1

Abstract The performance of matrix converters is strictly

related to the modulation technique adopted to control the state
of the switches. The best results are obtained by means of the
Space Vector Modulation technique, which allows the full con-
trol of output voltages and input power factor. It can be also
verified that this technique achieves the highest voltage transfer
ratio and an increase of the input current quality. In this paper
the most common switching strategies for Space Vector Modu-
lation are analyzed, and their effects on the line current distor-
tion are compared. The analysis demonstrates that a suitable
distribution of the zero voltage vectors within the switching
pattern can sensibly reduce the rms value of the line current
ripple. Numerical results are given to confirm the validity of the
proposed approach.

Index TermsACAC power conversion, harmonic analysis,
harmonic distortion, matrix converters, power quality, power
system harmonics.
I. NOMENCLATURE
j 1 .
s
v Line-to-neutral supply voltage space vector.
i
j
i i
e v v

= Line-to-neutral input voltage space vector.
o
j
o o
e v v

= Line-to-neutral output voltage space vector.
s
i Line current space vector.
i
j
i i
e i i

= Input current space vector.
o
j
o o
e i i

= Output current space vector.
i o
v v q = Voltage transfer ratio.
Scalar product.
II. INTRODUCTION
HREE-PHASE matrix converters have received consid-
erable attention in recent years because they may be-
come a good alternative to PWM-VSI topology. This be-
cause the matrix converter provides bi-directional power
flow, sinusoidal input/output waveforms, and controllable
input power factor. Furthermore, the matrix converter allows
a compact design, due to the lack of dc-link capacitors for
energy storage. On the other hand, the complexity of the
matrix converter topology makes the study and the determi-
nation of suitable modulation strategies a hard task.
Different approaches have been considered in the past to
face this problem [1] - [5].
The Space Vector Modulation (SVM) approach, initially
proposed in [6] to control only the output voltages, has been
successively developed in [7] [11] in order to completely

The authors are with the Department of Electrical Engineering Uni-
versity of Bologna - Italy. e-mail: (domenico.casadei, giovanni.serra, an-
gelo.tani, luca.zarri) @mail.ing.unibo.it.
exploit the possibility of matrix converters to control the
input power factor regardless the output power factor, to
fully utilize the input voltages, and to reduce the number of
switch commutations in each cycle period. Furthermore, this
strategy allows an immediate comprehension of the modula-
tion process, without the need for a fictitious DC link, and
avoiding the addition of the third harmonic components.
In this paper, by using the SVM approach, different switch-
ing strategies are compared with reference to the rms value
of the line current and the number of commutations. It will
be shown that a suitable disposition of the zero configura-
tions within the switching pattern, allows to reduce the line
current distortions.
III. SPACE VECTOR MODULATION
Reference is made to the scheme shown in Fig. 1. The Space
Vector Modulation is based on the instantaneous space vec-
tor representation of input and output voltages and currents.
Among the 27 possible switching configurations available in
three-phase matrix converters, 21 only can be usefully em-
ployed in the SVM algorithm, and can be represented as
shown in Tab. I.
The first 18 switching configurations determine an output
voltage vector
o
v , and an input current vector
i
i , having
fixed directions, as represented in Figs. 2a and 2b, and will
be named active configurations. The magnitude of these
vectors depends upon the instantaneous values of the input
line-to-line voltages and output line currents respectively.
The last 3 switching configurations determine zero input
current and output voltage vectors and will be named zero
configurations.
The remaining 6 switching configurations have each output
phase connected to a different input phase. In this case the
output voltage and input current vectors have variable direc-
tions and cannot be usefully used to synthesize the reference
vectors.
The SVM algorithm for matrix converters presented in this
section has the inherent capability to achieve the full control
of both output voltage vector and instantaneous input current
displacement angle [7]-[11].
At any sampling instant, the output voltage vector
o
v and
the input current displacement angle
i
are known as refer-
Domenico Casadei, Associate Member, IEEE, Giovanni Serra, Associate Member, IEEE,
Angelo Tani, and Luca Zarri
Effects of the Switching Strategy on the Input
Power Quality of Matrix Converters
T
3 i
i
0
3 o
i
2 i
i
3 i
v
1 o
i
2 o
i
1 i
i
3 o
v
2 i
v
33
S
1 o
v
2 o
v
1 i
v
11
S
12
S
13
S
21
S
22
S
23
S
31
S
32
S

Fig. 1 Basic scheme of matrix converters.
0-7803-7967-5/03/$17.00 2003 IEEE
Paper accepted for presentation at 2003 IEEE Bologna Power Tech Conference, June 23th-26th, Bologna, Italy
ence quantities. The input line-to-neutral voltage vector
i
v is
imposed by the source voltages and is known by measure-
ments. Then, the control of
i
can be achieved controlling
the phase angle
i
of the input current vector.
In principle, the SVM algorithm is based on the selection of
4 active configurations that are applied for suitable time in-
tervals within each cycle period
c
T . The zero configurations
are applied to complete
c
T .
The reference voltage vector
o
v is resolved into the compo-
nents
'
o
v and
"
o
v along the two adjacent vector directions.
The
'
o
v component can be synthesized using two voltage
vectors having the same direction of
'
o
v . With reference to
Figs. 2a and 2b, among the six possible switching configura-
tions (7, 8, 9), the ones that allow also the modulation of
the input current direction must be selected. It is verified that
this constraint allows the elimination of two switching con-
figurations (+8 and -8 in this case). Among the remaining
four, we assume to apply the positive switching configura-
tions (+7 and +9). The meaning of this assumption will be
discussed later in this Section. With similar considerations
the switching configurations required to synthesize the
' '
o
v
component can be selected (+1 and +3, in this case).
Using the same procedure it is possible to determine the four
switching configurations related to any possible combination
of output voltage and input current sectors, leading to the
results summarized in Tab. II.
Four symbols (I, II, III, IV) are also introduced in the last
row of Tab. II to identify the four general switching configu-
rations, valid for any combination of input and output sec-
tors.
Now it is possible to write, in a general form, the four basic
equations of the SVM algorithm, which satisfy, at the same
time, to the requirements of the reference output voltage
vector and input current displacement angle. With reference
to the output voltage vector, the following two equations can
be written
] 3 / 3 / ) 1 [( '
)
3
~
cos(
3
2
+

= + =
v
K j
o o
II II
o
I I
o o
e v v v v (1)
TAB. I - SWITCHING CONFIGURATIONS USED IN THE SVM ALGORITHM.

Switching
configuration
list
Switches On v
o

o
i
i

i

+1 S
11
S
22
S
32
2/3 v
12i
0 2/3 i
o1
-/6
1 S
12
S
21
S
31
-2/3 v
12i
0 -2/3 i
o1
-/6
+2 S
12
S
23
S
33
2/3 v
23i
0 2/3 i
o1
/2
2 S
13
S
22
S
32
-2/3 v
23i
0 -2/3 i
o1
/2
+3 S
13
S
21
S
31
2/3 v
31i
0 2/3 i
o1
7/6
3 S
11
S
23
S
33
-2/3 v
31i
0 -2/3 i
o1
7/6
+4 S
12
S
21
S
32
2/3 v
12i
2/3 2/3 i
o2
-/6
4 S
11
S
22
S
31
-2/3 v
12i
2/3 -2/3 i
o2
-/6
+5 S
13
S
22
S
33
2/3 v
23i
2/3 2/3 i
o2
/2
5 S
12
S
23
S
32
-2/3 v
23i
2/3 -2/3 i
o2
/2
+6 S
11
S
23
S
31
2/3 v
31i
2/3 2/3 i
o2
7/6
6 S
13
S
21
S
33
-2/3 v
31i
2/3 -2/3 i
o2
7/6
+7 S
12
S
22
S
31
2/3 v
12i
4/3 2/3 i
o3
-/6
7 S
11
S
21
S
32
-2/3 v
12i
4/3 -2/3 i
o3
-/6
+8 S
13
S
23
S
32
2/3 v
23i
4/3 2/3 i
o3
/2
8 S
12
S
22
S
33
-2/3 v
23i
4/3 -2/3 i
o3
/2
+9 S
11
S
21
S
33
2/3 v
31i
4/3 2/3 i
o3
7/6
9 S
13
S
23
S
31
-2/3 v
31i
4/3 -2/3 i
o3
7/6
0
1
S
11
S
21
S
31
0 - 0 -
0
2
S
12
S
22
S
32
0 - 0 -
0
3
S
13
S
23
S
33
0 - 0 -


o
v

1,2,3
4,5,6
7,8,9

o
v
o
sector

o
v

Fig. 2a Sectors and directions of the output line-to-neutral
voltage vectors.

2,5,8
3,6,9 1,4,7
i
i

i
sector






Fig. 2b Sectors and directions of the input current vectors.

] 3 / ) 1 [( "
)
3
~
cos(
3
2

+ = + =
v
K j
o o
IV IV
o
III III
o o
e v v v v (2)
With reference to the input current displacement angle, two
equations are obtained by imposing to the vectors
( )
II II
i
I I
i
i i + and ( )
IV IV
i
III III
i
i i + to have the direction
defined by
i
. This can be achieved by imposing a null
value to the two vector components along the direction per-
pendicular to
i
j
e

(i.e.
i
j
e j

), leading to
( )
( )
0
3 1
~
= +

i i
K j j II II
i
I I
i
e e j i i (3)
( )
( )
0
3 1
~
= +

i i
K j j IV IV
i
III III
i
e e j i i (4)
In (1)-(4)
o

~
and
i

~
are the output voltage and input cur-
rent phase angle measured with respect to the bisecting line
of the corresponding sector, and differ from
o
and
i
rep-
resented in Fig. 2. In these equations the following angle
limits apply
6
~
6

+ < <

o
,
6
~
6

+ < <

i
. (5)
I
,
II
,
III
,
IV
are the duty-cycles (i.e.
I
=t
I
/T
c
) of the 4
switching configurations, K
v
=1,2,..,6 represents the output
voltage sector and K
i
=1,2,,6 represents the input current
sector.
IV
o
III
o
II
o
I
o
v v v v , , , are the output voltage vectors associ-
ated respectively with the switching configurations I, II, III,
IV given in Tab. II. The same formalism is used for the input
current vectors.
Solving (1)-(4) with respect to the duty-cycles, after some
tedious manipulations, leads to the following relationships
[11]:
i
i o K K I
cos
) /
~
cos( ) /
~
cos(
q ) (
i v

=
+
3 3
3
2
1 (6)
i
i o K K II
cos
) /
~
cos( ) /
~
cos(
q ) (
i v

+
=
+ +
3 3
3
2
1
1
(7)
i
i o K K III
cos
) /
~
cos( ) /
~
cos(
q ) (
i v

+
=
+ +
3 3
3
2
1
1
(8)
i
i o K K IV
cos
) /
~
cos( ) /
~
cos(
q ) (
i v

+ +
=
+
3 3
3
2
1 . (9)
Equations (6)-(9) have a general validity and can be applied
for any combination of output voltage sector K
v
and input
current sector K
i
.
It should be noted that, for any sector combinations, two of
the duty-cycles calculated by (6)-(9) assume negative values.
This is due to the assumption made of using only the posi-
tive switching configurations in writing the basic equations
(1)-(4). A negative value of the duty-cycle means that the
corresponding negative switching configuration has to be
selected instead of the positive one.
Furthermore, for the feasibility of the control strategy, the
sum of the absolute values of the four duty-cycles must be
lower than unity
1 + + +
IV III II I
. (10)
The zero configurations are applied to complete the cycle
period.
IV. SWITCHING PATTERNS
Using the SVM technique, the switching pattern is defined
by the switching configuration sequence. With reference to
the particular case of output voltage vector lying in sector 1
and input current vector lying in sector 1, the selected
switching configurations are 0
1
, 0
2
, 0
3
, +1, -3, -7, +9. Among
all possible switching sequences, it can be verified that there
is only one switching configuration sequence utilizing the
three zero configurations, which is characterized by only one
switch commutation for each switching configuration
change, that is 0
3
, -3, +9, 0
1
, -7, +1, 0
2
. The corresponding
general double-sided switching pattern is shown in Fig. 3,
where
1
,
3
,
7
,
9
are duty-cycles corresponding to the
configurations +1, -3, -7 and +9 respectively.
It is a general principle that, once given the desired output
voltage vector and the input current displacement angle, the
configurations sequence is univocally determined through a
procedure similar to the one stated above. For this reason, in
the following, reference will be made to the case shown in
Fig. 3.
The use of the three zero configurations leads to 12 switch
commutations in each cycle period. It should be noted that
the possibility to select the duty-cycles of three zero configu-
rations gives two degrees of freedom, being
9 7 3 1 03 02 01
1 = + + . These two degrees
of freedom can be utilized to define different switching pat-
terns, characterized by different behavior in terms of ripple
of the input and output quantities. In particular, the two de-
grees of freedom might be utilized to eliminate one or two
zero configurations, affecting also the number of commuta-
TAB. II SELECTION OF THE SWITCHING CONFIGURATIONS FOR EACH COMBINATION
OF OUTPUT VOLTAGE AND INPUT CURRENT SECTORS

Sector of the output voltage vector (K
v
)

1 or 4 2 or 5 3 or 6
1 or 4 +9 +7 +3 +1 +6 +4 +9 +7 +3 +1 +6 +4
2 or 5 +8 +9 +2 +3 +5 +6 +8 +9 +2 +3 +5 +6
S
e
c
t
o
r

o
f

t
h
e


i
n
p
u
t

c
u
r
r
e
n
t


v
e
c
t
o
r

(
K
i
)

3 or 6 +7 +8 +1 +2 +4 +5 +7 +8 +1 +2 +4 +5

I II III IV I II III IV I II III IV

tions in each cycle period.
In the following the switching frequency
s
f is defined as the
number of state changes of the nine bi-directional switches
per second.
The switching strategies that will be compared in this paper
and their switching frequencies are reported in Tab. III.
Strategies 1-3 are characterized by only one zero configura-
tion for each half of the switching pattern. Using these
strategies the number of switch commutations in each cycle
period is 8. A particular case is the Strategy 1, which uses
only the zero configuration located in the middle of each
half of the switching pattern (configuration 0
1
in Fig. 3). In
this way, the switches of one column (in this case the first
one of Fig. 1) of the matrix converter do not change their
state (
11
S is always on,
12
S and
13
S are always off in Fig.
3).
Strategies 46 use two zero configurations, determining 10
commutations in a cycle period.
Finally, Strategy 7 is characterized by the fact that all zero
configurations are used with equal duty-cycles. As a conse-
quence, 12 switch commutations occur in each cycle period.
V. ANALYSIS OF THE LINE CURRENT RIPPLE
In order to satisfy the EMI requirements, an input filter for
the matrix converter is generally required, so the system is
composed of four parts: a non ideal voltage supply, an input
filter, a matrix converter and a load. In steady-state operat-
ing conditions the line current ripple is related to several
parameters, i.e. the input filter, the switching strategy, the
load current displacement angle, the voltage transfer ratio
and the output current. Furthermore, the line current ripple
depends on the instantaneous values of the phase angle of
the output voltage (
o
) and input current (
i
), and then its
waveform will change during the time. This makes a very
hard task the comparison of different switching strategies in
terms of line current ripple.
In order to develop a methodology useful for the compari-
son, as a first step, the analysis will be focused on the deter-
mination of the rms value of the line current ripple within a
cycle period, as a function of
o
and
i
. It is worth noting
that the rms value of the line current ripple depends on the
switching strategy, whereas the rms value of the input cur-
rent ripple does not.
With reference to Fig. 4, at the input side of the matrix con-
verter the following equations can be written in terms of
space vectors:
s
T
i
T
s
T
s s
v
L
v
L
i
L
R
dt
i d 1 1
+ = (11)
i
f
s
f
i
i
C
i
C dt
v d 1 1
= (12)
being
f s T
L L L + = the total inductance.
It is possible to define the current and voltage ripple compo-
nents of the input quantities as follows
i1 i i
i - i i = (13)
s1 s s
i - i i = (14)
s1 s s
v - v v = (15)
i1 i i
v v v = (16)
being
i1
i ,
s1
i ,
s1
v and
i1
v the fundamental components of
i
i ,
s
i ,
s
v and
i
v respectively.
Substituting (13)-(16) in (11) and (12) leads to the following
equations:

2 1

) on ( S
11
) on ( S
12
) on ( S
13

) on ( S
21
) on ( S
22
) on ( S
23

) on ( S
32
) on ( S
31
) on ( S
33

2 1

) on ( S
11
) on ( S
12
) on ( S
13

) on ( S
21
) on ( S
22
) on ( S
23

) on ( S
32
) on ( S
31
) on ( S
33

3
0
3 9 +
1
0

7
2
0
1 +
2
0
1 +
7
1
0

9 + 3
3
0
2
03

2
3

2
9

2
01

2
7

2
02

2
1

2
02

2
7

2
01

2
9

2
3

2
03

2
1

Fig. 3 Double-sided switching pattern in a cycle period Tc, with reference to the case shown in Figs. 2a and 2b.
TAB. III SWITCHING STRATEGIES
Strategy
number
Defining equation Switching
frequency
1
0
03 02
= =
c
s
T
f
8
=
2
0
03 01
= =
c
s
T
f
8
=
3
0
02 01
= =
c
s
T
f
8
=
4
0
01 03 02
= = ,
c
s
T
f
10
=
5
0
02 03 01
= = ,
c
s
T
f
10
=
6
0
03 02 01
= = ,
c
s
T
f
10
=
7
03 02 01
= =
c
s
T
f
12
=
i
T
s
T
s s
v
L
i
L
R
dt
i d
=
1
(17)
i
f
s
f
i
i
C
i
C dt
v d
=
1 1
. (18)
Equations (17) and (18) have been derived under the as-
sumption of balanced and sinusoidal supply voltages ( i.e.
0 =
s
v ).
Substituting (18) in the time derivative of (17) leads to
i s
s
s f
s
f T
i i
dt
i d
R C
dt
i d
C L = +

2
2
(19)
and the corresponding Laplace transform is
) s ( i
s R C s C L
) s ( i
i
s f f T
s

+ +
=
1
1
2
. (20)
Equation (19) is a second order linear differential equation
linking the line current ripple to the converter input current
ripple. In this equation the input current ripple can be as-
sumed as a given quantity. In fact, under the assumption that
the switching frequency is much higher than the input and
output frequencies, the mean value of the input quantities in
a cycle period is approximately equal to the value of the
corresponding fundamental components. As a consequence,
for any output current value and switching pattern,
i
i can be
determined using the fourth column of Tab. I and equations
(6)-(9). Finally,
i
i can be calculated by (13).
The assumption of high switching frequency allows (20) to
be simplified as follows:
) s ( i
s C L
) s ( i
i
f T
s

2
1
(21)
The previous equation can be rewritten in the time domain
leading to
i
f T
s
i
C L dt
i d
=
1
2
2
. (22)
Being (22) a second order linear differential equation, two
boundary conditions are needed in order to obtain the solu-
tion. For this purpose the following considerations can be
made.
According to the principle of operation of the SVM tech-
nique, it is verified that the mean value of
i
i during a cycle
period is zero. In addition, as the fundamental components
of the input quantities vary slowly with respect to the ripple
components,
s
i and
i
v can be considered having the
same value at the beginning and at the end of the cycle pe-
riod. Then, from (17) and (18) it follows that the mean val-
ues of
s
i and
i
v in a cycle period are also zero.
With the constraints of
s
i having the same value at the
beginning and at the end of the cycle period, and a null mean
value, the solution of (22) in a cycle period can be deter-
mined.
In order to reduce the number of parameters that should be
adopted for comparing the different switching strategies, the
following p.u. variables are introduced:
o
i
i
i
x

= (23)
s
o c
f T
i
i T
C L
y =
2
(24)
c
T
t
= . (25)
Using (23)-(25), (22) can be rewritten as
x
d
y d
=

2
2
. (26)
It should be noted that the p.u. variables x and y depend
only on four parameters, i.e. q,
o
,
o
,
i
, but they are
independent of
T
L ,
f
C ,
o
i and
c
T . Therefore, the behavior
of y , for given values of q,
o
,
o
,
i
and a specified
switching strategy, can be numerically determined through
the following steps:
1. determination of the sectors of output voltage and input
current vectors, K
v
and K
i
respectively, and selection of
the switching configurations on the basis of Tab. II;
2. determination of the duty-cycles from (6)-(9);
3. determination of the input current vectors related to the
selected switching configurations by means of Tab. I,
evaluation of the mean value of the input current vector
in a cycle period and determination of the waveform of
) t ( x from (13) and (23);
4. determination of the waveform of ) t ( y solving (26)
taking into account the two boundary conditions above
discussed.
Finally, after having determined the behavior of y , one can
obtain the line current ripple
s
i through (24).
VI. COMPARISON OF DIFFERENT SWITCHING STRATEGIES
The parameter that is commonly used to represent the distor-
tion of a waveform is the rms value of the ripple, defined as
2
3
2
2
2
1 RMS , s RMS , s RMS , s RMS , s
i i i I + + = (27)

3 i
i
0
2 i
i
3 s
v
1 i
i
2 s
v
1 s
v
.
.
.
Matrix
Converter
Power supply Input L-C filter
3 s
i
2 s
i
1 s
i
3 i
v
2 i
v
1 i
v
f
C
f
L
s
L
s
R
3 o
i
2 o
i
1 o
i
Load

Fig. 4 Schematic drawn of the analyzed system.
By means of the space vector notation, (27) can be rewritten
in the following form:

=
c
C
T
s T RMS s
dt i I
0
2 1
2
3
,
(28)
being
s
i the magnitude of
s
i .
Solving (24) for
s
i and substituting in (28) leads to the
following expression:
RMS
f T
o c
RMS , s
Y
C L
i T
I
2
= (29)
being
RMS
Y the rms value of y within the cycle period.
In order to compare the performance of the different switch-
ing strategies, the waveform of y and the corresponding
rms value related to each switching strategy should be de-
termined for any operating condition.
As already mentioned, y depends on q,
o
,
o
,
i
and
the switching pattern. In particular the dependence from
o

and
i
should be analyzed in the range , +. Owing to the
symmetries exhibited by
RMS
Y , it can be verified that the
analysis can be limited to two generic pairs of input current
and output voltage sectors, without missing generality, with
values of
o
~
and
i
~
varying in the range /6, +/6. The
first generic pair of input current and output voltage sectors
is characterized by K
v
+K
i
= odd number (odd sectors
pairs), whereas the second one by K
v
+K
i
= even number
(even sectors pairs).
With the aim to clarify the methodology of the analysis, ref-
erence is made to the contour lines of the surface represent-
ing Y
RMS
as function of
o
~
and
i
~
. As an example, Figs. 5a
and 5b show the results obtained using Strategy 4, with q =
0.5 and = 30
o
.
As can be seen,
RMS
Y shows large variations ranging from
about 3 to 6, thus making difficult a comparison between the
different switching strategies.
In order to overcome this problem, it is opportune to define
a new quantity, able to describe the behavior of the line cur-
rent ripple in a long period. For this purpose an integral ef-
fect must be introduced for considering all the possible
combinations of
o
~
and
i
~
.
This new quantity, named
RMS
Y

, is defined as:
2
2 2
e , RMS o , RMS
RMS
Y

+
= (30)
where
9
2
6
6
6
6
2

i o o , RMS
o , RMS
~
d
~
d Y
Y

(31)
9
2
6
6
6
6
2

i o e , RMS
e , RMS
~
d
~
d Y
Y

(32)
being Y
RMS,o
and Y
RMS,e
the rms value of y for odd and
even sectors pairs, respectively.
The quantity
RMS
Y

defined by (30) depends only on two

parameters, i.e. the transfer ratio q and the output current
displacement angle
o
. Therefore an effective comparison
of different switching strategies can be easily carried out
using
RMS
Y

as a quality index.
Figs. 6a, 6b and 6c show the
RMS
Y

behavior for the seven

switching strategies defined in Tab. III as function of q, for
three different values of
o
(Fig. 6a for
o
=15, Fig. 6b for
o
=30 and Fig. 6c for
o
=60).
As can be noted, Strategy 4 exhibits the best performance for
each value of q and
o
, whereas Strategies 1, 2 and 3 give
the worst results in terms of line current ripple.
It is worth noting that Strategy 1 is widely used because all
the switches of one column (for instance S
11
, S
12
and S
13
in
Fig. 1) do not commutate in a cycle period, thus reducing the
switching frequency. Strategies 2 and 3, having the same
switching frequency as Strategy 1, determine a lower ripple
for q > 0.6.
Strategy 7, which uses 3 zero configurations for each cycle
period, has a switching frequency higher than Strategy 4, 5
and 6, which use only 2 zero configurations. Therefore the
rms value of the line current ripple for Strategy 7 is expected
to be lower. Actually, Strategy 7 performs better than Strate-
gies 5 and 6 but determines a current ripple greater than that
of Strategy 4. This demonstrates that the use of the zero con-
figuration located in the middle of each half of the switching
pattern increases the line current ripple.
[degrees]
~
i

[degrees]
~
o

a)

[degrees]
~
i

[degrees]
~
o

b )

Fig. 5 Contour line of the surface representing YRMS x 10
3
using Strategy 4, with q = 0.5 and = 30
o
.
a) Kv+Ki is an odd number; b) Kv+Ki is an even number.

In order to emphasize the considerable dependence of the
line current ripple on different switching strategies, some
numerical simulations have been performed. The simulation
takes into account the switching behavior of the matrix con-
verter and the delay related to the digital implementation of
the control algorithm. Figs. 7a and 7b show the input current
waveforms obtained with Strategy 2 and Strategy 4 respec-
tively. As can be noted, the current ripple determined by
Strategy 2 is sensibly greater than that of Strategy 4.
As each strategy involves a different number of commuta-
tions in the cycle period, a further interesting comparison
can be performed assuming the same switching frequency
for each switching strategy.
In order to carry out this comparison, the switching fre-
quency should appear in (29) instead of T
C
, yielding
RMS
s f T
o
RMS s
Y
f C L
i
I =
2
,
(33)
The values of
RMS
Y can be easily calculated through Tab. IV
once the switching strategy is specified.
Figs. 8a, 8b and 8c illustrate the behavior of
RMS
Y

as func-
tion of q, for different value of
o
and assuming the same
switching frequency. It can be noted that in this case Strat-
egy 4 determines the lowest ripple for q < 0.7, whereas for
high values of the voltage transfer ratio Strategy 2 and Strat-
egy 3 perform better.
VII. CONCLUSIONS
Space Vector Modulation technique allows the full control
of the output voltages and the input power factor of matrix
converters. Several switching strategies based on SVM tech-
nique, differing from one another in the use of the zero con-
figurations, can be adopted. The possibility to distribute the
zero configurations within the switching pattern provides
two degrees of freedom that can be utilized to reduce the
number of commutations in each cycle period or to improve
the line current quality.
In this paper seven switching strategies have been compared
in terms of the rms value of the line current ripple, assuming
the same cycle period or the same switching frequency.
For this purpose, an analytical approach has been firstly de-
veloped, showing that, using p.u. variables, the number of
independent parameters affecting the current ripple can be
sensibly reduced. Then, an integral quantity representing the
rms value of the line current ripple in a long period has been
introduced, allowing an effective comparison among the
seven switching strategies.
Using the proposed method, it is possible to avoid several
tedious and time consuming numerical simulations, related
to the infinite combinations of supply and input filter pa-
rameters, output current and cycle period.
On the basis of the same cycle period, Strategy 4, which uses
the zero configurations located at the sides of each half of
the switching pattern, exhibits the best performance. Despite
of the higher switching frequency, Strategy 7, which uses the
three zero configurations equally distributed, may determine
a higher line current distortion.
On the basis of the same switching frequency, Strategy 4
performs better than the other strategies, except for high
values of the voltage transfer ratio.
Numerical simulations have confirmed the validity of the
TAB. IV RELATIONSHIP BETWEEN
RMS
Y AND
RMS
Y
Strategy number Relationship
1, 2, 3
RMS RMS
Y Y 64 =
4, 5, 6
RMS RMS
Y Y 100 =
7
RMS RMS
Y Y 144 =
a) b)
Fig 7 - Line current waveforms for Strategy 2 (a) and Strategy 4 (b).
System parameters: Rs = 0.25 , Ls = 0.4 mH, Lf = 0.6 mH, Cf = 10 F, q= 0.5, io= 15 A, TC = 250 s.
a)
0
2
4
6
8
10
12
14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
q
b)
0
2
4
6
8
10
12
14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
q

c)
0
2
4
6
8
10
12
14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
q
Strat. 1 Strat. 2,3 Strat. 4 Strat. 5,6 Strat. 7

Fig. 6 - Behavior of
RMS
Y

for the seven switching strategies as a function

of q, for o=15 (a), o=30 (b) and o=60 (c), on the basis of the same
cycle period.
analytical approach.
The results obtained can be useful for the filter design and
for the optimization of the converter control.
VIII. REFERENCES
[1] A. Alesina, M. G. B. Venturini: Solid-state power conversion: a Fou-
rier analysis approach to generalized transformer synthesis. IEEE
Trans. on Circuit and Systems. Vol. CAS-28, No. 4, April 1981, pp.
319-330.
[2] M.J. Maytum, D. Colman: The Implementation and Future Potential
of the Venturini Converter, in Proc. Drives, Motors and Controls,
1983, pp. 108-117.
[3] A. Alesina, M. G. B. Venturini: Analysis and design of optimum-
amplitude nine-switch direct AC-AC converters. IEEE Trans. on
Power Electronics. Vol. 4, No. 1, January 1989, pp. 101-112.
[4] G. Roy, G.E. April: Cycloconverter Operation Under a New Scalar
Control Algorithm, in Proc. IEEE PESC. Conf., Milwaukee (WI),
1989, pp. 368-375.
[5] P.D. Ziogas, S.I. Khan, M.H. Rashid: Analysis and Design of Forced
Commutated Cycloconverter Structures with Improved Transfer
Characteristics, IEEE Trans. on Industrial Electronics, vol. 1E-33,
No. 3, pp. 271-280, August 1986.
[6] L. Huber, D. Borojevic: Space Vector Modulator for Forced Commu-
tated Cycloconverters, in Proc. IEEE PESC Conf., , San Diego,
(USA) pp. 871-876, 1989.
[7] L. Huber, D. Borojevic: Space Vector Modulation with Unity Input
Power Factor for Forced Commutated Cycloconverters, in Proc.
IEEE IAS Conf., pp. 1032-1041, 1991.
[8] L. Huber, D. Borojevic: Space Vector Modulated Three-Phase to
Three-Phase Matrix Converter with Input Power Factor Correction,
IEEE Trans. Industry Applications, Vol. 31, No. 6, pp. 1234-1246,
November/December 1995.
[9] D. Casadei, G. Grandi, G. Serra, A. Tani: Space Vector Control of
Matrix Converters with Unity Input Power factor and Sinusoidal In-
put/output Waveforms. Proc. EPE Conference, Brighton (UK), 13-16
September 1993, vol. 7, pp. 170-175.
[10] D. Casadei, G. Serra, A. Tani: Reduction of the Input Current Har-
monic Content in Matrix Converters Under Input/Output Unbalance.
IEEE Trans. on Industrial Electronics, vol. 45, No. 3, pp. 401-411,
June 1998.
[11] D. Casadei, G. Serra, A. Tani, L. Zarri: Matrix Converter Modulation
Strategies: a new General Approach Based on Space Vector Represen-
tation of the Switches State. IEEE Transactions on Industrial
Electronics, Special Issue on Matrix Converters, vol. 49, no. 2, April
2002, pp. 370-381.
IX. BIOGRAPHIES
Domenico Casadei received the Laurea degree
with honors in Electrical Engineering from the
University of Bologna, Italy, in 1974.
He joined the Electrical Engineering Department,
University of Bologna, in 1975. He is currently
Professor of Electrical Drives. His scientific work
is related to electrical machines and drives, linear
motors, Maglevs, and power electronics. He has
published extensively in technical journals and
conference proceedings. His present research
interests include direct torque control of induction motors, brushless mo-
tors, matrix converters and power quality. Since 1994 he has been a mem-
ber of the international editorial board of the International Journal
ELECTROMOTION.
Dr. Casadei is a member of the IEEE Industrial Electronics and IEEE
Power Electronics Societies and the Italian Electrotechnical and Electronic
Association (AEI). He is a Registered Professional Engineer in Italy.

Giovanni Serra was born in Bologna, Italy, in
1950. He received the Laurea degree, with
honors, in electrical engineering from the Univer-
sity of Bologna, Bologna, Italy, in 1975. Follow-
ing service in the italian army, he joined the
Department of Electrical Engineering, University
of Bologna, first as a recipient of a Fellowship of
the Consiglio Nazionale delle Ricerche, then as a
Research Associate and, since 1987, as Associate
Professor. He is currently professor of Electrical
Machines at the Department of Electrical Engineering, University of Bolo-
gna. His fields of interests are electrical machines, electrical drives and
power electronic converters. His current activities include, direct torque
control of ac machines, linear motors, ac/ac matrix converters. He has
published more than 90 papers in technical journals and conference pro-
ceedings.
Dr. Serra is a member of the IEEE Industry Applications and IEEE Dielec-
trics and Electrical Insulation Societies and the Italian Electrotechnical and
Electronic Association (AEI). He is a Registered Professional Engineer in
Italy.

Angelo Tani was born in Faenza, Ravenna, Italy,
in 1963. He received the Laurea degree, with
honors, in electrical engineering from the Univer-
sity of Bologna, Bologna, Italy, in 1988.
He joined the Department of Electrical Engineer-
ing, University of Bologna, in 1990, where he is
currently a Research Associate. His scientific
work is related to electrical machines and drives,
power electronics and Maglevs. He has published
more than 50 papers in technical journals and
conference proceedings. His current activities include ac/ac matrix con-
verters and direct torque control of induction motors.

Luca Zarri was born in Bologna, Italy, in 1972.
He received the electrical engineering degree with
honors from the University of Bologna in 1998.
He worked as plant designer until 2002 for a civil
engineering company. In 2003 he has joined the
Department of Electrical Engineering, University
of Bologna, with which he has been collaborating
since 1998. His scientific work is related to
electrical drives and power electronics. His
research interests include ac/ac matrix converters
and direct torque control of induction motors. He is a registered profes-
sional engineer in Italy.
a)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
q

b)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
q

c)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
q
Strat. 1 Strat. 2,3 Strat. 4 Strat. 5,6 Strat. 7

Fig. 8 - Behavior of
RMS
Y

for the seven switching strategies as a function of

q, for o=15 (a), o=30 (b) and o=60 (c), on the basis of the same
switching frequency.