A Simple Space-Vector PWM Algorithm for VSI-fed AC Motor Drives

Razvan C. Panaitescu and Ned Mohan

Department of Electrical and Computer Engineering
University of Minnesota, Minneapolis, MN 55455
Email: [email protected], [email protected]

Abstract – Space-Vector Pulse Width Modulation (SV-PWM) is 1 1 vcont , a 1 van

a well-known modulation strategy in electric drives (also used in da (t ) = + = + ( etc.) (4)
power electronics). However, a simple explanation of its analysis 2 2 Vˆtri 2 Vd
and its hardware implementation is lacking, which is the focus Since vaN = d a ⋅ Vd , vbN = d b ⋅ Vd , vcN = d c ⋅ Vd , expressions
of this paper.
for vaN , vbN , vcN can be similarly written, using (4):
Key Words – Space-Vector PWM, AC Drives, Voltage Source
Inverters ia
a

I. INTRODUCTION + van
ib
In electric drives and other power electronic applications, Vd b vbn n
such as in switch-mode rectifiers, the intent is to supply −
vcn
three-phase sinusoidal currents. In the system of Fig.1a, the ic
qa (t ) c
neutral “n” may be hypothetical in case of delta-connected
analysis. At switching frequencies relatively high compared
N
to the fundamental frequency of synthesis, it is possible to vcont , a (t )
represent the switching circuit of Fig.1a by means of ideal
transformers in terms of switching average quantities, as vcont ,b (t) qb (t )
shown in Fig.1b [1]. Average variables with a bar (“—“) on
top do not contain switching frequency ripple. vcont ,c (t ) qc (t )
The intent to supply balanced sinusoidal currents into a vtri (a)
balanced ac motor implies that the line-to-line voltages across
the motor terminals be sinusoidal in steady-state. In a
balanced ac motor, it can be shown (Appendix 1) that the ida (t ) idb ( t ) idc (t ) a van
motor phase voltages (van, vbn and vvn) are also sinusoidal and, b
similarly to the motor currents, the three phase voltages sum vbn n
+ c vcn
to zero on an instantaneous basis (see [2] and [3]). + + +
van ( t ) + vbn ( t ) + vcn ( t ) = 0 (1) Vd 1 d a (t ) 1 d b (t ) 1 dc (t )
Of course, the above equation is also valid on an average − vaN (t ) vbN ( t ) vcN (t )
basis: − − −
van ( t ) + vbn ( t ) + vcn ( t ) = 0 (2) N
when,
ia ( t ) + ib ( t ) + ic ( t ) = 0 (3) vcont, a (t ) 1 +
2Vˆtri
+
II. SINUSOIDAL PULSE-WIDTH MODULATION
vcont,b (t ) 1 +
A traditional approach that is discussed in most textbooks
is sinusoidal pulse width modulation, where sinusoidal 2Vˆtri +
control voltages are compared with a triangular carrier signal
(see Fig.2a) to generate the switching functions qa,b,c in vcont, c ( t ) 1 +
Fig.1a. The resulting duty factors (transformer turn-ratios in 2Vˆtri 1/ 2
(b)
Fig.1b) are shown in Fig.2b, based on the following +
expression: Fig. 1. (a) Switch-mode inverter for ac drives;
(b) Average representation of the three phase converter

0-7803-7404-5/02/$17.00 (c) 2002 IEEE 72

Authorized licensed use limited to: IEEE Xplore. Downloaded on April 15, 2009 at 14:12 from IEEE Xplore. Restrictions apply.
 Vˆtri vcont ,a vcont ,b vcont ,c Vˆtri
vtri vcont , a vcont ,b vcont , c
vtri
vtri

vtri

(a) (a)

−Vˆtri −Vˆtri
viN v
di , d i , iN
Vd Vd
da db dc
da db dc

(b) (b)

van vbn vcn van vbn vcn

Vd 2 Vd 2 Vd 2 Vd 2 Vd 2 Vd 2

(c) (c)

Fig. 2. Control voltages, duty-factors and phase voltages Fig. 3. Control voltages, duty-factors and phase voltages
waveforms in Sinusoidal PWM waveforms in Space-Vector PWM

Vd 1  Vd  K = 0.5 ⋅ Vd Vˆtri (6)

vaN = + ⋅  ⋅ vcont , a ( etc.) (5)
2 2  Vˆtri  At the limit, in sinusoidal PWM, control voltages and the
corresponding duty-factors reach their peak values of 1 when
The inverter terminal voltages, with respect to the negative
dc-bus N, contain a dc-offset Vd/2 which is of zero-sequence. the required motor phase voltage peaks at Vd 2 . In this
This zero-sequence cancels out in the motor voltages van , situation, the maximum value of the line-to-line voltage, for
example vac (without overmodulation) is:
vbn , vcn (see Fig.2c).
V
Therefore, the switch-mode converter functions as an VˆLL ,max = 3 d = 0.866 ⋅ Vd (7)
amplifier, amplifying the control voltage by a gain K defined 2
as:

73

Authorized licensed use limited to: IEEE Xplore. Downloaded on April 15, 2009 at 14:12 from IEEE Xplore. Restrictions apply.
 However, in Space-Vector PWM (Fig.3), as discussed in
the next section, it is possible to achieve a 13.4% higher ac
line voltage from the same dc link voltage, thus fully utilize
the inverter capabilites without overmodulation.

III. SPACE-VECTOR PULSE WIDTH MODULATION

To fully utilize the inverter voltage capability, Fig. 4. Simulink and dSPACE implementation of the
ˆ
VLL ,max should be equal to Vd. As seen from (5), the zero- Space Vector PWM algorithm
sequence voltage dc offset of Vd / 2 is cancelled out from the
motor terminal voltages (line-to-line, as well as phase-to- vz ( t )
neutral “n” if present). This implies that it is possible to −1
vz ( t ) =  min ( van , vbn , vcn ) + max ( van , vbn , vcn ) 
choose from an infinite arrangements of zero-sequence 2 
voltages (see [4]) to be inserted into (5). The choice must be
based on keeping the line-to-line voltages balanced sinusoids,
simplicity of generating these control voltages and the impact
on the switching pattern which influences the switching
losses within the inverter. One such zero-sequence voltage
can be generated from the three motor voltages in either of
the two following equivalent forms (see Appendix 2): Fig. 5. Zero sequence voltage obtained with the
Space Vector PWM algorithm
1
vz = min  abs ( van ) , abs ( vbn ) , abs ( vcn )  (8) developed in our laboratory. The digital FOC and SV-PWM
2
or, were compiled in and executed by a dSPACE rapid
prototyping system, using the DS1102 DSP board and
−1
vz =  max ( van , vbn , vcn ) + min ( van , vbn , vcn )  (9) Simulink for algorithm implementation.
2  Since the dSPACE generates the PWM signals by
This implies that the zero-sequence voltage is made up of comparing the duty-factors with a sawtooth waveform, we
appropriately selected segments of the motor voltages. In have implemented a simple algorithm to obtain symmetrical
Simulink, the algorithm can be simply realized [2] using a switching pulses over one switching period, thus
limited number of blocks (Fig.4). A plot of the zero-sequence transforming the sawtooth in a triangular carrier ([5]). Thus,
voltage added to obtain the control voltages is represented in we could decrease the current ripple and the switching losses
Fig. 5. in the inverter.
Therefore in the comparator of Fig.1a, the control voltages The algorithm implemented permitted to use a lower dc-
in sinusoidal steady state will be, using (6): bus voltage equal to the motor peak line-to-line voltage,
vcont , a = ( van + vz ) ⋅ K −1 ( etc.) (10) without overmodulating the switches, and obtaining a very
The control voltages and the corresponding duty-factors good dynamical performance in terms of speed and torque
were plotted in Fig.3, for the same dc-bus voltage and phase control.
voltage amplitudes as for sinusoidal PWM. Note that the peak
value of the duty-factors is smaller in SV-PWM, thus a better V. CONCLUSION
utilization of the inverter can be achieved at higher voltage As discussed in this paper, Space-Vector PWM control of a
requirements, without entering into the overmodulation. voltage source inverter can be easily explained and
Usually, van , vbn , vcn in (9) are derived from the average implemented using the classical “carrier-based” approach.
voltage space-vector components in the (α , β ) stator oriented Since the method can be applied both to electric drives and
switch-mode power converters, its derivation can be
reference system, as in the most field oriented control (FOC)
straightforward, without any use of additional definitions
algorithms for ac motors.
such as “sector calculations”, “hexagon of states”, “vector
decomposition” etc.
IV. IMPLEMENTATION AND EXPERIMENTAL RESULTS
The control voltages, which depend on the average phase
The SV-PWM strategy was implemented and tested in a voltages, can be easily explained and implemented.
complete field oriented control of a 1hp, 230/460 V, 3 phase
induction motor. The inverter, built around the POWEREX
integrated power module, and all the analog circuitry were

74

Authorized licensed use limited to: IEEE Xplore. Downloaded on April 15, 2009 at 14:12 from IEEE Xplore. Restrictions apply.
 APPENDIX 1 vin
van vbn vcn
Derivation of van (t ) + vbn (t ) + vcn (t ) = 0 [2]

In a balanced induction machine, in terms of phase

resistance, currents and flux linkages:
d d d
van = Rs ia + λa ; vbn = Rs ib + λb ; vcn = Rs ic + λc (11)
dt dt dt
For a three-wire connection where the motor neutral is
isolated,
ia + ib + ic = 0 . (12)
Moreover, λa + λb + λc = 0 , which can be proven in terms
of the stator winding and equivalent rotor winding currents as ωt
follows:
λa = Ls ia + Lm iA cos θ m 60o 60o 60o 60o 60o 60o
λb = Ls ib + LmiB cos θ m (13)
Fig. 6. Sector determination in terms of phase voltages
λc = Ls ic + Lm iC cos θ m a
where A, B, C correspond to equivalent rotor phase b n
windings and θ m is the rotor angle. c
In a balanced rotor where all the bar currents sum to zero, van vbn vcn
iA + iB + iC = 0 . Adding the three equations in (13) and using
the condition of (12) results in Vd Vd Vd
λa + λb + λc = 0 (14) 2 2 2
Taking the time-derivative of (14) on both sides,
d d d vz vz vz
λa + λb + λc = 0 (15)
dt dt dt N
Adding the three equations in (11) and utilizing (12) and
Fig. 7. Inverter voltages
(15) proves that van (t ) + vbn (t ) + vcn (t ) = 0 .
From (10) and (18) we can write:
v v −v
APPENDIX 2 van + v z = ac = an cn (19)
Derivation of zero-sequence voltage 2 2
Solving for the zero-sequence voltage, using (2):
To prove the expressions in (8) and (9), observe that every
1 1
60°, two of the three phase voltages reach their extreme vz = − ( van + vcn ) = vbn (20)
values with respect to the others. Consider the first 60° sector 2 2
(Fig. 6), where van is maximum and vcn is minimum. At the We observe that in the first 60° sector considered here:
voltage limit, we require that the peak amplitude of the line- van = max ( van , vbn , vcn ) , vcn = min ( van , vbn , vcn ) (21)
to-line voltage equals Vd. In this case, the resulting sinusoidal and
a-c line-to-line voltage, in terms of the stator frequency ω is: vbn = min ( van , vbn , vcn ) (22)
vac = van − vcn = Vd cos (ω t − 30o ) (16) For all the other sectors the control voltage is obtained by
Thus, from (4) and (6), the control function for phase-a is: simple permutations, while conditions (21) and (22) change
V accordingly, showing that the zero-sequence voltage should
vcont , a = K −1 ⋅ d ( 2d a − 1) (17) be of the form in (8) or (9).
2
Full utilization of the inverter requires that, at the limit
REFERENCES
when vac is Vd, the duty-factor in phase-a is also maximum,
[1] N. Mohan, “Electric Drives – An Integrative Approach”, MNPERE,
i.e. d a = 1 (and d c = 0 ). To achieve this, in the sector Minneapolis, 2001, Ch.4, pp. 18-23.
considered, the control voltage must be in phase with the [2] N. Mohan, “Advanced Electric Drives – Analysis, Control and
Modeling using Simulink”, MNPERE, Minneapolis, 2001.
line-to line voltage, and with a peak value of K −1 ⋅ Vd 2 : [3] W. Leonhard, “Control of Electric Drives”, Narosa Publishing, 1992.
vcont , a = K ⋅Vd 2 ⋅ cos (ω t − 30 ) = K
−1 o −1
2 ⋅ vac (18) [4] J. Holtz, “Pulse Width Modulation for Electronic Power Conversion”,
in “Power Electronics and Variable Frequency Drives” edited by B.K.
In (10) we assumed that the control voltage would be the Bose, IEEE Press, 1997, pp. 154-161.
[5] A.K. Jain, “Implementation of Vector Control for Permanent Magnet
result of a zero-sequence voltage superimposed over the Synchronous Motors”, University of Minnesota, July, 2000.
required phase voltages, as in Fig.7.

75

Authorized licensed use limited to: IEEE Xplore. Downloaded on April 15, 2009 at 14:12 from IEEE Xplore. Restrictions apply.