.13 with each other and the results in Figures 4.12 and 4.14 with each other. Thus, it is
possible to conclude that the simulations with the same simulation programs are comparable
but the results obtained by different programs are not. In addition, the Simplorer-modulator-
Matrix Converter Modelling 55
based simulations give more realistic total THD values even though their frequency
distributions differ from real distributions.

Figure 4.11 Simulated spectra of DMC model in Simplorer with the ideal switches,
ref
= 0.8, f
o
= 40 Hz:
(a) output line-to-line voltage u
AB
, (b) output current i
A
, (c) supply current i
a
.

Figure 4.12 Simulated spectra of DMC model where modulator is in Simulink and main circuit in
Simplorer with the ideal switches,
ref
= 0.8, f
o
= 40 Hz: (a) output line-to-line voltage u
AB
, (b)
output current i
A
, (c) supply current i
a
.

Figure 4.13 Simulated spectra of IMC model in Simplorer with the ideal switches and without dc link
components,
ref
= 0.8, f
o
= 40 Hz: (a) output line-to-line voltage u
AB
, (b) output current i
A
, (c)
supply current i
a
.

Figure 4.14 Simulated spectra of IMC model where modulator is in Simulink and main circuit in Simplorer
with the ideal switches and without dc link components,
o,ref
= 0.8, f
o
= 40 Hz: (a) output line-
to-line voltage u
AB
, (b) output current i
A
, (c) supply current i
a
.
4.3 Induction Machine Modelling and Control System
This section describes briefly the modelling of an induction machine (IM) and presents the
control systems of the IM drive applied in [P1][P2] and [P7]. In this thesis, a cage induction
machine is the only type of induction machine considered and thus the abbreviation IM
denotes purely the cage induction machine below.
4.3.1 Cage Induction Machine Model
The next presentation is based on [Kov62], [Vas90], [Vas92] and [Nov96] and is applied in
[P7]. The stator and rotor voltage space vectors of a cage induction machine are u
s
and u
r
,
(a) (b) (c)
(a) (b) (c)
(a) (b) (c)
(a) (b) (c)
56 Chapter 4

respectively. The machine voltages in the general reference frame, rotating at angular speed

k
and denoted by subscript k, are
k s,
k
k s,
k s, s k s,
j
d

+ + =
t
d
i R u , (4.3-1a)
( )
k r,
r k
k r,
k r, r k r,
j
d
0

+ + = =
t
d
i R u , (4.3-1b)
where R
s
and R
r
are stator and rotor resistances, respectively,
s
and
r
are the stator and rotor
flux linkages, respectively, i
s
and i
r
are the stator and rotor currents, respectively, and
r
is the
electric rotor angular speed.
Flux linkages
(s,r)
depend on the stator and rotor self inductances L
s
and L
r
, respectively, and
on the magnetising inductance L
m
:
( )
k r, m k s, s m k r, m k s, s
k s,
i L i L L i L i L + + = + = , (4.3-2a)
( )
k s, m k r, r m k s, m k r, r
k r,
i L i L L i L i L + + = + = , (4.3-2a)
where L
s
and L
r
are the leakage inductances of the stator and the rotor, respectively.
The equivalent circuit of an IM, based on (4.3-1)(4.3-2), is presented in Figure 4.15.
u
s,k
i
s,k
R
s
L
s
j
k

s,k
L
m
L
r
j(
k
-
r
)
r,k
d
s,k
/dt d
r,k
/dt
R
r
u
r,k
= 0
i
r,k

Figure 4.15 Space vector equivalent circuit of a cage induction machine (4.3-1)(4.3-2).
The IM produces electromagnetic torque T
e
, which can be expressed with space vectors and
two-axis presentation, introduced in Section 3.1 [Kov62], [Vas90]:
( )
k s, k s, k s, k s, p k s,
k s,
p e
2
3
2
3
i i p i p T = = , (4.3-3)
where p
p
denotes the number of pairs of poles. In the ideal case, the electromagnetic torque T
e

is equal to the mechanical torques of the rotating system:
mech r,
mech r,
load e
d
d

B
t
J T T + + = , (4.3-4)
where T
load
is the mechanical load torque, J is the moment of inertia,
r,mech
is the mechanical
rotor angular speed and B is the friction constant.
The IM model following (4.3-1)(4.3-4) is implemented in Matlab Simulink. The space
vectors of the model are employed in a stationary reference frame using two-axis presentation.
Matrix Converter Modelling 57
The parameters used are presented in Table 4.2 and they are based on the IM drive used in the
experimental tests and described in Chapter 5.
Table 4.2 Induction motor and IM drive parameters used in simulations.
Parameter Value Parameter Value
Rated shaft power 2.2 kW Magnetisation inductance, L
m
155 mH
Rated rotor rotation speed 940 rpm Stator leakage inductance, L
s
7.2 mH
Rated stator frequency 50 Hz Rotor leakage inductance, L
r
7.2 mH
Number of pairs of poles, p
p
3 Stator resistance, R
s
2.3
Rated stator line-to-line rms voltage 380 V Rotor resistance, R
r
2.16
Rated stator rms current 6.2 A Moment of inertia, J 0.116 kgm
2
Rated shaft torque 22 Nm
Drive
parameters Friction constant, B 0.011 Nm/s
4.3.2 Cage Induction Machine Control System
Control System
The IM control is implemented by rotor-flux-oriented control, which can be considered a
standard IM control method [Vas90], [Kaz02]. It is applied in simulations in [P7] and in the
experiments in [P1][P2] and [P7]. The block diagram of the control system with converter
system and IM is presented in Figure 4.16, where the two-axis components of the rotor-flux-
oriented reference frame are denoted by subscripts x and y and: stator current i
s,mr
= i
sx
+ ji
sy
,
their reference i
s,mr,ref
= i
sx,ref
+ ji
sy,ref
and stator voltage reference u
s,mr,ref
= u
sx,ref
+ ju
sy,ref
.

r,mech,ref
denotes the reference of the mechanical rotor angular speed
r,mech
. Due to the
reference frame chosen, Im{
r
} = 0, i.e.
r
= Re{
r
} and the rotor flux linkage can be
considered to be simply a real value |
r
|. The outer control loops are speed and flux controllers
and the inner loops are the controllers of current components i
sx
and i
sy
.
MC
u
A
u
B
u
C
i
sy,ref
|
r,ref
|
Modulator

r,mech
i
A
, i
B
A,B,C
> x,y
i
sy
i
sx
|
r
|
+
-
i
sx,ref
+
-

r,mech,ref
+
-
+
-
u
sy,ref
u
sx,ref
i
sx
i
sy

mr
Flux model

ua
LC-
filter

u
a
u
b
u
c
Speed
controller
Flux
controller
Current
controller

mr
PLL
IM

Figure 4.16 Block diagram of rotor-flux-oriented control system of MC-supplied IM drive.
The PI (proportional-integral) controllers have been chosen as IM controllers because they can
be considered to be the basic control strategy in electric drives. In all cases, the feedback
control loop is based on discrete-time anti-windup PI controllers, which are transformed to
discrete form using backward Euler approximation [st97]:
58 Chapter 4

) 1 ( ) 1 ( ) ( ) ( ) (
I
s
p
+

+ = k y k e k e
T
T
k e K k y , (4.3-5)
where y is the controller output variable, e is the error variable, K
P
is the proportional action
coefficient, T
I
is the integral action time, T
s
is the sampling time and k is the present sample.
The block diagram of the PI controller is presented in Figure 4.17 where 1/z describes the
delay of single T
s
. The PI controller described by (4.3-5) is also used in the PLL system for the
synchronisation presented in Section 3.3.3.
Zero order
hold (T
s
)
K
P
1/z
1/z
+
+
+
T
s
/T
I
e(t)
y(k)
e(k)
y(k-1)

Figure 4.17 The block diagram of a discrete-time anti-windup PI-type controller.
The block Flux model in the block diagram in Figure 4.16 estimates the angle
mr
of the
rotor-flux-oriented reference frame and the magnitude |
r
| of rotor flux linkage [Vas90] by
applying the magnitude |i
mr
| of magnetisation current:
sx
r
mr
1
1
| | i
s T
i
+
= , (4.3-6) | | | |
mr m
r
i L = , (4.3-7)
| |
mr r
sy
slip
i T
i
= , (4.3-8) ( )
r slip mr
1
+ =
s
, (4.3-9)
where the rotor time constant T
r
= L
r
/R
r
and
slip
denotes electric angular slip speed. Electric
angular rotor speed
r
in (4.3-9) can be calculated from the measured mechanical rotor angular
speed
r,mech
when the number of pairs of poles p
p
is known:
r
= p
p

r,mech
. Equations (4.3-6)
and (4.3-9) are used in the simulations and experiments in discrete form given by backward
Euler approximation [st97]:
) 1 ( | | ) ( ) ( | |
mr
FM r
r
sx
FM r
FM
mr

+
+
+
= k i
T T
T
k i
T T
T
k i , (4.3-10)
) 1 ( ) ( ) ( ) (
mr r C slip C mr
+ + = k k T k T k , (4.3-11)
where T
FM
and T
C
are the sample times.
4.3.3 Induction Motor Drive Simulations
Simulation Model
The system presented in Figure 4.16 has been simulated with the ideal Matlab Simulink IM
model. As shown in Section 4.2.3, there is no sense in using different DMC and IMC
simulations in the case of ideal circuits. Thus, simulations presented here are performed using
Matrix Converter Modelling 59
only the IMC Simulink model due to its slightly faster computation compared to the DMC
Simulink model with IM load, and the general abbreviation MC is used instead of the DMC or
the IMC in this section.
All controllers of the system are discrete-time PI-type controllers, as presented above (4.3-5).
As presented in Section 4.2, the modulation period T
m
is 200 s and the modulator is updated
twice in it so that the calculation time T
C
is 100 s. Current control is executed before every
modulator updating so that its sampling time is T
C
, i.e. 100 s. Speed and flux controls are
performed alternately every 400 s, i.e. both are executed every 800 s. That part of the block
Flux model which contains (4.3-7)(4.3-8) and (4.3-10) is updated every 1.6 ms, which is
the value of sample time T
FM
. The rest of the block Flux model, i.e. (4.3-11), is calculated
before current control at every T
C
. Controller parameters are presented in Table 4.3 and the
current controllers of both the x and y components are equal.
The model parameters are presented in Table 4.2. The rotor flux linkage reference |
r,ref
| is set
at 0.9 Wb, which is the flux linkage value calculated assuming rated conditions, i.e. it is the
rated flux linkage without field weakening. Thus, the rated rotation speed of 940 rpm cannot
be achieved because the maximum stator line-to-line voltage in an MC drive is 346 V with
400 V supply voltage. The starting speed of the field-weakening operation has been set at
800 rpm, which is achieved when the input-to-output voltage transfer ratio 0.85, i.e.
0.85

940

rpm

800

rpm.
Table 4.3 Induction motor controller parameters.
Controller Parameter Value 3 dB cut-off frequency
Proportional action coefficient, K
P
10
Integral action time, T
I
2.85 ms
Sample time, T
C
100 s
Current controller
Output limit 280 V
120 Hz
Proportional action coefficient, K
P
0.25
Integral action time, T
I
65 ms
Sample time 800 s
Speed controller
Output limit 13.15 A
2 Hz
Proportional action coefficient, K
P
11
Integral action time, T
I
80 ms
Sample time 800 s
Flux controller
Output limit 13.15 A
10 Hz
Simulation Example
The CSVM-modulated IM model with the control system presented above has been
implemented in Matlab Simulink. Steady state simulation results are presented in Figure 4.18,
where the load torque T
load
is rated 22 Nm and the mechanical rotation speed n
r
is 800 rpm,
resulting in the output frequency f
o
42 Hz. The frequency resolutions f
res
of spectra in Figures
4.18d4.18f are the fundamental frequencies of each quantity, so that their fundamental
component magnitudes are denoted by A1. The results in Figure 4.18 are the same as
presented in [P7].
60 Chapter 4

The MC drive dynamics is shown by Figures 4.194.21, where the operations with nominal
load torque of 22 Nm are presented. The simulated cases in Figures 4.194.21 are speed
reference step, speed reference ramp and acceleration using the field weakening, respectively.
The waveforms in Figures 4.19a, 4.20a and 4.21a are the same as presented in [P7].


Figure 4.18 Steady state simulation results of IM drive when T
load
= 22 Nm and n
r
= 800 rpm: (a) supply
voltage u
a
and current i
a
, (b) output/stator line-to-line voltage u
AB
, (c) output/stator current i
A
,
(d) spectrum of i
a
, (e) spectrum of u
AB
, (f) spectrum of i
A
.

Figure 4.19 Speed reference step simulation when T
load
= 22 Nm: (a) rotation speed n
r
and its reference
n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator current component
i
sy
and its reference i
sy,ref
.

Figure 4.20 Speed reference ramp simulation when T
load
= 22 Nm: (a) rotation speed n
r
and its reference
n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator current component
i
sy
and its reference i
sy,ref
.

Figure 4.21 Field-weakening acceleration simulation when T
load
= 22 Nm: (a) rotation speed n
r
and its
reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator current
components i
sx
and i
sy
and reference i
sy,ref
.
As can be seen, the MC does not restrict the IM drive dynamic operation and the MC-fed IM
drive with rotor-flux-oriented control produces the waveforms assumed. In the field
weakening, the flux linkage reference |
r,ref
| is decreased from the rated 0.9 Wb inversely
proportionally to the speed reference n
r,ref
via the stator current x-direction component i
x
in
Figure 4.21. The acceleration from 800 to 1200 rpm applying the field weakening shows that
the rated operation speed of 940 rpm and higher can be achieved with the rated torque.
However, the rated conditions of machines designed for a direct grid connection cannot be
(a) (b) (c)
(a) (b) (c)
(d) (e) (f)
(a) (b) (c)
(a) (b) (c)
Matrix Converter Modelling 61
achieved without field weakening. Thus, the stator currents increase, leading to increased
machine losses compared with the VSI because stator current increase means increased copper
loss. However, that does not necessarily lead to increased drive losses in total and the results
depend e.g. on the modulation frequency [Ber02], [Bla03]. A more extensive comparison of
MCs and conventional converters is beyond the scope of this thesis.
4.4 Conclusion
The basics of modelling and simulation models of the DMC and the IMC have been presented
in this chapter. In addition, the modelling and the control system of the MC supplied IM drive
have been presented.
The modelling of the DMC and IMC main circuits has included the introduction of the system
parameters and the design and the modelling of the LC-type supply filter, which is identical in
both converters. The LC filter model contains resistances, which models the natural damping
of inductors in practice. The modelling of the IMC dc link has also been presented so that the
stray inductances of the link bars are taken into account together with the capacitor used to
minimise the effects of the stray inductance. As a result, it has been found that the IMC dc link
has no effects on the frequency components of the voltage below 150 kHz.
The structures of both the Simplorer and the Simulink modulator models of both the DMC and
the IMC have been presented. In all cases, the CSVM modulator presented in Chapter 3 has
been applied but the structures of the models are slightly different due to the different main
circuits and simulation programs. The Simulink models of the DMC and IMC modulators
produces identical results in ideal case and the same holds also for the Simplorer models.
However, the modulator models in different programs produce different results: the modulator
models in Simplorer produce more distorted output voltages than the modulator models in
Simulink. Thus, it has been concluded that the analyses based on the comparison of the results
produced by different simulation programs are not reliable and they are never used to compare
the DMC and the IMC in this thesis [P1][P7].
The modelling of the IM has been based on the space vector equivalent circuit. The space
vector model of an IM has also been the basis for the rotor-flux-oriented space vector control
system used to control the MC-supplied IM drive. The parameters of the IM drive have also
been introduced and an ideal MC circuit has been used in the simulations. The simulation
results have shown that the MC does not restrict the IM drive dynamic operation and the MC-
fed IM drive produces sinusoidal output and input current waveforms as assumed. In addition,
the simulations have shown that the rated rotation speed and load torque of the IM can be
achieved with the MC when field weakening operation has been used. The use of field
weakening increases the stator currents which increases the losses in the IM compared with
the conventional VSI supplied IM drive. However, that does not necessarily increase the total
losses of the drive when the converter losses are also taken into account.
62
5 Experimental Setup
This chapter presents the basics of prototype implementations and experimental test setups
applied in [P1][P7]. The prototype of the IMC was built in 2002 [RP2] and the prototype of
the DMC in 2004 [RP8], [Jus05]. The IMC prototype has been used in [P1][P7] and the
DMC prototype in [P2] and [P4][P7]. First, their hardware and software implementations are
described briefly. After that, measurement equipment and practices are introduced. Finally, the
prototypes are confirmed with some experimental results. More extensive presentation of the
prototype implementations can be found in [Jus05].
5.1 Prototype Implementations
The block diagrams of the MC prototypes are presented in Figure 5.1. The main circuit
structure of the IMC prototype is the same as in Figure 2.16. The ISB consists of Semikron
SKM75GAL123D IGBT modules containing an IGBT with an antiparallel connected diode
(rated current 75 A and voltage 1200 V). The ILB consists of a Semikron SKM40GD123D
IGBT module containing a VSI bridge (rated current 40 A and voltage 1200 V), i.e. a six-pack
module. The main circuit of the DMC prototype is basically as in Figure 2.14 with the
exception that all switches are not common-emitter-connected but some of them are common-
collector-connected, as presented in Figure 5.2. The reason for this DMC structure is the
minimised number of IGBT modules when conventional inverter leg IGBT modules
SKM75GB123D by Semikron are applied. These modules are circled in Figure 5.2. The
switches not consisting of them are formed with Semikron SKM75GAL123D modules, of
which the IGBTs and diodes are identical to those in SKM75GB123D modules [Sem97].
Photographs of the prototypes are presented in Figure 5.3.
The parameter values of the supply filter and the overvoltage clamp are identical in the DMC
and the IMC prototypes, and the filter parameters are the same as those presented in Section
4.1. The supply filter inductor value is 2.3 mH (rated rms current 10 A, manufactured by
Trafomic) and the supply filter capacitor value in Y-connection is 10 F (rated ac rms voltage
250 V, manufactured by Arcotronics). The DMC has a single overvoltage clamp circuit with
twelve diodes, as in Figures 2.14 and 5.2. The IMC has two overvoltage clamps, both having
ten diodes as shown in Figure 2.16. The diodes in the clamp circuits are type RHRG50120 by
Fairchild Semiconductor (rated current 50 A and voltage 1200 V). The clamp capacitor value
is 7.3 F with 1200 V rated voltage (three 22-F and 400-V capacitors in series). The varistors
used are type V480LA80B by Littelfuse with 750 V varistor voltage. As presented in Sections
2.1 and 4.1, the main difference between the DMC and the IMC is the dc link, the inductance
of which was approximated [Kre98] and measured to be around 1 H. It was attempted to
Experimental Setup 63
minimise the effects of that inductance by mounting a 0.1-F capacitor in the ILB terminals
(rated dc voltage 750 V, manufactured by Arcotronics).
(a)
Microcontroller board
Three driver boards
Overvoltage clamp
Three logic boards
Supply
i
a
, i
b
u
a,
u
b
, u
c
i
A
, i
B
, i
C
Supply filter
DMC
i
A
, i
B
Load
Clamp voltage
n
r
Commands
Data reading
Personal computer
u
a
18
11
6 6 6

(b)
Microcontroller board
Two driver boards
Overvoltage clamp
Logic board
Supply
i
a
, i
b
u
a
, u
b
, u
c
i
dc
Supply filter
IMC
i
A
, i
B
Load
Clamp voltage
n
r
Commands
Data reading
Personal computer
u
a
ILB ISB
Link
12
12
Driver board
6
6 6

Figure 5.1 Block diagrams of the prototypes: (a) DMC, (b) IMC.
A
B C
a b c

Figure 5.2 The main circuit of the DMC prototype where six inverter leg IGBT modules and six single
IGBT modules are applied.
In the DMC, all load currents are measured to define their directions for the four-step
commutation, as presented in Figure 5.1a. Both in the DMC and the IMC, two supply and two
64 Chapter 5

load currents are measured for overcurrent protection. The load currents are also used for the
current control in [P1][P3], [P5] and [P7]. In addition, the dc link current is measured for
overcurrent protection in the IMC and the direction of the dc link current is also the basis for
the four-step commutations of the ISB. All current measurements in the prototypes are
performed with LA55-P current sensors by LEM-Heme. Both prototypes contain the
differential measurements of clamp capacitor voltages for overvoltage protection. The supply
voltage u
a
is measured by applying resistance scaling for the synchronisation with the PLL,
described in Section 3.3.3. Resistance scaling is also used for the supply voltage magnitude
measurement in [P3]. Measurement of the rotor speed n
r
by tachometer is used in [P1][P2]
and [P7] with the IM and with the PMSM in [P5], in which the rotor position angle is also
measured.
(a)
(b)
Figure 5.3 Prototypes (a) DMC with IM drive and (b) IMC.
The microcontroller boards of both prototypes are identical. In addition to the Motorola
MPC555 microcontroller, the board contains auxiliary circuits as overcurrent and overvoltage
sensing and protection, logic circuits and measurement signal processing circuits. The board
has been designed in the Institute of Power Electronics at Tampere University of Technology
Experimental Setup 65
for the control of three-phase converters in general. The same holds for the driver boards, each
of them containing seven voltage supplies for the electronics (15, 12, 8, 12, 15 and 24 V and
regulated 8 V) and six isolated gate control voltages (levels 15 V). Thus, three driver boards
are required in both prototypes. 22- gate resistors are used with all IGBTs [Sem97].
In addition to the identical microcontroller boards, the software of the microcontrollers is the
same in both prototypes. The software is based on the interruption service program performed
after every 100 s. The only task of the main program is to communicate with the personal
computer via serial bus RS-232. The main program reads the commands given by the user via
Matlab and writes the values of chosen variables into the bus when required by the user. The
flow chart of the interruption service program is presented in Figure 5.4.
Check faults
(rise Fault)
Set up flag
Converter fault
Update phase angle
ua
of the supply voltage u
a

Set up flag
Angle control
Check flag
Converter fault
Prevent converter
operation
Define the angle of output
voltage reference frame
Update modulator
Check time level 12 ms
and flag Angle control
Control
ua
increment
[Fault]
[not Fault]
[angle > 360
0
]
[time > 12 ms
and flag up]
[no control]
[flag up] [flag down]
Read command line
(if required)

Figure 5.4 Flow chart of interruption service program used in the prototypes.
The interruption service program presented in Figure 5.4 1) monitors the converter state and
performs a protection operation, 2) defines the angle of supply voltage u
a
, i.e. synchronises the
66 Chapter 5

system with the PLL, presented in Section 3.3.3, and 3) calculates the variable values of the
modulator executing the CSVM, presented in Section 3.3.2. In more complex applications, the
software also performs feedback control, etc. The data reading possibility of the
microcontroller is used only in applications which are more complex than the basic modulator.
5.1.1 Modulator Implementations
Next, the modulator implementation including the modulator logic in addition to the software
is presented. The flow chart of the software performing the calculations required in the CSVM
implementation is presented in Figure 5.5 and it is the same in both prototypes. The system is
based on the CSVM presented in Section 3.3.2 and the structure is quite similar to the
simulation models presented in Section 4.2. In the flow chart, TPUA and TPUB are the time
processor unit (TPU) A and B of the microcontroller. TPU outputs change their states as ruled
by the values of their registers. TPU output voltage levels are 0 and 5 V.
The block diagrams of the modulator implementations are presented in Figure 5.6. The system
is designed so that eight TPUA outputs are the inputs of GAL1 (GAL, gate array logic), the
logic of which creates five on-time signals of the five states (Tables 3.23.4, Appendix B) in
each half of the 200-s modulation period T
m
and a signal summing the first and the second
time. The latter is used only in the IMC modulator (Table 3.4), as shown in Figure 5.6. TPUB
outputs define the sectors of the reference vectors (Figures 3.43.5) and are the inputs of
GAL2, which forwards them to the logic board(s) if neither overcurrent nor overvoltage is
detected in the block Fault detection. The voltage levels of the GAL circuit inputs and
outputs are 0 and 5 V.
The differences between the prototype modulators occur in logic boards. The DMC has three
logic boards, each of them generating the control signals of a single output phase. The board
circuits are identical and the different logics in GAL3 are the only differences. Circuit GAL3
generates the sequence codes, which are presented in Appendix B. The same number code
denotes the same state with every output phase so that the logics in GAL4s are identical. In the
IMC, only a single logic board is used and it contains both the ISB and ILB control signal
generations. The even parity of the sectors is defined in GAL4 with the help of the input
reference sector and the last bit of the output reference sector, which denotes its own even
parity. As a result, the DMC modulator generates nine and the IMC modulator generates
twelve switch control signals, so that both modulators generate theoretically equal output
voltages.
In the DMC and in the ISB, the four-step commutation is used to generate the individual IGBT
gate control signals from the switch control signals. The four-step commutation is
implemented by generating three differently delayed signals from the original signals using
RC circuits and Schmitt triggers. The correct outputs are then chosen by means of a
multiplexer controlled by the current direction signal and the non-delayed control signal of the
switch. As a result, fixed delays of 400 ns between each step are generated (Figure 2.11). The
blanking times of 1

s for the ILB IGBTs are generated using RC circuits and Schmitt triggers.
Experimental Setup 67
Calculate the angle
uoref,k
and magnitude |u
o,ref,k
| of u
o,ref,k
in the reference frame used
Calculate the angle
swi
of input reference vector i
i,ref

Calculate the angle of output voltage reference vector u
o,ref
in the stationary reference frame
Calculate modulation index m
Prevent converter operation
[flag Converter
fault up]
Calculate the sum of the sector numbers
Calculate duty cycles
Update timer registers (TPUA)
Update voltage sector register (TPUB)
Define the sector number of i
i,ref
and its angle
i
inside the sector
[odd sector sum] [even sector sum]
Update current sector register (TPUB)
Define the sector number of u
o,ref
and its angle
o
inside the sector
Calculate duty cycles
Update timer registers (TPUA)
[flag Converter
fault down]
[flag Converter
fault down]
[flag Converter
fault up]

Figure 5.5 Flow chart of modulator update program.
(a)
Microcontroller board
Logic board for phase A Microcontroller
TPUA
TPUB
8
GAL1
GAL2
GAL4
State
times
i
i,ref
u
o,ref
Generation of
four-step
commutation
u
SC(a,b,c)(1,2)
u
SA(a,b,c)(1,2)
Driver 3
Driver 2
Driver 1
u
SB(a,b,c)(1,2)
GAL3
Sector
Duty cycles
Sequence
M
o
d
u
l
a
t
o
r
u
p
d
a
t
e
p
r
o
g
r
a
m
Fault detection
codes
Sector
codes
6
6
6
6
6
6
6
Logic board for phase B
Logic board for phase C
Output current direction
code

(b)
Logic board
TPUA
TPUB
GAL1
GAL2
GAL3
GAL4
Generation of blanking
time commutation
Generation of four-
step commutation
u
S(A,B,C)(p,n)
u
Sn(a,b,c)(1,2)
Driver 1
Driver 2
Driver 3
u
Sp(a,b,c)(1,2)
sector sum even parity
Microcontroller board
Microcontroller
8
M
o
d
u
l
a
t
o
r
u
p
d
a
t
e
p
r
o
g
r
a
m
Duty cycles
State
times
Fault detection
Sector
codes
Sector
codes
i
i,ref
u
o,ref
6
6
6
6
6
6
6
Link current direction

Figure 5.6 Modulator implementations in (a) the DMC prototype, (b) the IMC prototype.
68 Chapter 5

5.1.2 Induction Motor Drive
The flow chart of the IM control system implementation is presented in Figure 5.7. As can be
seen, it is basically the same as the chart in Figure 5.4 but has additional blocks for IM control.
The control system implements the rotor-flux-oriented control of IM drives, presented in
Section 4.3. The controllers are discrete PI controllers, presented in (4.3-5) and in Figure 4.17.
Update modulator
Update
mr
and
execute current control
Check time level 1.6 ms
(rise FE)
Check time level 400 s
(rise F or S by turns)
Read & send data
(if required)
[F] [S]
Execute speed control Execute flux control
[FE]
Estimate |
r
| and
slip

Read command line
(if required)
Start A/D converter
[Neither F nor S]
[Not FE]
Check faults
(rise Fault)
Set up flag
Converter fault
Update phase angle
ua
of the supply voltage u
a

Set up flag
Angle control
Check flag
Converter fault
Prevent converter
operation
[Fault]
[not Fault]
[angle > 360
0
]
[flag up] [flag down]
Check time level 12 ms
and flag Angle control
Control
ua
increment
[time > 12 ms
and flag up]
[no control]

Figure 5.7 Flow chart of the IM control system software implementation.
Experimental Setup 69
The IM drive is shown in Figure 5.3a with the DMC prototype. The IM is HXA 112 6 B3,
manufactured by ABB. That is a 2.2-kW IM the parameters of which were presented in Table
4.2. The load machine of the test bench is a 6-kW separately excited dc machine
GNA 3210 33, manufactured by Strmberg and driven by a four-quadrant dc drive Simoreg
D400/30, manufactured by Siemens. The dynamo tachometer used is a Radio-Energie REO-
444 producing 0.06 V/rpm. The tachometer signal is scaled to range between 0 and 5 V with a
separate board for the A/D (analogue to digital) converters of the microcontrollers.
5.2 Measurement Equipment
In all experiments, a four-channel digital oscilloscope LeCroy LT554M has been used. The
currents have been measured with Tektronix TCPA300 probes with 1

A 0.1

V measurement
range. Voltage measurements from phase to ground have been performed with LeCroy
PPE 2kV probes having the attenuation of 1:100. Output line-to-line voltage measurements
have been performed with Tektronix P5200 differential voltage probes with the attenuation of
1:500. In most cases, Finnish 400-V supply was used, but in [P3][P5] ideal sinusoidal or
distorted supply voltage is generated by voltage source Elgar SW5250A.
In more complex applications than those in the basic modulator, measurement data has also
been gathered by reading microcontroller variable values via the serial bus, as presented e.g. in
Figures 5.1 and 5.7. The total length of the data vector is always 1000 samples, but its duration
depends on the case, i.e. on how long a time is needed. The maximum sampling frequency
with the microcontroller is 10 kHz. With the IM drive, the sampling frequency ranges from
500 to 1000 Hz, depending on the sequence duration. In the oscilloscope measurements, the
sampling frequency has been 1 MHz in [P2][P7] and 50 kHz in [P1].
5.3 Experimental Results
This section presents the experimental results from the operation of both prototypes both with
passive RL-type load and with IM drive. RL load results have been presented originally in
[Jus05] and IM drive results in [P7].
5.3.1 Passive Load
Experiments, the results of which are presented in this section, were performed with an RL
load of 20 and 10 mH and a 400-V network served as supply voltage [Jus05]. The
measurement accuracy of the supply voltage was 1.6 V, the accuracy of the current
measurement was 0.16 A and the sampling frequency was 50 kHz. Measurement results with
input-to-output voltage transfer ratio reference
ref
= 0.8 and output frequency f
o
= 40 Hz are
presented in Figures 5.85.9. Waveforms show that both the DMC and the IMC produce
sinusoidal supply currents and PWM output voltages, which do not contain significant
frequency distortion. However, the supply current distortion is slightly higher in the IMC,
which is considered further in Section 7.1. The supply voltage THD up to 2 kHz was about 1%
70 Chapter 5

in all cases. The results with fixed output frequency (f
o
= 40 Hz) and different output voltage
values are presented in Tables 5.15.2. The results with fixed input-to-output voltage transfer
ratio reference (
ref
= 0.8) and varying output frequency are presented in Tables 5.35.4.


Figure 5.8 Experimental results of the DMC with RL load when
ref
= 0.8 and f
o
= 40 Hz: (a) supply
voltage u
a
and current i
a
, (b) output voltage u
AB
, (c) output current i
A
, (d) spectrum of i
a
, (e)
spectrum of u
AB
, (f) spectrum of i
A
.


Figure 5.9 Experimental results of the IMC with RL load when
ref
= 0.8 and f
o
= 40 Hz: (a) supply
voltage u
a
and current i
a
, (b) output voltage u
AB
, (c) output current i
A
, (d) spectrum of i
a
, (e)
spectrum of u
AB
, (f) spectrum of i
A
.
Table 5.1 Experimental results of the DMC with RL load when f
o
= 40 Hz.

o,ref
(V)
a,1
(V)
a,1
(A) THD-i
a
(%)
A,1
(A) THD-i
A
(%)
o,1
(V) THD-u
AB
(%)
ref

20 335 0.9 6.5 0.9 6.6 21 41.1 0.06 0.06
40 335 1.0 4.9 1.9 3.1 40 21.4 0.12 0.12
80 334 1.3 4.3 3.8 1.7 80 11.5 0.25 0.24
120 334 2.4 2.0 5.8 1.4 120 10.6 0.37 0.36
160 334 3.9 2.2 7.8 1.2 161 5.4 0.49 0.48
200 334 6.0 2.3 9.7 1.0 201 5.7 0.61 0.60
220 334 7.2 1.6 10.7 1.0 220 5.7 0.67 0.66
240 336 8.6 1.6 11.7 1.1 242 5.6 0.74 0.72
260 335 10.1 1.4 12.6 1.1 264 4.2 0.80 0.79
280 334 11.7 1.4 13.6 1.0 283 4.5 0.86 0.85
300 334 13.1 3.8 14.3 2.0 300 4.8 0.92 0.90
As can be seen, the phase voltage magnitude of the supply is from 6 to 9 V higher than the
ideal. Thus, neither the DMC nor the IMC attain the voltage transfer ratio reference
ref
. The
voltage loss in the DMC is lower than in the IMC, as shown in Figure 5.10 with the ratio
/
ref
. In Figure 5.10, the two lowest transfer ratio values signify such low output voltages
that the measurement accuracy and the rounding of the results together lead to an identical
(a) (b) (c)
(d) (e) (f)
(a) (b) (c)
(d) (e) (f)
Experimental Setup 71
situation for the DMC and the IMC. Voltage transfer characteristics are considered further in
Section 7.2.
Table 5.2 Experimental results of the IMC with RL load when f
o
= 40 Hz.

o,ref
(V)
a,1
(V)
a,1
(A) THD-i
a
(%)
A,1
(A) THD-i
A
(%)
o,1
(V) THD-u
AB
(%)
ref

20 334 1.0 14.2 0.9 3.3 20 46.9 0.06 0.06
40 334 1.1 15.4 1.8 2.8 40 26.9 0.12 0.12
80 334 1.4 13.2 3.7 2.2 78 14.0 0.25 0.23
120 333 2.4 8.1 5.6 1.6 118 8.7 0.37 0.35
160 333 3.9 5.4 7.6 1.6 157 7.0 0.49 0.47
200 332 5.9 4.0 9.5 1.5 197 5.6 0.61 0.59
220 332 7.1 3.5 10.4 1.4 217 5.7 0.67 0.65
240 332 8.4 3.2 11.4 1.4 236 4.9 0.74 0.71
260 332 9.8 2.9 12.3 1.3 255 4.3 0.80 0.77
280 331 11.4 2.9 13.3 1.3 275 4.4 0.86 0.83
300 331 12.8 4.1 14.1 2.1 292 4.4 0.92 0.88
Table 5.3 Experimental results of the DMC with RL load when
o,ref
= 260 V, i.e.
ref
= 0.8.
f
o
(Hz)
a,1
(V)
a,1
(A) THD-i
a
(%)
A,1
(A) THD-i
A
(%)
o,1
(V) THD-u
AB
(%)
5 335 10.5 1.7 13.0 4.1 264 8.2 0.79
10 335 10.5 1.4 12.8 1.4 263 3.9 0.79
20 334 10.3 1.5 12.8 1.0 263 3.7 0.79
30 335 10.3 1.5 12.7 1.4 264 3.8 0.79
40 335 10.1 1.4 12.6 1.1 264 4.2 0.79
50 335 9.9 1.6 12.5 1.4 264 4.6 0.79
60 335 9.7 1.6 12.3 1.0 264 4.4 0.79
70 334 9.5 1.6 12.1 1.6 264 4.4 0.79
80 334 9.2 1.7 11.9 1.1 263 4.5 0.79
90 335 9.0 2.2 11.8 1.9 264 4.7 0.79
100 335 8.8 2.0 11.6 1.4 263 3.8 0.79
Table 5.4 Experimental results of the IMC with RL load when
o,ref
= 260 V, i.e.
ref
= 0.8.
f
o
(Hz)
a,1
(V)
a,1
(A) THD-i
a
(%)
A,1
(A) THD-i
A
(%)
o,1
(V) THD-u
AB
(%)
5 332 10.0 3.0 12.6 4.2 258 9.9 0.78
10 332 10.0 3.0 12.5 2.2 256 4.7 0.77
20 332 9.9 3.1 12.4 1.4 256 4.2 0.77
30 332 9.9 2.9 12.4 1.9 256 4.8 0.77
40 332 9.8 2.9 12.3 1.3 255 4.3 0.77
50 332 9.7 3.3 12.2 1.8 256 4.1 0.77
60 332 9.6 3.0 12.1 1.2 256 4.5 0.77
70 332 9.4 3.1 12.0 1.9 256 4.6 0.77
80 332 9.3 3.3 11.9 1.2 256 4.2 0.77
90 332 9.1 3.1 11.8 1.9 256 4.3 0.77
100 332 9.0 3.3 11.7 1.2 256 4.2 0.77


Figure 5.10 Measured voltage transfer ratios () of DMC and IMC compared with their reference value
(
ref
), i.e. values of /
ref
as a function of
ref
with RL load when f
o
= 40 Hz.
72 Chapter 5

5.3.2 Induction Motor Load
This section presents the experiments performed with the 2.2-kW IM drive described in
Section 5.1. The controller parameters are shown in Table 4.3. The results were originally
presented in [P7]. The stationary measurement results with n
r
= 800 rpm and T
load
= 22 Nm are
presented in Figures 5.115.12. Basically, the DMC and the IMC are identical with identical
controller parameters. However, the same controller parameters lead to slightly different
behaviour because of the different gain of the processes [Esk06], i.e. the converter prototypes
have different voltage losses, as presented above with RL load [P7]. As above, the IMC
waveforms are slightly more distorted than the DMC waveforms.


Figure 5.11 Experimental results of the DMC-supplied IM drive when n
r
= 800 rpm and T
load
= 22 Nm:
(a) supply voltage u
a
and current i
a
, (b) output voltage u
AB
, (c) output current i
A
, (d) spectrum
of i
a
, (e) spectrum of u
AB
, (f) spectrum of i
A
.


Figure 5.12 Experimental results of the IMC-supplied IM drive when n
r
= 800 rpm and T
load
= 22 Nm:
(a) supply voltage u
a
and current i
a
, (b) output voltage u
AB
, (c) output current i
A
, (d) spectrum
of i
a
, (e) spectrum of u
AB
, (f) spectrum of i
A
.
The dynamic situations with full load torque T
load
of 22 Nm, presented in Figures 5.135.18
and originally in [P7], are speed reference step, speed reference ramp and acceleration using
the field weakening, respectively. These dynamic results do not show any significant
differences between the DMC and IMC. In addition, both prototypes operate as in the ideal
simulations presented in Figures 4.194.21. Unlike the acceleration using field weakening
with the IMC in [P1], the results here show that the field weakening works with both the DMC
and the IMC [P7]. The reason for this is that different rotor resistance values are used in the
flux model. The value of 1.8 , used in [P1], describes the resistance of the cold rotor, which
does not hold true long after start-up in practice. As presented in Table 4.2, the rotor resistance
value of 2.16 , used in [P7], describes the value of a warm machine but is too low with a hot
(a) (b) (c)
(a) (b) (c)
(d) (e) (f)
(d) (e) (f)
Experimental Setup 73
machine. This shows clearly the importance of parameter tolerance with the flux model, and a
more robust and complex model should be applied in commercial applications [Kaz02].

Figure 5.13 Measured speed reference step operation of the DMC when T
load
= 22 Nm: (a) rotation speed n
r

and its reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator
current component i
sy
and its reference i
sy,ref
.

Figure 5.14 Measured speed reference step operation of the IMC when T
load
= 22 Nm: (a) rotation speed n
r

and its reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator
current component i
sy
and its reference i
sy,ref
.

Figure 5.15 Measured speed reference ramp operation of the DMC when T
load
= 22 Nm: (a) rotation speed
n
r
and its reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator
current component i
sy
and its reference i
sy,ref
.

Figure 5.16 Measured speed reference ramp operation of the IMC when T
load
= 22 Nm: (a) rotation speed n
r

and its reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator
current component i
sy
and its reference i
sy,ref
.

Figure 5.17 Measured field-weakening acceleration of the DMC when T
load
= 22 Nm: (a) rotation speed n
r

and its reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator
current components i
sx
, i
sy
and reference i
sy,ref
.

Figure 5.18 Measured field-weakening acceleration of the IMC when T
load
= 22 Nm: (a) rotation speed n
r

and its reference n
r,ref
, (b) output/stator voltage references u
sx,ref
and u
sy,ref
, (c) output/stator
current components i
sx
, i
sy
and reference i
sy,ref
.
(a) (b) (c)
(a) (b) (c)
(a) (b) (c)
(a) (b) (c)
(a) (b) (c)
(a) (b) (c)
74 Chapter 5

5.4 Conclusion
The basics of prototype implementations and experimental test setups have been presented in
this chapter. Experimental results of both prototypes with passive RL-type load have been
presented to show that they operate as assumed. In addition, both prototypes have supplied the
IM load in the same conditions as in the simulations presented in Chapter 4.
The main circuit implementations of the DMC and the IMC prototypes have been based on the
basics presented in Chapter 2. To assure a reliable comparison, the prototypes have been
implemented as identically as possible. Thus, the control software and the first part of the
modulator logic are identical in both prototypes. Due to the different main circuits, the second
part of the modulator logic implementations and the generation of the safe commutations are
different in the prototypes. The structures of the modulator implementations have been based
on the structures of the simulation models presented in Chapter 4. As in the simulations in
Chapter 4, the modulation method used has been the CSVM with PLL-based synchronisation
to the supply voltage, which has been presented in Chapter 3.
The prototypes have been tested in the laboratory and the experimental results show that they
both produce sinusoidal supply currents and PWM output voltages, which do not contain
significant frequency distortion. The experimental results have also shown that the DMC and
the IMC are basically identical in practice but some differences also exist. With RL load, the
DMC has had higher input-to-output voltage transfer ratio than the IMC. In addition, the IMC
has produced slightly more distorted supply currents than the DMC. These differences have
not been found in the simulations with ideal simulation models. Thus, the differences in the
voltage transfer ratio and in the supply current distortion are caused by the non-ideal
characteristics of real circuits, which are studied in Chapters 6 and 7.
In the experiments with the IM load, the control system in the DMC and the IMC has been the
rotor-flux-oriented space vector control presented in Chapter 4. The same IM test bench and
the same tachometer system have been used with both converters. In addition, the software
implementations and the control parameters have been identical. The measurement results of
stationary and dynamic operations of both converters have also been presented. The
experimental results are in agreement with the simulation results presented in Chapter 4 and
they verify again that the DMC and the IMC operate as assumed. However, the similar
differences between the DMC and the IMC as in the experimental test with the RL load have
also been found to occur in the experimental tests with the IM load.
75
6 Matrix Converter under Distorted Conditions
The MC modulation presented in Chapter 3 assumed ideal operation conditions, i.e. sinusoidal
supply voltage and load current waveforms. However, that is not the case in practice. Neither
matrix converters (MCs) nor their operation environments are ideal. Thus, it is important to
have knowledge of MCs under non-ideal operation conditions. The literature concerning non-
ideal operation conditions has concentrated mostly on unbalanced supply voltages or load
currents, as e.g. [Nie96a], [Nie96b], [Cas98a], [Bla02] and [Kan02], whereas other distortion
frequencies are taken into account more rarely, as e.g. in [Cas98b], [Zha01], [Ohg04] and
[P3][P4]. However, the unbalanced situation is only a special case of a harmonic distortion,
as shown in [P3].
This chapter presents the analysis of the migration of supply voltage distortion to load, which
has been presented in [P3], and the analysis of the migration of load current distortion to
supply, which has been presented in [P4]. This chapter also compares how four simple control
methods mitigate the migration of supply voltage distortion, which has been presented in [P3].
6.1 Space Vector Analysis of Distortion Migration in Matrix Converters
As presented in [P3], supply voltage distortion is carried through MCs to the load, even though
not necessarily entirely. The same holds also for the other input-to-output and output-to-input
disturbances. Compared to random temporary disturbances, continuous repetitive distortion
can be considered a more severe and general problem. The block diagram of distortion
migration in the MCs is presented in Figure 6.1.
Output voltage distortion Output current distortion
Input current distortion
Distorted supply voltage Supply current distortion
(With weak mains)
Distortive load
Filter capacitor
voltage distortion

Figure 6.1 Block diagram of distortion migration in MCs.
76 Chapter 6

The possible distortion sources are distorted supply voltage and distortive load. Supply voltage
distortion can be caused by e.g. non-linear loads such as thyristor and diode rectifiers or the
asymmetric loading of the supply, which all cause voltage distortion in the point of common
coupling (PCC) of the MC. The distortive load can be any load distorting currents or requiring
non-sinusoidal currents. As presented in Figure 6.1, it is possible that the load current
distortion alone can cause voltage distortion in the PCC if the impedance of the mains is high,
i.e. Z
sup
in the DMC and IMC circuits in Figure 4.2.
6.1.1 Input Voltage Distortion Migration
In [P3], the harmonic distortion in balanced three-phase supply voltages was considered. That
kind of voltage contains only odd-order harmonics not divisible by three. Instead of that, the
distortion migration analysis is presented here in general form to attain the maximum
usability. Thus, the analysis presented next is an extension of the analysis presented in [P3].
As in Section 3.3, the modulation frequency distortion is also ignored here and only low-
frequency components are considered. In addition, the effect of a supply filter is taken into
account separately later and the analysis begins from input voltage, i.e. the filter capacitor
voltage in Figure 6.1.
The input voltage space vector u
i
consists of fundamental component u
i,1
and the sum of the
distortion components:
( )
( )
x x
t
x
u u u a u a u u
, i , i
j
i, 1 , i ic
2
ib ia i
e
3
2
+