Induction motor variable speed drives and the matrix converter

Material Information

Induction motor variable speed drives and the matrix converter
Day, Brian
Publication Date:
Physical Description:
vii, 117 leaves : illustrations ; 29 cm


Subjects / Keywords:
Electric current converters ( lcsh )
Thyristor control ( lcsh )
Thyristor converters ( lcsh )
Silicon-controlled rectifiers ( lcsh )
Electric current converters ( fast )
Silicon-controlled rectifiers ( fast )
Thyristor control ( fast )
Thyristor converters ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 114-117).
General Note:
Submitted in partial fulfillment of the requirements for the degree, Master of Science, Department of Electrical Engineering, Department of Computer Science and Engineering
Statement of Responsibility:
by Brian Day.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
23066101 ( OCLC )
LD1190.E54 1990m .D38 ( lcc )

Full Text
Brian Day
B.S.E.E., University of Colorado, 1987
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Science
Department of Electrical Engineering and Computer Science

This Thesis for the Master of Science degree by
Brian Day
has been approved for the
Department of Electrical Engineering and Computer Scien
William R. Roemish
Marvin F. Anderson

Day, Brian ( M.S., Electrical Engineering )
Induction Motor Variable Speed Drives and the Matrix Converter
Thesis directed by Associate Professor Pankaj K. Sen
Over the past thirty years, since the introduction of
the silicon controlled rectifier (SCR), many innovative
topologies have been developed for the speed control of AC
motor drives. This thesis will review the more commonly
encountered of these converters in tutorial form, while
presenting several 'new' schemes which have remained dormant
over the years for want of suitable switching and control
hardware. The latest trends in power electronics and
control technology as associated with variable speed
drives are discussed in detail, proceeding finally to
explore two drive elements in some depth.
The Mos-Controlied Thyristor (MCT) semiconductor
device, only recently introduced, has generated great
interest due to its unique turn-off capability, while the
Matrix Converter with its simple circuit configuration shows
promise for future drives. The thesis will combine these
two devices into a realisable unit.
The form and content of this abstract are approved. I
recommend its publication.

1.0 Introduction......................................1
2.0 Basic Considerations in Speed Control.............3
3.0 Converters.......................................13
3.1 Fundamentals.................................13
3.2 DC Link Converters...........................15
3.2.1 Voltage Source Inverter...............19
3.2.2 Current Source Inverter...............24
3.2.3 Pulse-Width Modulation Inverter.......28
3.3 Cycloconverter............................ 33
4.0 Advances in Technology......................... 42
4.1 Converters...................................42
4.1.1 Voltage Source Inverter...............43
4.1.2 Current Source Inverter...............45
4.1.3 Pulse-Width Modulation................47
4.1.4 Resonant Link Inverter................49
4.1.5 Matrix Converter......................54
4.2 Power Electronic Devices.....................55
4.3 Control Technology...........................69
4.3.1 Constant Volts/Hertz.................69
4.3.2 Field Orientation....................71
4.3.3 Microcomputer........................75
5.0 Matrix Converter................................77

5.1 Fundamentals
5.2 Switching...................................82
5.3 Control and Simulation......................93
5.4 Analysis...................................103
5.5 Presentation of Matrix Converter Circuit... 105
6.0 Conclusions....................................Ill
7.0 Bibliography...................................114

2.1 Torque Speed Characteristics..............7
2.2 Voltage Frequency Relationship............8
3.1 Block Diagram of Converters................14
3.2 AC Voltage Control Using Transformers......17
3.3 DC Link Level Control......................18
3.4 DC Link (VSI) Converter....................20
3.5 VSI Output Voltage and Current Waveforms..21
3.6 DC Link (CSI) Converter................... 25
3.7 Three-Phase Bridge ASCI Inverter...........27
3.8 PWM Converter............................ 30
3.9 Principle of Sinusoidal PWM................32
3.10 Typical Line and Phase Voltages of PWM....32
3.11 Single Phase Cycloconverter............... 35
3.12 18-Thyristor Cycloconverter................38
3.13 Cycloconverter Output Voltage Waveforms...39
3.14 36-Thyristor Cycloconverter................40
4.1 VSI Converter with BJT Output Devices......44
4.2 DC Current Link Converter with PWM.........46
4.3 DC Link Converter with PWM.................48
4.4 Series Resonant Link Converter.............50
4.5 Parallel Resonant Link Converter...........50
4.6 Parallel Resonant DC Link Converter........53

4.7 Operating Characteristics............... 57
4.8 Power Semiconductor Device Comparison.....58
4.9 MCT Thyristor: Specifications.............64
4.10 MCT Forward Voltage Drop Characteristics..65
4.11 Open-loop volts/hertz method of control...70
4.12 Indirect vector control...................72
5.1 PWM Matrix Converter......................79
5.2 Matrix Converter Waveforms................81
5.3 Matrix Converter Switch Realization.......84
5.4 Bidirectional Switch Realization..........85
5.5 Switch Arrangement for Commutation........86
5.6 Matrix Converter Filter Arrangement.......87
5.7 Bidirectional Switch Arrangement..........89
5.8 State Transition Diagram..................92
5.9 Input Vs Output Voltage - Vo = 0.5 Vin....95
5.10 Input Vs Output Voltage - Vo = 0.75 Vin...97
5.11 Input Vs Output Voltage - Vo = 0.866 Vin..98
5.12 Input Vs Output Voltage - Vo = 0.9 Vin...l00
5.13 Input Vs Output Voltage modif'd by Eq.8..101
5.14 Input Vs Output Voltage modif'd by Eq.9..102
5.15 Gate Logic Circuit.......................107
5.16 Block Diagram of Matrix Converter........109

1.0 Introduction
The development of modern AC drives can be traced
back to 1957 with the introduction of the Silicon
Controlled Rectifier (SCR). Since that time technological
advances, both in components and design philosophy, have
been made at a remarkable pace. Before the advent of the
SCR, the vast majority of variable speed drives were of
the DC variety because of their relative simplicity and
the much higher cost of AC drives. Today it has been
estimated that about 70% of the cost of an AC drive is in
the power electronics portion of that drive, compared
with 30% for the DC drive [18]. Technological advances
are quite clearly on the side of the AC drive, which can
be expected to overtake DC drives as the system of choice
in nearly all industrial applications as power
electronics technology continues to evolve.
Converter concepts and topologies have been
continually changing over the past 30 years, some of the
older ones being discarded as newer more efficient
circuits are developed. Similarly, some circuit designs

which have for years languished in text books are now
becoming realisable due to the emergence of new classes
of power electronic devices. This thesis will review some
of the more common converter designs, highlight several
new developments and areas of technical advancement, and
discuss in detail one such 'new' design, namely the
"matrix converter", as applied to an AC motor
variable-speed drive.

2.0 Basic Considerations in Speed Control
From the following well-known equation, induction
motor speed can be calculated.
Nr = 120 f where Nr = rotor speed
f = stator supply frequency
P = number of poles
s = slip
Review of equation (1) suggests that speed control
can be obtained by varying either:
i) frequency
ii) slip
iii ) number of poles
Speed variation by changing the number of poles can

be achieved using pole adjustment modulation (PAM) or
similar switching techniques [23]. However this method
results in discrete step speed control.
Slip variation can be accomplished by changes in
stator voltage or rotor resistance. Both methods realise
reduced efficiency and poor speed regulation and hence
will not be considered any further. Other methods of slip
energy recovery as applied to wound rotor machines are
also beyond the scope of this thesis.
A search of the available literature reveals that the
majority of variable-speed AC drives are operated from a
variable-voltage, variable-frequency power source. The
reasons for this choice are shown as follows.
Approaching the induction motor speed control problem
from the torque equation, which can be shown as [10]:
T = - (P) ttDI Bsr Fr Sin Sr (2)
2 2
where P = number of poles
D = Average diameter of air gap, m
1 = axial length of air gap, m

Bsr = peak value of stator rotor flux
density wave
Fr = rotor frequency
Sin Sr = sine of angle between mmf waves
An alternative form can be shown as
T = - it (P) fisr Fr Sin Sr (3)
2 2
where j&sr = flux produced by the combined
effects of stator-rotor mmf's
The fundamental equation for voltage induced (Erms)
in an AC machine
Erms = 4.44 f kw Nph ft Vrms/phase (4)
or $ = Erms Erms (5)
4.44 f kw Nph f
where kw = winding constant
Nph = number of turns per phase in
stator windings
ft = flux

If we assume idealised conditions, that is, no stator
winding resistance or leakage reactance, then airgap flux
and hence rotor torque are proportional to the terminal
volts/hertz ratio.
As shown in equations (1) and (5), if the stator
supply frequency is increased to raise the machine speed
then air-gap flux is reduced resulting in a reduction in
developed torque. If the machine speed control is
required with full torque capability then air-gap flux
must be held constant, implying that the stator
volts/frequency (V/Hz) ratio be fixed over the total
speed range.
One advantage of this type of V/Hz speed control is
that fast transient response is obtained. This is due to
torque sensitivity per unit of current being maintained
at maximum.
A family of typical induction motor torque-speed
characteristic curves is shown in Fig.2.1.
Below base speed (1.0 pu frequency), the
characteristic is defined as the constant torque region.
In this area torque is controlled by maintaining a
constant volts/hertz ratio. This results in a constant
pull-out torque capability. It should be noticed in
Fig.2.2 that at low frequencies stator resistance becomes

Torque Torque
Torque Speed curves at variable frequency/ constant
Fig.2.1 Torque Speed Characteristics

Fig-2.2 Induction Motor Voltage Frequency Relationship

increasingly significant and additional voltage must be
injected to compensate for the increase in stator
resistance voltage drop.
Above base speed, the frequency can still be
increased. This puts the machine into the constant power
region. However, if the frequency is increased above 1.0
pu (stator voltage limited to 1.0 pu) then the air-gap
flux will decrease with a resulting decrease in motor
torque. Slip remains relatively constant so the speed
will still increase. In this region the upper speed limit
is usually held to 2.5 pu maximum. Operation above this
figure results in higher copper losses with no
significant change in torque output.
Converter fed AC machines are subject to
non-sinusoidal voltage and current waves. This effect is
produced when converting the constant input AC power into
a variable output AC source. In this conversion process
a rectifier effectively chops the alternating current
waveform by allowing current flow only during a section
of the cycle, resulting in a waveform that can be rich
in harmonic content. For example, suppose we have a
converter that produces a 6-pulse output wave as shown
below, then using standard Fourier analysis techniques
we can determine the harmonics associated with the

converter output as follows.

For an odd function with quarter wave symmetry
*k = T
0.5 I Sin (kWQt ) dt + / (I)Sin(kW0t ) dt
b. for all k odd = 8 I f(t) Sin (ktf t) dt
K > O
T/6 T/4
u o
= -ar[-
5 I + 0.5 I Cos (k Tr/3) for k odd
= f(t) = Sin (nWQt ) = ^
k=l k=l
^ ( 1 + Cos (k tt/3 ))
i (t)= ( 1 + Cos (n tt/3) ) Sin (nWQt )

i(t) = 2L(sin (WQt) + i Sin (5WQt) + ^ Sin (7WQt) +
............ ... )
This shows that for this particular 6-pulse example,
associated with the fundamental frequency are multiple
frequencies or 5th, 7th, 11th, 13thjetc.harmonics, which
have to be taken into account.
These harmonics have two detrimental effects on the
rotating machine.
a) Harmonic Heating
b) Torque Pulsations
All converters generate harmonics in their output;
however^the contributions of these harmonics to the
power developed by the motor is negligable. The harmonic
currents developed, simply add to the rms value of the
motor stator current and tend to distort the flux
developed. Consequently the copper loss and core iron
losses are increased giving rise to an increase in motor
temperature. To offset this temperature rise requires
that the motor be derated whenever harmonics are

If the machine is fed from a sinusoidal source, the
air-gap flux and machine mmf will be sinusoidal. Equation
(3) predicts, therefore, that the developed torque will
be smooth. Harmonic torque results from the interaction
of any air-gap flux and rotor mmf produced at different
harmonic frequencies. This torque, when produced by
higher order harmonics, can usually be ignored as it is
effectively damped out by the system inertia. Pulsating
torque produced at lower frequencies, in a practical
system, needs to be examined. Torque pulsations occuring
at frequencies close to the system mechanical resonant
frequency could result in shaft fatigue, gearbox chatter
or unsatisfactory system performance.


3.0 Converters
3.1 Fundamentals
Converter is a general term covering rectifiers,
inverters or any other device for changing power from one
form to another. It is, however, most commonly used to
denote a device able to function both as a rectifier and
an inverter.
A static power converter represents the controllable
link between the primary power source and the rotating
machine or load. Modern converters can be classified as
either DC link converters or cycloconverters.
Block diagrams of both are shown below in Fig.3.1
The DC link converter as shown is a two stage
conversion device. The first stage (rectifier) converts
the connected AC source, with frequency fi, to a variable
level DC output called the link. The second stage

AC to D.C.LINK DC to
Rectif ier
Fig.3.1 Block Diagrams of Converters

(inverter) transforms this DC link into a variable
frequency AC output (fo).
Unlike the DC link converter, the cycloconverter is a
single-stage device. The cycloconverter transforms AC
line power from one frequency to another through a single
stage of conversion. The cycloconverter is just one of a
family of static frequency changers, of which several
excellent references are listed in the bibliography
3.2 DC Link Converters
As discussed earlier, to provide full torque
capability to an induction motor load a constant
volts/hertz ratio source is required for speed less than
the rated speed. Variable frequency output is obtained
from the DC link converter by sequential triggering of
the inverter section.
A variable output voltage level can be obtained by
manipulation of either : 1 2 3
1. AC input or output voltages.
2. DC link level
3. Inverter voltage.

AC Voltage Control
Voltage control can be accomplished simply with an
input or output variable transformer. Block diagrams of
these systems are shown in Fig.3.2
DC Link Level Control
The DC link level can be controlled by use of either
a phase-controlled rectifier or a rectifier-chopper
arrangement. Block diagrams for both systems are shown
below in Fig.3.3
The phase controlled rectifier uses SCR's, or other
switchable power electronic devices, firing angle delay
control and sequential triggering to produce a variable
DC link voltage.
The rectifier/chopper arrangement uses power diodes
in the rectifier section to produce a fixed DC output.
The chopper, which is a DC DC conversion device,
converts the input into a series of unidirectional
constant amplitude pulses. Proper pulse-width-modulation
control of the chopper allows the effective DC output
voltage level to be varied.

Fig.3.2 AC Voltage Control Using Transformers

Fig.3.3 DC Link Level Control

Inverter. Voltage Control
Using a specific control algorithm within the
inverter triggering logic system allows the inverter to
produce in one stage a variable-frequency,
variable-amplitude output from the DC link. This requires
the use of pulse-width modulation techniques which will
be discussed later.
The DC link converter has developed into three
distinct types
1. Voltage Source Inverter (VSI)
also known as Variable Voltage
Inverter (VVI)
2. Current Source Inverter
3. Pulse Width Modulation Inverter
3.2.1 Voltage Source Inverter
The voltage source inverter is shown in Fig.3.4
In this system a three phase bridge controlled

controlled filter inverter
Fig.3.4. DC Link (VSl) Converter



/ L


Fig.3.5 VSI Output Voltage and Current Waveforms. [5]

rectifier converts the AC source to a variable DC
voltage, which is impressed upon a forced commutated
bridge inverter. The inverter output is then a
variable-voltage, variable-frequency supply which can
control the speed of a motor.
Forced commutation, which can be achieved in several
ways, is a method by which the thyristor forward current
is forced to zero by some external circuitry. This allows
the thyristor to be turned-off at any point during its
conducting cycle, as opposed to natural commutation which
relies on the natural passage of current and voltage
waves through zero to extinguish conduction.
Normally each thyristor of an inverter leg conducts
for a full 180. This generates a six-step square wave
output voltage on each phase of the machine. The LC
filter is used to provide a stiff DC voltage source
(ideally zero Thevenin impedance) and aid in commutation.
Since an induction motor constitutes a lagging power
factor load, the inverter thyristors require forced
commutation to minimize current shoot-through.
The VSI has several desirable characteristics, which
are given below:
. VSI configuration allows dynamic braking

. VSI configuration has faster dynamic
response than the CSI; however, the large
filter capacitor limits this response
. The use of PWM techniques with VSI drives
allow smooth operation relatively free
from torque pulsations
. As the VSI is driven by a stiff DC
voltage, several inverters can be
supplied from one rectifier section
. Frequency range is higher than the CSI,
and so has greater speed range.
Since commutation is provided by the filter capacitor
whose charge is determined by the link voltage,
commutation capability decreases with reduced voltage
levels. Many different commutation schemes have been
developed to overcome this problem, most of which can be
found in several of the references listed at the end of
this thesis.
The bypass diodes serve several functions as well as
being part of the commutation process. They allow reverse

current to flow during reactive power flow and
regeneration, and clamp the load voltage to that of the
DC link.
The output voltage waveform is a function of the
circuit configuration and switching parameters and is
unaffected by the load. The current waveforms are,
however, affected by the load and hence, in the case of
an inductive load, tend to be smoother than the voltage
due to load filtering effects.
3.2.2 Current Source Inverter
The current source inverter (CSI) shown in Fig.3.6
likes to see a stiff DC current source (ideally infinite
Thevenin impedence) at the input. A variable DC link
voltage generated by a phase controlled rectifier is
converted to a current source by the series inductance.
Current magnitude can be controlled by adjusting the
rectifier output.
Current switched in the inverter section produces a
six-step current wave through the load terminals. Since
the source is stiff, the load has no effect on the output
current wave. It can be shown that the machine terminal
voltage is nearly sinusoidal with superimposed

*; (\i
b ,
c n-
Fig.3.6 DC Link (CSl) Converter

commutation spikes.
As with the VSI, because an induction motor
represents a lagging power factor load, some method of
forced commutation is required.The Auto-Sequential
Commutated Inverter (ASCI), as shown in Fig.3.7, is
presently the most popular form of forced-commutated CSI.
Operation is quite simple, in that the conducting SCR is
switched off when the second SCR is fired, by applying
reversed biased capacitor voltage across the outgoing
The CSI has many good features which can be
summerized as follows:
. It is rugged and reliable, with no
possibility of shoot-through fault
. momentary load short circuit and
thyristor misfiring are acceptable
. faults on the inverter side cause a slow
rise of fault current which can be
cleared by rectifier section gate
. It has inherent four-quadrant operational

Spl S.03 S
SL 3. ZL

Si -2 .2.
AZ.Q4 S.Q6 S.
Fig.3.7 Three-phase bridge ASCI inverter

capability without need for additional
Despite the features listed above the CSI has several
limitations. The frequency range of the inverter is
somewhat lower than the VSI, and the CSI can't be
operated at no load for obvious reasons. If
auto-sequential commutation is used, then the inverter
section becomes complex, bulky and expensive with
increased losses. Also commutation is affected by a
machine's subtransient reactance which can lead to
serious transient overvoltages across the machine
terminals. Dynamic response is sluggish with possible
stability problems under low-load, high-frequency
conditions. Finally, multi-machine operations with a CSI
converter are difficult to accomplish. These limitations,
along with its high component cost, have combined to
limit the acceptability of this converter which is
predicted to be slowly superseded by other designs [18].
3.2.3 Pulse-Width Modulation Inverter
The pulse-width modulation (PWM) converter as shown
in Fig.3.8 is configured similarly to the VSI converter.

It has, however, been designed to overcome many of the
problems associated with the VSI.
The PWM inverter is supplied by a uncontrolled bridge
diode rectifier front end and an LC filter. The
fundamental output frequency and output voltage can be
controlled by switching the inverter elements using a
pulse-width modulation technique. With PWM the output
devices are switched on and off many times within a cycle
to control the output voltage. One advantage of this
method for obtaining a variable output voltage is that it
is normally low in harmonic content.
Several PWM types have been described in the
literature [5]:
1. Sinusoidal PWM
2. Harmonic Elimination PWM
3. Adaptive Control PWM
4. Phase Shift PWM
Sinusoidal PWM is the most popular of these
techniques, for industrial applications, and therefore
will be the only one described here.


5 5 5

ii v.ii
Fig.3.8 PWM Converter

With sinusoidal PWM an isoceles triangle carrier wave
is compared with a fundamental frequency sine wave. The
two wave crossover points determine the switching points
of the power devices. This technique is also known as the
triangulation, subharmonic or suboscillation methods [5].
The general principle is shown in Fig.3.9 with
typical line and phase voltage waves shown in Fig.3.10.
In Fig.3.9 voltage Vd is the DC link voltage, while
voltages Vp & Vt define the upper limits of the
fundamental sine wave and the triangular carrier waves.
Although the PWM inverter output has reduced harmonic
content, hence lower motor harmonic losses, the inverter
efficiency is lower due to the multiple commutations of
the output devices per half cycle. Good design dictates
that the commutation frequency balance the increased
inverter losses with the decrease in machine harmonic
Since the DC link is relatively constant, commutation
problems like those associated with the VSI converter are
not encountered. Also since the DC link is not
controlled, several independantly controlled inverter
sections can be coupled to the rectifier section. This
results in a lower component count, hence a large
reduction in system losses and costs.

output voltage w.r.t center tap
of dc voltage.
Fig.3.9 Principle of Sinusoidal PWM.
Fig.3.10 Typical Line and Phase Voltages of PWM Inverter.
( a ) Line Voltage, ( b ) Phase Voltage.

PWM techniques have received wide attention over the
past few years, because of their ready adaptability to
implementation by microprocessor. With computer methods
the amplitudes of a per unit sine wave at regular angular
intervals can be stored in ROM. At each trough of the
triangular wave, the desired voltage is sampled then
multiplied by the per unit sine wave value. The resulting
digital word can then be converted to a pulse using some
type of counter, and supplied to the output device
triggering circuit [3].
3.3 Cycloconverter
At the present time the majority of industrially
applied cycloconverters rely on natural commutation to
aid in output voltage control. Natural commutation relies
on the natural passage of current and voltage waves
through zero to extinguish conduction.
Natural commutation limits the output frequency
range, consequently the maximum usable output frequency
range is typically somewhere between 1/3 and 2/3 of the
input frequency.

This fraction depends upon [20] :
1. The particular converter circuit used.
2. The amount of distortion tolerated at
the output.
As a result the major application of cycloconverters
has been low speed, high power AC drive systems.
The fundamental principle behind cycloconversion can
be explained with the help of the single phase circuit
shown in Fig.3.11. Here a center tap single phase full
wave converter utilizes two sets of antiparallel
thyristors so that voltage and current of either polarity
can be supplied to the load, assuming the load is
resistive for simplicity. Fig.3.lib shows the output
waveform generated for a firing angle Of = 0. The
fundamental frequency (fo) is found from :
fo = 1/n fi
where n Humber of input half-cvcles
half-cycles of output

Fig.3.11 Single Phase Cycloconverter
(b) with no firing angle modulation.
(c) with firing angle modulation.

approximation of the desired output sine wave can be
synthesised using PWM techniques.
Gyugyi and Pelly [14] suggest other construction
methods based on both the half-cycle and PWM methods,
called consecutive composite and concurrent composite
wave methods. Output sinewave synthesis is better, but
the switching control becomes more complex.
Both these switching characteristics are discussed
extensively in the reference quoted.
Cycloconverters can be produced with many different
circuit configurations depending upon the required
Fig.3.14 shows a 36-thyristor six-pulse bridge
circuit. With this particular cycloconverter
configuration the output voltage magnitude is double the
input. Alternatively it can be said that for a particular
output voltage then the input voltage magnitude required
is only half of that needed for an 18-thyristor circuit.
This has the advantage of requiring that the thyristor
voltage rating be only half of that required for the
18-thyristor configuration. The six-pulse output gives
smooth voltage and current reproduction with reduced
harmonic content, however the control mechanism for 36
thyristors becomes very complex.

If the firing angle x is varied then the output
harmonic content can be controlled along with the
voltage. This result is shown in Fig.3.11c. In both the
previous cases frequency reduction is observed.
It should be pointed out here that the average
positive voltage is equal to the average negative
voltage; however, the instantaneous positive and negative
voltages are not equal. This inequality results in
circulating currents which can be prevented either by
blocking the nonconducting converter or by the use of an
intergroup reactor (IGR).
As would be imagined the single-phase converter is
rarely used, but the described method of operation can be
translated to the three-phase domain manifesting itself
as the 18-thyristor, three pulse cycloconverter shown in
Fig.3.12. The circuit consists of three identical phase
groups and is shown coupled to a wye-connected load.
Sequential switching, of the positive and negative groups
so that the instantaneously most positive input voltage
is connected to the output during the positive
half-cycle, and the most negative input voltage is
connected during the negative half-cycle will produce a
output voltage wave that approximates a sine-wave. This
output voltage wave, of fo < fi, will be a crude
approximation as shown in Fig.3.13. A better

Fig.3.12 18-thyristor cycloconverter.

Fig.3.13 Cycloconverter Output Voltage Waveforms [14]
(a) using sequential switching
(b) using PWM switching techniques

O BJ >

Because of the large number of circuit components,
cycloconverters at present are only cost effective when
coupled with large horsepower drives requiring relatively
low speed ranges.
With the inceasing availability of new high power
electronic switching devices, forced commutation
techniques are becoming the subject of renewed research.
These commutation techniques will allow the output
frequency range of the cycloconverter to approach that of
the DC link converter. Forced commutation also has the
advantage of allowing power factor control, however at
the expense of increased complexity and system losses.

4.0 Advances in Technology
4.1 Converters
Recent advances in AC drive technology have focused
almost exclusively on power electronic devices, switching
control and induction motor loads, converter topologies
having remained relatively constant over the past 25
years. In each case a major goal of these advancements
has been to increase the switching frequency of the
various circuits. A higher switching frequency means that
more circuitry can be squeezed into a smaller space
allowing more efficient use of power conditioning
systems. This arises from the fact that the inductors,
capacitors and transformers required for filtering and
energy storage shrink as the frequency is increased. The
high switching frequency also brings about smoother
operation with faster system response than that obtained
from conventional Silicon-Controlled Rectifier (SCR)
driven circuits; an absolute necessity if variable speed

drives are to be considered for anything other than high
power industrial machinery.
4.1.1 Voltage Source Inverter
The original VSI converters, see Fig.3.4, used a
resonant LC circuit for forced commutation purposes. The
commutation circuit utilized auxilliary SCR's connected
in a inverse parallel relationship with the main devices
to divert current flow and switch the inverter off. This
increased the converter component count and complexity,
increasing losses and reducing overall efficiency.
Improvements in the characteristics of the present
generation of SCR's have helped partially overcome many
of these problems. SCR's with high switching speeds,
greater di/dt and dv/dt limits have greatly increased the
reliability, speed and power handling capabilities of the
VSI drive.
With the advent of high power bipolar transistors
(BJT), many of the problems associated with commutation
have been eliminated. A typical VSI converter circuit
using power transistors (BJT's) as the output device is
shown in Fig.4.1. This type of converter has now almost
completely replaced the SCR converter in sizes up to 100

Fig.4.1 VSI Converter withBipolar Junction Transistor
Output Devices.

HP because of its advantages of high frequency switching
and simple control. Power transistors are now becoming
available which will continue to push this 100 HP limit
even higher, and research is being conducted which will
help overcome the problems associated with the switching
of high power transistors [21].
4.1.2 Current Source Inverter
The Current Source Inverter (CSI) converter shown in
Fig.3.6, like the VSI, has remained basically unchanged
since its inception. As was discussed before, the CSI
design has many good points such as good starting torque,
wide speed range and most importantly built-in protection
against overloads and short circuits. However, it was not
capable of multiple machine operation and with its static
design it has suffered a cost disadvantage as component
parts have steadily increased in price. It was predicted
that the CSI converter would slowly be replaced [18],
however recent developments suggest that its demise is
not yet imminent [29].
Fig.4.2 shows the CSI converter with pulse-width
modulated input and output bridges. These bridges are
configured using BJTs, which allow much faster switching

Fig.4.2 DC Current Link Converter with PWM
of both Input and Output Bridges.

speeds and simpler control. Several other semiconductor
devices could be utilized to perform the same functions
as the transistors, for example Metal-Oxide Semiconductor
(MOS) transistors or Gate Turn-off Thyristors (GTOs),
with the choice of element really depending upon the
particular application involved.
4.1.3 Pulse-Width Modulation
Pulse-Width Modulation (PWM) is one subject that is
receiving a great deal of attention. PWM switching
algorithms coupled with high power BJTrs have made
possible much cleaner output current waveforms.
Fig.4.3 shows a DC voltage link converter with
double-ended PWM modulated bridges. The PWM control of
the DC link allows the input current to be maintained
almost sinusoidal. Hence, the output inverter, which is
also PWM controlled, can supply the load with an almost
sinusoidal current wave. The advantages of this topology,
apart from sinusoidal input/output current, are four
quadrant operation and unity fundamental power factor at
the input. It can be seen that with this particular
converter, because forced commutation is used, no filter
inductor is required, which reduces cost and size.

Fig.4.3 DC Link Converter with FWM of both
Input and Output Bridges.

A double-ended PWM bridge attached to a CSI converter
was shown in Fig.4.2. Its advantages are nearly
sinusoidal current at both the input and output, unity
fundamental power factor and four quadrant operation.
Since a DC current link is used, the switches used are
required to carry only unidirectional current.However,
the switches must be capable of blocking voltage of
either polarity when switched off. The converter shown
employs bipolar junction transistors (BJT's) for speed
and turn-on/turn-off capabilities. It should be noted
that since eight devices are in series with the load at
any switching instant the system losses are considerable.
4.1.4 Resonant Link Inverter
Two similar converter topologies suitable for motor
drives which are receiving attention due to their many
advantages over conventional PWM switched converters are
the series and parallel resonant link converters shown in
Figs.4.4 and 4.5 [8]. These advantages include lower
switching losses at high frequency, higher efficiency,
reduced component stress due to low di/dt and dv/dt, and
easier Electro-Magnetic Interference (EMI) filtering.
With this type of converter, unlike the conventional DC

Resonant Filter
Fig.4.4 Series Resonant Link Converter [18]
Fig.4.5 Parallel Resonant Link Converter [18]

link converter, the link voltage resonates at high
frequency. Typically this resonant frequency can be from
20 kHz, to be above the audio range for high-power motor
drives, up to 500 kHz for a switching power supply. With
the series resonant converter, the load is connected in
series with the circuit and the controlled output voltage
is obtained from the link current. This normally requires
the use of capacitors on the input and output to smooth
the current pulses.
The parallel resonant converter has the load
connected in parallel with the circuit, and the output
voltage is taken from across the resonant capacitor. A
feature of this circuit are the inductors which are
required to filter the voltage pulses.
In both circuits the rectifier and inverter stages
are PWM driven which allows the switches to be forced to
operate at the zero crossing point of the voltage wave.
Because of this zero voltage switching, the output
devices can operate without snubbers, little heat sinking
is required, device stress is reduced, and device
switching losses are greatly reduced from those of
conventional DC link units. In addition, because the link
components resonate at 20 kHz and above, their component
size is minimized. A major advantage acheived with these
types of converters is that the high resonant frequency

implies a high PWM frequency so that the output function
is produced with reduced harmonics and little distortion.
Some principal characteristics of the series resonant
converter include, inherent overload protection, load
sensitivity, and poor operation at no load. While those
characteristics of the parallel resonant converter are,
load insensitivity, no load operation, but a need for
overload protection.
Although the series resonant converter has seen some
application in the aerospace industry, the parallel
resonant converter is prefered for applications having
wide load variation such as switching power supplies.
Resonant link converters have seen little general
high-power application because the output devices are
required to have bidirectional current conduction and
voltage blocking capabilities. This can be accomplished
with a simple antiparallel arrangement of thyristors;
although even with fast turn-off thyristors, the maximum
attainable link frequency is compromised. A recent
solution which partially overcomes the problem of
bidirectional switches is the resonant DC link converter,
shown in Fig.4.6. With this scheme the resonant ac link
is biased in such a way as to ride on top of a dc signal.
This allows the link current to be considered as

-6 H5 -Q
Fig.4.6 Parallel Resonant DC Link Converter [is]

unidirectional while oscillating at high frequency.
Careful triggering control still allows the voltage or
current wave to be switched at the zero point with all
the advantages of the ac resonant link converter
described earlier.
4.1.5 Matrix Converter
The matrix converter, which will be discussed in
detail later, is another converter topology that has
suffered for lack of an adequate bidirectional switch.
However, like the resonant converters just discussed, it
also is receiving renewed attention due to recent
advances in power semiconductor devices.

4.2 Power Electronic Devices
If the "state of the art" of power converters have,
over the past few years, gone through some remarkable
changes then it can truely be said that the field of
power electronic devices has experienced a revolution.
The following is a list of power electronic devices
which have become available since the advent of the
Phase-Control-Type Thyristor, more commonly know as the
Silicon Controlled Rectifier (SCR) [6][22].
Phase Control Thyristor (SCR)
Bidirectional Triode Thyristor (Triac)
Gate Turn Off Thyristor (GTO)
Fast Switching Thyristor (SCR)
Silicon Controlled Switch (SCS)
Light Activated SCR (LASCR)
Asymmetrical Thyristor (ASCR)
Reverse Conducting Thyristor (RCT)
Gate-Assisted Turn-Off Thyristor (GATT)
Power Metal-Oxide-Semiconductor
-Field-Effect Transistor (MOSFET)
Power Transistor (BJT)
Insulated Gate Bipolar Transistor (IGBT)

Static Induction Transistor
Static Induction Thyristor !
MOS Controlled Thyristor j
Since many of these devices are well known they will
only be discussed briefly in this| chapter, however
several of them are of recent devjelopment and worthy of
detailed comment. Fig.4.7 shows in tabular form operating
characteristics of some common semiconductor devices,
while Fig.4.8 gives a graphical overview of the voltage
and current extremes of some presently available power
electronic devices [8][11]. j
Although the SCR still reignsj supreme in the areas of
very high power drives and HVDC transmission, its two
disabilities being slow switching! speed and turn-off
difficulties, have required device manufacturers to turn
to other sources to overcome these limitations. This has
resulted in the development of high-power bipolar
junction transistors with ratings that would have been
considered impossible just a few years ago. For example
the DT800 npn transistor from Marconi Electronic Devices
passes continuous collector current of 1000 A (1200 A
peak) and withstands 400 V of collector-emitter voltage.
The device, although packaged in a 54 mm diameter capsule
dissipates 3 kW, has a respectable 2 uS fall-time and a

Forward Blocking Voltage,
Fig.4.8 Power Semiconductor Device Comparison [8]

Device Major applications Forward block range,v Reverse block voltage Max device current Max operating frequency dv/dt range, V/jiS di/dt range, A/jiS Max operating temperature
BJT switching power supply; motor drives, UPS; 50-1400 50 1000 200 kHz 3000 10000 500 2000 150 C
IGBT motor drives; power supply; 400-1200 200 100 50 kHz 50000- 100000 2000 5000 200C
SIT induction heating & welding; RF generators; 50-1000 0 100 50000 kHz 100000 300000 10000 30000 200 C
10SFET switching power supply, RF generator: automotive power supply; 50-1000 0 100 25000 kHz 50000 200000 10000 30000 200 C
GTO high power drives;UPS; industrial process controllers; 800-4500 200 4000 20 kHz 300-1500 100-300 125 C
SCR very high power switchgear; HVDC;high power drives 600-5000 0-5000 5000 50 kHz 20-500 50-1200 125 C
Fig.4.7 Operating Characteristics of some common Semiconductor Devices [8]

minimum hfe of 7. Hfe is an h-parameter derived from the
Hybrid-Pi model of the transistorj, which roughly
approximates to the gain of the transistor.
Another manufacturer, Westcode, has several bipolar
transistors with collector-emitte|r voltage ratings of
1000 V. The WT56XX series can switch between 70 and 80 kW
while dissipating 1.25 kW within ja 61 mm square device.
The device hfe is between 5 and 6 for a collector current
of 100 A, with 1.7 jiS turn-on time, 1.5 jiS fall time and
a storage time of 7 jiS.
Despite their high current anjd voltage ratings,
bipolar transistors do have disadvantages: they need a
high base drive, suffer from secondary breakdown, and
like the SCR have serious frequency limitations. The
upper frequency limit for most power BJT's appears to be
around 40 kHz. Newer generations of transistors can
increase the operating frequency range to 100 kHz. Even
so, BJT's are still no match for Metal-Oxide
Semiconductor (MOS) devices which
with swiching times of
100 nS can operate at 2 MHz and above.
Power MOSFETs, being voltage-controlled, require only
a relatively small current pulse to turn them on or off.
They can easily be paralleled to carry more current, are
high efficiency devices, but require large chip areas if

that efficiency is to be maintained at high voltage
levels. For example, a power MOSFET rated at 400 V may be
four times as large as a corresponding BJT, and their
practical upper voltage limit stinds at about 1000 V.
Presently available units have characteristics ranging
between 1000 V @ 12 A to 60 V @ 100 A, with both being
manufactured by the Ixys Corporation. Due to their high
switching frequency and continually lowering cost MOSFETs
are seeing application in motor driven appliances such as
washers, dryers, blenders, etc. MOSFETs are also finding
wide acceptance in the automotive industry where their
ability to operate at low voltagej and high current and
their ease of control make them ideal for automobile
electronic packages. :
The insulated-gate-bipolar-trjansistor (IGBT),
sometimes called a conductivity- modulated field-effect
transistor (COMFET), combines features of both the
biploar transistor and MOSFET. Voltage pulses to the MOS
gates switch the device on or off easily while the
bipolar characteristic ensures low forward conduction
drop. The IGBT device is usually smaller in size than the
MOSFET by up to a factor of three, while its switching
frequency, of not more than 50 kHz, is much lower. IGBT
ratings are presently in the 100 V @ 50 A or 500 V @ 100

A range. Continuing research is expected to push these
values higher. Due to their low conducting resistance and
ease of control, the IGBT is finding wide usage in 240 V
or 480 V motor drives. It is predicted that they will
eventually replace BJTs in many applications between 400
V and 1200 V, and GTO's in their
lower voltage range.
Although the GTO has been around since the late
1950's it has continued to be refined, mainly by the
Japanese, until present models a^e available with ratings
up to 4.5 kV, 4000 A [6]. The GTO thyristor, constructed
much like the SCR, is a four-layer device which can be
turned-off anywhere within the conduction cycle. Turn-on
is accomplished in the usual manner but turn-off requires
a negative pulse to the gate that! reverses the voltage
across it, or by commutation circuitry reminisent of the
SCR. The GTO typically suffers frjam low turn-off gain,
typically between 3 and 6, thus requiring a high current
level pulse lasting a few microseconds to ensure full
turn-off. As their ratings continue to be expanded, the
GTO, because of its simpler control characteristics, is
replacing the SCR in such applications as railroad
locomotives and heavy machinery. 1
Japan has also been an innovator in the development

and commercial introduction of both the static induction
transistor (SIT) and the static induction thyristor
(SITH). The SIT is essentially alJFET with
characteristics like those of a vacuum triode. Its main
advantage is a combination of high voltage control, high
current capability and a very fast switching frequency.
Commercial units are already available with ratings of
1200 V, 80 A, which switch at over 2 MHz. Demand for the
SIT has been slow for two reasons; first it suffers a
large conduction voltage drop, and second, because they
are built in a normally-on configuration they may present
difficulties in ensuring fail-safe start-up. Because of
the SITs high operating frequency range, this device is
finding application in such diverjse fields as induction
heating and welding, and high powjer radio frequency
generators. Closely associated with the Static-induction
transistor, the SITH is an asymme
with characteristics similar to those of the 6T0 for
trical blocking device
drop. Switching speeds
up to 100 kHz with
turn-on, turn-off, and conduction
are higher than those of the 6T0,
present models available in the 2500 V @ 300 A range.
Typical applications for the static induction thyristor
are expected to be found in medium power converters
having a frequency of several hundred kHz such as
induction heating and R.F. generators.

The MOS-controlled Thyristor (MCT), recently
introduced by General Electric [26], is a device which is
showing much promise for reviving several dormant
converter topologies. It is a device that is based on the
optimal combination of both MOS and Thyristor elements.
This new device overcomes the inherent disadvantages of
both the SCR and the GTO, being a! turn-on/turn-off
thyristor with high turn-off gain and high di/dt and
dv/dt protection. Turn-on is accomplished with a -5 V
signal to the gate. Conduction dejlay is just 200 nS.
Turn-off requires at least a +7 Vj signal although 14 V is
recommended. Turn-off storage time is given as 0.5 ^iS
with an actual turn-off time of 2 jiS. These short times
make it possible to consider high switching frequencies.
As the switches are MOS devices, switching currents are
in the 100-200 mA range, for both turn-on and turn-off, a
major improvement over existing turn-off devices such as
the GTO.
Listed below in Fig.4.9b are some presently available
MCT thyristor specifications [12].
Fig.4.10 shows the forward voltage drop plotted
against current density for various power devices. It can
be seen that the MCT offers considerable advantage over
other semiconductor devices in terms of forward drop. A

Fig. 4.
I IK 200 ' 300 400
Peak onMabie e*rr*M (A)
( a ) ;
rf Specification S: Units tasiba TA9789B TA9836A TA9836B
Breakdown voltage (a) V an 1000 500 ' 1000 :
i. Unsnubbered SOA (b) V 300 GOO 300 600
Peak controllable current (c) A 9 SO 100 100
Peak current A an 500 1000 1000
Die aize mils 170X227 170 X 227 260 X 390 260 X 390
,(*> V LI 1.1 1.1 1.1
Input capacitance pF TWO 7000 14,000 14,000
di/dt A/jis 2000 2000 2000 2000
dv/dt W/is 20000 20,000 20,000 20,000
Turn-on time ns 200 200 200 200
Storage time (e) ns 90 500 500 500
Turn-off time ns 2000 2000 2000 2000
. Gate-to-anode voltage Maximum v 23 20 20 20
Turn-on V V -5b)-1S 5to15 -5 to-15 -5to-15
Turn-off +10 to+15 +10 to ^15 +10 to+15 +10 to+15
Package (f) Five-jeadTO-218
Tj = 150*C
(a) Breakdown voltage with a diode-resiskr-capacttor snubber between anode and cathode
(see Fig. 1b). j
(b) SOA = safe operating area (see Fig. 4).
(c) Current that can be turned off at the gate.
(d) Forward voltage drop at peak conhoflahiecurrent.
(e) Time between application of tum-pff poise and when the MCTs anode-to-cathode current
. starts to drop. !
( b )
9 MCT Thyristor Specifications and Characteristics [12]

-V p Anode
1000 -
600-V MOS-controlled thyristor
600-V insulator-gate bipolar transistor
600-V, 10-,*$ SCR
600-V Darlington
600-V bipolar transistor
or 300-V OMOSFET
1 2 3
Forward dnp(V)
Fig.4.10 MCT Forward Voltage Drop Characteristics

600-V MCT with a forward drop of
1 V operates at 10 times
the current density of a 600-V IGBT, a 600-V BJT, or a
300-V FET. It works at about 30 times the density of a
600-V Darlington, 100 times that of a 600-V FET, and
approximately 500 times that of a low-technology SCR.
The MCT promises to be an extremely rugged device.
Its dv/dt of 20,000 V/pS will alljow it to withstand
transients found in electrically noisy environments.
Similarly its 2000 A/jiS di/dt will permit the MCT to
turn-on into a short without failing. For comparison,
typical SCR ratings will be in the range: 20 1000 V/jiS
and 20 500 A/pS. j
MCT's are specified with a junction temperature of
150 C, however they will operate over a much wider
temperature range with the appropriate packaging. The
practical upper temperature is unlikely to be much above
O ! O .
250 C, as reverse leakage current doubles every 10-12 C
o !
above 150 C. It has also been found that the peak
controllable current decreases with increasing junction
temperature. i
Fig.4.9a shows the MCT safe operating area (SOA).
Attempting to turn-off the device operating at current
levels to the right of the characteristic leads to the
risk of device destruction. The device itself will not
fail at these current levels, if no off-gate voltage is

applied, and so needs to be protected from faults by
circuit breakers or fuses. j
As of the time of writing this paper limited
engineering prototype MCT's were just becoming available
for evaluation. Controllable current levels for these
units were limited to 100-A at 1000 V.
As the MCT is still under development, it is it's
theoretical possibilities which fuel the imagination and
I .
suggest a bright future. Devices jwith 400-500 A
capability appear to be readily available as die sizes
grow. It has been suggested that an MCT with a 500 A
capability could service 95 X of the AC drive market
[18]. Another interesting aspect of the MCT is that
because it uses MOSFET gates for switching, then it is
also possible to gate the device with a single polarity
either positive or negative. Thisl is because the on-FET
and the off-FET can be chosen to be an enhancement or
depletion device. Turn-on/turn-off from a single polarity
source makes the MCT an attractive device for interface
with a microprocessor to control a simplified switching
A recent paper presented by the developers of the MCT
[15] has reaffirmed many of the basic characteristics of
this device, and listed several modifications to
previously published data. For example:

Measurements confirm the! low forward voltage
drop at high current densities.
(1.1 V @ 200A/cm 150 C)
MCT's can be effectively' paralleled to
switch high current levels with less than
10% device derating. Experiments were
performed with eight 80 Ja matched
devices which were paralleled to switch
600 Amps at 275 V, representing a 6%
current derating. j
Switching times have been measured at
0.3 rise-time, and 0.5 jiS fall-time
for a described module.
Originally indicated di/dt ratings have
to suggest tnat some modules can reach
Improvements in MCT performance characteristics are
still being sought, especially in |the areas of di/dt
yet to be realised. Present devices can
with evidence

ratings, SOA expansion, and voltage and current ratings.
Presently available MCT thyristors are
turn-on/turn-off devices with blocking capability in both
directions. Bidirectional current conduction is not yet
available, but is the goal of future development efforts.
4.3 Control Technology
4.3.1 Constant Volts/Hertz
No discussion on the advances in converter technology
would be complete without a short section on changes in
control theory. Thus far all discussion has revolved
around the open-loop constant volts/hertz method for the
control of induction motor loads, a representative block
diagram of which is shown in Fig.4.11. While this method
is still most widely used for commercial drive systems,
the method of field orientation has been derived for use
on high performance drives. Because of the complex
nonlinear relationship between induction motor stator
current and the resultant output torque, the use of
microcomputers, with their enormous computational powers,
have figured prominently in this new control scheme.

Fig.4.11 Open-loop volts/hertz method of
induction motor speed control.

4.3.2 Field Orientation
Field-orientation (also known as vector control, and
transvector control) provides a means by which that
portion of the stator current that goes into producing
the air-gap flux, can be effectively decoupled from that
portion that produces torque. This, theoretically, allows
the independant control of both flux and torque. Field
orientation has developed into two general methods, the
direct method and the indirect method, with the basic
difference between the two being the way in which the
internal vector control signals are generated [5]. For
reasons beyond the scope of this thesis the concept of
indirect field orientation has emerged as the consensus
choice for high performance drives [18].
The indirect field-orientation control method
utilizes a feedforward scheme where the load slip
frequency is actually calculated from estimates of the
stator current and rotor time constant. The desired slip
frequency is added to the rotor speed and the resultant
is used as the control signal for the inverter. This can
be seen quite clearly in Fig.4.12 [6], where the flux
component of the stator current Ids is kept constant,

Fig.4.12 Indirect vector control for a motor load. [6]
coordinate 2PH/3PH

while the torque component of the current Iqs is
controlled by the speed loop. The slip signal Wsl, which
is a function of the stator current Iqs, is added with
the rotor speed and the unit vectors are generated from
the resulting signal.
With this method of control torque response can
exceed that of an equivalent DC drive; however, speed
control is limited by the accuracy of the estimated rotor
time constant. Unfortunately the rotor time constant,
varies with both temperature and air-gap flux, so
research into ways of overcoming this problem is
continuing. The availability of high speed
microprocessors has made possible the use of the extended
Kalman filter algorithm to estimate the rotor time
constant. In this approach, a sample of voltages and
currents are taken and stored in memory. The Kalman
filter algorithm is then used to compute and estimate the
current rotor time constant. Current computational time
for this estimation method is given as approximately 30
sec [18]. Because of the time delay, this method of
on-line parameter identification is presently limited to
drives with slowly changing output load conditions.
A lower cost approach to indirect field-orientation
is the method of off-line parameter identification. This
procedure requires that the inverter control mechanism

perform a series of diagnostic tests at initial
commissioning. These data are stored in memory and used
by the microprocessor as the basis for the field-oriented
control algorithm generated outputs to the inverter.
Much time and effort is being put into development of
the field-orientation method of control, with the
ultimate aim being the development of a self-learning,
self-tuning adaptive controller for industrial AC drives.
Concurrent with the development of field-oriented
control have emerged innovations for maximizing energy
savings brought about by speed control. It has been
shown, by Nola [19], that energy can be saved in
constant-frequency machine drives if the voltage is
regulated in such a way as to keep the machine load power
factor constant. Similar savings are possible through the
judicious control of the flux and torque producing
elements of the stator current; however, the greatest
gains appear to be associated with reducing the harmonic
content of the output current and voltage waveforms.
Additional benefits can be realised at low
frequencies by increasing the motor flux level. This can
be traced to the fact that the optimum mix of copper and
iron losses change at low frequency, due to the frequency
dependance of the iron losses. Lipo [18] suggests that

savings of between 20% and 40% of these losses can be
obtained for both light and heavy loads at load speed.
Finally, when adaptive controllers for field-oriented
control systems become available power minimization
techniques should become a common sight. Power
minimization allows the system to monitor both input
power and available output power, thus allowing the
continuous calculation of system losses. An incorrect
rotor time constant estimate will increase motor losses,
hence this scheme should allow identification of a
minimum value of time constant simply by adjusting the
control system until a minimum input power condition is
4.3.3 Microcomputer
Many of the control concepts so far described can all
be implemented using dedicated analog and digital
circuitry. The application of complex control algorithms
such as the Kalman estimation routine require the
extensive use of microcomputers. Microcomputers also
permit the application of modern optimal and self-tuning
adaptive control stratagies to power electronic systems
including variable speed drives.

Continuing developments in the area of microcomputer
technology have resulted in the simplification of control
hardware with the benefit of reduced size and cost,
elimination of drift, improved reliability and reduced
problems of EMI. Standard hardware configurations are
easily coupled with flexible software control algorithms
to reduce development and production costs for noncomplex
topologies. Microcomputer control also permits on-line
monitoring and diagnostics, complex control and command
processing, and facilitates multimachine control from a
single host computer.
As microcomputer technology continues its evolution
several other fields of opportunity are becoming
available to designers of variable speed drives.
Computer-Aided-Design (CAD) and Artificial Intelligence
(AI) can be expected to play a large roll in the
development and advancement of power electronics. CAD/CAE
will allow design and simulation of new control schemes
without the need for initially building prototypes,
helping reduce costs and speeding up development time. AI
is expected to play a large roll in the development and
testing of power electronic systems by automating
simulation studies and system modeling, testing and
advanced diagnostics, and control design.

5.0 Matrix Converter
Over the years research attention has focused on a
wide variety of converter schemes for the control of
induction motor speed. Amongst the best known, and most
widely used industrially, are the VSI, CSI and PWM
inverters which were discussed earlier. Recently, with
the introduction of several new power electronic devices,
some converter topologies which to date have recieved
little practical attention are becoming the subject of
renewed research. Most of these converters have laid
dormant for want of a suitable bidirectional power
switch. Of these schemes, only the general class of AC-AC
converters, generally known as static frequency changers,
with their many desirable features will be considered
5.1 Fundamentals
The cycloconverter, which is possibly the best known

of all frequency changers, can convert AC power at one
input frequency to output power at a different frequency
with a single stage of conversion. The distinguishing
feature of the cycloconverter is that it has no 'DC Link'
and so requires no large capacitor or inductor, to
support the link voltage. This allows the converter to be
produced in a much smaller, more compact package with
fewer components, hence reduced losses; while the single
conversion stage allows a simpler trigger control scheme
to be used.
In general cycloconverters have the following
desirable characteristics:
1. Sinusoidal input and output waveforms
with minimal higher order harmonics.
2. Bidirectional energy flow capability.
3. Minimal energy storage requirements.
4. Controllable power factor.
Of all static frequency circuits, the so called
"matrix" converter whose block diagram is shown in
Fig.5.1 [2] is showing the most promise for development

Fig.5.1 General Circuit Diagram of a
PWM Type Matrix Converter. [2]

when coupled to one of several new switchable
semiconductor power devices such as the Mos-controlled
Thyristor (MCT). The matrix converter gets its name from
the nine switches that are arranged in a matrix
configuration, such that at any instant, any one of the
three input phases can be connected to any one of the
three output phases. This can be accomplished with either
polarity of input voltage or output current.
The input and output voltage and current waveforms
for a typical matrix converter are shown in Fig.5.2 [18].
Fig.5.2a shows the generated output current waveform,
while the output voltage is described in Fig.5.2b.
Minimal distortion appears on the supply voltage and
current waveforms to the converter as can be seen in
Fig.5.2c. Input and output waveforms with such close
approximations to pure sinusoids are what make this
particular converter topology so desirable.
The self-commutating, PWM, matrix converter is
endowed with all the desirable cycloconverter
characteristics listed above. Unfortunately, the
configuration has not found wide utilization because of a
number of practical problems related to the bidirectional
switching elements, commutation, switch synchronization
and protection, as well as some intrinsic theoretical


Matrix converter waveforms, (a) Output current (30 Hz), (b) Out-
put voltage, (c) Supply voltage and input current (60 Hz). Chopping
frequency > 960 Hz.
Fig.5.2 Matrix Converter Waveforms [181

These limitations can be listed as follows:
1. Output frequencies are limited to a
fraction of the input frequency due
to the use of natural commutation.
2. The output voltage amplitude Vo
cannot exceed approximately 1/2 the
input voltage Vin, resulting in poor
semiconductor device utilization.
3. Input power factor cannot be lower
than the output power factor.
4. Switching instants need careful
control as the circuit is sensitive
to timing inaccuracies.
5.2 Switching
At present, until a bidirectional power switching
device is actually produced, each of the bidirectional
switches shown in Fig.5.1, must be realised by one of a

variety of switch arrangements. Several generalised
switch configurations are shown in Figs.5.3 and 5.4 [14].
From these figures it can clearly be seen that the high
device count, resulting in increased losses and greater
production costs, have worked against this particular
Another major problem associated with the
bidirectional switches of the matrix converter is that of
safe device switching. As was mentioned earlier the
cycloconverter has no DC link, and so is not protected
from AC line surges by the DC link inductor or capacitor.
This problem can be somewhat overcome by the use of a
diode clamping circuit. Fig.5.6 shows the general
arrangement of filters associated with the matrix
converter. Line side filter (Fin) and Load side filter
(Fout) are line conditioning filters whose size turns out
to be inversely proportional to the P.W.M frequency -
Fout can usually be omitted for inductive loads [18].
More importantly however, the cycloconverter has no
built-in protection to line surges produced internally by
a mismatch of the output device timing pulses. Overlap or
gaps in each timing instant will produce a current or
voltage surge across the device being switched, hence
requiring that the circuit employ complex timing controls

Fig.5.3 Matrix Converter Switch Realization [14]
using Bipolar Junction Transistors.

Fig.5.4 Bidirectional Switch Realization using [14]
(a) Gate Controlled Switches
(b) Thyristors with Forced Commutation Circuits

or energy dissipating snubber networks.
To more clearly understand this problem, a simple
circuit consisting of two switches is shown below.
^ v. C sw2
R-L Load
11 VJ
Fig.5.5 Switch Arrangement for Commutation Example
Suppose that switch sw^ is conducting the load
current i^as shown. If we then require that SW2 conduct
the load current, then we must know the exact instant at
which swj. opens and sw^ closes. In an ideal world switch
opening and closing could be considered to be an
instantaneous event, however practical realizations
dictate that a finite time, associated with switching and
circuit delays, be accounted for.
If swj. is still conducting when sw.2 is turned-on then
a short circuit current path will be established around
the loop. This short circuit can generate a large current
spike which could destroy the switches if they were not

5 25

sufficiently sized. Similarly if sw ^ is turned-off before
s2 is turned-on then there will be no path for the
inductive load current and potentially destructive
voltage spikes will be produced.
One way of overcoming this voltage and current spike
problem has been reported in [32]. The method discussed
in this reference relies on the intentional introduction
of deadtime between the conduction cycles of the devices
while a voltage limiting circuit dissipates energy
returned to the circuit by the load during the switching
Another recent paper [2], approaches the problem in a
more elegant manner using a technique called "staggered
commutation". This commutation method can only be applied
to converters using bidirectional switches realized from
antiparallel connected devices. Staggered commutation, as
the name implies, utilizes sequential commutation of the
segments of the bidirectional switches in such a way as
to allow only portions of the device to be switched at
any particular time. An explanatory example, shown below
in Fig.5.7, is based upon a recently presented paper
which approached the commutation problem in a very
similar manner to the staggered commutation approach [7].

Fig.5.7 Bidirectional Switch Arrangement
If the two switch segments 1A and IB are in the 'on'
position, then load current will flow through one of the
switches depending upon the currents direction. Both
segments of the other switch will normally remain 'off'.
However, if after some controlled interval we attempt to
turn off 1A and IB, and turn-on 2A and 2B, then the
commutation problem previously described is encountered.
To prevent the associated voltage and current anomolies
requires that the switches be switched in such a way as
to always present a non-hazardous switch combination to
the output. In this instance a non-hazardous combination

is simply one which will not allow or lead to a short
circuit or open circuit condition as indicated above.
A list of non-hazardous switch combinations is
presented below.
The various states are arbitrarily defined as
The 'on' state is denoted by 1 and the 'off state
0. Current iLis considered as +' for iL> 0 and flowing
into the source
1 1 1 0 0 + -
2 0 0 1 1 + -
3 1 0 0 0 +
4 0 1 0 0 -
5 0 0 1 0 +
6 0 0 0 1 -
7 1 0 1 0 +
8 0 1 0 1
Table 1. Hon-Hazardous Switching Combinations
As before, assuming that both switches 1A and IB are

initially 'on' and the load current i > 0 then the
swiching algorithm will be:
1. Step 1 .... 1 1 0 0
2. Step 3 Turn-off SW1B 1 0 0 0
3. Step 7 Turn-on SW2A .... 1 0 1 0
4. Step 5 Turn-off StflA 0 0 1 0
5. Step 2 Tun-on SW2B .... 0 0 1 1
Similarly if i <0 then the switching algorithm
1. Step 1 .... 1 1 0 0
2. Step 4 Turn-off SW1A 0 1 0 0
3. Step 8 Turn-on SW2B 0 1 0 1
4. Step 6 Turn-off SW1B 0 0 0 1
5. Step 2 Turn-on SW2A 0 0 1 1
The state transition diagram of the switching moments
for changing from switch 1 to switch 2 is shown below in

Fig.5.8 State Transition Diagram
A matrix converter circuit with this type of
commutation arrangement does not require snubbers simply
because it is impossible for the load current to change
sign during a single switching moment. The current may
fall to zero as dictated by the load or source but it
cannot change sign, hence there is no energy that needs
to be absorbed.
When a fixed delay is expected, as that associated
with transistor saturation, an introduced delay must be
programmed into the switching algorithm. This delay must
be slightly greater than the maximum possible delay
associated with the circuit and device to ensure correct
switching operation.

5.3 Control and Simulation
Another major problem associated with the matrix
converter is that of switching device utilization.
This concern is brought about by the realisation that
under a normal switching operation, the output voltage
amplitude (Vo) cannot exceed one-half the input voltage
amplitude (Vin) due to the lack of a supporting link
voltage. A graphical simulation of this problem is
shown in Fig.5.9. This under-utilization of the
electronic switching device power handling capabilities
simply compounds the high device count costs described
A review of the recent literature reveals however, that
this problem has to a great extent been overcome [2].
For the matrix converter circuit shown in Fig.5.1, it
can been shown in fact that the maximum amplitude
limitation for a nine switch three phase to three phase
converter is Vo = Y3 Vin for sinusoidal input and
output waveforms.
A nonrigorous mathematical justification is given below.
If a set of three phase input voltages are defined as
Vi = { vi Cos (w t + (n-1) _2n ) }
i 3 n=l