Professional Documents
Culture Documents
jIm
0
2 u
3 b
0
u(t)
t ua
2
0 0
Re 3
0
0 Figure 6.24
2 0
3 u Three-phase coordinate vectors and
c space vector u(t).
where 2/3 is a scaling factor. Equation (6.57) can be written in real and imaginary
components in the x–y domain as
u1t2 = u x + ju y (6.58)
Using Eqs. (6.57) and (6.58), we can obtain the coordinate transformation from the
a–b–c-axis to the x–y axis as given by
-1 -1
1 ua
ux 2 2 2
a b = § ¥ £ u b ≥ (6.59)
uy 3 13 - 13
0 uc
2 2
which can also be written as
2
ux = [v - 0.51vb + vc 2] (6.60a)
3 a
13
uy = 1vb - vc 2 (6.60b)
3
The transformation from the x–y axis to the α-β axis, which is rotating with an
angular velocity of ω, can be obtained by rotating the x–y-axis with ωt as given by
(Appendix G)
π
cos(ωt) cos ¢ + ωt≤
uα 2 ux cos(ωt) -sin(ωt) u x
a b = § ¥ a b = a ba b (6.61)
uβ π uy sin(ωt) cos(ωt) u y
sin(ωt) sin ¢ + ωt≤
2
Using Eq. (6.57), we can find the inverse transform as
u a = Re1u 2 (6.62a)
u b = Re1ue -j12/32π 2 (6.62b)
j12/32π
u c = Re1ue 2 (6.62c)
For example, if ua, ub, and uc are the three-phase voltages of a balanced supply with a
peak value of Vm, we can write
u a = Vm cos 1ωt2 (6.63a)
u b = Vm cos 1ωt - 2π/32 (6.63b)
u c = Vm cos 1ωt + 2π/32 (6.63c)
1 2 3 4
001 101 111 000
5 6 7 8
Figure 6.25
The on and off states of the inverter switches. [Ref. 13]
Space vector (SV). The switching states of the inverter can be represented by
binary values q1, q2, q3, q4, q5, and q6; that is, qk = 1 when a switch is turned on and
qk = 0 when a switch is turned off. The pairs q1q4, q3q6, and q5q2 are complementary.
Therefore, q4 = 1 - q1, q6 = 1 - q3, and q2 = 1 - q5. The switch on and off states
are shown in Figure 6.25 [13]. Using the relation of trigonometry e jθ = cos θ + j sin θ
for θ = 0, 2π/3, or 4π/3, Eq. (6.57) gives the output phase voltage in the switching state
(100) as
2 -1 -1
va 1t2 = V ; vb 1t2 = V ; vc 1t2 = V (6.65)
3 S 3 S 3 S
The corresponding space vector V1 can be obtained by substituting Eq. (6.65) into
Eq. (6.57) as
2
V1 = V e j0 (6.66)
3 S
Similarly, we can derive all six vectors as
2 π
Vn = VSe
j1n - 12
3 for n = 1, 2, c 6 (6.67)
3
The zero-vector has two switching states (111) and (000), one of which is redundant.
The redundant switching state can be utilized to optimize the operation of the inverter
such as minimizing the switching frequency. The relationship between the space vec-
tors and their corresponding switching states is given in Table 6.2. It should be noted
that these vectors do not move in space, and thus they are referred to as stationary
vectors. Whereas the vector u(t) in Figure 6.24 and in Eq. (6.64) rotates at an angular
velocity of
ω = 2πf (6.68)
where f is the fundamental frequency of the inverter output voltage.
Using the three-phase to two-phase transformation in Eq. (6.59) and the line
voltage (U3 phase voltage) as the reference, the α-β components of the rms output
voltage (peak value/U2) vectors can be expressed as functions of q1, q3, and q5.
-1 -1
1 q1
VLα 2 3 2 2
a b = V§ ¥ £ q3 ≥ (6.69)
VLβ 3A2 s 13 - 13
0 q5
2 2
Using the factor U2 for converting the rms voltage to its peak value, the peak value of
the line voltage is VL(pak) = 2VS/U3 and that of the phase voltage is Vp(peak) = Vs/U3.
Using the phase voltage Va as the reference, which is usually the case, the line voltage
vector Vab leads the phase vector by π/6. The normalized peak value of the nth line
voltage vector can be found from
12 * 12 j12n - 12π/6 2 12n - 12π 12n - 12π
Vn = e = c cos a b + j sin a bd
13 13 6 6
for n = 0, 1, 2, 6 (6.70)
There are six nonzero vectors, V1 -V6, and two zero vectors, V0 and V7, as shown in
Figure 6.26. Let us define a performance vector U as the time integral function of Vn
such that
L
U =Vn dt + U0 (6.71)
where U0 is the initial condition. According to Eq. (6.71), U draws a hexagon locus
that is determined by the magnitude and the time period of voltage vectors. If the out-
put voltages are purely sinusoidal, then the performance vector U becomes
U* = Me jθ = Me jωt (6.72)
where M is the modulation index 10 6 M 6 12 for controlling the amplitude of the
output voltage and ω is the output frequency in rads per second. U* draws a pure circle
locus as shown in Figure 6.26 by a dotted circle of radius M = 1 and it becomes the
reference vector Vr. The locus U can be controlled by selecting Vn and adjusting the
time width of Vn to follow the U* locus as closely as possible. This is called the quasi-
circular locus method. The loci of U and U* 1= Vr 2 are also shown in Figure 6.26.
The angular displacement between reference vector Vr and the α-axis of the α-β
frame can be obtained by
t
L0
θ1t2 = ω1t2dt + θo (6.73)
When the reference (or modulating) vector Vr passes through the sectors one by one, dif-
ferent sets of switches will be turned on or off according to the switching states as listed in
Table 6.2. As a result, when Vr rotates one revolution in space, the inverter output voltage
completes one cycle over time. The inverter output frequency corresponds to the rotating
speed of Vr and its output voltage can be adjusted by varying the magnitude of Vr.
Modulating (rotating)
vector Vr [vr] Sector number
V2 Vi1
110
2 1
Stationary
state
V3 vr V1 Vi
010 100
o
3 V7, 8 6
vc
‹ 1
V4 V6
011 101
4 5
V5 011
Figure 6.26
The space vector representation.
Modulating reference vectors. Using Eqs. (6.59) and (6.60), the vectors of
three-phase line modulating signals [vr]abc = [vravrbvrc]T can be represented by the
complex vector U* = Vr = [vr]αβ = [vrαvrβ]T as given by
2
vrα = [v - 0.51vrb + vcr 2] (6.74)
3 ra
13
vrβ = 1vrb - vrc 2 (6.75)
3
If the line modulating signals [vr]abc are three balanced sinusoidal waveforms with an
amplitude of Ac = 1 and an angular frequency ω, the resulting modulating signals in
the α-β stationary frame Vc = [vr]αβ becomes a vector of fixed amplitude M Ac 1 =M 2
that rotates at frequency ω. This is also shown in Figure 6.26 by a dotted circle of
radius M.
modulating signal Vr is laying between the arbitrary vectors Vn and Vn + 1, then the
two nonzero vectors (Vn and Vn + 1) and one zero SV (Vz = V0 or V7) should be used
to obtain the maximum load line voltage and to minimize the switching frequency. As
an example, a voltage vector Vr in section 1 can be realized by the V1 and V2 vectors
and one of the two null vectors (V0 or V7). In other words, V1 state is active for time
T1, V2 is active for T2, and one of the null vectors (V0 or V7) is active for Tz. For a suf-
ficiently high-switching frequency, the reference vector Vr can be assumed constant
during one switching period. Because the vectors V1 and V2 are constant and Vz = 0,
we can equate the volt time of the reference vector to the SVs as
Vr * Ts = V1 * T1 + V2 * T2 + Vz * Tz (6.76a)
Ts = T1 + T2 + Tz (6.76b)
which is defined as the SVM. T1, T2, and Tz are the dwell times for vectors V1, V2, and
Vz, respectively. Equation (6.67) gives the space vectors in sector 1 as
2 2 π
V1 = VS ; V2 = VS e j 3 ; Vz = 0 ; Vr = Vr e jθ (6.77)
3 3
where Vr is the magnitude of the reference vector and θ is the angle of Vr.
This is achieved by using two adjacent SVs with the appropriate duty cycle
[15–18]. The vector diagram is shown in Figure 6.27.
Substituting Eq. (6.77) into Eq. (6.76a) gives
2 2 π
Ts Vr e jθ = T1 VS + T2 VS e j 3 + Tz * 0
3 3
V2
Q
T2V2
Ts Vr
V2T2
Ts
Table 6.3 Relationship between the Dwell Times and the Space Vector Angle θ for Sector 1
where M is given by
13 Vr
M = (6.82)
VS
Let Va1 equal the rms value of the fundamental component of the inverter output
phase (phase-a) voltage. Vr, which is the peak reference value, is related to Va1 by
Vr = 12Va1
which, after substituting in Eq. (6.82), gives M as
13 Vr 16 Va1
M = = (6.83)
VS VS
which shows that the rms output voltage Va1 is proportional to the modulation index
M. Since the hexagon in Figure 6.26 is formed by six stationary vectors having a length
of 2VS/3, the maximum value of the reference vector is given by
2 13 VS
Vr1max2 = V * = (6.84)
3 S 2 13
Substituting Vr(max) into Eq. (6.82) gives the maximum modulation index Mmax as
13 VS
M max = * = 1 (6.85)
VS 13
which gives the range of the modulation index for SVM as
0 … M max … 1 (6.86)
SV sequence. The SV sequence should assure that the load line voltages have
the quarter-wave symmetry to reduce even harmonics in their spectra. To reduce the
switching frequency, it is also necessary to arrange the switching sequence in such
a way that the transition from one to the next is performed by switching only one
inverter leg at a time. That is, one is switched on and the other one is switched off. The
transition for moving from one sector in the space vector diagram to the next requires
no or a minimum number of switching. Although there is not a systematic approach
to generate an SV sequence, these conditions are met by the sequence Vz, Vn, Vn + 1 Vz
(where Vz is alternately chosen between V0 and V7). If, for example, the reference
vector falls in section 1, the switching sequence is V0, V1, V2, V7, V2, V1, V0. The time
interval Tz 1 =T0 = T7 2 can be split and distributed at the beginning and at the end
of the sampling period Ts. Figure 6.28 shows both the sequence and the segments of
three-phase output voltages during two sampling periods. In general, the time intervals
of the null vectors are equally distributed, as shown in Figure 6.28, with Tz/2 at the
beginning and Tz/2 at the end.
The SVM pattern in Figure 6.28 has the following characteristics:
1. The pattern in Figure 6.28 has a quarter-wave symmetry.
2. The dwell times for the seven segments add up to the sampling period
1 Ts = T1 + T2 + Tz 2 or a multiple of Ts.
Ts Ts
vaN
V0 V1 V2 V7 V7 V2 V1 V0
vbN
0
t
vcN
0
t
Figure 6.28
Pattern of SVM.
3. The transition from state (000) to state (100) involves only two switches and is
accomplished by turning Q1 on and Q4 off.
4. The switching state (111) is selected for the Tz/2 segment in the center to reduce
the number of switching per sampling period. The switching state (000) is selected
for the Tz/2 segments on both sides.
5. Each of the switches in the inverter turns on and off once per sampling period.
The switching frequency fsw of the devices is thus equal to the sampling frequency
fs = 1/Ts or its multiple.
6. The pattern of waveform as shown in Figure 6.28 can be produced for a duration
of nTs that is a multiple (n) or a fraction (1/n) of the sampling period Ts by either
multiplying or dividing the dwell times by n. That is, if we multiply by 2, the seg-
ments will cover two sampling periods.
The instantaneous phase voltages can be found by time averaging of the SVs dur-
ing one switching period for sector 1 as given by
Vs -Tz Tz Vs π
vaN = a + T1 + T2 + b = sin a + θb (6.87a)
2Ts 2 2 2 3
Vs -Tz Tz 13 π
vbN = a - T1 + T2 + b = Vs sin aθ - b (6.87b)
2Ts 2 2 2 6
Vs -Tz Tz
vcN = a - T1 - T2 + b = -VaN (6.87c)
Ts 2 2
To minimize uncharacteristic harmonics in SV modulation, the normalized sampling
frequency fsn should be an integer multiple of 6; that is, T Ú 6nTs for n = 1, 2, 3, c.
This is due to the fact that all the six sectors should be equally used in one period for
producing symmetric line output voltages. As an example, Figure 6.29 shows typical
waveforms of an SV modulation for fsn = 18 and M = 0.8.
vc vc
t
90 180 270 360
S1
on
t
0 90 180 270 360
(b) Switch S1 state
S3
on
t
0 90 180 270 360
(c) Switch S3 state
vab vab1
vi
t
Figure 6.29 0 90 180 270 360
Three-phase waveforms for space
vector modulation 1M = 0.8, fsn = 18 2 . (d) Ac output voltage spectrum
3 cos() sin()
T1 T .
3 cos() sin()
. 2 sin( )
v3 010 v2 110 T2 T 3 cos() sin()
T0 0
(T T1 T2)
v1 100
v4 011
T1 m.T.sin(60 )
T2 m.T.sin( )
T0 T (T1 T2)
v5 001 v6 101
Figure 6.30
Overmodulation. [Ref. 20, R. Valentine]
and Tz in Eq. (6.81). However, the portions of the circle outside the hexagon are limited
by the boundaries of the hexagon, as shown in Figure 6.30, and the corresponding time
states Tn and Tn + 1 can be found from [20]:
The maximum modulation index M for SVM is Mmax = 2/U3. For 0 6 M … 1, the
inverter operates in the normal SVM, and for M Ú 2/U3, the inverter operates com-
pletely in the six-step output mode. Six-step operation switches the inverter only to the
six vectors shown in Table 6.2, thereby minimizing the number of switching at one time.
For 1 6 m 6 2/U3, the inverter operates in overmodulation, which is normally used
as a transitioning step from the SVM techniques into a six-step operation. Although
overmodulation allows more utilization of the dc input voltage than the standard SVM
techniques, it results in nonsinusoidal output voltages with a high degree of distortion,
especially at a low-output frequency.
Sector Segment 1 2 3 4 5 6 7
1 Vector V0 V1 V2 V7 V2 V1 V0
State 000 100 110 111 110 100 000
2 Vector V0 V3 V2 V7 V2 V3 V0
State 000 010 110 111 110 010 000
3 Vector V0 V3 V4 V7 V4 V3 V0
State 000 010 011 111 011 010 000
4 Vector V0 V5 V4 V7 V4 V5 V0
State 000 001 011 111 011 001 000
5 Vector V0 V5 V6 V7 V6 V5 V0
State 000 001 101 111 101 001 000
6 Vector V0 V1 V6 V7 V6 V1 V0
State 000 100 101 111 101 100 000
SVM implementation. Figure 6.28 shows the switching sequence for sector 1
only. The practical requires the switching sequence for all six segments as listed in
Table 6.4. The block diagram for digital implementation of the SVM algorithm is
shown in Figure 6.31. The implementation involves the following steps:
Gate
signals
Sector # 1–6
vr θ Sector g1
va* calculator
vr θk g2
tan−1 v
r Switching
vb* abc/ g3
Dwell sequence
v2 + v2 vr
M generator g4
vc* ÷ time
3
calculator g5
Reference vr Vs
Ts g6
signals
Figure 6.31
Block diagram for digital implementation of the SVM algorithm.
5. Calculate the dwell times T1, T2, and Tz from Eq. (6.81).
6. Determine the gating signals and their sequence according to Table 6.4.
Key Points of Section 6.7 Sinusoidal, harmonic injection, and SVM modulation tech-
niques are usually used for three-phase inverters. Due to the flexibility of manipula-
tion and digital implementation, SVM has increasing applications in power converters
and motor control.