Multi-phase Inverter Analysis

John William Kelly

Elias G. Strangas

John Michael Miller

Electrical and Computer Engineering

Electrical and Computer Engineering

Michigan State University
East Lansing, MI 48824
Email:strangas@egr.msu.edu

Ford Motor Company

Vehicle Electronics Systems Department
Dearborn, MI 48121

Michigan State University

East Lansing, MI 48824
Email:kellyjo2@egr.msu.edu

Abstract- Pole-phase modulation adjusts the pole-phase

ratio of an induction machine. This requires a multi-leg,
multi-phase inverter.This paper analyzes an n-leg, n-phase
inverter and presents several techniques for n-phase Space
Vector Pulse Width Modulation.
The n-leg inverter is examined in terms of its (n - 1)/2 load
circuit configurations. Each load configuration is analyzed
to determined the ratio of phase to source current. Next,
the waveforms resulting from n-step mode operation for each
configuration are presented.
There are 2n space-vectors for a n-phase system. These
space-vectors can divided into (n - 1)/2 sets which correspond t o the inverters load circuit configurations. Within
each set, there exists a maximum magnitude subset of spacevectors. This paper presents several SVPWM techniques
that are based on the use of these vectors.
KeywordInverter, Space Vector Pulse Width Modulation, Pole Phase Modulation

I. INTRODUCTION
Although a 3-phase 3-leg inverter is most commonly used
to power AC machines, there are applications were a nleg, n-phase (n > 3) inverter is needed. One such application is where pole-phase modulation is used to control
an induction motor. Pole-phase modulation extends the
speed-torque range of an induction motor by adjusting the
pole-phase ratio. For example, a nine-phase symmetrically
wound induction machine can operate as either a four- or
twelve-pole machine. This is accomplished by reconfiguring
the nine-leg inverter as either a nine-phase or a three-phase
voltage source. Extended speed-torque range is desired for
several automotive induction motor applications. The purpose of this paper is to provide a comprehensive analysis
of a multi-leg, multi-phase inverter and propose techniques
for its implementation.
First, the n-leg, inverter will be examined in terms of
the load equivalent circuit during a 2?/n switching period.
Unlike the three-phase inverter, which has only one load
equivalent circuit, a n-leg inverter has (n - 1)/2 load equivalent circuits. Each one of these equivalent circuits can be
operated in a n-step mode to produce a unique discrete ac
or stair-step voltage waveform. These waveforms differ
in terms of harmonic content and magnitude of the fundamental component. As shown by Holtz [3] the fundamental component of the three-phase 6-step voltage waveform
is 2vdc/7r. On the other hand, a nine-phase inverter has
four different such voltage waveforms, with fundamental
components of O.64Vdc, O.56Vdc, 0.41Vdc, O.22Vdc.
For most electrical machines, the windings are assumed
sinusoidally distributed in the stator. Voltage space-vectors
are defined based on the voltages applied to the windings.
For the conventional three-phase inverter, connected to an

0-7803-709
1-O/O1/$ 10@200
1 I EEE

electrical machine, there are eight space-vectors possible.

For a n-leg, inverter connected to a n-phase machine, there
are 2 possible space-vectors. The relationship between the
(n - 1)/2 load equivalent circuits and the 2n space vectors
will be established. This relationship will aid in the Pulse
Width Modulation implementation of the multi-leg, multiphase inverter.
The implementation of PWM in a multi-leg, multi-phase
inverter is examined. In particular, a generalized implementation method for Space-Vector PWM is presented.
The theory of Space-Vector PWM and its implementation
through the Unified theory are extended to many phases.
Additional novel algorithms, particular to multi-phase machines are presented and compared.
Through out this paper, a nine-leg, nine-phase inverter
will be used as a practical example. In the final section,
experimental data from a nine-phase SVPWM system will
be presented.

11. ANALYSIS
OF NINE-LEG
INVERTER
The nine-leg inverter of figure 1 requires eighteen gating
signals, gl-g18. The gating combination determines the
phase voltage across each of the nine separate phases of
the load, which are connected at a point to form a star.

Fig. 1. Nine-leg inverter

Depending on the gating pattern, the nine phase loads

can form four different equivalent circuits. For example,
when the gating pattern is [gl, g3,95,97,g18,92,94, g6,g8]
the load equivalent circuit consists of a group of four parallel impedances in series with a group of five parallel
impedances as shown in Figure 2-a.
Likewise, the 3-6 equivalent circuit configuration, figure 2-b, is formed by three parallel impedances in series with a group of six parallel impedances. One of
the possible inverter gating patterns for this case is
[gl, g3,95, g16,g18,92,94, g8] or, in terms of energized inverter legs, [11100000]. In this switching state 1 represents
the upper switch of the inverter leg closed, on, and the
147

phase current and the smallest ratio of phase current to

source current. For the 1-8 configuration, one inverter leg
carries the entire source current, while the other eight legs
carry only 1/8th of it.

1I-

B. n-Step Inverter Operation

a) 4-5 Coifigiration

. . .
c) 2-7 Configuration

In a three-phase inverter operating in six-step mode,

switching states are changed every sixth of a cycle. The
four potential phase voltage levels are fVdc/3 and &2V&/3.
The sequence of the six switching states is such that only
one switch changes state. The sequence of switching states
for six-step operation is shown below.

b) 3-6 Configuration

d) 1-8 Configuration

Fig. 2. The four equivalent load configurations of a nine-leg inverter

lower switch open, 'off'. The other two possibilities are the
2-7 configuration and the 1-8 configuration, figure 2-c,d.

A. Load Circuit Analysis

For an impedance of 2 for each leg, the total equivalent

1
1
1

0
0
1

1
0
0

0
0

1
0

1
1

This sequence produces three phase voltages that are displaced 120'and that appear in 'stair-step' shape, as shown
in figure 3. For the nine-leg inverter a phase can attain

impedance for the 4-5 configurations is:

-Vd3

-2vs

For a voltage supply of

vdc,

the source current is:

-v,/3
-2V,/3

The maximum phase current possible in any one leg is:

(3)
Table I shows the equivalent impedance, source current
and the maximum phase current for the four different load
circuits. The corresponding quantities for a three-phase
inverter were included for comparison. Of the four possible

TABLE I
EQUIVALENT
CIRCUIT
COMPARISON

load circuit configurations, the 4-5 one results in maximum

148

Fig. 3. Phase voltages of a three-phase inverter

on of sixteen possible voltage levels. However, each of the

four possible load circuit configurations produces only four
distinct voltage levels at the phases. For example the 4-5
configuration results in phase voltage levels of f4Vdc/9 and
f5Vdc/9.
Because only four voltage levels can exist in one conduction interval, eighteen-step operation requires the inverter to operate using only one circuit configuration for
the entire cycle. Figure 4 shows the upper gating signals
for eighteen-step operation in the 4 5 configuration, dong
with the resulting nine phase voltages.
The numbers above the gating signals represent the
switching state. The switching state for the first 20" conduction interval is [100000111]. Note, that only one switch
changes states between consecutive conduction intervals.
Also shown in figure 4 are the resulting phase voltages
(Va...K). Similarly for the other three load circuit configurations, figures 5 6 and 7 show the respective gating
patterns and phase voltage waveforms for eighteen-step operation.

Fig. 4. 4-5 eighteen-step operation: gating signals(upper switches)

and the resulting phase voltages
Fig. 6. 2-7 eighteen-step operation: gating signals (upper switches)
and one of the resulting phase voltages

Fig. 5. 3-6 eighteen-step operation: gating signals (upper switches)

and one of the resulting phase voltages

C. Inverter Switching Eficiency

In the mode of figure 5, three switches change state between conduction intervals. In the mode shown in figure 6,
five switches change states. Likewise, in the mode of figure 7, seven switches change state. An inverter operating
continuously in the 4-5 switching set would have greater
switching efficiency than one operating in the 1-8 switching mode. More switches have t o change states to change
voltage levels from 1Vdc/9 to 8Vdc/9 than from 4Vdc/9 to
5vdc/9.

D. Harmonic Content
For a three-phase inverter, switching every 60, the phase
voltage has a fundamental of 2/7rVdc. Table I1 shows the
result of a fundamental magnitude of the waveforms of figures 4,5, 6 and 7.
For a nine-leg inverter operating in eighteen-step mode,
the 4 5 switching set waveform has the highest fundamental
content. The 1-8 switching set has the lowest fundamental
content.

Fig. 7. 1-8 eighteen-step operation: gating signals (upper switches)

and one of the resulting phase voltages

111. EXTENSION
OF SPACE VECTOR THEORY
TO 12
PHASES
Symmetric time-varying electrical quantities operating
on spatially distributed functions (e.g. windings) give rise
to the concept of space-vectors. In an AC machine stator, voltages are assigned a direction (that of the winding
spatial distribution) and their vectorial sum defines the corresponding voltage space-vector.

A . Three Phase Space- Vectors

Figure 8 shows the voltage space-vector directions for
each phase and the resultant voltage space-vector $ for a
three-phase system.
For symmetric three-phase sinusoidal voltages, the
space-vector
sweeps a continuous circular path. It rotates at the frequency of the terminal voltages. If, however,
a three-leg inverter supplies the voltage, there are only six
angular positions the voltage space-vector can take at any
instance. The three-leg inverter results in 23 possible space

149

Fundamental
n=l

3" order
n=3

n=5

3 phase

0.6366=2/pi

0.1273

0.0909

Four-Five
Three-Six
Two-Seven
One-eight

0.636bUpi

0.2122
0
0.2122

0.1273
0.1715

0.0909

0.5598
0.4155

0.221 1

0.2122

Configuration

5IhOrder

7* Order

NineQhaseVdlageSpace V m o e

./

/.

I.

;.

/.

/.

!.

0.2303

0.0677

0.2619

0.1951

0.1709

TABLE I1
HARMONIC
CONTENT

OF 18-STEP MODE VOLTAGE WAVEFORMS

Fig. 10. Nine-phase voltage space-vectors

Fig. 8. Phase voltages of a three-phase inverter and resultant spacevector voltage

vector positions. The six non-zero vectors form a hexagon,

shown in figure 9.

to switching states from the [4-51 switching set. Specifically, these space-vectors correspond to the 18 switching
states of the (4'5 muz} subset. This subset consists of
only those switching states where the switches operate 40"
out of phase. The {4,5 muz} switching states are shown
below.

{ 0 0 0 0 0 1 1 1
0 0 0 0 1 1 1 1

1
0
(5)

P-'

'
8

'

0 1 1 1 1 1 0 0 0
1 1 1 1 1 0 0 0 0 }

f .

A space-vectors of figure 10 is the resultant vector of

phase space-vectors. The phase space-vectors are displaced 40 with respect to each other. Therefore, each of
these space-vectors contributes positively t o the resultant
voltage space-vector. For the inverter switching state of
[000011111], the resultant voltage space-vector is derived
from (6).
These six space-vectors, are from (4).

E] 2
=

[e::]
e 3

-1 -1
1
1
-1
1 -1
1 -1
1 -1 -1 -1
1

-:I

(4)

B. Nine Phase Space- Vector

For a nine-phase machine fed by a nine-leg inverter, there
are 2' possible voltage space-vectors. Figure 10 shows the
possible positions of the resultant voltage space vectors.
The 512 resultant space-vectors of figure 10 can be divided into four sets,[45],[3-6],[2-7],[1-81,
based on the load
circuit configuration as discussed in section 11. These four
sets can further be divided based on the number of adjacent 'on' inverter switches. For example, the eighteen
outermost resultant space-vectors of figure 10 are the result of switching combinations of either four or five 'on7
switches. Thus these eighteen space-vectors correspond
150

where Q = e j % . Figure 11-a, shows the resultant voltage

space-vector for the switching state [000011111] and the
addition of phase voltage space-vectors. Figure 11-b shows
the resultant voltage space-vector for the non-adjacent
switching state [000101111]. The magnitude of this resultant voltage space-vector is less than the magnitude of the
voltage space vector of switching state [000011111].
Space-vectors corresponding to the four maximum magnitude subsets form four 18-sided polygons shown in figure

-are presented.

A . Three-phase Space-Vector PWM

-

Three-phase SVP-WM, [3], increases the DC bus utilization over PWM based on Sine-Triangle comparison [l]. By
injecting third-harmonic into the reference voltage signal,
the fundamental of the phase. voltage can be increased.
SVPWM is: implemented by averaging-the. time spent in
adjacent switching' states.
-

[I101

7. .. ..

,,

a)[O0001 1 1 1 111

UW

+--

b)-[OO010 1 1 1 1 11

Fig. 11. Vector addition of voltage phase space-vectors

Fig. 13. Three phase Space Vector PWM

12. The polygons enclose all the subsets of the four switching sets.

In figure 13 the desired voltage v* is realized by spending

t l time in switching state [loo], t2 time in switching state
[110] and to in the zero switching state.
The drawback of this algorithm is the requirement for
trigonometric calculations. A method proposed by Chung
[2] simplifies the implementation of SVPWM. Chung
showed that the third harmonic injection can be accomplished by adding a 'voltage offset' t o the voltage reference signals. This voltage offset is the average between
the maximum and minimum voltage reference signal. The
addition of this offset, centers the three voltage reference
signal about the dc axis, allowing for all three signals to
be increased. The operation requires only addition and
multiplication operations. This algorithm is extended for
n-phase systems in section IV-B.

B. Extension of three-phase SVPWM to n-Phase SVPWM

Fig. 12. The four bounding voltage space-vector subsets of a ninephase system

In general, the number of polygons is equal t o the number of equivalent load circuits. In section IV it will be
shown how vectors that form these concentric polygons are
used in a multi-phase Space Vector Pulse Width Modulation algorithm.
IV.

SPACE-VECTOR
PWM

For a n-phase system, nth harmonic signal can be injected by adding V, o f f s e t to the voltage reference.
tending the Unified method, this offset signal is

v, offset = max

[".I

vdc

- min
vdc

[".$I

vdc

(7)

Figure 14shows the block diagram for the n-phase Space

Vector PWM implementation.

IMPLEMENTATION

Pulse Width Modulation is a general technique for realizing a voltage reference via an inverter. The duration
of the gating pulses controls the frequency and magnitude of the inverter output voltage. Space Vector Pulse
Width Modulation (SVPWM) is a PWM technique which
uses space-vectors to generate these gating duration. Section IV-A reviews three-phase SVPWM. In section IV-B
a proposed general algorithm is presented, which extends
present three-phase SVPWM to n-phase SVPWM. In section IV-C,several SVPWM techniques, specific to n-phase

Fig. 14. Block diagram of Space Vector PWM Algorithm

Realizing a voltage space-vector by applying this a l p

rithm to a nine-phase inverter, results in the switching
151

states shown in figure 15. Also shown in figure 15 are the

nine upper gating pulses.

. d l 1

I 1

1 1

1 1

1 1

l l e

1 1 1

1 1

1 1 1 1 1 1 1

111

1 1

11

v
3 6 max

V*-

K[(ts + t 3 e j g ) ]

VTotal

Fig. 15. The Space Vector PWM Realization of the vector l.0G300

The switching states generated by this algorithm are exclusively from the four maximum subsets described in section 111. Also apparent in figure 15 is that only one switch
changes state at a time. Figure 16 shows the switching
state sequence, for half a sampling period, in terms of corresponding space-vectors.

In (15), fs is the sampling frequency and Vdc is the dc

bus voltage. The term 1/3 adjusts for the number of turns
per phase. A winding of a nine-phase system has a third
the number of turns per phase than a winding in an equivalent three-phase system. We define as equivalent systems
those that produce equal air-gap flux for the same voltage. Figures 17a,b illustrate the placement of windings for
a machine designed to operate as either a nine-phase or
three-phase system. This machine will produce the same
flux in either configuration.

Fig. 16. Vector sequence for the realization of the voltage space
vector 1 . 0 ~ 3 0 ~

The algorithm of figure 14 can also be described in

terms of switching times derived from space-vector averaging. In the three-phase system there are three switching durations that have t o be determined, [to tl t z ] . For
a nine-phase system there are nine such switching durations,. [to tl t 2 t 3 t 4 t5 t 6 t 7 ] . The voltage space-vector
V e j a , with 0 < a 5 ~ / 9 can
,
be realized by solving the
following:
V*

=2fs [Vi,s m a z ( t l e a 2
+

v3,6
max(t7ea2

+ tge@) + ~ 2 , 7m a x ( t 3 e a 2 + t g e j a l )

+tzeja1) +v ~ m
, a~x ( + 5 e ~ 2+ t 4 e 3 a i ) ]

(8)
1

to = - - tl
2fs

- t 8 - t 2 - t7

-t 3 - t6

- t4 - t5

T~ solve (refeq:a16)we add the constraints which

guarantee the use of space-vectors from the subsets
{ h ,m~
ax},{V3,6 max},{V2,7 maz},{&,8
Two constraints are that cy1 = 0" and a2 = 20". This results in
utilizing only space-vectors
that are linearly proportional
to the space-vectors of { V 4 , 5 m a x } - . - { V 1 , 8 m a x } . A third
constraint is the time spent at each switching state will
be proportional to the amplitude of the space-vector from
152

Fig. 17. a)Windings of a 3/9 phase machine in 3-phase operation b)9phase operation c)Windings of a conventional 3-phase machine

Figure 17c shows the winding configuration of a conventional three-phase machine with the same number of
turns. The resultant winding distribution factor for one
phase in figure 17a will be less than the winding distribution factor of figure 17c. Therefore, the resulting voltage
space-vectors will have lower magnitude. For a conventional three-phase machine the six possible non-zero volt-

y,

age space-vectors have a magnitude of

while for the
three-phase machine Of figure 17a the voltage vectors have
a
of2/9(2cos(2O0) -+ l>Vdc 2 / r *
We define the modulation index as the ratio of the fundamental of the output voltage, V I ,to the dc input voltage.
m = -Vl
vdc

(16)

When SVPWM is applied to the conventional three-phase

machine of figure 17c, the maximum possible modulation
index is
m3q1maz

..fi
=

(17)

For the equivalent system of figure 17a, the maximum possible modulation is

When the same machine is operated'in nine-phase, using

the method described in (8), the maximum modulation is
mg+ maz M

0.507

4
0.02

0.04

0.06

Fig. 19. Reference Voltage signals and respective voltage offset for
nine and three phase SVPWM

(19)

The extended SVPWM scheme of figure 14 was implemented using a dsp processor, a field programmable gate
array (FPGA), a nine-phase inverter and a nine-phase load.
The dsp supplied the FPGA with nine voltage commands.
The actual SVPWM algorithm was implemented on the
FPGA, which supplied eighteen pulses to gating circuitry.
Figure 18 is a plot of four of the nine phase voltages of an
RC load.

0.08

0.1

seconds

Fig. 18. Phase voltage of a 9-phase load generated by a 9-phase

SVPWM scheme

As shown in figure 18, the ninth order harmonic is

present in the phase voltages. This harmonic injection increases the fundamental content of the waveform and is
filtered in a nine-phase load. However, the increase in fundamental content is not as great as third order harmonic
injection is in the three-phase system, 0.507 vs. 0.577. In
figure 19 are shown the reference phase voltage switching
times and the voltage offsets (normalized switching times)
for the three and nine-phase systems.
As the number of phases increases, the magnitude of the
voltage offset decreases t o zero and SVPWM converges to
Sine-Triangle PWM.

C. Proposed n-phase SVPWM Techniques

In order to increase the phase voltage fundamental in
a n-phase SVPWM system, we propose several SVPWM
techniques. The first technique allows for lower order harmonic injection into a system with many phases. Another
proposed SVPWM technique uses only those vectors that

form the concentric n-sided polygons described in section

I11 and the zero space-vectors. The third technique is related to the second one, as it uses the same vectors; however, the zero space-vectors is not always used. Eliminating
the zero space-vector results in decreased switching losses.
C.1 Low order Harmonic injection

The proposed technique allows for the fi order harmonic to be injected into a n-phase system. The advantage
of this technique is the increase fundamental content of the
output voltage due t o the lower order harmonic injection.
Howerver, f i separate neutrals have t o be used.
In the case of nine-phase system, the nine-phase symmetrical load is treated as three, three-phase systems displaced
40, each producing one third of the total voltage spacevector. Conventional three-phase SVPWM is applied to
each of the three systems to realize the voltage command.
When operating with one neutral, experimental results
showed the phase current had a significant amount of harmonic distortion. Figure 20 shows one of the nine phase
currents.

Fig. 20. One of the nine distorted phase currents

This harmonic distortion is caused by circulating currents in the load. These currents are a result of unbalanced
3rd order harmonic voltages. Each of the three three-phase
SVPWM algorithms injected a 3rd harmonic component
that was forty degrees with respect t o each other. Figure
21 shows three of the nine voltage reference signals and
their respective offset.
As seen in figure 21, the three voltage offsets are displaced 40, resulting in the three injected 3rd harmonic
voltages being out of phase, resulting in circulating currents. This problem can be solved by separating the ninephase load neutral into three isolated neutrals.
153

The switching times for this scheme can be determined

by reducing (8) to (13) with V T o t a l = V4,5 ma5.
The advantage of using only vectors of the subset
{ V 4 , 5},,,
is that the inverter operates in the 4-5 equivalent load configuration which results in a greater modulation factor.
Continuously operating in the 4-5 configuration increases
harmonic content, see tableII. If switching states from the
{ v 3 , 6},,,
subset are included, the harmonic content is
reduced. However, the maximum modulation is also reduced. The switching times are determined by solving (13)
and (131, with VTotal = v 4 , 5 max v 3 , 6 m a x .

Fig. 21. Reference and offset voltages for a three and nine-phase
SVPWM systems.

C.2 SVPWM Using Maximum Subsets

Equation 8 is based on the use of a limited number of
switching states, using only those space-vectors located at
angles e ( ~ ~ ) , e ( ~ ~ ) Projecting
, . . . ~ ( the
~ ~desired
) .
voltage
space-vector onto those vectors from the maximum magnitude subsets, described in 111-B further reduces the number
of switching states. This technique reduces the number of
possible space-vectors from 512 t o 74. Within a 20' sector
two space-vectors from each maximum subset, along with
two zero space-vectors, are used to realized a voltage. However, the desired voltage space-vector can also be realized
by using just two non-zero and two zero vectors. Figure 22
shows how the same voltage space-vector is realized using
four such vectors.
24

Fig. 22.
{V4,}5
,,,

C.3 Modified SVPWM with adjusted V,,,,

In previously mentioned n-phase SVPWM schemes the

zero space-vectors are used. This is a necessity for threephase SVPWM, because at least three switching states are
needed t o realize a desired space-vector. The use of zero
voltages in the SVPWM scheme of section IV-C.2 result
in four switches being simultaneously changed. In order
t o avoid this and improve switching efficiency, a non-zero
vector can be used. In general, the closer in proximity the
space-vectors are, the greater the switching efficiency.
The previous SVPWM techniques have eighteen switching sectors. This technique subdivides each sector into four
regions as shown in figure 10. The angle of the desired voltage space-vector determines the sector, while its magnitude
determines the region within that sector, for a possible 74
switching sequences. Within each region the voltage spacevector is realized by the space-vectors located at the indices
of that region.

Voltage space-vector realization using vectors from

subset and zero vectors

The maximum modulation in this case is

2
= (-)(COS(lOo))
= 0.63
71

mg$

(20)

The switching combination is shown below. During one

sampling period 18 switches change state. This is the same
number of switches that change state for the extended
three-phase SVPWM technique discussed in section IV-A.

0
1
1
1
1
1
0

154

0
0
0
1
0
0
0

0
0
0
1
0
0
0

0
0
0
1
0
0
0

0
0
0
1
0
0
0

0
0
1
1
1
0
0

0
1
1
1
1
1
0

0
1
1
1
1
1
0

0
1
1
1
1
1
0

Fig. 23. Switching schemes for adjacent sector regions

(21)

For the shaded region of figure 23, a desired voltage

space-vector can be realized by the four vectors located

V. CONCLUSIONS

at the indices. The switching sequence is

0
1
1
1
1
1
1
0

0
0
0
1
1
0
0
0

0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0

0
0
1
1
1
1
0
0

1
1
1
1
1
1
1
1

1
1
1
1
1
1
1
1

1
1
1
1
1
1
1
1

Figure 24 depicts the corresponding space-vector sequence.

/
i'

20

Fig. 24. The vector sequence for realizing V'

Only six switches change state, compared t o eighteen

switches for the SVPWM schemes of sections IV-C.2 and
IV-B. The switching sequences of figure 23 result in a
maximum of two switching changing between adjacent regions. For example, one switch changes states when the
desired vo.ltage vector changes from from the region with indkes of [v4,5max ,v4,5 m a x , v3,6 .ma", v3,6 rnaz] t o a region
of [v2,7m a z , v 2 - 7 max, &,6 max, v3,6 ma,].

Pole-phase modulation requires the use of a multi-phase

system since the phase order is adjusted as well as the number of poles. For example, the phase order is increased from
three phases to nine phases when switching from twelve
to four poles. The nine phases result in a smoother airgap flux distribution. The necessity of implementing nine
phases via a nine-leg inverter resulted in the proposal of
several novel n-phase SVPWM techniques.
The first technique was the extension of conventional
three-phase SVPWM t o high order phase systems. However, this resulted in lower utilization of the dc bus voltage. The second technique treated the multi-phase system
as a sum of lower order phase systems. This technique resulted in higher dc bus utilization through the injection of
harmonics. However, it is no longer possible t o have one
neutral.
The third proposed technique utilizes the fact that a
multi-phase system has more than one equivalent load configuration. Along with the zero space-vectors, this technique uses only space-vectors from a maximum subset (or
a combination of maximum subsets) to realized a desired
voltage.
Like the third technique, the fourth technique uses only
a limited number of switching states, however, it does not
necessarily use the zero space-vectors, improving switching
efficiency.

REFERENCES
[l] A.B.Plunkett. A current-controlled pwm transistor inverter drive.
Conference Recording IEEE/IAS Annual Meetmg, pages 786-

792, Oct 1979.

[2] D. W. Chung, J. S. King, and S. K . Sul. Unified voltage
modulation technique for real-time three-phase power conversion. IEEE h n s a c t r o n s on Industry Applzcatrons, 34(2):374380, March/April 1997.
[3] J . Holtz, P. Lammert, and W. Lotzkat. High-speed drive system
with ultrasonic mosfet pwm inverter and single-chip microprocessor. IEEE Ransactaons on Industry Applzcatzons, 23(6):10101015, November/December 1987.

155