Modeling and Simulation of Direct Torque Controlled

PMSM Drive System Incorporating Structural and

Saturation Saliencies

Ying Yan, Jianguo Zhu, Youguang Guo and Haiwei Lu
Faculty of Engineering
University of Technology, Sydney
PO Box 123, Broadway, NSW 2007, Australia
yingyan@eng.uts.edu.au


AbstractThe direct torque controlled (DTC) permanent magnet
synchronous motor (PMSM) drive has become competitive
compared with other types of drive systems because of its simple
and sensorless control algorithm. The application of the system,
however, is handicapped by the difficulty of starting under full
load due to the unknown initial rotor position. This paper
presents a nonlinear model of PMSMs which incorporates both
the structural and saturation saliencies to enable the numerical
simulation of initial rotor position detection algorithms. In this
model, the phase inductances are expressed by Fourier series as
functions of the stator current and rotor position. The
inductances of a surface mounted PMSM is measured with
different rotor positions and DC offset currents, which emulate
the effect of the three phase stator currents. By using the
proposed model, the DTC PMSM is simulated and the results are
compared with those obtained by the PMSM model in the
Simulink library. With the model, an initial rotor position
estimation scheme using voltage pulses is investigated by
numerical simulation. The scheme is also experimentally tested
and the results are compared with the inductance variation to
verify the validity of the method. The effectiveness of the scheme
to estimate the initial rotor position for the testing SPMSM is
analyzed and verified by numerical simulation before physical
implementation.
Keywords- PMSM model; structural saliency; saturation
saliency; inductance; direct torque control; initial rotor position
estimation
I. INTRODUCTION
The direct torque controlled permanent magnet
synchronous motor (PMSM) drive system has become
competitive compared with other types of drive systems,
because of its simple and sensorless control algorithm. The
practical application of the system, however, is handicapped by
the difficulty of starting under full load due to the unknown
initial rotor position. A lot of efforts have been made to detect
the initial rotor position. Among them, the most versatile
method is to make use of the structural and magnetic saturation
saliencies which exist in the PMSM. The structural saliency
could be employed to acquire the position of the rotor axis,
while the saturation saliency, which is generated by the rotor
permanent magnets, can be used to detect the magnetic
polarity. Since the conventional PMSM model does not
incorporate the saturation saliency, when developing a new
method for the initial rotor position detection, it is not possible
to numerically simulate the method, and the experimental trial
and error method is widely used.
This paper presents a nonlinear model of PMSMs which
incorporates both the structural and saturation saliencies to
enable the numerical simulation of new rotor position detection
algorithms. In this model, the self and mutual differential
inductances of the phase windings are expressed by Fourier
series as functions of the rotor position and stator current. A
method to measure the differential inductances on a surface
mounted PMSM is outlined in the paper. Based on the model,
the direct torque control (DTC) scheme is simulated within the
Matlab/Simulink environment. The validity of the new model
is verified by comparing the numerical simulation results with
those obtained by using the model in the Simulink library.
With the model, an initial rotor position estimation scheme
using voltage pulses is investigated by numerical simulation.
The scheme is also experimentally tested and the results are
compared with the inductance variation to verify the validity of
the method. The effectiveness of the scheme to estimate the
initial rotor position for the testing PMSM is analyzed and
verified by numerical simulation before being implemented in
the physical system.
II. NONLINEAR MODEL OF PMSM WITH SALIENCIES
A. Flux Linkage and Circuit Equations
In a PMSM, the field produced by the rotor magnets is the
dominant component which determines the operating point of
the magnetic core. Although the characteristics of the magnetic
core is nonlinear, the magnetic circuit can be considered as
piecewisely linearized around the operating point P at a given
rotor position. Thus, the total flux linkage of phase a,
a
, can
be separated into two components [1-3]:
( ) ( ) ,
a as ms af
i = + (1)
where is the rotor position,
as
and
af
the flux linkage
components produced by the magnetization component of the
1-4244-0365-0/06/$20.00 (c) 2006 IEEE
76
stator current i
ms
, and rotor magnets, respectively. The flux
linkage
as
can be further separated into components attributed
by individual stator phase currents,
aa
,
ab
, and
ac
.
a aa ab ac af
= + + + (2)
where
aa
,
ab
, and
ac
are the flux linkage components of
phase a corresponding to each phase current. Under the
assumption of linearization, the flux linkage components,
aa
,

ab
, and
ac
can be considered as proportional to the
corresponding phase currents:
( ) ,
aa aa a a
L i i = (3)
( ) ,
ab ab b b
L i i = (4)
( ) ,
ac ac c c
L i i = (5)
The proportionality coefficients L
aa
, L
ab
, and L
ac
, which are
determined by the gradient of the magnetization curve at the
operating point, are defined as the self and mutual incremental
inductances of the phase a winding. The flux linkages of
phases b and c can be obtained similarly. The voltage equations
of the three-phase stator windings can be written as:
[ ]
0 0
0 0
0 0
a a a a af a
b b b T b bf b
c c c c cf c
v R i i e e
d
v R i L i e e
dt
v R i i e e

( + ( ( ( (
(
( ( ( (
= + + +
(
( ( ( (
(
( ( ( ( +


(6)
where
[ ]
aa ab ac
aa a ab b ac c
a b c
ab bb bc
T ab a bb b bc c
a b c
ac bc cc
ac a bc b cc c
a b c
L L L
L i L i L i
i i i
L L L
L L i L i L i
i i i
L L L
L i L i L i
i i i
(
+ + +
(

(
(

= + + +
(

(
(

+ + +
(


(7)
v
a
, v
b
, and v
c
are the terminal voltages, R
a
, R
b
, and R
c
the
winding resistances, i
a
, i
b
, and i
c
the currents of phases a, b, and
c, respectively, e
af
=
r
d
af
/d, e
bf
=
r
d
bf
/d, and e
cf
=
r
d
cf
/d
are the electromotive forces (emfs) induced by the rotor
magnets,
r
=d/dt is the mechanical angular speed of the rotor,
and

aa ab ac
a r a b c
L L L
e i i i

| |
= + +
|

\ .
(8)

ab bb bc
b r a b c
L L L
e i i i

| |
= + +
|

\ .
(9)

ac bc cc
c r a b c
L L L
e i i i

| |
= + +
|

\ .
(10)
are the three phase emfs induced by the variation of flux
linkage due to the saliencies.
B. Electromagnetic Torque
The electromagnetic torque of the PMSM can be obtained
by taking the partial derivative of the system co-energy, W
f
,
with respect to the rotor position angle , i.e. T=W
f
/, and it
can be derived as:
2 2 2
2 2 2
af bf cf
aa a bb b cc c
a b c
L i L i L i
T i i i




= + + + + +


2 2 2
ab ac bc
a b a c b c
L L L
i i i i i i


+ + +

(11)
The electromagnetic torque has two components: one
produced by the stator currents and rotor magnets, and the
other by the saliencies.
III. NONLINEAR INCREMENTAL INDUCTANCE
A. Analytical Expression of Nonlinear Inductance
As discussed above, the structural and magnetic saturation
saliencies can be reflected by the variation of the self and
mutual inductances of the stator windings. These inductances
are nonlinear functions of stator phase winding currents and the
rotor position, which require a large amount of data to describe
numerically. Use of analytical expressions of these nonlinear
inductances could simplify the implementation of the
numerical simulation model and reduce significantly the
amount of parameters.
Since the phase inductance is a periodic function of rotor
angular position, the relationship between the phase inductance
and the rotor position can be expressed by a Fourier series [4]:
( ) ( )
0
1
cos sin
n
m m
m
L a a m b m
=
= + +

(12)
The number of terms of the Fourier series can be chosen
through the best curve fitting of the experimentally measured
phase inductances.
Due to the nonlinear characteristics of the magnetic core,
the inductances also vary with the stator current. Consequently,
different sets of coefficients, a
0
, a
m
, and b
m
(m=1, 2, , n), can
be obtained with different currents, and expressed as functions
of corresponding currents. Therefore, we have:
( ) ( ) ( ) ( ) ( )
0
1
, cos sin
n
m m
m
L i a i a i m b i m
=
= + +

(13)
For a group of known values of phase inductance, L(
j
,i
k
),
the coefficients of the corresponding Fourier series can be
obtained by nonlinear curve fitting with sufficient accuracy,
where j and k refer to the various experimental rotor positions
and currents, respectively. The nonlinear model of the self and
mutual inductances can be readily incorporated into the PMSM
model presented in Section II.
B. Experimental Measurement of Phase Inductances
In its normal operating state, the total stator flux linkage of
a PMSM has two components due to the stator and rotor fluxes,
or
t
=
r
+
s
, and therefore, the phase inductances are related
77
to both of them. In order to estimate the saturation effect at
various stator fluxes, DC offsets, i
dc
, are used to emulate the
effect of the three-phase currents. Based on the analysis of the
relationship between the inductance variation and the saturation
of magnetic core, a method for measuring the phase
inductances is designed and the saturation effect can be
reflected in the inductance expression measured by this
method.
The experiment is carried out on a 6-pole surface mounted
PMSM, and the rotor of the motor is locked by a dividing head
as shown in Fig.1. Various DC offsets are injected to one of the
stator windings. Meanwhile, a small AC current is applied to
one phase while the other two are open-circuited. The power,
voltage, and current of the excited phase are measured and the
voltages of the other two phases are also recorded. The self and
mutual inductances of phases a, b and c at different rotor
positions and stator fluxes can then be calculated via circuit
analysis. Fig. 2 illustrates the self inductances of phase a, L
ssa
,
at various rotor positions with a DC offset of 6A and without
DC offset.
According to the magnetic flux distribution in a PMSM, it
can be concluded that the inductances by considering the
structural and saturation saliencies are functions of amplitudes
and angles of both stator and rotor fluxes, L(
s
,
r
,
s
,
r
),
where
s
is the position of the stator rotating flux,
r
the
position of rotor flux, and
s
and
r
are the amplitudes
of the stator and rotor fluxes, respectively. Since the amplitude
of the rotor flux produced by the permanent magnets can be
considered as a constant, the inductances can be simplified as
functions of
s
,
r
, and
s
, and rewritten as L(
s
,
r
,
s
).
Because of the symmetrical distribution of the three phase
stator windings, the three phase inductances have the same
profile with a 120 phase shift.
In order to simplify the inductance expressions, the angle
between the stator flux,
s
, and the stator winding axis of
phase a, denoted by , and the angle between the stator and
rotor fluxes, denoted by , as shown in Fig. 3, are used in the
inductance model instead of
s
and
r
. Because of the
symmetrical motor structure, the inductances at the following
positions can be obtained by injecting various DC currents to
phase a according to the previous experiment: (1) =0, =0-
360, and
s
corresponding to i
ms
=0~I
rated
(the rated
current); (2) =120, =0~360, and
s
corresponding to
i
ms
=0~I
rated
; (3) =240, =0~360, and
s
corresponding
to i
ms
=0~I
rated
. The phase a inductance, L
saa
(, , i
ms
), at any
stator and rotor fluxes (=0~360, =0~360, and i
ms
=0~I
rated
)
can be estimated by linear interpolation of the measured
inductance expression. Due to the symmetry of the three phase
windings, the inductances of phases b and c at any rotor
position and stator current can be estimated based on the phase
a inductance expression by shifting the rotor position by 120,
respectively. For a given set of and i
ms
, we have L
sbb
(,,i
ms
)=
L
saa
(-120
o
,,i
ms
), and L
scc
(,,i
ms
)= L
saa
(+120
o
,,i
ms
). Similar
procedure can be applied to estimate the mutual inductances.
Therefore, the structural and saturation saliencies of the motor
magnetic filed could be mapped out in terms of the three phase
self and mutual inductances.

U
W
V
N
S
Dividing
Head
VU
VV
VW
Neutral
Function
Generator
Amplifier
C
Power Analyzer
CH1 CH2 CH3
A V A V A V
L

Figure 1. Experimental setup for inductance measurement
0 30 60 90 120 150 180 210 240 270 300 330 360
0.0092
0.0094
0.0096
0.0098
0.0100
0.0102
0.0104
FittingResult without DC
TestingResult without DC
FittingResult withDC
FittingResult withDC
S
e
l
f

i
n
d
u
c
t
a
n
c
e

o
f

p
h
a
s
e

a

(
H
)
Rotor position(Electrical Angle)

Figure 2. Measured and fitted curves of self inductances
with and without DC offset

N
S
r
A X
s


Phase a axis
Phase b axis Phase c axis
B
Y C
Z

N
S
r
A X
s

b
Phase a axis
Phase b axis
Phase c axis
B
Y
C
Z

(a) (b)

N
S
r
A X s

c
Phase a axis
Phase b axis
Phase c axis
B
Y
Z
C

(c)
Figure 3. (a) Rotor and stator fluxes linking phase a with a DC offset in
phase a; (b) Rotor and stator fluxes linking phase b with a DC offset in phase
a; (c) Rotor and stator fluxes linking phase c with a DC offset in phase a.
IV. NUMERICAL SIMULATION OF DTC DRIVE
A. Implementation of PMSM DTC System
Fig.4 shows a schematic block diagram of the conventional
DTC [5]. In the DTC of PMSM, the electromagnetic torque can
be controlled by keeping the amplitude of the stator flux
linkage constant and controlling the load angle by applying a
proper stator voltage vector. The actual flux and torque values
are firstly calculated based on the proposed PMSM model by
considering the saliencies, and in the hysteresis comparators,
the calculated torque/flux are compared with the torque/flux
references. The outputs of the comparators are used to select
78
the proper stator voltage vector based on the optimal voltage
vector switching logic so as to minimize the errors between the
actual and reference values of flux and torque. In the
conventional DTC, only six voltage vectors and two null
voltage vectors can be produced by a two level voltage-source
inverter.
The PMSM DTC scheme shown in Fig.4, which is a closed
loop system with an outer speed control loop and an inner
torque control loop, is implemented numerically in Simulink.
In the simulation model, the two phase stationary - reference
frame is used. The stator currents i
a,
i
b
and i
c
and the DC link
voltage v
dc
are detected by current and voltage transducers. The
currents are then transformed by the Clark transformation into
components (i

, i

), and the u

and u

are deduced from v

dc
and
the switching state in the previous sampling period. The stator
flux linkage and the electromagnetic torque are obtained as
follows:

( ) ( ) ( ) ( ) ( )
1 n n s n s n
t u t R i t T t


+
= + (14)

( ) ( ) ( ) ( ) ( )
1 n n s n s n
t u t R i t T t


+
= + (15)

( ) ( ) ( )
2 2
s n n n
t t t

= + (16)

( ) ( )
3
2
em n
T t P i i

= (17)
B. Simulation by the Proposed Model
Based on the proposed model and the parameters identified
by experiments, the dynamic performance of the surface
mounted PMSM DTC drive system with known initial rotor
position is simulated. Figs.5 and 6 show the simulated stator
current, and the stator flux trajectory of the system
corresponding to a speed command of 1500 rpm and a load
torque of 4.5 Nm, with the parameters of the speed loop PI
controller carefully tuned for optimal performance.
C. Comparison with PMSM Model in Similink Library
The simulation results of the proposed PMSM model are
compared with that of the PMSM model in the Simulink library
with constant inductances under the same conditions. Figs.7
and 8 show the simulated electromagnetic torque and rotor
speed of these two models in the same time frame.

Figure 4. Control block diagram of conventional DTC
0.00 0.05 0.10 0.15 0.20 0.25 0.30
-15
-12
-9
-6
-3
0
3
6
9
12
15
C
u
r
r
e
n
t

(
A
)
Time (s)
Ia
Ib
Ic

Figure 5. Stator currents by the proposed model
-0.20 -0.15 -0.10 -0.05 0.00 0.05 0.10 0.15 0.20
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20

Figure 6. Stator flux trajectory (Wb)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0
1
2
3
4
5
6
7
8
9
10
E
l
e
c
t
r
o
m
a
g
n
e
t
i
c

t
o
r
q
u
e

(
N
m
)
Time (s)
Tembased on the Simulink model
Tembased onthe proposed model

Figure 7. Torque comparison between the two models
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0
200
400
600
800
1000
1200
1400
1600
Rotor Speed basedon the Simulink model
Rotor Speed basedon the proposed model
R
o
t
o
r

S
p
e
e
d

(
r
p
m
)
Time (s)

Figure 8. Speed comparison between the two models
79
V. ESTIMATION OF INITIAL ROTOR POSITION
A. Experimental Procedure
In the DTC scheme, the real time rotor position is derived
by integrating the back emf against time. However, the starting
procedure under full load is very difficult because no current or
emf information is available for determining the rotor position
before starting. Therefore, it is essential to develop a sensorless
method to determine the initial rotor position for full load
starting of PMSM drive systems. Numerous approaches have
been proposed to estimate the initial rotor position from
motors terminal properties. Most of them are effective for the
motor with large structural saliency, e.g. the interior PMSM.
However, in a surface mounted PMSM, it becomes difficult
due to the small saliencies. The saturation saliency may have to
be used because the structural saliency is not significant.
It has been discussed that the winding inductance is a
function of rotor position and stator current. With this feature,
the initial rotor position can be estimated by injecting two
kinds of voltage pulses with high and low amplitudes, as
shown in Fig.9, to the stator winding terminals, respectively
[6], [7]. At an unknown initial rotor position, one of the three
phase terminals, say phase a, is applied with a positive pulse
while the terminals of the other two phases are applied with a
negative pulse. After a certain time interval, the positive pulse
is followed by a negative pulse and the negative pulse by a
positive pulse to bring down the currents in the phase windings
to zero.
In the experiment, these pulse voltages are produced by the
PWM inverters, and the gate signals of the inverters are
controlled by a dSPACE DS1104 system. For example, when
the positive pulse is applied to phase a, the gate signals of 100
are created and applied to the inverter and the stator winding
terminals, as shown in Fig.10.
B. Experimental Results
The high and low voltage pulses are applied to the testing
surface mounted PMSM through a PWM inverter. The
corresponding phase currents are measured by current probes
and an oscilloscope. Figs. 11-12 show the experimental current
waveforms. Fig.11 depicts the three phase current response
recorded for high voltage pulses with gate signals 100, 010 and
001, respectively. Fig.12 plots the current response obtained for
low voltage pulses with gate signal 100. For each phase, the
measurement is taken at every locked rotor position and the
motor is mechanically rotated from one position to the next.
The measured three phase peak currents at different rotor
positions with high and low voltage pulses are plotted in Figs.
13 (a) and (b), respectively.
C. Comparison of Peak Current and Inductance Variation
Compared with the almost sinusoidal waveform under the
non-saturated condition, it can be seen the magnetic saturation
due to the stator currents could be neglected with the low
voltage pulse. However, with the high voltage pulse, the
magnetic saturation becomes evident.
-0.00050.00000.0005 0.0010 0.00150.00200.00250.0030
-210
-180
-150
-120
-90
-60
-30
0
30
60
90
120
150
180
210
V
o
l
t
a
g
e

P
u
l
s
e


(
V
)
Time (s)
Low voltage pulse
High voltage pulse

Figure 9. Low and high voltage pulses

Figure 10. Equivalent circuit of the motor at standstill
0.0000 0.0005 0.0010 0.0015 0.0020 0.0025 0.0030 0.0035
-3
-2
-1
0
1
2
3
4
5
6
A for 100
B for 100
C for 100
A for 010
B for 010
C for 010
A for 001
B for 001
C for 001
C
u
r
r
e
n
t

(
A
)
Time (s)

Figure 11. Experimental current response to high voltage pulse
-0.0005 0.0000 0.0005 0.0010
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
C
u
r
r
e
n
t

(
A
)
Time (s)
A
B
C

Figure 12. Experimental current response to low voltage pulse
V
dc

Phase a
Phase b
Phase c
i
a
R
a
R
b
R
c

N
e
a

i
b
i
c

v
n

80
0 20 40 60 80 100 120
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
A
B
C P
e
a
k

c
u
r
r
e
n
t

(
A
)
Rotor position (Degree)

(a)
0 20 40 60 80 100 120
3.0
3.5
4.0
4.5
5.0
5.5
6.0
P
e
a
k

c
u
r
r
e
n
t

(
A
)
Rotor position (Degree)
A
B
C

(b)
Figure 13. Peak current versus rotor position with: (a) low voltage pulse, (b)
high voltage pulse
At some positions when the stator flux aids the rotor flux,
the magnetic circuit becomes more saturated resulting in a
smaller stator phase winding inductance. When the stator flux
is in the opposite direction of the rotor flux, however, the
magnetic circuit becomes less saturated, resulting in a larger
stator phase winding inductance. Since the inductance variation
has been measured, it is possible to find the rotor position by
analyzing the relationship between the inductance variation and
peak current variation versus the rotor position.
From the experimental study, it is known that the effect of
the saturation saliency is sufficiently evident in the phase self
inductance variation with a DC offset of 6A, while the effect of
the structural saliency is solely demonstrated in the inductance
variation without the DC offset (Fig.2). Both the inductance
variations and the peak current variations with high voltage
pulses are plotted together in Figs.14-16 in order to determine
the d- and q-axes of the rotor. Since the DC offset has only
been applied in the phase a winding, the effect of the saturation
saliency in the self inductances of phases b and c are not as
apparent as that in the phase a inductance. It can also be seen
that the structural saliency is very small due to the surface
mounted structure of the testing motor.
As shown in Figs. 14-16, for the surface mounted PMSM,
since the q-axis inductance is higher than the d-axis inductance,
the positions of the d- and q-axes can be clearly seen in both
profiles of the inductance measured with and without the DC
offset while only the inductance profile with DC offset can
reveal the rotor polarity (north or south pole). Because of the
structural saliency, the stator current produced by the high
voltage pulse reaches a valley when the magnetic field of the
stator current is on the q-axis, and a peak when the stator field
is on the d-axis.
Once the position of the d-axis is determined, the rotor
polarity can be identified by comparing the magnitudes of the
stator currents on the d-axis. If the stator field aids the rotor
field (same polarity) on the d-axis, the magnetic circuit
becomes more saturated, and thus the magnitude of the stator
current produced by the high pulse voltage would be higher
than that when the stator field opposes the rotor field (opposite
polarity) and the magnetic circuit becomes less saturated [8].
Therefore, detection of the initial rotor position by means of
the structural and saturation saliencies is feasible.

Figure 14. Phase a inductance and peak current variation

Figure 15. Phase b inductance and peak current variation

Figure 16. Phase c inductance and peak current variation
81
D. Comparison of Simulated and Measured Peak Currents
Based on the equivalent circuit in Fig.10, a Simulink model
of the PMSM at standstill is built and the voltage pulse
injection algorithm outlined above is numerically simulated
with the measured inductances as the parameters. Fig.17 plots
both the simulated and measured peak currents of phase a
versus the rotor position when a high voltage pulse of 200 V in
magnitude and 0.6 ms pulse width is applied.
As shown in Fig.17, the general profiles of the simulated
and experimental peak currents are similar. As mentioned
previously, the self and mutual inductances were measured
only at limited rotor positions with DC offsets applied only in
phase a winding, and the linear interpolation is performed to
deduce the inductances at other rotor and DC offset field
positions. This means the inductance values at these positions
are inaccurate and therefore, some error would appear in the
simulated results at those specific positions.
E. An Algorithm for Detecting Initial Rotor Position
The above analysis shows that it is quite feasible to identify
the positions of the d- and q-axes and the rotor polarity from
the measured peak current versus rotor position curve. The
peak current variation with the low voltage pulse in Fig.13 (a)
is very small due to the surface mounted PM structure. The
peak current distribution can be approximately regarded as sine
waves versus the rotor position and modeled by:

( )
0 0
cos 2
a
I I I = + (18)

0 0
2
cos 2
3
b
I I I

| |
= + +
|
\ .
(19)

0 0
2
cos 2
3
c
I I I

| |
= +
|
\ .
(20)
where I
a
, I
b
and I
c
are peak currents and rotor position,
I
0
=(I
a
+I
b
+I
c
)/3, I
0
is the magnitude of sine waves and can be
obtained by curve fitting the experimental results shown in Fig.
18(a). As discussed in [6], the signs of I
a
, I
b
, and I
c
can be
used to determine the region where the rotor is. Although the
region of rotor position could be estimated by the simulation
results of I
a
, I
b
, and I
c
, it is almost impossible to implement
it in the real time system for a surface mounted PMSM because
the saturation saliency is too small to detect.
0
1
2
3
4
5
6
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96 102 108 114 120
Rotor Position (Degree)
P
e
a
k

c
u
r
r
e
n
t

(
A
)
Simulated peak current
Testing peak current

Figure 17. Comparsion of the simulated and measured peak currents
As for the peak current variation measured with high
voltage pulses, the phase currents shown in Fig.13 (b) can also
be modeled by an average value, I
0
, plus some offset values,
I
01
and I
02
, as a function of rotor position, , as the
following:

( ) ( )
0 01 02
cos cos 2
a
I I I I = + + (21)

0 01 02
2 2
cos cos 2
3 3
b
I I I I


| | | |
= + + +
| |
\ . \ .
(22)

0 01 02
2 2
cos cos 2
3 3
c
I I I I


| | | |
= + + +
| |
\ . \ .
(23)
Fig.18 (b) plots the curve fitting of the experimental data
with the above model.
Based on the above analysis, a very simple algorithm is
composed for detecting the initial rotor position. The measured
peak current curves of phases a, b and c at different rotor
positions are stored the memory of the microprocessor in the
form of a look up table. When the rotor is standstill, send 100,
010, and 001 gate signals to the PWM inverter to apply high
voltage pulses to the three phase stator winding terminals, and
record the peak phase currents. Possible rotor positions can be
readily found by solving inversely (21), (22), and (23) using a
sectional interpolation method. For phase a, the possible rotor
positions corresponding to the peak current I
ap
can be obtained
as
a1

a2
,
a3
and so on, and similarly, for phases b and c, the
possible rotor positions corresponding to I
bp
and I
cp
can be
obtained as
b1

b2
,
b3
,
c1

c2
,
c3
and so on. Among all these
possible rotor positions, the real rotor position is given by the
common angle obtained all three phases. Fig. 19 compares the
simulated rotor position against the real rotor position. It can be
seen that this algorithm can produce satisfactory results.
0.2
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96 102 108 114 120
Rotor position (Degree)
P
e
a
k

c
u
r
r
e
n
t

(
A
)
A B C

(a)
3
3.5
4
4.5
5
5.5
6
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96 102 108 114 120
Rotor position (Degree)
P
e
a
k

c
u
r
r
e
n
t

(
A
)
A
B
C

(b)
Figure 18. Fitting peak current with (a) low and (b) high voltage pulses
82
0
6
12
18
24
30
36
42
48
54
60
66
72
78
84
90
96
102
108
114
120
0 6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96 102 108 114 120
Actual position (Degree)
E
s
t
i
m
a
t
e
d

p
o
s
i
t
i
o
n

(
D
e
g
r
e
e
)

Figure 19. Simulation result of initial rotor position estimation
VI. CONCLUSIONS
A nonlinear PMSM model incorporating both the structural
and magnetic saturation saliencies has been presented in this
paper. In the model, the saliencies are reflected by the variation
of the stator winding inductances with respect to the rotor
position and stator phase currents. The nonlinear inductances of
a surface mounted PMSM at various rotor positions and stator
currents are experimentally measured and discussed. The
Fourier series are employed to express the inductances as
functions of the rotor position and stator currents. The
structural saliency though small in the surface mounted PMSM
is revealed in the self and mutual inductance profile versus
rotor position without any DC offset, and when DC offset
currents are applied in the stator phase windings, the effect of
magnetic saturation becomes evident.
The performance of the surface mounted PMSM drive
system with the conventional DTC scheme is simulated by
using the proposed model and the validity of the model is
confirmed by the comparison of results with those obtained by
the PMSM model in the Simulink library.
Two schemes for detecting the initial rotor position by
injecting low and high voltage pulses are investigated by both
numerical simulation and experimental testing. It is found that
the low voltage pulse method is not suitable for surface
mounted PMSMs because the structural and saturation
saliencies are too small to detect. Corresponding to the high
voltage pulses, the peak stator currents are strong enough to
produce reasonably large magnetic saturation saliency for
identifying the rotor position and polarity.
A simple numerical algorithm for implementing the high
voltage pulse method is proposed. The validity and accuracy of
this algorithm is confirmed by a numerical simulation before it
is implemented in real time system.
REFERENCES
[1] Y. Yan, J.G. Zhu, and H.W. Lu, A PMSM Model Incorporating
Structural and Saturation Saliencies, in Proc. 8
th
International
Conference on Electrical Machines and Systems, Nanjing, China,
Sept.27-29, 2005, pp.194-199.
[2] Y. Yan, J.G. Zhu, and H.W. Lu, Direct Torque Control of a Surface-
Mounted Permanent Magnet Synchronous Motor based on Accurate
Modelling, in Proc. Australasian Universities and Power Engineering
Conference, Hobart, Australia., Sept.25-28, 2005, pp.302-307.
[3] Y. Yan, J.G. Zhu, H.W. Lu, Y.G. Guo, and S.H. Wang, Study of a
PMSM model incorporating structural and saturation saliencies, in
Proc. 6
th
IEEE International Conference on Power Electronics and
Drive Systems, Kuala Lumpur, Malaysia, Nov. 28 - Dec. 1, 2005, pp.
575-580.
[4] P. Cui, J.G. Zhu, Q.P. Ha, G.P. Hunter, and V.S. Ramsden, Simulation
of non-linear switched reluctance motor drive with PSIM, in Proc. 5th
International Conference on Electrical Machines and Systems, Aug.
2001, vol. 1, pp. 1061-1064.
[5] D. Sun, High performance direct torque control for permanent magnet
synchronous motors, Ph.D. Thesis, Zhejiang University, P.R. China,
March, 2004.
[6] N. Matsui and T. Takeshita, A novel starting method of sensorless
salient-pole brushless motor, Conference Record of the 1994 IEEE
Industry Applications Society Annual Meeting, Oct. 2-6, 1994, vol.1, pp.
386392.
[7] P.B. Schmidt, M.L. Gasperi, G. Ray, and A.H. Wijenayake, Initial rotor
angle detection of a nonsalient pole permanent magnet synchronous
machine, Conference Record of the 1997 IEEE Industry Applications
Society Annual Meeting, Oct. 5-9, 1997, Vol.1, pp. 459-463.
[8] S. Ostlund, and M. Brokemper, Initial rotor position detections for an
integrated PM synchronous motor drive, Conference Record of the
1995 IEEE Industry Applications Society Annual Meeting, Oct. 8-12
1995, Vol.1, pp.741747.
.

83
Conference Record of the
2006 IEEE Industry Applications Conference
Forty-First lAS Annual Meeting
Copyright 2006 by the Institute of Electrical and Electronics Engineers, Inc.
All rights reserved.
Copyright and Reprint Permission
Abstracting is permitted with credit to the source. Libraries are permitted to photocopy beyond the limit of U.S. copyright law, for
private use of patrons, those articles in this volume that carry a code at the bottom of the first page, provided that the per-copy fee
indicated in the code is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923.
Other copying, reprint, or reproduction requests should be addressed to:
IEEE Copyrights Manager, IEEE Service Center, 445 Hoes Lane, P.O. Box 1331, Piscataway, NJ 08855-1331.
IEEE Catalog Number 06CH37801
06CH37801C (CD-ROM)
ISBN 1-4244-0365-0
ISSN 0197-2618
Additional copies of this publication are available from
IEEE Operations Center
P.O. Box 1331
445 Hoes Lane
Piscataway, NJ 08855-1331 USA
1-800-678-IEEE
1-732-981-1393
1-732-981-9667 (FAX)
email: customer.services@ieee.org