Paper

A Novel Voltage Modulation Technique of the Space Vector PWM

Non-member Joohn-Sheok Kim (Seoul National University, Korea)

Member Seung-Ki Sul (Seoul National University, Korea)

In this paper, a novel voltage modulation scheme is described. With the effective voltage concept, the actual
switching times for each inverter arm are deduced directly as a simple form. With the help of the presented zero
sequence voltage allocation algorithm, the proposed PWM method has the high performance voltage generation
capability exactly same as the conventional space vector PWM method with reduced effort. As well as giving a
detailed explanation of the new PWM algorithm, the paper presents the comparison results with the conventional
method such as space vector PWM and sinusoidal PWM method. In the proposed PWM method, the execution
time can be reduced more than 25% as compared with the conventional space vector PWM method, and memory
size will be minimized to 15% of the conventional PWM method.

Key word: Space Vector PWM.

1. Introduction
Due to the improvement of fast-switching power
semiconductor devices and machine control algorithm, more
precise PWM(Pulse Width Modulation) method finds
particularly growing interest. A large variety of methods for
PWM exists on which a survey was recently given.(1) For
the ac machine drive application, full utilization of the dc
bus voltage is extremely important in order to obtain the
maximum output torque under all operating conditions. In
this aspect, compared with any other PWM method for the
voltage source inverter, the PWM method based on voltage
space vectors results in excellent dc bus utilization. (1,2)
Moreover, compared with the sine-triangle PWM method,
the current ripple in steady state operation can be greatly
reduced by this method. (2,3)
However, there is one disadvantage in the space vector
PWM method, which decreases the applicability of this
powerful modulation technique. In the conventional space
vector PWM method, the actual switching times are
determined on the foundation of the eight switching states of
the inverter. Thus, in order to generate the actual switching
pattern, two nearest active voltage vector should be
predetermined according to the location of the reference
voltage vector. Then, using the information of the reference
vector's location once more, the actual switching times is
calculated through the recombination task for the effective
vectors. Therefore, with the point of view of the practical
implementation, the conventional space vector PWM method
is very complex and needs longer calculation time than any
other methods.
In this paper, a novel voltage modulation technique is
described. From the concept of 'effective voltage', the actual
switching times for each inverter arm are deduced directly as
a simple form. Moreover, with the help of the presented zero
sequence voltage relocation algorithm, the proposed PWM
method has the high performance voltage generation
capability exactly same as the conventional space vector
PWM method. Using the proposed PWM algorithm to ac
machinedrives,the accurate switchingpattern can be easily
implementedwith reducedeffort.Moreover,in the caseof a
slowercontroller,and the limitationof executiontime canbe
removedto somedegree.
As well as giving a detail explanation of the new PWM
algorithm,the paper presentsthe comparisonresultswiththe
conventional method such as space vector PWM and
sinusoidalPWM method. The performanceof the proposed
PWMmethodhas beenprovedby laboratorytests.
2. Conventional Space Vector PWM
In Fig.1(a), the typical power stage of the three phase
inverter and the equivalent circuit of a machine are
presented. And, in Fig.1(b), the available eight different
switchingvectors of the three phase inverter are depicted
with spacevectorconcept.(1'2)If the upper switchingdevice
of one arm is turn on, this state representedby '1', and in the
lower device turn-on case this state representedby '0'. For
example,the (100) switchingstate called by voltagevector
'Vi' means that upper switchingdevice of A-phase arm is
turn-on and the lower switchingdevicesof B-phase,C-phase
arms are turn-off. Note that all the machine terminals are
connectedto each other electricallyand no effectivevoltages
are applied to machinewhen the 'zero vector'presentedby
V0 and V7 is selected.When one of the other six voltage
vectors is selected, an 'effective voltage' is applied to
machine,and these vectorsare calledas 'effectivevectors'in
this paper.
If a constant referencevoltage vector V* is given in one
sampling period, this vector can be generated using zero
vector(V0 or V7) in combination with only two nearest
effectivevectors(Vn and Vn+i). From the average voltage
conceptduringone sampling period, the referencevectorcan
be written as follows.

820 T. IEE Japan, Vol. 116-D, No .8, '96

 A Novel Voltage Modulation Technique of the Space Vector PWM

(Superscript's'denotes stationary frame value. And, m:sector

number (1,2....6), if m=1 then m-1=6. And θ is the voltage

vector angle on the stationary frame. So θ=α+ ?? (m-1).)

Note that, in fact, the effective times doesn't imply the

actual switching times. The switching time complies with
the time delay from the initial point of one sampling period
to the activation time of a switching device. Therefore, in
order to evaluate the active switching times, the effective
times should be recombined with the zero voltage applied
time, To, according to the reference vector location.
In Fig. 2, the relationship between the effective times and
the actual gating times is depicted when the reference vector
is located in the Sector-1. The A+gatingof the figure means
the gating signal for the upper switching device of A-phase
arm, and B+gating,C+gating mean the signals for related
phases, respectively. In the case of Fig.2, the V1 vector is
applied to the inverter during T1 interval, and consequently
V2 vector is applied during T2 interval. In the three phase
symmetry modulation method, the zero voltage vector is
distributed symmetrically in one sampling period to reduce
the current ripple. Thus, in general, the switching sequence
is given by 0-1-2-7-7-2-1-0 within two sampling periods. In
Fig. 1. Switching state diagram of the inverter. the point of view of the upper switching device of one
(a) Three phase pulse inverter. inverter arm, the former sequence(0-1-2-7 sequence) is
(b) Space vector diagram of the Available called 'ON' sequence, and the latter(7-2-1-0) is called 'OFF'
switching vectors of the inverter. sequence in this paper. Therefore, the actual switching times
corresponding to the case of Sector-1 can be written as,
(1) On gating sequence, Off gating sequence

(where T1, T2 are the applied effective times corresponding to (4)

the active vectors, and TS is sampling time.)

And, the 'effective times' can be deduced as,

Upper ON sequence Upper OFF sequence

(2)

(Where Vdc is the DC link Voltage, α is the voltage angle

between reference vector and nearest right side effective vector

in the related sector So,0 ?? α<60°.)

Usually, in order to reduce the calculation effort, the

reference voltages of the stationary frame d-q axis are
associated as followings. (4)
Fig.2. Actual gating signal pattern of the conventional
space vector PWM. (in the case of the Sector-l.)

From this analysis, the conventional space vector PWM

task can be solved into following steps to make the actual
PWM pattern.
Step 1) Sector Identification: By comparing the stationary
frame d-q components of the reference voltage vector,
the sector where the reference vector is located is
identified.
where (3) Step 2) Calculating the Effective Times: Using the d-q
components of reference vector, a sine lookup table
and the DC link voltage information, the effective
times T1,T2 are calculated by (2). Or, to reduce the
電 学 論D, 116巻8号,平 成8年 821
 calculation time, another look-up table which
contains the coefficients corresponding to each sector
number may be used to calculate T1, T2 by (3).
Step 3) Determining the Switching Times: Using the
corresponding sector information, the actual
switching times for each inverter arm are deduced
from the combination of the effective times and zero
sequence time. In this step, another look-up table may
be referred to simplify the calculation process.
that the effective time evaluated in the conventional
SVPWM just means the difference value between two
applied times corresponding to the phase voltage. Therefore,
regardless of the sector location of the reference vector, the
applied times for each phase voltage can be defined as
followings and named as 'imaginary times' for phase
voltages in this paper.

Briefly speaking, in the space vector PWM (SVPWM) (7)

algorithm, the reference vector is partitioned to effective
vectors corresponding to the sectors and these effective
vectors are recombined to generate actual PWM switching
pattern. Therefore, owing to this basic approaching manner
of the conventional SVPWM method, the overall calculation
process is so complex that the practical implementation is
formidable. However, using the effective voltage vector
concept of the conventional SVPWM in different way, it is
possible to reconstruct the actual gating time without
partitioning and recombination task.
3. Novel Voltage Modulation
3.1 Evaluation of the PWM pattern
By the d-q transformation theory, the reference voltage
for each phase are found from the stationary frame reference
voltages.
In order to determine the actual switching time directly
from the phase voltage, the stationary frame reference
voltages utilized in the equation about effective times are
transformed to the phase voltages as follows.
From (5), in the case of Sector-1,
Note that the time intervals between the 'imaginary
times' are meaningful only and some of the imaginary times
will have negative value according to the phase reference
voltage. An imaginary switching pulse pattern calculated
from (7) is depicted in the Fig. 3 (a). The effective time Teff
will be defined as the time duration between Tmax and Tmin,
and the effective voltage is supplied to the machine this
Technique during this time interval.
Another important thing to be mentioned is that there is
one degree of freedom that the effective time can be located
anywhere within one sampling interval. From this useful
freedom. the actual switching pattern which preserves the
effective time can be simply generated by shifting operation
for 'imaginary times' in the time domain as shown in
Fig. 3(b). The zero voltage time is subjoined to the
(5) imaginary phase voltage times, and will be symmetrically
distributed at the beginning and end of one sampling period
in order to locate the effective time to the center of the
sampling interval. The voltage vector sequence that

(6-1)

(6-2)

As shown in the above equations, the effective times

T1, T2 can be represented by the time difference between the
times Tas ,Tbs and Tcs, corresponding to the phase voltages.
Also, in the case of the remaining sectors, the effective times Fig. 3. Switching pulse pattern of proposed PWM method .
can be replaced with the phase voltage times in the same (a) Imaginary switching pulse pattern.
method described above. From this result, it can be known (b) Actual switching pulse pattern for inverter.

822 T. IEE Japan , Vol. 116-D, No.8, '96

 A Novel Voltage Modulation Technique of the Space Vector PWM

minimizes the current harmonics has been discussed

extensively in the literature. (2,3)
Therefore, the actual switching times for each inverter
arm can be obtained as follows.

(8)

To distribute the zero voltage symmetrically during one

sampling period, the offset time To et is calculated using a
simple 3-element sorting algorithm. In this sorting
algorithm, only the maximum and minimum value is
required among the three imaginary phase voltage times.
From Fig. 3 (a) and (b),

and, (9)

Thus, the actual gating times can be determined simply

from (7),(9) and (8).
The whole gating time determination process mentioned
above is related to the 'OFF' switching sequence. In the case
of 'ON' switching sequence, the actual switching time should
be replaced by the subtraction value with sampling time in
order to generate a symmetrical switching pulse pattern
within two sampling intervals.
(10)

Now, with the effective vector concept, the actual

switching time can be directly obtained from the stationary
reference frame voltage To implement this novel space
vector PWM, only a 3-element simple sorting algorithm is
required. Therefore, the calculation efforts of the proposed
PWM method is greatly reduced compared with the
conventional SVPWM method.

3.2 Comparison Study

In Fig. 4, the actual switching pulse patterns are
illustrated to compare the proposed PWM technique with the
conventional SVPWM method and the sinusoidal PWM
method. In this paper, it is shown that the pulse pattern
exactly same as the conventional space vector PWM can be
generated with the sinusoidal modulation fashion. As widely
known, the sinusoidal PWM method has decisive
disadvantage that the linear range of controllable voltage is
limited to the modulation index 1. This is only 78 percent of
the value that would be reached by square-wave operation. (3)
This demerit is basically caused by the improper usage of the Fig.4. Actual implementations of the PWM methods
zero voltage. In the sinusoidal PWM, the maximum phase (a) Proposed PWM method
voltage is limited to Vdc/2 in linear modulation range and (b) Conventional space vector PWM
the zero voltage would be distributed asymmetrically as (c) Sinusoidal PWM method
shown in Fig. 4 (c). Thus, the effective voltages can not be applied during entire one sampling interval by the proposed
applied during the whole sampling interval in the linear zero voltage relocation algorithm. Therefore, in the view
modulation range.
point of the effective time, the proposed PWM algorithm has
As shown in Fig. 4(b), in the conventional SVPWM not only the same PWM wave form but also the same linear
method, the effective times should be combined with the zero controllable voltage range of the conventional SVPWM.
voltage time to generate the switching time. However, in the Note that the controllable range of the phase voltage
proposed method in Fig. 4 (a), both the effective time and the reference is not significant because the effective time
actual switching time can be directly obtained from the calculated from the voltage reference will be shifted on the
reference voltage. Furthermore, the effective time can be time domain by this zero voltage relocation algorithm. Thus,

電学 論D, 116巻8号,平 成8年 823

when the effective time exceeds the sampling time, an over
modulation strategy should be applied to the PWM
algorithm as the case of the conventional SVPWM.

3.3 The Over Modulation Technique

As widely known, the maximum magnitude of the
inverter output voltage is restricted by the DC link voltage.
Therefore, to guarantee the linearity of the inverter output
voltage, the reference vector should reside in the hexagon
region as shown in Fig. 1(b). However, in the transient state
of the machine control system with current regulator, the
reference voltage vector may eventually exceed the hexagon
area as shown in Fig. 5. In this case, a proper over
modulation technique should be implemented because the
actual output vector selected from the over-modulation
scheme will determine the transient system dynamics. In
order to apply the effective voltages to the machine during
the whole sampling time, a voltage vector located on the
hexagon surface is normally selected in the over-modulation
process.
One commonly used simple over-modulation strategy is
adopted in the proposed PWM algorithm. In that strategy,
the available voltage
vector on the point 'd' is
selected for an actual
output vector instead of
original vector on the
point 'c' as shown in
Fig. 5.(5)
From the voltage
vector concepts, the
modified effective
times, T1',T2', can be Fig.6. Overall algorithm of the proposed novel
simply calculated using Fig. 5. Simple over-modulation PWM method.
the original effective strategy
are needed in the proposed algorithm. Therefore, in the
times, T1, T2 as follows.
proposed PWM method, the execution time is reduced more
than 25% compared with the conventional space vector
(11) PWM method, and 15% memory size of the conventional
PWM method is required. Furthermore, due to the simple
Therefore,
structure of the proposed PWM strategy, this powerful
algorithm can be easily implemented on a slower controller.
(12)
In table 1, the comparison result of the calculation burden is

The comfortable conversion forms of this over presented. In the DSP controller, every arithmetic/logical
modulation technique are presented in the over-modulation operation task are executed within one state except the
dividing operation. Therefore, in the case of a slower and
part of Fig.6. For the sake of explanation, whole algorithm
cheaper controller, it is more compatible to compare the
of the proposed PWM method based on the space vector
theory is depicted in the C-language base in Fig. 6. execution performance according to the numbers of each
operation task. Especially, the number of times of the
4. Implementation Results multiplying or dividing operations is very important in the

Experiments are conducted to evaluate the performance slower controller because longer executionthey need
time
of the proposed PWM algorithm. To compare the proposed than any other operations, As shown
in table 1 , the number

algorithm with the conventional SVPWM algorithm, These of times for multiplying operation of the proposed SVPWM
two algorithms are implemented on the TMS320C30 is much smaller than that of the conventional . Therefore, the
floating point DSP (Digital Signal Processor) controller in overall execution times can be greatly reduced when the

the assembly language base. The sampling time is settled to proposed algorithm is adopted to the slower controller
60µSeC for the 8.33kHz switching frequency. When the instead of the conventional method.
Fig. 7 shows the actual upper switching pulse of the
conventional SVPWM algorithm is maximally optimized
proposed method with the voltage angle Į. To display the
using about 250 double words lookup table, it has needed 58
detailed pulse traces, the interval between two arrow points
states calculation time (1 state means 60nsec in this DSP ).
of the trace (a) is enlarged. The voltage angle of the arrow
On the contrary, only 40 state times and no additional tables
point is about 30•‹. Thus, the reference voltage vector is

824 T. IEE Japan, Vol. 116-D, No .8, '96


Fig.7. Actual switching pulse train of the proposed
method (θ ≡ ?? ,￨v*￨=100[V]・Vdc=3[V]).
located around the center of the Sector-1 and the effective
times T1, T2 will have near same value. As shown
gating pulse trains are generated with 0-1-2-7-7-2-1-0
sequence and the zero voltages are distributed symmetrically.
Fig. 8. The FFT result for the steady state phase current.
(a) A-phase(sampled) & B-phase(measured) currents.
(b) The FFT result for the B-phase current.
In Fig.8, some servo application wave-forms using the
proposed PWM for a 0.47[kW] permanent magnet
synchronous motor drive system are presented. In Fig. 8 (a),
the phase currents of the machine are shown. The A-phase
current is displayed by D/A converter using the sampled data
in the controller, and B-phase current is measured directly
by Tecktronics current measurement system. In Fig.8 (b), the
FFT(Fast Fourier Transform) result for the measured B
phase current is shown. From these results, it can be seen
that high performance voltage generation technique for a
current regulation system is easily implemented using the
proposed PWM algorithm.
5. Conclusion
In this paper, a novel voltage modulation technique based
on the space voltage vector has been described. From the
effective voltage concept, the actual switching times can be
deduced simply from the reference phase voltage without
sector identification, calculation for effective times and
recombination. To implement the proposed novel space
vector PWM, only a simple three-element sorting algorithm
is needed. Therefore, the calculation effort of the proposed
PWM method is greatly reduced as compared with the
電 学 論D, 116巻8号,平 成8年
A Novel Voltage Modulation Technique of the Space Vector PWM
conventional space vector PWM method.
Using the proposed method for the ac drives, an accurate
switching pattern can be simply obtained. The performance
of the proposed PWM method has been proved by laboratory
tests. (Manuscript received June I, 1995,revised November6,
1995)A
cknowledgment
The authors would like to thank Engineering Research
Center (ERC) in Seoul national university, Korea, for their
financial support.
Table 1. Calculation burden comparison between the
conventional SVPWM and the proposed PWM.
PWM
Fig.7, the
References
(1) J. Holtz, "Pulsewidth Modulation-A survey", Conf.
Record of IEEE, PESC'92, pp.11-18, 1992
(2) H.W. Van der Broeck and H.C. Skudelny," Analysis and
Realization of a Pulse Width Modulator Based on
Voltage Space Vectors", IEEE Trans. on Ind. Appl.
vol.24, no.1, pp. 142-150, 1988
(3) H.W. Van der Broeck, "Analysis of the Harmonics in
Voltage Fed Inverter Drives Caused by PWM Schemes
with Discontinuous Switching Operation", Conf. Record
of EPE'91, no. 3, pp.261-266, 1991
(4) J.F. Moynihan, et alii "Indirect Phase Current Detection
for Field Oriented Control of a Permanent Magnet
Synchronous Motor Drive", Conf. Record of EPE'91,
no.3, pp.611-616, 1991
(5) T.G. Habetler, F. Profumo, M. Pastorelli, and L.M.
Tolbert, "Direct Torque Control of Induction Machines
Using Space Vector Modulation", IEEE Trans. on Ind.
Appl., vol.28, no. 5, pp. 1045-1053, 1991
Joohn-SheokKim was born in Korea,in 1965. He received
B.S., M.S. and Ph.D degrees in all electrical engineering
from SeoulNational University,Korea,in 1989, 1992 and
1995,respectively.Andhe is workingasa researchengineer
of the engineeringcenterin SeoulNationalUniv. His major
interestingis powerelectronics,especiallyon the adjustable
speedac drivesand static power converter.
Seung-KiSul (Member)received the B.S., M.S., and Ph.D.
degrees in electrical engineering from Seoul National
University,Seoul, in 1980, 1983, and 1986,respectively.
He was with the Departmentof Electrical and Computer
Engineering,Universityof Wisconsin-Madisonas a research
associate from 1986 to 1988. He joined the Gold-Star
IndustrialSystemsCompanyas a principalresearch engineer
in 1988, where he remained until 1990. Since 1991, he has been with the
Departmentof Electrical Engineering,Seoul National University. His present
research interestsare in high-performanceelectric machine controlusing power
electronics.He is performingvariousresearchprojectsfor industrial systemsand
some of the results are applied to the fields of industrialhigh-power electric
machinecontrol.
825