SPECTRUM CHARACTERIZATION OF AC/AC POWER CONVERSION SYSTEM

By
Emad Eldin Fathi Sherif

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Electrical Engineering
2012

ABSTRACT
SPECTRUM CHARACHTERIZATION OF AC/AC POWER CONVERSION SYSTEM
By
Emad Eldin Fathi Sherif
This Thesis proposes an analytical method based on three-dimensional Fourier integral to
obtain accurate spectra of both the switching functions and the synthesized output voltages and
input currents of matrix converter. The principle of this analytical is carried out in two steps. The
first step is that the spectra of the switching functions are derived based on three-dimensional
Fourier integral. The second step is that the spectra of the output voltages and input currents are
evaluated using a convolution operation in the frequency domain.
The challenges associated with the spectral analysis of matrix converter waveforms are
twofold. On one hand, the modulation signal contains both the input and output frequencies.
Unlike the third harmonic injection in the modulation functions, the input frequency and the
output frequency are typically independent from each other and will not form an integer ratio. On
the other hand, it is very common that the switching frequency or the carrier frequency is not
rational multiple of either the input frequency or the output frequency. These aforementioned
challenges make it a very challenge task to accurately characterize the spectra of matrix
converter waveforms through commonly resorted numerical methods such as fast Fourier
transform (FFT). The proposed analytical approach, which is an extension of the double Fourier
series expansion for the pulse width modulated waveforms of voltage source inverter will
provide an accurate solution to spectral analysis of matrix converters.

ACKNOWLEDGMENTS
I would like to express my most sincere gratitude to my advisor, Professor Bingsen
Wang, for his excellent guidance, encouragement, caring, patience and providing me with an
excellent atmosphere for doing research. His sparkling thoughts during our discussion not only
improved the quality of my research, but also helped me to develop my background and inspired
me to approach a new analytical method. I consider myself very fortunate for being able to work
with a very considerate and encouraging professor like him. I am sure that this research would
not have been possible without his support.
I would also like to express my deeply grateful to my committee members, Professor
Joydeep Mitra and Professor Feng Peng for participating in my final defense and taking their
precious time to consider my work.
In addition, I would like to express my gratitude and appreciation to all my friends who
have given me their support and encouragement during my studies at Michigan State University.
I am so grateful to the department of external studies in Libya for the generous
scholarship they provided for my master studies.
Finally, I am deeply indebted to my parents, brothers and sisters for their love and
support. I am very grateful to them for everything they have given me. This work is dedicated to
them.

iii

TABLE OF CONTENTS
LIST OF TABLES ....................................................................................................................... vi
LIST OF FIGURES .................................................................................................................... vii
CHAPTER 1: Introduction .......................................................................................................... 1
1.1
Background of AC/AC Power Conversion ...................................................................... 1
1.2

Matrix Converter Concept................................................................................................ 2

1.3

Modulation Techniques .................................................................................................... 4

CHAPTER 2: Matrix Converter Modulation Strategies .......................................................... 9
2.1
Existence Function ........................................................................................................... 9
2.2

Alesina and Venturini Method ....................................................................................... 11

2.3

Space Vector Modulation Method ................................................................................. 13

2.4

Duty-Cycle Space Vector Modulation Method .............................................................. 16

2.5

Summary ........................................................................................................................ 19

CHAPTER 3: Analytical Approach to Characterize Spectral Performance of Matrix
Converters ................................................................................................................................... 20
3.1
Fourier Series Analysis .................................................................................................. 21
3.2

Double Fourier Integral Analysis ................................................................................... 22

3.3

Triple Fourier Integral Analysis ..................................................................................... 27

3.4

Application of Triple Fourier Integral Analysis to PWM .............................................. 30

3.5

Summary ........................................................................................................................ 33

CHAPTER 4: Spectral Analysis of Matrix Converter Waveforms ....................................... 35
4.1
Modulation Process ........................................................................................................ 35
4.2

Analytical Solution to the Spectra of Switching Functions ........................................... 37

4.3

Analytical Spectra of Output Voltages and Input Currents............................................ 45

4.4

Analytical and Numerical Results .................................................................................. 47

4.5

Total Harmonic Distortion (THD) and Weighted Total Harmonic Distortion (WTHD)54

4.6

Summary ........................................................................................................................ 60

CHAPTER 5: Conclusion and Future Work ........................................................................... 62
5.1
Conclusions .................................................................................................................... 62
iv

5.2

Future Work ................................................................................................................... 62

BIBLIOGRAPHY ..................................................................................................................... 64

v

LIST OF TABLES
Table 2.1: Switching configuration for three-phase matrix converters. ....................................... 15
Table 4.1: Illustration multiple unit-cube for one switching cycle. .............................................. 47
Table 4.2: THD magnitudes of three phase output voltages. ........................................................ 57
Table 4.3: THD magnitudes of three phase input currents. .......................................................... 57
Table 4.4: WTHD magnitudes of three phase output voltages. .................................................... 60
Table 4.5: WTHD magnitudes of three phase input currents. ...................................................... 60

vi

LIST OF FIGURES
Figure 1.1: Schematic of direct matrix converter. .......................................................................... 2
Figure 1.2: Four-quadrant switch. ................................................................................................... 3
Figure 1.3: Schematic of indirect matrix converter. ....................................................................... 5
Figure 2.2: Geometrical representation for the validity for mu . .................................................. 17
Figure 3.1: Illustration of the unit cube. ....................................................................................... 32
Figure 4.1: Analytical and numerical spectra of switching functions h11 , h12 and h13 . ................. 48
Figure 4.2: Analytical and numerical spectra of switching functions h21 , h22 and h23 . ................ 49
Figure 4.3: Analytical and numerical spectra of switching functions h31 , h32 and h33 . ................ 50
Figure 4.4: Analytical and numerical spectra of switching functions Vo1 , Vo 2 and Vo 3 . ................ 52
Figure 4.5: Analytical and numerical spectra of switching functions I i1 , I i 2 and I i3 . ................... 53
Figure 4.6: (a) THD versus voltage gain of output voltages (phase A-C), (b) (phase B). ............ 55
Figure 4.7: (a) THD versus voltage gain of input currents (phase A-C), (b) (phase B). .............. 56
Figure 4.8: (a) WTHD versus voltage gain of output voltages (phase A-C), (b) (phase B). ........ 58
Figure 4.9: (a) WTHD versus voltage gain of input currents (phase A-C), (b) (phase B). .......... 59

vii

CHAPTER 1
Introduction
1.1

Background of AC/AC Power Conversion

The family of AC/AC power converters can be categorized into two types which are AC-toAC converters without DC-link (direct matrix converter) and AC-to-AC converters with DC-link
(indirect matrix converter).
a) AC-to- AC converter without DC-link:
This type of converter uses only one conversion stage. In other word, input AC voltages
are connected directly into output AC voltages through proper operation of four-quadrant
or voltage-and-current bidirectional switches.
b) AC-to- AC converter with DC-link:
This type of converter uses two conversion stages. First, the input AC voltages are
converted into intermediate DC voltage and this stage is called rectifier stage. Then, the
intermediate DC voltages are converted into output AC voltages and this stages is called
inverter stage.
Each converter topology has its own advantages and disadvantages and the choice of the
converter depends on the requirement of the application.

1

1.2

Matrix Converter Concept

The matrix converter that was first introduced is a direct AC/AC power converter. It consists
of nine four-quadrant or bidirectional switches which are used to connect directly the three-phase
power supply to a three-phase load without using energy storage elements or DC-link as shown
in Figure 1.1. Each four-quadrant switch employed in the matrix converter is typically
implemented from two two-quadrant switches as illustrated in Figure 1.2.

V i1
Vi 2

Vi3

I i1

S11

S 31

S22

S 32

S13

I i3

S 21

S12

Ii2

S 23

S 33

I o1

Io2

V o1
Figure 1.1 Schematic of direct matrix converter.

2

I o3

Vo 2

Vo 3

Figure 1.2 Four-quadrant switch.
The nine bidirectional switches are modulated to generate the desired output waveform based on
the input supply voltages and the demanded output voltages.
The matrix converter is usually fed on the input side by a three-phase voltage source. Each
phase is connected with an input filter capacitor to ensure continuous input voltage waveforms.
In addition, the matrix converter is connected to an inductive load which leads to continuous
output current waveforms.
The matrix converter has received an increased amount of interest in the last three decades.
Furthermore, it has been considered intensely as an alternative to conventional indirect power
converter systems due to its several desirable features such as:
•

Sinusoidal input and output current waveforms

•

Generation of load voltage with variable magnitude and frequency

•

Controllable input power factor for any load

•

Simple and compact power circuit

•

Inherent regeneration capability

3

Despite these advantages, there are some reasons for rejecting the matrix converter in
industrial applications. The major disadvantages is the limitation of the voltage gain ratio up to
0.866. Consequently, electrical motors or any standard device connected as load to the matrix
converter do not operate at their nominal rate voltage. Another drawback is that the number of
the semiconductor switches is more than the number used in a dc-link converter. Therefore, the
cost will be more expensive. In addition, with the absence of the dc-link, there is no decoupling
between the input and output sides. Hence, and distortion in the input voltage is reflected in the
output voltage at different frequencies. As a consequence, sub-harmonics can be generated.
Furthermore, the control of the bidirectional switches of the matrix converter is very
complicated.

1.3

Modulation Techniques

Pulse width modulation (PWM) strategies for matrix converters have received significant
research effort recently [1]- [23]. The modulation research on matrix converter starts with the
work of Venturini and Alesina [3]. The authors provided a mathematical approach to describe
how the low frequency behaviors of the voltages and currents are generated at the load and the
input. This strategy allows the full control of the output voltages and input power factor. This
approach is also known as a direct matrix converter (DMC). Unfortunately, the drawback of this
strategy is that the maximum voltage transfer ratio (q) which is known as the ratio of the
magnitude of the output voltage to the magnitude of the input voltage, was limited to one half.
Later, the same authors proposed an improved method in 1989 which increased the voltage
transfer ratio to 0.866 by utilizing the third harmonic injection technique in the input and output

4

voltage waveforms. The third harmonic injection has been extended with the input power factor
control which leading to a very powerful modulation strategy (optimum AV method) [4].
A well-known technique for controlling the fictitious rectifier and inverter was proposed in
[8], [24]. The authors split the matrix converter into a two-stage system, namely a three-phase
rectifier and a three-phase inverter connected together through a fictitious DC-link as shown in
Figure 1.3. First, the input voltage rectified to create a fictitious DC-link. Then, inverted at the
required output frequency. This technique was defined as indirect matrix converter (IMC).
Rectifier
Stage

Inverter
Stage

DC-link

Three
Phase
Supply

Three Phase Load

Figure 1.3 Schematic of indirect matrix converter.
Space vector modulation (SVM) approach for matrix converters is another technique which
was first proposed by Huber et al in 1989 [25], [26] to control only the output voltages. The
space vector modulation technique was adapted to the matrix converter by employing a basic
concept of indirect method. This technique was successively developed in order to achieve the

5

full control of the input power factor, to fully utilize the input voltages and to improve the
modulation performance. Furthermore, it allows the maximum gain ratio of 0.866.
A general solution of the modulation problem for matrix converters based on the concept of
Duty-Cycle Space Vector was presented in [19]. This method simplifies the study of the
modulation strategy. In addition, it has been demonstrated that three degrees of freedom are
available in defining the modulation strategy, allowing for the control of the instantaneous values
of the output voltages and input power factor to be obtained. Thus, this technique has been
considered as the best solution for achieving the highest voltage gain ratio.
In principle, the main task of PWM schemes is to create trains of switched pulses. The major
problem with these trains of switched pulses is that they have unwanted harmonic components
which should be reduced. Hence, for any PWM scheme, the main objective which can be
identified is to determine the most effective way of arranging the switching processes to reduce
unwanted harmonic distortion, switching losses, or any other performance criterion.
The analysis of PWM schemes has been the subject of intensive research for several
decades. Some of this work has investigated the harmonic components which are produced from
the modulation process. It is quite complex to accurately determine the spectrum of PWM
waveforms that are typically present in any converter and it is often resorted to a fast Fourier
transform (FFT) analysis. It is known that this approach does not yield accurate results if the
ratio between the switching frequency to fundamental frequency is not integer. In contrast, a
theoretical analysis which identifies the exact harmonic components when various PWM
strategies are compared against each other was presented in [33]. This analysis is based on

6

double Fourier integral analysis in two variables ( high frequency carrier signal with low
frequency sinusoidal signal ) has been well developed in [16], [29]- [31].
The challenges associated with the spectral analysis of matrix converter waveforms are
twofold. On one hand, the modulation signal contains input and output frequencies. Unlike the
intentional third harmonic injection in the modulation functions of voltage source inverters
(VSI), the input and output frequencies in the matrix converter are typically independent from
each other and will not form an integer ratio. On the other hand, it is very common that the
switching frequency or the carrier frequency is not rational multiple of either the input frequency
or output frequency [33]. These aforementioned challenges make it a very difficult task to obtain
an accurate spectrum of the matrix converter waveforms through commonly resorted numerical
methods such as fast Fourier transform (FFT).
Thus, the research work here proposes a new analytical method based on three dimensional
Fourier integral which exactly identifies accurate spectra of the switching functions and
synthesized terminal quantities of the matrix converter. The proposed method is an extension of
the double Fourier integral analysis of the output voltage waveforms of the voltage source
inverter [33]. The scope of the thesis is organized as follows:
•

Chapter 2, presents the concept of the existence function and summarizes the most
important modulation strategies for matrix converter.

•

In chapter 3, a general review of Fourier series analysis is given first. Then, the
formulation of double Fourier integral analysis is presented. Finally, the triple Fourier
integral analysis has been explained follow by its application to the pulse width
modulation.

7

•

In chapter 4, detailed spectral analysis of the matrix converter based on three
dimensional Fourier integral to characterize the accurate spectra of the switching
functions and synthesized output voltages and input currents of matrix converter which
was proposed by Alesina and Venturini in 1981 [3] and verified with FFT.

•

In chapter 5, the work presented in this thesis is summarized. Topics for further
investigation are suggested as future work.

8

CHAPTER 2
Matrix Converter Modulation Strategies
In this chapter, a brief introduction to the existence function which used to provide a
mathematical form for expression switching function is given first. Then, the main modulation
strategies which used in matrix converters are considered (the Alesina and Venturini modulation
methods are reviewed, SVM algorithm for matrix converter is presented and the concept of duty
cycle space vector is considered).

2.1

Existence Function

The existence functions which was proposed by Wood [32], provides a mathematical
expression for describing the switching patterns. The existence function for a single switch when
it is closed, the existence function has a value 1 and when it is opened, the existence function has
a value 0.
For the matrix converter which is shown in Figure 1.1, the existence function Suv (t ) for each
of the switches can be expressed as follows:

1 when Suv (t ) is closed
Suv (t ) = 
0 when S uv (t ) is opened

(2.1)

where u ∈{1,2,3} , v ∈ {1,2,3} .

9

In order to prevent the short circuit in the capacitive input as well as the open circuit in the
inductive output, only one switch on each output phase must be closed. Therefore,
3

3

∑ S1v = ∑ S 2v =

v =1

v =1

3

∑ S3v = 1

(2.2)

v =1

In Figure 1.1, the input voltages Viu (t ) and output currents I ov (t ) are assumed three-phase
balanced with peak value Vi and I o respectively as follows:
Vi1 (t ) 




V (t )  = V
Viu (t ) = i 2
i





V
 i3(t ) 
 I o1 (t ) 




I ov (t ) =  I o 2 (t )  = I o




I

 o3(t ) 

cos (ωi t )





cos (ωi t − 2π 3) 




cos (ωi t + 2π 3) 



(2.3)


cos (ωo t )




cos (ωo t − 2π 3) 




cos (ωo t + 2π 3) 



(2.4)

where ωi is the angular frequency of the input voltages and ωo is the angular frequency of the
output currents.
The instantaneous voltage and current relationships are related to the state of the nine switches
and can be written in a matrix form as:

10

Vo1 (t ) 




V (t )  =
 o2 



V
 o3(t ) 

 S11 (t )


 S 21 (t )


 S 31 (t )


S12 (t )

 I i1 (t )   S11 (t )

 

 
 I (t )  =  S (t )
 i 2   12

 
I
 
 i3(t )   S13 (t )

S 21 (t )

S 22 (t )
S 32 (t )

S 22 (t )
S 23 (t )

S13 (t ) 


S 23 (t ) 


S 33 (t ) 


S 31 (t ) 


S 32 (t ) 


S 33 (t ) 


Vi1 (t ) 




V (t ) 
 i2 



V
 i3(t ) 
 I o1 (t ) 




 I (t ) 
 o2 


I

 o3(t ) 

(2.5)

(2.6)

Equations (2.5) and (2.6) are the basic of all modulation methods which consist in selecting
appropriate combinations of switches to generate the desired output voltages.

2.2

Alesina and Venturini Method

Alesina and Venturini first proposed the high frequency synthesis technique for direct matrix
converter [3], [4]. The high frequency technique allows the utilization of low frequency to
calculate the existence functions for each switch in the matrix converter. Therefore, the aim of
using Alesina and Venturini methods is to find the modulation matrix which satisfies the
following equations:
Vo1 (t )  m11 (t )

 

 
V (t )  = m (t )
 o 2   21

 
V
 
 o3(t )  m31 (t )

m12 (t )
m22 (t )
m32 (t )

m13 (t ) 


m23 (t ) 


m33 (t ) 


Vi1 (t ) 




V (t ) 
 i2 


V

 i3(t ) 

11

(2.7)

 I i1 (t ) 




 I (t )  =
 i2 



I
 i3(t ) 

m11 (t )


m12 (t )


m13 (t )


m21 (t )
m22 (t )
m23 (t )

m31 (t ) 


m32 (t ) 


m33 (t ) 


 I o1 (t ) 




 I (t ) 
 o2 



I
 o3(t ) 

(2.8)

where muv (t ) is the duty cycle of the nine switches Suv (t ) with 0 ≤ muv (t ) ≤ 1 .
In order to prevent a short circuit from the input side, the duty cycles matrix muv (t ) must
satisfy the three following conditions:
mu1 (t ) + mu 2 (t ) + mu 3 (t ) = 1

(2.9)

Alesina and Venturini proposed a control method for a nine-switch DMC using a
mathematical approach to generate the desired output waveform. In [3], the first modulation
strategy allows the control of the output voltages and input power factor. This strategy can be
summarized in the following equation
1
2π 
2π


muv (t ) = 1 + 2q cos  ωo t − (u − 1)
 cos  ωi t − (v − 1)
3
3 
3







(2.10)

where q is the voltage transfer ratio. The solution given in (2.10) is valid for unity input power
factor and the maximum output to input voltage ratio is limited to one half. This value represents
the major drawback of this modulation strategy.
In later work by the same authors [4], it was shown that the maximum voltage transfer
ratio increased up to 0.866 by the third harmonic injection techniques. This value represents an
intrinsic limitation of three-phase to three-phase matrix converters balanced supply voltages. The

12

modulation strategy in this case is valid for unity input power factor and can be described in the
following relationship:

1
2π

muv (t ) = 1 + 2q cos  ωi t − (v − 1)
3
3



2π

 cos  ωo t − (u − 1)
3




1
 1
cos(3ωi t ) 
 − cos(3ωo t ) +
2 3
 6


1 2 
2π 
2π


− 
q cos  4ωi t − (v − 1)  − cos  2ωi t + (v − 1)
3 3 3 
3 
3



2.3

 
 
 

(2.11)

Space Vector Modulation Method

The first matrix converter using SVM technique was proposed by Huber et al in 1989 to
control three-phase to three-phase matrix convert [25], [26]. These methods divided the matrix
converter into two fictitious converters (rectifier and inverter) as shown in Figure 1.3. The
rectifier section is controlled as a full-bridge diode rectifier and the space vector modulation is
applied only to inverter section to control the output voltages. The application of space vector
modulation is presented in the literature [27], [28]. The space vector modulation shows a better
performance than the carried-based PWM control method in terms of output voltage harmonic
distortion.

The instantaneous space vector x representation of output voltage and input current can be
expressed as follows:

x=

[

2
x1 + x2 e j 2π 3 + x3 e j 4π 3
3

]

(2.12)

where x1 , x2 and x3 are three time-varying variables which represent the output phase voltages
and input line currents.
13

The zero space vector xo can be determined by the following relationship:

1
xo = ( x1 + x2 + x3 )
3

(2.13)

The inverse transformation of (2.12) and (2.13) can be written as follows:

x
xu = o + x e j (u −1)2π 3
3

(2.14)

where u ∈{1,2,3} .
In three-phase matrix converters, there are twenty seven possible switching configurations
and only twenty one can be applied in the space vector modulation algorithm as shown in Table
2.1 which can be categorized into three groups:

Group I. consists of eighteen switching combinations which determine the output voltage and
input current vectors that have fixed directions. The magnitude of these vectors depend on the
instantaneous values of input line-to-line voltages and output currents. These switching
combinations are called active configurations.

Group II. consists of three switching combinations which determine zero output voltage and
input current vectors. The three output phases are connected to the same input phase. These
switching combinations are named zero configurations.

Croup III. Consists of six switching combinations which have each output phase is connected to
a different input phase . The output voltage and input current vectors have variable directions,
but cannot be used to synthesis the reference vectors.

14

Table 2.1 Switching configuration for three-phase matrix converters.
Switching
configuration

vo

Switches On

αo

βi

ii

+1

S11

S 22

S 32

2v12i 3

0

2io1

−1

S12

S 21

S 31

− 2v12i 3

0

− 2io1

+2

S12

S 23

S 33

2v23i 3

0

2io1

−2

S13

S 22

S 32

− 2v23i 3

0

− 2io1

+3

S13

S 21

S 31

2v31i 3

0

2io1

−3

S11

S 23

S 33

− 2v31i 3

0

− 2io1

3

7π 6

+4

S12

S 21

S 32

2v12i 3

2π 3

2io 2

3

−π 6

−4

S11

S 22

S 31

− 2v12i 3

2π 3

− 2io2

+5

S13

S 22

S 33

2v23i 3

2π 3

2io 2

−5

S12

S 23

S 32

− 2v23i 3

2π 3

− 2io2

+6

S11

S 23

S 31

2v31i 3

2π 3

2io 2

−6

S13

S 21

S 33

− 2v31i 3

2π 3

− 2io2

+7

S12

S 22

S 31

2v12i 3

4π 3

2io3

−7

S11

S 21

S 32

− 2v12i 3

4π 3

− 2io3

+8

S13

S 23

S 32

2v23i 3

4π 3

2io3

−8

S12

S 22

S 33

− 2v23i 3

4π 3

− 2io3

+9

S11

S 21

S 33

2v31i 3

4π 3

2io3

−9

S13

S 23

S 31

− 2v31i 3

4π 3

− 2io3

01

S11

S 21

S 31

0

-

0

-

02

S12

S 22

S 23

0

-

0

-

03

S13

S 23

S 33

0

-

0

-

15

3

3
3
3
3

3
3
3
3
3
3
3
3
3
3
3

−π 6
−π 6

π 2
π 2
7π 6

−π 6

π 2
π 2
7π 6
7π 6
−π 6
−π 6

π 2
π 2
7π 6
7π 6

2.4

Duty-Cycle Space Vector Modulation Method

A general solution to the modulation problem for matrix converters based on the concept of
Duty Cycle Space Vector, was presented in [19]. The duty-cycle space vector can be developed
by utilizing space vector modulation. This method simplifies the study of the modulation
strategy. In addition, it has been demonstrated that three degrees of freedom are available in
defining the modulation strategy, allowing for the control of the instantaneous values of the
output voltages and input power factor. Thus, this technique has been considered as the best
solution for achieving the highest voltage gain ratio.
The nine duty cycle muv can be represented by the duty cycle space vector mu by the
following transformation:

mu =

[

2
mu1 + mu 2 e j 2π 3 + mu 3 e j 4π 3
3

]

(2.15)

Taking into account conditions in (2.9), the inverse transformation of the equation (2.15) is:

1
muv = + mu e j (v −1) 2π 3
3

(2.16)

By considering the constraints of the modulation function 0 ≤ muv ≤ 1 , the duty cycle space
vector mu is a geometrically illustrated in d-q plane and is always inside the equilateral triangle
which represents in Figure 2.1.

16

mu 2 = 1

mu3 = 0
mu1 = 1

12

mu1 = 0
mu 2 = 0

12

mu 3 = 1

−1 3

23

Figure 2.1 Geometrical representation for the validity for mu .

The position of the duty cycle space vector mu inside the triangle is determined by the switching
commutations of S uv in a cycle period.
It is noted that the output voltage and input current space vector can be written by using the
equation (2.12) as follows:

vo =

ii =

[

2
vo1 + vo 2 e j 2π 3 + vo3 e j 4π 3
3

[

2
ii1 + ii 2 e j 2π 3 + ii3 e j 4π 3
3

]

(2.17)

]

(2.18)

By substituting the equations (2.7), (2.8) and (2.16) into (2.17) and (2.18), the output voltage and
input current space vector can be rewritten as follows:

17

[

]

[

v
v*
*
vo = i m 1 + m * e j 2π 3 + m * e j 4π 3 + i m1 + m2 e j 2π 3 + m3 e j 4π 3
2
3
2
2

[

] [

i
i* *
ii = i m1 + m2 e j 4π 3 + m3 e j 2π 3 + i m 1 + m * e j 4π 3 + m * e j 2π 3
2
3
2
2

]

]

(2.19)

(2.20)

Equations (2.19) and (2.20) suggest three new variables md , mi and m0 which can be defined
as:

md =

[

1
m1 + m2 e j 2π 3 + m3 e j 4π 3
3

[

mi =

1
m1 + m2 e j 4π 3 + m3 e j 2π 3
3

m0 =

1
m1 + m2 + m3
3

[

]

(2.21)

]

(2.22)

]

(2.23)

where md , mi and m0 are considered as direct, inverse and zero components of the duty cycle
space vectors respectively.
The inverse transformation of (2.21), (2.22) and (2.23) can be expressed by:
mu = md e j (1− u ) 2π 3 + mi e j (u −1) 2π 3 + m0

(2.24)

By substituting (2.24) into (2.19) and (2.20), the output voltage and input current space vectors
can be defined as:

vo =

[

3
vi × m * + v *× md
i
i
2

]

(2.25)

18

ii =

[

3
io × mi + i * × md
o
2

]

(2.26)

Equations (2.25) and (2.26) represent the output voltage and input current space vector of threephase matrix converters in a compact form.

2.5

Summary

This chapter has reviewed the concept of the existence function and described the modulation
strategies of three-phase matrix converters.
Alesina and Venturini approach presented a general waveform synthesis technique in [3].
The major drawback of this approach is the maximum output to input voltages ratio was only
limited to one half. Later work by the same authors increased the ratio to 0.866 which is the
maximum theoretical limit using high frequency synthesis [4]. In addition, the space vector
modulation approach is based on the instantaneous output voltage and input current space
vectors. By using this approach, the switching pattern will be defined by the switching
configuration sequence. This approach was developed and introduced a new concept of duty
cycle space vector, which simplifies the study of modulation strategies of three-phase matrix
converters.

19

CHAPTER 3
Analytical Approach to Characterize Spectral
Performance of Matrix Converters
PWM is one of the cornerstone of switched converter systems. In addition, it has been
become the subject of intensive research for several decades. A vital part of this work has
investigated the harmonics that are inevitably produced as part of the modulation process.
In general, the harmonic frequency components of the PWM waveforms is quite
complex to characterize and it is often obtained by using FFT analysis to identify their
magnitudes. However, it is known that this approach does not yield accurate results if the ratio
between the fundamental frequency and carrier frequency is not integer. In contrast, to
characterize the exact harmonic components of the PWM waveforms without relying on the
numerical method, an analytical approach has been considered in [33]. This approach is termed
double Fourier integral analysis, which exactly identifies the correct harmonics magnitude when
various PWM strategies are compared against each other.
A new analytical approach for identifying the harmonic components of PWM waveforms
for AC/AC converters has been proposed. This analytical approach is termed triple Fourier
integral analysis, which is a natural extension of the double Fourier integral analysis that has
been developed for VSI waveforms [33]. However, in this chapter, the Fourier series analysis is

20

reviewed first. Then, the formulation of the double Fourier analysis is presented. Finally, the
triple Fourier analysis is proposed follow by its application to the pulse width modulation.

3.1

Fourier Series Analysis

The principle of Fourier series is that any periodic function f(t) can be expressed as an
infinite summation of sinusoidal harmonic components as follows:
a0 ∞
f (t ) =
+ ∑ [an cos( nωo t ) + bn sin( nωo t )]
2 n =1

(3.1)

where a0 , an and bn are called the Fourier coefficients and ωo is the fundamental frequency of
the periodic function.
The Fourier coefficients can be found as:

an =

bn =

1

π

1

π

π

∫ f (t ) cos(nωo t ) dωot

for

n=0,1,2,3…

(3.2)

for

n=1,2,3…

(3.3)

−π

π

∫ f (t ) sin(nωot ) dωot

−π

The cosine and sine functions in (3.2) and (3.3) can be expressed in Euler formula as follows:

e jnωo t + e − jnωo t
cos(nωo t ) =
2

(3.4)

e jnω o t − e − jnωo t
2j

(3.5)

sin(nωo t ) =

21

By substituting (3.4) and (3.5) into (3.2) and (3.3) respectively, the Fourier series can be
expressed in a complex form as follows:
∞

∑ cn × e jnωo t

f (t ) =

(3.6)

n = −∞

The complex coefficients are then given by:

cn =

1
2π

π

∫ f (t ) × e

− jnω o t

(3.7)

−π

The complex coefficient cn is related to the real-numbered an and bn as the following:

a − jbn
cn = n
2

(3.8)

OR

an = an + a(− n)
bn = j (bn − b(− n) )

3.2

(3.9)

Double Fourier Integral Analysis

The challenge associated with the non-periodicity of PWM waveform was solved by the
mathematical approach called double Fourier integral analysis. The double Fourier integral
analysis is utilized as an analytical approach to identify the harmonic components of carrierbased PWM. This analytical approach was originally proposed for communication system by
Bennet [34] and Black [35], later by Bowes and Bird [36] for power converters.

22

The analysis process assumes the existence of two time variables as follows:

x(t ) = ωct + θc
y(t ) = ωot + θo

(3.10)

where x(t ) and y(t ) being defined as the time variation of the high frequency carrier waveform
and low-frequency reference waveforms respectively. ωc is the angular frequency of the carrier
signal while ωo is the angular frequency of the low-frequency modulation signal. θc and θo are
the phase angles of the carrier and low-frequency signals respectively and are assumed to be
zero.
Let us assume that the triangular signal c(x) is considered for this analysis. Then, the
mathematical expression for the carrier signal c(x) is:

c( x) =

1

π

arccos(cos( x))

(3.11)

Furthermore, the modulation function m(y) is defined by

m( y ) =

1 + M cos( y )
2

(3.12)

where M is the modulation index with range 0 < M < 1 . Then, the switching function f ( x, y) can
be determined by the comparison of modulation function m(y) against carrier signal c(x) as
follows:

f ( x, y ) = Φ [m( y ) − c( x)]

(3.13)

where Φ(⋅) is defined by:

23

1 if u > 0
Φ (u ) = 
0 if u ≤ 0

(3.14)

It can be seen from (3.14) that the switching function f ( x, y) takes values of 1 or 0 for any
combination of x and y.
Let a double variable function f ( x, y) be periodic in both x and y directions. It is
further assumed that x and y are phase-angle variables and the period in both directions is 2π .

f ( x, y) = f ( x + 2π , y) = f ( x, y + 2π )

(3.15)

The Fourier expansion can be conducted in two steps. First, the function f ( x, y) is expanded in

x direction while y being kept a constant. In complex form, the Fourier series with coefficients
being functions of y will read
f ( x, y ) =

∑ [Fm ( y ) × e jmx ]
∞

(3.16)

m = −∞

where the complex coefficients are determined by

1
Fm ( y ) =
2π

π π

∫ ∫ f ( x, y ) × e

− jmx

(3.17)

dx

−π −π

It is obvious that Fm ( y) is periodic since f ( x, y ) is periodic in direction.
Fm ( y + 2π ) =

1
2π

1
Fm ( y + 2π ) =
2π

π

∫ f ( x, y + 2π ) × e

− jmx

dx

−π

π

∫ f ( x, y ) × e

− jmx

dx

−π

24

Fm ( y + 2π ) = Fm ( y )

(3.18)

Hence, the complex coefficient Fm ( y) can be expanded into

∑ [Fmn × e j (mx + ny) ]
∞

Fm ( y ) =

(3.19)

n = −∞

where
π

1
Fmn =
2π

∫ Fm ( y ) × e

− jny

(3.20)

dy

−π

Substituting the equation (3.19) into (3.16), the double function variable f ( x, y) can be rewritten
as following:

∑ ∑ [Fmn × e j (mx + ny) ]
∞

f ( x, y ) =

∞

(3.21)

m = −∞ n = −∞

where

Fmn =

π π

1
2

∫ ∫ f ( x, y ) × e

4π −π −π

− j ( mx + ny )

(3.22)

dxdy

Consequently, the double Fourier can be expressed in real-numbered coefficients as following:
∞
A
f ( x, y ) = 00 + ∑ [ A0n cos( ny ) + B0n sin( ny ) ]
2
n =1

+

∞

∑ [Am0 cos(mx) + Bm0 sin(mx)]

m =1

25

∞

+

∞

∑ ∑ [Amn cos(mx + ny) + Bmn sin(mx + ny)]

(3.22)

m =1 n = −∞
n≠0

where the Fourier coefficients Amn and Bmn are determined by the following double integral

Amn =

Bmn =

1

π π

∫ ∫ f ( x, y ) cos(mx + ny )dxdy

(3.23)

2π 2 −π −π

1

π π

∫ ∫ f ( x, y ) sin(mx + ny )dxdy

(3.24)

2π 2 −π −π

The complex coefficient Fmn is related to the real-numbered Amn and Bmn as the following

A − jBmn
Fmn = mn
2

(3.25)

OR

Amn = Fmn + F(−m)(− n)
Bmn = j ( Fmn − F(−m)(−n) )

(3.26)

Replacing x by ωct + θc and y by ωot + θo , the equation (3.22) can be expressed in the timevarying form as:
∞
A
f (t ) = 00 + ∑ [A0n cos( n(ωot + θ o )) + B0n sin( n(ωot + θ o ))]
2
n =1

26

+

∞

∑ [Am0 cos(m(ωct + θc )) + Bm0 sin(m(ωct + θc ))]

m =1

 Amn cos(m(ωc t + θ c )) + n(ωot + θ o )) 

+ ∑ ∑ 


m =1 n = −∞ + B sin(m(ω t + θ ) + n(ω t + θ ))
mn
c
c
o
o 
n≠0 
∞

∞

(3.27)

where m and n are the carrier and baseband index variable respectively.
In (3.27), the first term represents the dc components of PWM waveform.The first
summation term defines the output fundamental low-frequency waveform and it is called
baseband harmonics. The second summation term corresponds to the carrier waveform
harmonics and it is called carrier harmonics. The final double summation term defines the
sideband harmonics around the carrier harmonic components.

3.3

Triple Fourier Integral Analysis

The triple Fourier integral analysis is a natural extension of the double Fourier integral
analysis that can accurately characterize the spectrum of the PWM waveforms for the matrix
converters. Without loss of generality, the sine-triangle naturally sampled modulation is
considered for this analysis as well. The analysis process assumes the existence of three time
variables:

x(t ) = ωct + θc
y (t ) = ω1t + θ1
z (t ) = ω 2t + θ 2

(3.28)

27

where x(t ) , y (t ) and z (t ) being defined as the time variation of the high frequency carrier
waveform and two low-frequency reference waveforms respectively. ωc is the angular
frequency of the carrier signal whereas ω1 and ω2 are two independent angular frequencies of
the two low-frequency modulation signals. θc is the phase angle of the carrier frequency while

θ1 and θ 2 are the phase angles of the two low-frequency modulation signals and all the phase
angles are assumed to be zero.
Let a triple variable function f ( x, y, z ) be periodic in x , y and z directions. It is further
assumed that x , y and z are phase-angle variables and the period in all directions is 2π .
Consequently,

f ( x, y, z ) = f ( x + 2π , y, z ) = f ( x, y + 2π , z ) = f ( x, y, z + 2π )

(3.29)

With reference to the double Fourier expansion, the triple Fourier expansion is possible. Since
the assumption of f ( x, y, z ) be periodic in x , y and z directions, the triple Fourier expansion
can be written as the following:

∑ ∑ ∑ [Fkmn × e j (kx + my + nz ) ]
∞

f ( x, y, z ) =

∞

∞

(3.30)

k = −∞ m = −∞ n = −∞

where

Fkmn =

π π π

1
3

∫ ∫ ∫ f ( x, y , z ) × e

8π −π −π −π

− j ( kx + my + nz )

dxdydz

Therefore, the triple Fourier can be expressed in real-numbered coefficients as following:

28

(3.31)

∞
A
f ( x, y , z ) = 000 + ∑ [A0m 0 cos(my ) + B0m0 sin( my )]
2
m =1

∞

∑ [A00n cos(nz ) + B00n sin(nz )]

+

n =1

∞

∑ [Ak 00 cos(kx) + Bk 00 sin(kx)]

+

k =1

∞

+

∞

∞

∑ ∑ ∑ [Akmn cos(kx + my + nz) + Bkmn sin(kx + my + nz)]

(3.32)

k =1 m = −∞ n = −∞
m≠0 n≠0

where the Fourier coefficients Akmn and Bkmn are determined by the following triple integral

Akmn =

Bkmn =

1

π π π

∫ ∫ ∫ f ( x, y, z ) cos(kx + my + nz )dxdydz

8π 3 − π − π − π

1

(3.33)

π π π

∫ ∫ ∫ f ( x, y, z ) sin(kx + my + nz )dxdydz

8π 3 − π − π − π

(3.34)

The real coefficients Akmn and Bkmn are related to the complex coefficient Fkmn as the
following:

Akmn = Fkmn + F(−k )(−m)(−n)
Bkmn = j ( Fkmn − F(− k )(− m)(− n) )

(3.35)

A
− jBkmn
Fkmn = kmn
2

(3.36)

29

Replacing x by ωct + θc , y by ω1t + θ1 and z by ω2t + θ 2 the equation (3.32) can be expressed
in the time-varying form as:
∞
A
f (t ) = 000 + ∑ [A0 m0 cos( m(ω1t + θ1)) + B0 m0 sin( m(ω1t + θ1 ))]
2
m =1

∞

+ ∑ [A00n cos( n(ω2t + θ 2 )) + B00n sin( n(ω2t + θ 2 ))]
n =1

+

∞

∑ [Ak 00 cos(k (ωct + θc )) + Bk 00 sin(k (ωct + θc ))]

k =1

 Akmn cos(k (ωc t + θ c ) + m(ω1t + θ1 ) + n(ω2t + θ 2 )) 

+∑ ∑ ∑ 


k =1 m = −∞ n = −∞ + B
sin(k (ωc t + θ c ) + m(ω1t + θ1 ) + n(ω2t + θ 2 ))
kmn

m≠0 n≠0 
∞

∞

∞

(3.37)

where k is the carrier index variable while m and n are two baseband index variables.
The first term of the equation (3.37) represents the dc components of PWM waveform.The
first and second summation terms define the two output fundamental low-frequency waveforms
and they are called baseband harmonics. The third summation term corresponds to the carrier
waveform harmonics and it is called carrier harmonics. The final triple summation term defines
the sideband harmonics around the carrier harmonic components.

3.4

Application of Triple Fourier Integral Analysis to PWM

For the PWM process in matrix converters, there is one high-frequency signal called carrier
signal and two low-frequency components are named modulation signals. The analysis process
assumes the existence of three time variables:
30

x(t ) = ωct + θc
y(t ) = ωm1t + θ m1
z (t ) = ωm2t + θ m2

(3.38)

where x(t ) , y (t ) and z (t ) being defined as the time variation of the high frequency carrier
waveform and two low-frequency reference waveforms respectively. ωc is the angular
frequency of the carrier signal whereas ωm1 and ωm2 are two independent angular frequencies
of the two low-frequency modulation signals. θc is the phase angle of the carrier frequency
while θ m1 and θ m 2 are the phase angles of the two low-frequency modulation signals and all the
phase angles are assumed to be zero.
The comparison between the carrier signal c( x) and modulation function m( y, z ) will
give rise to the switching function. The switching function h( x, y, z ) takes the value one when
the modulation signal is greater than the carrier signal while it takes the value zero when the
modulation signal is less than the carrier signal. The mathematical description of the modulation
process can be written as follows:

h( x, y, z ) = Φ(m( y, z ) − c( x))

(3.39)

1 if u > 0
Φ (⋅) = 
0 if u ≤ 0

(3.40)

The triple Fourier series expansion of the nine switching functions can be given in the
complex form by the following expression:

31

∑ ∑ ∑ [H kmn × e j (kx + my + nz ) ]
∞

h( x, y, z ) =

∞

∞

(3.41)

k = −∞ m = −∞ n = −∞

H kmn =

π π π

1
3

∫ ∫ ∫ h ( x, y , z ) × e

8π − π − π − π

− j ( kx + my + nz )

dxdydz

(3.42)

If the modulation signal consists of two sinusoidal components of two independent
frequencies and the carrier signal is triangular, then the three-dimensional integral in (3.42) will
be conducted within the unit cube with each edge of the length of 2π as illustrated in Figure 3.1.

Figure 3.1 Illustration of the unit cube. For interpretation of the references to color in this and all
other figures, the reader is refered to the electronic version of this thesis.
32

where x –axis represents the phase angle of the carrier signal whereas y –axis and z –axis
represent the phase angles of the two components in the modulation function.
Within the unit cube, the switching function h( x, y, z ) is non-zero only in the space
between the two surfaces that are defined by (3.43).

m( y , z ) − c ( x ) = 0

(3.43)

In addition, the line which illustrated in Figure 3.1, is defined as follows:

y=



ω
ω m1
x + θ m1 − m1 θ c 
ωc
ωc



(3.44)

z=



ω
ω m2
x + θ m 2 − m 2 θ c 
ωc
ωc



(3.45)

The intersections of the line defined by (3.44), (3.45) and the surfaces defined by (3.43)
determine the instants when the value of the switching function h( x, y, z ) switches from 0 to 1
and vice versus.
Once the integration boundaries of the above equation are defined, the procedure of carrying out
the integration will be straightforward process as defined in [33]. This will be clearly explained
in the next chapter.

3.5

Summary

A new analytical approach has been presented in this chapter for identifying the harmonic
components of PWM waveforms for AC/AC converters. However, the analytical method based
on three-dimensional Fourier integral will be explained in detail for obtaining accurate spectra

33

of the switching functions and synthesized terminal quantities of the matrix converter in the next
chapter.

34

CHAPTER 4
Spectral Analysis of Matrix Converter Waveforms
This chapter presents an analytical approach based on three-dimensional Fourier integral to
characterize the accurate spectra of the switching functions and synthesized output voltages and
input currents of matrix converter which was proposed by Alesina and Venturini 1981[3]. The
principle of this analysis is carried out in two steps. The first step is that the spectra of the
switching functions are derived based on three-dimensional Fourier integral. The second step is
that the spectra of the output voltages and input currents are evaluated using a convolution
operation in the frequency domain. The analytical spectra have been compared with the
numerical spectra that are obtained from FFT. In addition, the total harmonic distortion (THD)
and weighted total harmonic distortion (WTHD) of the synthesized output voltages and input
currents matrix converter are considered.

4.1

Modulation Process

Even though the modulation process of the matrix converter is well understood, the purpose
of the section is just to introduce some notations which will be utilized in the subsequent
sections.
The modulation function proposed in [3], which is defined in (2.10), can be rewritten in the
following general form:

35

2π 
2π 


1 + q cos ωm1t − (u + v + 1)  + q cos ωm 2t − (u − v ) 
3 
3 


muv (t ) =
3

(4.1)

where 0 ≤ q ≤ 0.5 , ωm1 and ωm2 are related to the input frequency ωi and output frequency

ωo as follows:
ωm1 = ωi + ωo

(4.2)

ωm2 = ωi − ωo

(4.3)

However, in order to prevent a short circuit on the input side and ensure uninterrupted
load current flow, the switching function huv ( x, y, z ) which is determined by (3.39), must satisfy
the following constraint condition:
3

∑ [huv ( x, y, z)] = 1

for u ∈{1,2,3}

(4.4)

v =1

With 0 ≤ huv ( x, y, z ) ≤ 1 for u ∈{1,2,3} , v ∈{1,2,3} .
The switching functions are generated in the following expression:

Φ[mu1 ( y, z ) − c( x)]



huv ( x, y, z ) = Φ[mu1 ( y, z ) + mu 2 ( y, z ) − c( x)] − hu1 ( x, y, z )


1 − hu1 ( x, y, z ) − hu 2 ( x, y, z )


36

if v = 1
if v = 2
if v = 3

(4.5)

The synthesized output voltages (phase-to-neutral) and input currents can be determined by the
following expression:
3

Vou = ∑ [huv ( x, y, z ) × Viu ]

for u ∈{1,2,3}

(4.6)

for v ∈{1,2,3}

(4.7)

v =1

3

I iv =

∑ [huv ( x, y, z) × I ov ]

u =1

4.2

Analytical Solution to the Spectra of Switching Functions

The triple Fourier expansion of the switching functions given by (4.5) can be written in
complex form as follows:

huv ( x, y, z ) =

∑ ∑ ∑ [H kmn (u, v) × e j (kx+my+nz) ]
∞

∞

∞

(4.8)

k =−∞ m=−∞ n=−∞

where

H kmn (u , v ) =

π π π

1
3

∫ ∫ ∫ huv ( x, y, z ) × e

− j ( kx+ my + nz )

8π −π −π −π

(4.9)

dxdydz

Equation (4.9) can be proceed with consideration of (3.43), (4.1) and (4.5). Hence, the spectrum
of the nine switching functions can be determined in three different cases:
1) When v = 1 , by substituting (4.5) into (4.9), this will lead to

H kmn (u ,1) =

π π π

1
3

∫ ∫ ∫ Φ[mu1 ( y, z ) − c( x)]× e

− j ( kx + my + nz )

8π −π −π −π

37

dxdydz

π π a

1

=

3

∫ ∫ ∫e

− j ( kx + my + nz )

8π −π −π −a

dxdydz

(4.10)

where a = π × mu1 ( y, z) .
If k = 0 , equation (4.10) will become as:

H 0mn (u ,1) =

π π a

1
3

∫ ∫ ∫e

− j ( my + nz )

8π −π −π −a

1
3


 q − j (u − 2) 2π 3
 e
3

H 0mn (u ,1) = 
q
 e − j (u −1) 2π 3
3


0



dxdydz

for m = n = 0

for m = 1, n = 0
(4.11)
for m = 0, n = 1
for m, n ≠ 0

If k ≠ 0 , the equation (4.10) will become as:

H kmn (u,1) =

π π a

1
3

∫ ∫ ∫e

8π −π −π −a

− j ( kx+ my + nz )

dxdydz

 k m + n  
2 sin  +
π
2  
 3
 × J  kπ ⋅ q  × J  kπ ⋅ q  × e − j [m(u + 2) + n (u −1) ]2π 3
=
 n

m
kπ
 3 
 3 

38

(4.12)

where J m (⋅) and J n (⋅) are the m -th and n -th order Bessel functions of the first kind
respectively.
2) When v = 2 , by substituting (4.5) into (4.9), this will lead to two steps:
•

The first step is;

'
H kmn (u ,2) =

π π π

1
3

∫ ∫ ∫ Φ[mu1 ( y, z ) + mu 2 ( y, z ) − c( x)]× e

8π −π −π −π

=

π π b

1
3

∫ ∫ ∫e

− j ( kx+ my + nz )

8π −π −π −b

dxdydz

− j ( kx + my + nz )

dxdydz

(4.13)

where b = π (mu1 ( y, z ) + mu 2 ( y, z )) .
If k = 0 , the equation (4.13) will become as:

'
H 0 mn (u ,2) =

π π b

1
3

∫ ∫ ∫e

− j ( my + nz )

8π −π −π −b

2
3


 − q − j (u +1) 2π 3
e

 3

'
H 0mn (u ,2) = 
− q

e − j (u ) 2π 3
 3


0



dxdydz

for m = n = 0

for m = 1, n = 0
(4.14)
for m = 0, n = 1
for m, n ≠ 0

39

If k ≠ 0 , the equation (4.13) will become as:

'
H kmn (u ,2) =

π π b

1
3

∫ ∫ ∫e

− j ( kx + my + nz )

8π −π −π −b

dxdydz

 2k  m + n   
−
2 sin 
 π 


 3  2    × J  kπ ⋅ q  × J  kπ ⋅ q  × e − j [m(u +1) + n(u ) ]2π 3
=
 n

m
kπ
 3 
 3 
•

(4.15)

The second step is;

'
H kmn (u ,2) = H kmn (u ,2) − H kmn (u ,1)

(4.16)

The equation (4.16) is valid when k ≠ 0 . If k = 0 , the equation (4.16) will become:
1
3


 q − j (u ) 2π 3
 e
3

H 0 mn (u ,2) = 
q
 e − j (u −1) 2π 3
3


0



for m = n = 0

for m = 1, n = 0
(4.17)
for m = 0, n = 1
for m, n ≠ 0

3) When v = 3 , by substituting (4.5) into (4.9), this will lead to

H kmn (u ,3) =

π π c

1
3

∫ ∫ ∫e

8π −π −π −c

− j ( kx + my + nz )

dxdydz

where c = π × mu3 ( y, z ) .

40

(4.18)

If k = 0 , equation (4.18) will become:

H 0mn (u ,3) =

π π c

1
3

∫ ∫ ∫e

− j ( my + nz )

8π −π −π −c

1
3


 q − j (u +1) 2π 3
 e
3

H 0 mn (u ,3) = 
q
 e − j (u ) 2π 3
3


0



dxdydz

for m = n = 0

for m = 1, n = 0
(4.19)
for m = 0, n = 1
for m, n ≠ 0

If k ≠ 0 , equation (4.18) will become:

H kmn (u ,3) =

π π c

1
3

∫ ∫ ∫e

8π −π −π −c

− j ( kx + my + nz )

dxdydz

 k m + n  
2 sin  +
π
2  
 3
 × J  kπ ⋅ q  × J  kπ ⋅ q  × e − j [m(u +1) + n(u ) ]2π 3
=
 n

m
kπ
 3 
 3 

(4.20)

Equations (4.11), (4.12), (4.16), (4.17), (4.19) and (4.20) can be rearranged to formulate the
general form of the switching functions based on three-dimensional Fourier integral in (4.21),
(4.22) and (4.23).

41

If k = 0 ,
1
3


 q − j (u + v +1) 2π 3
 e
3

H kmn (u , v) = 
q
 e − j (u − v ) 2π 3
3


0



for m = n = 0

for m = 1, n = 0
(4.21)
for m = 0, n = 1
for m, n ≠ 0

If k ≠ 0 ,
When v = 1,3
 (3v − 1) k m + n  
+
2 sin 
π
6
2  

 J  kπ ⋅ q  J  kπ ⋅ q e − j [m (u +v +1)+ n (u −v ) ]2π 3 (4.22)
H kmn (u , v) =
 n

m
kπ
 3   3 

When v = 2
'
H kmn (u , v ) = H kmn (u ,2) − H kmn (u ,1)

(4.23)

Hence, the Fourier harmonic component forms of (4.21), (4.22) and (4.23) can be developed for
a triple variable controlled waveforms in time-varying huv (t ) as four different cases:
1) When v = 1 , u ∈{1,2,3}
∞ 
 
A
2π
hu1(t ) = 000 + ∑  A0 m0 cos m ωm1t − (u + 2)

2
3
 
m =1 

42

 
 

 

∞ 
 
2π
+ ∑  A00n cos n ωmt − (u + 2)

3
 
n =1

∞

∞

 
 

 

∞

(4.24)



2π  
2π

 + n ωm2t − (u + 2)
 Akmn cos kωct + m ωm1t − (u + 2)

3  
3


k =1 m = −∞ n = −∞ 

+∑

∑ ∑

m≠0 n≠ 0

 
 

 

2) When v = 2 , u ∈{1,3}
 
A000 ∞ 
2π
hu 2 (t ) =
+ ∑  A0m 0 cos m ωm1t − (u )

2
3
 
m =1

 
 

 

∞


 
2π
+ ∑  A00n cos n ωmt − (u + 1)

3
 
n =1

 
 

 

(4.25)

 '

2π  
2π   

 Akmn cos kωc t + m ω m1t − (u )  + n ω m 2t − (u + 1)   

3  
3  



∞
∞
∞ 

+∑ ∑ ∑ 


k =1 m = −∞ n = −∞

2π  
2π  

 '
m≠0 n≠0 + B


 kmn sin kωc t + m ω m1t − (u ) 3  + n ω m 2t − (u + 1) 3  

 
 


3) When v = 2 , u ∈{2}
∞ 
 
A
2π
hu 2 (t ) = 000 + ∑  A0m 0 cos m ωm1t − (u )

2
3
 
m =1

 
 

 

∞ 
 
2π
+ ∑  A00n cos n ωmt − (u + 1)

3
 
n =1

 
 

 

43

(4.26)

 ''

2π  
2π   

 Akmn cos kωc t + m ω m1t − (u )  + n ω m 2t − (u + 1)   

3  
3  



∞
∞
∞ 

+∑ ∑ ∑ 


k =1 m = −∞ n = −∞

2π  
2π  

 ''
m≠0 n≠0 + B


 kmn sin kωc t + m ω m1t − (u ) 3  + n ω m 2t − (u + 1) 3  

 
 


4) When v = 3 , u ∈{1,2,3}
∞ 
 
A
2π
hu 3 (t ) = 000 + ∑  A0m0 cos m ωm1t − (u + 1)

2
3
 
m =1

∞


 
2π
+ ∑  A00n cos n ωmt − (u )

3
 
n =1
∞

∞

 
 

 

 
 

 

∞



2π  
2π

 + n ωm 2t − (u )
 Akmn cos kωct + m ωm1t − (u + 1)

3  
3


k =1 m = −∞ n = −∞ 

+∑

∑ ∑

(4.27)

 
 

 

m≠0 n≠0

Where the spectral coefficients are defined by a triple Fourier integral analysis as follows:

A000 =

q
2
, A010 = A001 =
3
3

 kπ ⋅ q   kπ ⋅ q 
2 × jm 
 jn 

 3   3  × sin  k + m + n π 
Akmn =

 3
kπ
2  


  k m + n  
2π     2k m n  

'
Akmn = Akmn × cos ( m + n)




 × sin   3 − 2 − 2 π  − sin   3 + 2 π 
3   

 
 

  k m + n  
2π     2k m n  

'
Bkmn = Akmn × sin ( − m − n)
× sin  
 3 − 2 − 2 π  + sin   3 + 2 π 



3   


 
 


44

   2k m n  
 k m + n  
4π 
2π


''
Akmn = Akmn sin  
 3 − 2 − 2 π  × cos ( m + n) 3  − sin   3 + 2 π  × cos ( m + n) 3



 

 



 






  2k m n  
 k m + n  
4π 
2π


''
Bkmn = Akmn − sin  
 3 − 2 − 2 π  × sin  ( −m − n) 3  + sin   3 + 2 π  × sin  ( −m − n) 3



 


 





4.3





Analytical Spectra of Output Voltages and Input Currents

Let f1(t ) and f 2 (t ) be two functions of t . The product of the two time-domain functions is
denoted by

f (t ) = f1(t ) × f 2 (t )

(4.28)

The convolution process can be illustrated by transforming the product defined in (4.28) into
frequency domain F (ω ) , which can be obtained to a convolution integral of individual spectrum
in the frequency domain as follows

F (ω ) = F1 (ω ) ⊗ F2 (ω )
∞

=

∫ F1(σ ) × F2 (ω − σ )dσ

(4.29)

−∞

where F1 (ω ) and F2 (ω ) are Fourier transforms of f1(t ) and f 2 (t ) , respectively.
This principle can be applied to obtain the spectra of the synthesized output voltages due
to the fact that they are the summation of the products of the time-domain switching functions
and the input voltages as described in (4.6). In a similar manner, the spectra of the synthesized
input currents as described in (4.7) can be determined.
45

For the case where the input voltage waveforms are sinusoid defined by (2.3), the
corresponding spectrum can be calculated by

[

V
Viu (ω ) = i e − j (u −1) 2π 3δ (ω − ωi ) + e j (u −1) 2π 3δ (ω + ωi )
2

]

(4.30)

if ω = 0

1
where δ (ω ) = 
0

(4.31)

otherwise

Also, the spectrum of the switching functions huv ( x, y, z ) that is defined by (4.8), can be
represented as a function of ω .

H uv (ω ) =

∞

∞

∞

∑ ∑ ∑ [H kmn (u, v) × δ (ω − (kωc + mωm1 + nωm2 ))]

(4.32)

k = −∞ m = −∞ n = −∞

where H kmn(u, v) is defined by (4.21), (4.22) and (4.23).
Therefore, the spectra of the synthesized output voltages Vou (ω) can be easily determined
since the spectrum of the input voltages switching functions and switching functions are known
in (4.30) and (4.32) respectively.
∞

Vou (ω ) = ∑  ∫ H uv (σ ) × Viu (ω − σ )dσ 

v =1 − ∞


3

[

V 3
= i ∑ e− j (u −1) 2π 3 H uv (ω − ωi ) + e j (u −1) 2π 3 H uv (ω + ωi )
2 v =1

]

(4.33)

Since the output current waveforms are sinusoid defined by (2.4), the spectra of the synthesized
input currents as described in (4.7) can be determined as follows:

46

∞

Iiv (ω ) = ∑  ∫ H uv (σ ) × I ov (ω − σ )dσ 

u =1− ∞


3

[

I o 3 − j (v −1) 2π 3
=
H uv (ω − ωo ) + e j (v −1) 2π 3 H uv (ω + ωo )
∑e
2 u =1

]

(4.34)

It can be seen that the equations (4.33) and (4.34) are very important results because they
illustrate the spectral of the synthesized output voltages and input currents that are simply the
superposition of the nine switching functions spectrum with frequency shifts of ±ωi .

4.4

Analytical and Numerical Results

The analytical results have been verified against the numerical simulation based on FFT.
Table 4.1 illustrates the list of parameters used in the simulation.

Table 4.1 Illustration multiple unit-cube for one switching cycle.
Input frequency

fi

60

Hz

Output frequency

fo

50

Hz

Input voltage amplitude

Vi

100

V

Output current amplitude

Io

50

A

Carrier frequency

fc

1260 Hz

q

Voltage gain ratio

0.5

The analytical spectra of the nine switching functions have been compared with the
numerical spectra that obtained from FFT in Figure 4.1to Figure 4.3, which clearly show a very
good agreement between the analytical and numerical results.
47

Figure 4.1 Analytical and numerical spectra of switching functions h11 , h12 and h13 .

48

Figure 4.2 Analytical and numerical spectra of switching functions h21 , h22 and h23 .

49

Figure 4.3 Analytical and numerical spectra of switching functions h31 , h32 and h33 .

50

Figure 4.4 illustrates the analytical and numerical spectra of phase-to- neutral output
voltages whereas the analytical and numerical spectra of synthesized input currents shows in
Figure 4.5. Again, a very good agreement between analytical and numerical spectra has been
observed for both the synthesized output voltages and input currents.

Figure 4.4 Analytical and numerical spectra of switching functions Vo1 , Vo 2 and Vo 3 .

51

Figure 4.4 (cont’d)

52

Figure 4.5 Analytical and numerical spectra of switching functions I i1 , I i 2 and I i3 .

53

4.5

Total Harmonic Distortion (THD) and Weighted Total

Harmonic Distortion (WTHD)
THD is the summation of all the harmonic components of the voltage or current waveform
compared against the fundamental components of the voltage or current waveform which can be
obtained as follows:

THD =

[ ]

∞
1
∑ Vn2
V1 n = 2,3,...

(4.35)

On the other hand, WTHD is needed to compare harmonic distortions that caused mainly
by lower order harmonics and it is defined as follows:

WTHD =

∞  2
1
V
∑  nn2 
V1 n = 2,3,...  
 

(4.36)

The two formula above show the calculation for THD and WTHD on a voltage waveform
where V1 is the amplitude of fundamental component n is the order of harmonic, and Vn is the
amplitude of nth harmonic components. Equations (4.35) and (4.36) can be also utilized to
calculate THD and WTHD for a input current waveform.
Figure 4.6 and Figure 4.7 show the THD versus the voltage gain ratio of output voltages
and input currents respectively. It can be seen from both figures how the harmonic significance
reduces with increasing the voltage ratio gain for both output voltages and input currents.

54

Figure 4.6 (a) THD versus voltage gain of output voltages (phase A-C), (b) (phase B).

55

Figure 4.7 (a) THD versus voltage gain of input currents (phase A-C), (b) (phase B).

56

Table 4.2 and Table 4.3 describe the THD magnitudes of Figure 4.6 and Figure 4.7
respectively. It can be noted from both tables that as the voltage gain increases, the of the output
voltages and input currents will decrease.

Table 4.2 THD magnitudes of three phase output voltages.
Voltage gain ratio q

THD (Phases A - C)

THD (Phase B)

0.1

22.242

22.353

0.2

11.339

11.514

0.3

7.701

7.879

0.4

5.837

5.982

0.5

4.683

4.793

Table 4.3 THD magnitudes of three phase input currents.
Voltage gain ratio q

THD (Phases A - C)

THD (Phase B)

0.1

36.973

32.446

0.2

18.415

15.337

0.3

12.118

9.38

0.4

8.877

6.355

0.5

6.899

4.633

The WTHD versus the voltage ratio are summarized for the output voltages and input
currents in Figure 4.8 and Figure 4.9, respectively. It can also be noted that the harmonic
presented in Figure 4.8 and Figure 4.9 reduces as the voltage ratio (q) increases for both output
voltages and input currents.

57

Figure 4.8 (a) WTHD versus voltage gain of output voltages (phase A-C), (b) (phase B).

58

Figure 4.9 (a) WTHD versus voltage gain of input currents (phase A-C), (b) (phase B).

59

Table 4.4 describes the WTHD magnitudes of Figure 4.8. It can be seen that the output
voltages decrease as the voltage gain decreases. Whereas, Table 4.5 represents the WTHD
magnitudes of Figure 4.9. The input currents also decrease with increasing in voltage gain.

Table 4.4 WTHD magnitudes of three phase output voltages.
Voltage gain ratio q

WTHD (Phases A - C)

WTHD (Phase B)

0.1

0.876

0.878

0.2

0.436

0.439

0.3

0.288

0.292

0.4

0.213

0.218

0.5

0.168

0.174

Table 4.5 WTHD magnitudes of three phase input currents.
Voltage gain ratio q

WTHD (Phase B)

0.1

1.62

0.773

0.2

0.802

0.365

0.3

0.525

0.224

0.4

0.385

0.151

0.5

4.6

WTHD (Phases A - C)

0.3

0.11

Summary

This chapter has presented an analytical approach based on three-dimensional Fourier
integral for characterizing the spectra of the synthesized output voltages and input currents of the
matrix converter which was proposed by Alesina and Venturini 1981 [3]. The analytical spectra
of the synthesized output voltages and input currents has been observed a very good agreement

60

with the numerical spectra obtained from FFT. The analytical and numerical spectra of the nine
switching functions has been shown a very good result as well.

61

CHAPTER 5
Conclusions and Future Work
5.1

Conclusions

This thesis has presented an analytical method based on three-dimensional Fourier integral to
obtain accurate spectra of the switching functions and synthesized terminal qualities of matrix
converters.
The high frequency modulation of the matrix converter which was proposed by Alesina and
Venturini 1981 has been considered in this analysis [3]. The analytical spectra of the nine
switching functions have been compared first with the numerical spectra obtained from FFT. The
results shows very good agreement between analytical and numerical spectra for each switching
functions. Then, the spectra of the synthesized output voltages and input currents have been
verified against the numerical spectra based on FFT. Again, a very good agreement between the
analytical and numerical spectra has been observed for both the synthesized output voltages and
input currents.

5.2

Future Work

PWM strategies for matrix converters have been received increasing attention recently. The
analytical approach of the three-dimensional Fourier integral which is presented in this thesis, is

62

demonstrated through the carrier based pulse width modulation of a conventional matrix
converter.
It is worth nothing that the analytical approach is equally applicable to space vector based
modulation once the equivalent modulation functions are determined. In addition, the modulation
process of indirect matrix converters can also be analytically characterized using the proposed
approach in this thesis.

63

BIBLIOGRAPHY

64

BIBLIOGRAPHY
[1] P. W. Wheeler, J. Rodriguez, J. C. Clare, L. Empringham, and A.Weinstein, "Matrix
converters: A technology review," IEEE Trans. Ind. Electron, vol. 49, no. 2, p. 276–288,
Apr. 2002.
[2] M.Venturini, "A new sine wave in sine wave out, conversion technique which eliminates
reactive elements," in Proc. POWERCON 7, p. E3_1–E3_15, 1980.
[3] A. Alesina and M. Venturini, "Solid-state power conversion: A Fourier analysis approach to
generalized transformer synthesis," IEEE Trans. Circuits Syst, vol. CAS-28, p. 319–330,
Apr. 1981.
[4] A. Alesina and M. G. B. Venturini, "Analysis and design of optimum amplitude nine-switch
direct ac–ac converters," IEEE Trans. Power Electron, vol. 4, no. 1, p. 101–112, Jan. 1989.
[5] G. Roy and G. E. April, "Cycloconverter operation under a new scalar control algorithm," in
, 1989, p. 368–375.
[6] H. Hojabri, H. Mokhtari, and L. Chang, "A generalized technique of modeling, analysis, and
control of a matrix converter using SVD," IEEE Trans. Ind. Electron, vol. 58, no. 3, p. 949–
959, Mar. 2011.
[7] G. Clos, "Straight forward control of the matrix converter," in , 2005, p. P1–P8.

[8] P. Ziogas, S. Khan, and M. Rashid, "Some improved forced commutated cycloconverter
structures," IEEE Trans. Ind. Appl, vol. IA-21, no. 5, p. 1242–1253, Sep. 1985.
[9] A. Odaka, I. Sato, H. Ohguchi, Y. Tamai, H. Mine, and J.-I. Itoh, "A PAM control method
for matrix converter based on virtual AC/DC/AC conversion method," IEEE Trans. Ind.
Appl, vol. 126, no. 9, p. 1185–1192, 2006.
[10] A. G. H. Accioly, F. Bradaschia, M. C. Cavalcanti, F. A. S. Neves, and V. N. Lima,
"Generalized modulation strategy for matrix converters— Part I," in Proc. IEEE Power
Electron. Spec. Conf, Orlando, FL, Jun. 17-21,2007, p. 645–652.
[11] F. Bradaschia, M. C. Cavalcanti, F. A. S. Neves, V. N. Lima, and A. G. H. Accioly,
"Generalized modulation strategy for matrix converters—Part II," in in Proc. IEEE Power
Electron. Spec. Conf, Orlando, FL, Jun. 17–21, 2007, p. 665–671.
[12] F. Bradaschia, M. C. Cavalcanti, F. Neves, and H. de Souza, "A modulation technique to
reduce switching losses in matrix converters," IEEE Trans. Ind. Electron, vol. 56, no. 4, p.
1186–1195, Apr. 2009.

65

[13] L. Huber and D. Borojevic, "Space vector modulated three-phase to threephase matrix
converter with input power factor correction," IEEE Trans. Ind. Appl, vol. 31, no. 6, p.
1234–1246, Nov. 1995.
[14] P. Nielsen, F. Blaabjerg, and J. K. Pedersen, "Space vector modulated matrix converter with
minimized number of switchings and a feed forward compensation of input voltage
unbalance," in Proc. PEDES, 1996, p. 833–839.
[15] H. J. Cha and P. N. Enjeti, "An approach to reduce common-mode voltage in matrix
converter," IEEE Trans. Ind. Appl, vol. 39, no. 4, p. 1151–1159, Jul. 2003.
[16] L. Helle, K. B. Larsen, A. H. Jorgensen, S. Munk-Nielsen, and F. Blaabjerg, "Evaluation of
modulation schemes for three-phase to three phase matrix converters," IEEE Trans. Ind.
Electron, vol. 51, no. 1, p. 158–171, Feb. 2004.
[17] Y. Tadano, S. Urushibata, M. Nomura, Y. Sato, and M. Ishida, "Direct space vector PWM
strategies for three-phase to three-phase matrix converter," in Proc. PCC-Nagoya, 2007, p.
1064–1071.
[18] M.Braun, "Ein dreiphasiger direktumrichter mit pulsbreitenmodulation zur getrennten
steuerung der ausgangsspannung und der eingangsblindleistung," in Ph.D. dissertation,
Technische Hochschule Darmstadt, Darmstadt, Germany, 1983.
[19] D. Casadei, G. Serra, A. Tani, and L. Zarri, "Matrix converter modulation strategies: A new
general approach based on space-vector representation of the switch state," IEEE Trans.
Ind. Electron, vol. 49, no. 2, p. 370–381, Apr. 2002.
[20] J. Igney and M. Braun, "A new matrix converter modulation strategy maximizing the
control range," in Proc. IEEE Power Electron. Spec. Conf, 2004, p. 2875–2880.
[21] D. Casadei, G. Serra, A. Tani, and L. Zarri, "A novel modulation strategy for matrix
converter with reduced switching frequency based on output current sensing," in Proc. IEEE
Power Electron. Spec. Conf, 2004, p. 2373–2379.
[22] J. Igney and M. Braun, "Space vector modulation strategy for conventional and indirect
matrix converters," in Proc. EPE Conf, 2005, p. P1–P10.
[23] B. Velaerts, P. Mathys, E. Tatakis, and G. Bingen, "A novel approach to the generation and
optimization of three-level PWM waveforms for three-level inverters," in Proc. IEEE
Power Electron. Spec. Conf, 1988, p. 1255–1262.
[24] P. D. Ziogas, S.I. Khan, M.H. Rashid, "Analysis and design of forced commutated
cycloconverter structures with improved transfer characteristics," IEEE Trans. Industrial
Electronics, vol. 1E-33, No. 3, pp. 271-280, Aug. 1986.

66

[25] L.Huber and D. Borojevic, "Space vector modulator for forced commutated
cycloconverters," in Conference Record of the IEEE Industry Applications Society Annual
Meeting, 1989, vol. 1, p. 871–6.
[26] L. Huber, D. Borojevic, and N. Bur´any, "Voltage space vector based pwm control of forced
commutated cycloconverters," in IECON ’89. 15th Annual Conference of IEEE Industrial
Electronics Society, 1989, vol. 1, p. 106–11.
[27] L. Huber and D. Borojevic, "Space vector modulated three-phase to three-phase matrix
converter with input power factor correction," IEEE Trans. Ind. Applicat, vol. 31, p. 1234–
1246, Nov. 1995.
[28] L. Huber, D. BorojeviC, and N. Burany, "Analysis, design and implementation of the spacevector modulator for forced-commutated cycloconverters," IEE Proc.-B, vol. 139, no. 2, pp.
103-113, Mar. 1992.
[29] C. Klumpner, F. Blaabjerg, I. Boldea, and P. Nielsen, "New modulation method for matrix
converters," IEEE Transactions on Industry Applications, vol. 42, no. 3, p. 797–806, 2006.
[30] T. D. Nguyen and H.-H. Lee, "Modulation strategies to reduce common-mode voltage for
indirect matrix converters," IEEE Transactions on Industrial Electronics, vol. 59, no. 1, p.
129–140, 2012.
[31] S. Kim, Y.-D. Yoon, and S.-K. Sul, "Pulse width modulation method of matrix converter for
reducing output current ripple," Power Electronics, IEEE Transactions on, vol. 25, no. 10,
p. 2620–2629, 2010.
[32] P. Wood, "General theory of switching power converters," IEEE Power Electronics
Specialists Conference, PESC'79, pp. 3-10, Jun. 1979.
[33] D. G. Holmes and T. A. Lipo, Pulse Width Modulation for Power Converters. John Wiley &
Sons, 2003.
[34] W. R. Bennett, "New results in the calculation of modulation products," The Bell System
Technical Journal, vol. 12, p. 228–243, 1933.
[35] H. S. Black, Modulation Theory. New York, Van Nostrand, 1993.
[36] S. R. Bowes and B. M. Bird, "Novel approach to the analysis and synthesis of modulation
processes in power converters," IEEE Proceedings, vol. 122, no. 5, p. 507–513, 1975.

67