Transactions of the Institute of Measurement and Control 33, 5 (2011) pp.

522–541

A composite speed controller based

on a second-order model of
permanent magnet synchronous
motor system
Shihua Li, Kai Zong and Huixian Liu
School of Automation, Southeast University, Nanjing, P.R. China

The speed-regulation problem for a permanent magnet synchronous motor (PMSM) servo
system is studied in this paper. In traditional control design for a speed loop, a first-order
model is used to approximately describe the relationship between the reference quadrature axis
current and the speed output, i.e. the reference quadrature axis current is regarded as the same
as the quadrature axis current. This approximation degrades the closed-loop performance
of PMSM system when using a vector control scheme. To this end, a second-order model is
built to describe the relationship between the reference quadrature axis current and the speed
output for a PMSM system. Based on this second-order model, a standard sliding mode
controller is designed for the speed loop. Then, to reduce the chattering and improve the
performance of the system, a composite controller consisting of a sliding mode feedback part
and a disturbance compensation part based on an extended state observer are developed.
Simulation and experiment results and comparisons are given to show the effectiveness of the
proposed method.

Key words: permanent magnet synchronous motor (PMSM); composite control; second-order
model; sliding mode control; extended state observer.

1. Introduction

Owing to its high power density, low inertia, almost no need for maintenance and
many other advantages, permanent magnet synchronous motors (PMSM) are widely
used in industrial applications (Pillay and Krishnan, 1989). A PMSM is a non-linear

Address for correspondence: Shihua Li, School of Automation, Southeast University, Nanjing
210096, P.R. China. E-mail: lsh@seu.edu.cn

ß 2010 The Institute of Measurement and Control 10.1177/0142331210371814

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 523

system with multiple coupled states, and time-varying parameters. So the control of
the PMSM servo system is not an easy task due to its non-linearities and uncertainties.
Classical linear controllers such as PI controllers cannot provide a satisfactory
performance. Hence, many kinds of advanced non-linear control algorithms have been
reported, such as adaptive control (Jin and Lee, 2009; Li and Liu, 2009; Lin et al., 1998;
Mohamed, 2007; Mohamed and El-Saadany, 2008), robust control (Hsien et al., 1997; Lin
et al., 2008), sliding mode control (SMC) Baik et al. (2000); Lai and Shyu (2005); Lin et al.
(1998); Utkin (1993); Wai (2001), input–output linearization control Baik et al. (2000);
Grcar et al. (1996); Lin et al. (2008), backstepping control Vilathgamuwa et al. (2003);
Zhou and Wang (2002), finite-time control Li et al. (2010), and intelligent control Jan
et al. (2008); Kung and Tsai (2007); Wai (2001); Wang et al. (2001). These control
algorithms have enhanced the performance of the servo system from different aspects.
Compared with a direct torque control frame, the vector control scheme has been
well developed and received much more attention. Under this scheme, the torque and
flux producing components of the stator current are decoupled so that the
independent torque and flux controls are possible as in dc motors. A structure of
cascade control loops is usually employed. For example, for the speed-regulation
problem, the control scheme includes a speed loop and two current loops. The cascade
architecture offers obvious advantages including the ability to resist disturbances and
to improve set point response performance.
Under the framework of a vector control and cascade structure, in the control
design for a speed loop, a first-order model is usually used to approximately describe
the relationship between the reference quadrature axis current and the speed output,
i.e. the reference quadrature axis current iq is regarded as the same as the quadrature
axis current iq (Achour et al., 2009; Li and Liu, 2009; Li et al., 2010). Conventionally, the
torque (current) is controlled with a response time faster than that of the speed in
order to control the speed. When the control period of the speed loop is notably larger,
e.g., a factor of 10 or more, than the control periods of current loops, this
approximation error can be ignored. However, considering the developing trend in
high-performance servo systems, the control periods of the speed and current loops
are all becoming smaller and the relative differences in control periods between the
speed loop and current loops are becoming smaller or even vanishing. In this case, this
approximation degrades the closed-loop performance of a PMSM system.
To guarantee the high performance of PMSM systems, another important factor that
should be considered is system disturbances. In real industrial applications, PMSM
systems are always faced with different disturbances. These disturbances may arise
internally, e.g. friction force and unmodelled dynamics, or externally, e.g. load
disturbances. Among conventional feedback-based control methods, the SMC method
is an efficient example. It shows a very good ability to reject disturbances (Hung et al.,
1993; Utkin, 1977). Utkin (1993) designed a sliding mode controller for the speed-
regulation problem of a PMSM system based on the second-order equation between
the q-axis voltage uq and the speed output, where the acceleration signals are needed.

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

524 Speed-regulation problem for a PMSM servo system

However, the acceleration is a high-bandwidth signal and, in general, it is difficult to

acquire or estimate accurately (Ho and Sen, 1991; Nandam and Sen, 1989). Lai and
Shyu (2005) designed a multi-segment SMC-based controller for the position control
problem of PMSM system. According to the standard sliding mode theory, if the
switching control gain is selected to be larger than the upper bound of disturbances, the
disturbances can be completely rejected. Note that for practical applications, the upper
bound of disturbances is difficult to obtain in advance which often results in a
conservative control law with large control effort, causing a large amount of chattering.
To this end, some improvements are made to reduce the conservativeness of
selecting switching control gain. Lin et al. (1998) proposed an adaptive SMC method
for the position control problem, where a simple adaptive algorithm is utilized to
estimate the bound of disturbances. Wai (2001) proposed a total sliding mode
controller for the position control problem of a PMSM system, where a recurrent fuzzy
neural network is adopted as a bound observer to facilitate adaptive control gain
adjustment.
The above-mentioned methods face an unavoidable application problem: chatter-
ing. Some improved methods, e.g. the boundary layer method (Slotine and Li, 1991),
can alleviate or avoid the chattering. Baik et al. (2000) proposed a non-linear controller
using a feedback linearization technique and a boundary layer integral SMC
technique. The chattering phenomenon is avoided. However, to do this, the
disturbance rejection properties are sacrificed to some extent.
Another efficient way of improving the disturbance rejection performance of a
system in such cases is to introduce a feed-forward compensation part into the
controller in addition to the conventional feedback part. Thus, a composite control
method is obtained. Since in real PMSM applications it is usually impossible to
measure the disturbances directly, disturbance estimation techniques have to be
developed. The disturbance observer (DOB) technique was originally presented by
Ohnishi (1987). Many DOB-based control schemes for linear and non-linear systems
have been put forward, (e.g. Chen, 2004; Guo and Chen, 2005; Wei et al., 2009a; Wei
and Guo, 2009; Wei et al., 2009b). In particular, different types of DOB-based SMC
methods have been firstly proposed by Wei et al. where rigorous proofs have been
presented to ensure the stability of the whole closed-loop systems (Wei and Guo, 2009;
Wei et al., 2009a,b). Another kind of disturbance estimation techniques is the extended
state observer (ESO) (Han, 1995, 2009). This method has also been applied in some
industrial control problems, e.g. robotic systems (Su et al., 2004), machining processes
(Wu et al., 2007), power converters (Sun and Gao, 2005), and motor systems (Feng
et al., 2004; Pan et al., 2006). The DOB-based control method and ESO-based
control method are two kinds of popular solutions for PMSM systems. Application
examples can be found in the literature, e.g. Kim et al. (1999), Kim and Youn (2002),
Li and Liu (2009), Li et al. (2010) and Su et al. (2005).
In this paper, considering the above-mentioned problem, the speed-regulation
problem is addressed. First, a second model is built to describe the relationship

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 525

between the reference quadrature axis current and the speed output for the PMSM
system. Second, a standard sliding mode controller is designed for the speed
loop. To ensure that the closed-loop system possesses a good disturbance rejection
property, the switching gain of the sliding mode controller needs to be larger than the
bound of the lumped disturbances. However, simulation and experimental results
show that the steady speed of the closed-loop system has obvious fluctuations. Third,
an efficient disturbance estimation technique, i.e. ESO, is introduced to estimate the
disturbances of the system. After disturbance compensation based on ESO, the
switching gain only needs to be larger than the bound of the disturbance
compensation error which is usually much smaller than that of the lumped
disturbances. Hence, the chattering will be much reduced while the disturbance
rejection property of the closed-loop PMSM system can still be maintained. Thus, a
composite speed controller (SMC þ ESO), using a combination of the sliding mode
feedback part and disturbance compensation part based on ESO, is developed.
Simulation and experimental results and comparisons are given to show the
effectiveness of the proposed method.
This paper is organized in the following manner. The mathematical model
of PMSM is introduced and a second model is built to describe the relationship
between the reference quadrature axis current and the speed output for the
PMSM system in Section 2. In Section 3, first the design of a standard SMC
scheme is given in detail. An improved control scheme, called SMC þ ESO, is
then developed. Simulation and experimental results on a PMSM speed-regulation
system are presented in Section 4. Finally, the conclusions are given in Section 5.

2. Model description and analysis

2.1 A first-order model of a PMSM speed-regulation system

Assume that the magnetic circuit is unsaturated, hysteresis and eddy current loss are
ignored and the distribution of the magnetic field is in sine space. Under this
condition, the mathematical model of a surface-mounted PMSM in the d–q frame can
be described as follows (Krishnan, 2001):
0 1 0 Rs 10 1 0 u 1
i_d  L np ! 0 id L
d

@ i_q A ¼ B np  f C B q C
@ np !  RLs  L A@ iq A þ @ L A,
u
ð1Þ
Kt B TL
!_ 0 J J !  J

where ud and uq are the stator voltages of the d- and q-axes, id and iq are the stator
currents of the d- and q-axes, ! is the rotor speed in angular frequency, np is the
number of poles pairs, L is the stator inductance, Rs is the stator resistance, f is the

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

526 Speed-regulation problem for a PMSM servo system

rotor flux linkage, TL is the load torque, J is the moment of inertia, B is the viscous
friction coefficient, and Kt ¼ 32 np f is the torque constant.
Figure 1 shows the general control scheme of the PMSM speed-regulation system
based on vector control. A cascade structure which includes a speed loop and two
current loops is employed. In order to decouple the speed and the currents, the
reference current id is set to be zero. Here two standard PI controllers, which are used
to stabilize the d–q axes current errors, are adopted in the two current loops. In this
paper, we concentrate on the design of the speed controller.
From (1), the motor dynamic equation can be rewritten as
!_ ¼ biq þ aðtÞ, ð2Þ
where
Kt B TL
b¼ , aðtÞ ¼  ! 
J J J
can be considered as the system disturbances including friction and external load
disturbances. The Laplace transformation of (2) is
sðsÞ ¼ bIq ðsÞ þ AðsÞ: ð3Þ
For the control design of the speed loop, the q-axis stator current iq is usually
approximately replaced by the q-axis reference current iq , i.e.
!_  biq þ aðtÞ: ð4Þ
Thus, a first-order model is used to approximately describe the relationship between
the reference quadrature axis current and the speed output. Some controllers have
been constructed according to this model (Achour et al., 2009; Li and Liu, 2009; Li et al.,
2010).

wr i*q ua
uq
Speed
controller PI
Inverse
park SV 3-phase
iq
ud transformation ub PWM inverter PMSM
w i*d = 0
PI

id
ia

Park
Clark Sensor
transformation
transformation ib (encoder)

Speed and angle

calculation

Figure 1 A general vector control scheme of a PMSM speed-

regulation system

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 527

2.2 A second-order model of a PMSM speed-regulation system

According to previous analyses, this approximation degrades the closed-loop
performance of the PMSM system. Next, we try to have a more precise model for
the reference quadrature axis current and the speed output.
Observing the input and output of the PI controller of the current loop iq from
Figure 1, the following expression can be obtained:
Uq ðsÞ ki
¼ kp þ , ð5Þ
Iq ðsÞ
 Iq ðsÞ s

where kp and ki are the proportional and integral gains of the PI controller in the
current loop of iq, respectively.
From (5), we have
Uq ðsÞ
Iq ðsÞ ¼ Iq ðsÞ  : ð6Þ
kp þ ki =s
Substituting (6) into (3) yields
   
2 ki ki  b
s þ s ðsÞ ¼ b s þ Iq ðsÞ  sUq ðsÞ
kp kp kp
 
ki
þ sþ AðsÞ: ð7Þ
kp
Thus, a second-order model between the reference quadrature axis current and the
speed output for the PMSM speed-regulation system is obtained.
For simplicity of design, defining
 
ki 
UðsÞ ¼ b s þ I ðsÞ, ð8Þ
kp q
i.e.
kp
Iq ðsÞ ¼ UðsÞ,
bðkp s þ ki Þ
then (7) can be rewritten as
   
2 ki b ki
s þ s ðsÞ ¼ UðsÞ  sUq ðsÞ þ s þ AðsÞ: ð9Þ
kp kp kp
The inverse Laplace transformation of (9) is

ki
!€ ¼  !_ þ d ðtÞ þ u, ð10Þ
kp

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

528 Speed-regulation problem for a PMSM servo system

where
b ki
d ðtÞ ¼  u_ q þ a_ ðtÞ þ aðtÞ
kp kp
can be considered as the lumped disturbances of the system. Thus, the equivalent
controlled dynamics are second order.

3. Controller design

3.1 A controller design using standard SMC

Let !r denote the reference speed signal. The speed tracking error is defined
as follows

e ¼ !r  !: ð11Þ
Substituting (10) into the second derivative of (11), the following error equation
can be obtained
ki
e€ ¼ !€ r  !€ ¼ !€ r þ !_  d ðtÞ  u: ð12Þ
kp
The sliding surface is designed as

S ¼ ce þ e_ , c40 ð13Þ
and the speed controller can be designed as

ki
u ¼ ce_ þ !€ r þ !_ þ k  sign ðSÞ, ð14Þ
kp
where sign() is defined as follows
8
< þ1, S 4 0,
sign ðSÞ ¼ 0, S ¼ 0,
:
1, S 5 0:
For the PMSM speed-regulation system, the q-axis reference current iq is actually the
controller output for the speed loop. From (8) and (14), iq can be obtained from the
following expression

 
_i þ ki i ¼ 1 ce_ þ !€ r þ ki !_ þ k  sign ðSÞ : ð15Þ
q
kp q b kp

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 529

Assumption 1. We assume that d(t) is bounded and there exists a constant l140, such that
|d(t)|  l1, for all t  0.
Theorem 1. If Assumption 1 holds, for PMSM system (1) under control law (15), the speed
tracking error will converge to zero if k4l1.
Proof. According to the above analysis, the error equation (12) is obtained. Choosing
the Lyapunov function V ¼ 12 S2 , and taking the derivative of it along the error
system (12), yields
 
_ _ ki
_ € _
V ¼ SS ¼ S½ce þ e ¼ S ce þ !€ r þ !_  dðtÞ  u : ð16Þ
kp
Substituting (14) into (16) yields
_ ¼ SS_ ¼ S½k  signðSÞ  dðtÞ:
V ð17Þ
According to Assumption 1, |d(t)|  l1, hence (17) can be rewritten as follows:
 
_ ¼ SS_ ¼ S k sign ðSÞ  dðtÞ
V
 kjSj þ jdðtÞjjSj
¼ ½k  jdðtÞjjSj
 ½k  l1 jSj
pffiffiffi
¼  2ðk  l1 ÞV 1=2 : ð18Þ

From above inequality, it can be verified that if k4l1, the speed error will arrive at the
sliding surface S ¼ 0 in finite time and stay there. When the speed error states reach
the sliding surface, i.e. S ¼ ce þe_¼ 0, since c40, the speed error will converge to zero
along the sliding surface S ¼ 0. Thus, the theorem is proved. h
According to the above design procedure, a control scheme based on SMC is
obtained, as shown in Figure 2. Note that the generalized PMSM in Figure 2 represents
the two current loops which include the PMSM and other components as in Figure 1.
To ensure that the closed-loop system possesses a good disturbance rejection property,
the switching gain k of sliding mode controller (15) need to be taken larger than the
bound of the lumped disturbances. If there is no precise estimation on the bound of
disturbances, the switching gain k will be selected to be high enough. This imprecision
worsens the chattering caused by SMC. The subsequent simulation and experimental
results show that under this standard sliding mode controller (15), the steady speed of
the closed-loop system has obvious fluctuations.

3.2 An improved sliding mode controller design based on ESO

We now introduce an efficient disturbance estimation technique, i.e. ESO, to estimate

the disturbances of the system. This estimate of disturbances will be employed as

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

530 Speed-regulation problem for a PMSM servo system

Speed controller

d ki
dt kp

d2
dt 2
wr u kp i *q Generalized w
C K.sign(.)
b(kps +ki) PMSM
d
dt C

Figure 2 Simplified control scheme of the second-order model of

the PMSM speed-regulation system

a compensation item in the speed controller. After disturbance compensation based on

ESO, the switching gain k only needs to be taken larger than the bound of the
disturbance compensation error which is usually much smaller than that of the
lumped disturbances.

3.2.1 Design of ESO: The ESO (Han, 1995, 2009) regards the lumped disturbances
of the system, which consists of internal dynamics and external disturbances, as a new
state of the system. This observer is one order of magnitude higher than the usual state
observer. It can estimate both the states and the disturbances. Based on the ESO,
a feed-forward compensation for the disturbances can also be employed in the
control design.
Let us consider the disturbance item here, i.e.
b ki
dðtÞ ¼  u_ q þ a_ ðtÞ þ aðtÞ:
kp kp
Since a second model between iq and ! is obtained from (9) and (10), according to ESO
theory, a three-order ESO can be employed to estimate the states and the disturbances
simultaneously. However, here we only use a second-order ESO to estimate a(t), i.e. a
combination term of viscous friction and external load, instead of d(t). The reason is
that, actually, it is very difficult to have good estimates of the derivatives of uq and a(t).
The three-order ESO has been realized in experiments. However, it does not show a
good tracking effect. Instead, it causes a poor effect on feed-forward compensation.
Further investigation should be carried out.
Hence, a second-order linear ESO is designed for the system (4) as follows
(Miklosovic and Gau, 2004):

z_ 1 ¼ z2  2pðz1  !Þ þ biq ,
ð19Þ
z_ 2 ¼ p2 ðz1  !Þ,

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 531

where p is the desired double pole of ESO with p40. Here, according to
the analysis in Miklosovic and Gao (2004), z1(t) is an estimate of the speed
output ! while z2(t) is an estimate of the disturbances a(t), i.e. z1(t) ! !(t) and
z2(t) ! a(t).

3.2.2 Sliding mode controller based on ESO feed-forward compensation: For the
system (10), a composite sliding mode controller for the speed loop can be designed as
follows

ki ki
u ¼ ce_ þ !€ r þ !_ þ k  sign ðSÞ  z2 , ð20Þ
kp kp

i.e.
 
_i þ ki i ¼ 1 ce_ þ !€ r þ ki !_ þ k  sign ðSÞ  ki z2 : ð21Þ
q
kp q b kp kp

Note that in the above composite controller, (ki/kp)z2 is a newly

added feed-forward compensation part in addition to the sliding mode feedback
part.
Assumption 2. Assume that d(t)  (ki/kp)z2 is bounded and there exists a constant l240,
such that

ki
dðtÞ  z2  l2 , for all t  0:
kp

Theorem 2. If Assumption 2 holds for the PMSM system (1) under control law (21), the
speed tracking error will converge to zero if k4l2.
Proof. Choosing a Lyapunov function V ¼ 12 S2 , and taking the derivative of it along
the error system (12), yields

 
_ _ ki
V ¼ SS ¼ S½ce_ þ e€  ¼ S ce_ þ !€ r þ !_  dðtÞ  u : ð22Þ
kp

Substituting (20) into (22) yields

 
_ ¼ SS_ ¼ S k  sign ðSÞ  dðtÞ þ ki z2 :
V ð23Þ
kp

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

532 Speed-regulation problem for a PMSM servo system

According to Assumption 2, |d(t)  (ki/kp)z2|  l2, hence (23) can be rewritten

as follows:
 
_ _ ki
V ¼ SS ¼ S k sign ðSÞ  dðtÞ þ z2
kp
ki
 kjSj þ dðtÞ  z2 jSj
kp
" #
ki
¼  k  dðtÞ  z2 jSj
kp
5  ½k  l2 jSj ð24Þ

pffiffiffi
¼  2ðk  l2 ÞV 1=2 : ð25Þ
From above inequality, it can be verified that if k4l2, the speed error will arrive at and
stay in the sliding surface S ¼ 0 in finite time. When the speed error states reach the
sliding surface, i.e. S ¼ ce þe_¼ 0, since c40, the speed error will converge to zero along
the sliding surface. Thus, the theorem is proved. h
Remark 1. In Theorem 2, we have a similar conclusion with Theorem 1. The difference
is that in Theorem 2, a feed-forward compensation is introduced into the control law.
When the parameter selection of the ESO is appropriate, z2 will be a good estimate of
a(t) and the bound of
ki b ki
dðtÞ  z2 ¼  u_ q þ a_ ðtÞ þ ðaðtÞ  z2 Þ
kp kp kp
can be less than d(t). In this case, the SMC þ ESO composite controller may take a
smaller value for the switching gain k without sacrificing disturbance rejection
performance, which helps to reduce large chattering caused by high control gains.
Figure 3 shows the controller scheme of SMC based on ESO.

4. Simulation and experimental results

To demonstrate the performance of the proposed composite control method,

simulations and experiments have been performed on a PMSM servo system.
Two control methods, i.e. SMC and SMC þ ESO, are applied to the PMSM servo
system. To have a fair comparison, the control inputs of two algorithms have the same
saturation limit.

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 533

Speed controller
d ki
dt kp
d2
dt 2
wr u kp i *q Generalized w
C K.sign(.) b(kps +ki) PMSM
d ki
dt C b
kp
z2
Second-order
ESO (19)

Figure 3 Controller scheme of SMCþESO

The parameters of a PMSM used in the simulation and experiment are

given as: rated power P ¼ 750 W, rated voltage U ¼ 200 V, number of poles np ¼ 4,
armature resistance Rs ¼ 1.74 , stator inductances Ld ¼ Lq ¼ 0.004 H, viscous
damping B ¼ 7.403 105 N  m  s/rad, moment of inertia J ¼ 1.74 104 kg  m2,
rated speed N ¼ 3000 rpm, rotor flux linkage f ¼ 0.1167 wb, rated torque TN ¼ 2.4 N  m.

4.1 Simulation results

The PMSM speed-regulation system under the control schemes SMC þ ESO and SMC
are simulated using Matlab/Simulink. The control gains of SMC and SMC þ ESO
methods are both selected as k ¼ 300. The pole of ESO is p ¼ 10,000. The PI
parameters of both current loops are the same, i.e. the proportional gains are kp ¼ 80,
the integral gains are ki ¼ 5000. The saturation limit of iq is 12 A. The results of the
simulation are shown in Figures 4 and 5.
From Figure 4, SMC and SMC þ ESO both show smaller overshoots and shorter
settling times. From Figure 5, it is obviously shown that the chattering of SMC
becomes larger when the control gain becomes larger. Meanwhile, the composite
controller, SMC þ ESO, with the smaller control gain can experience less speed
fluctuations while the disturbance rejection property is retained. The chattering is
reduced successfully.

4.2 Experimental results

An experimental test setup has been built for the speed control for a PMSM. The
configuration of the experimental test setup is shown in Figures 6 and 7. The whole
speed control algorithm includes the SVPWM, which is implemented by the
program of the DSP TMS320F2808 with a clock frequency of 100 MHz. The control
algorithm is implemented using C. The speed-loop and current-loop sampling

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

534 Speed-regulation problem for a PMSM servo system

(a) (b) 10
1000
i*q
8
iq
Speed (rpm)
800

Current (A)
6
600

400 4

200 SMC k = 300

2
SMC+ESO k = 300
0 0
0 0.005 0.01 0.015 0 0.005 0.01 0.015
t (s) t (s)

(c) 10
i*q
8
iq
Current (A)

0
0 0.005 0.01 0.015
t (s)

Figure 4 Step responses (simulation): (a) speed; (b) q-axis current

(SMC k ¼ 300); (c) q-axis current (SMCþESO k ¼ 300)

periods are 250 s and 60 s, respectively. The saturation limit of the q-axis
reference current is 11.78 A. The PMSM is driven by a three-phase PWM inverter
with an intelligent power module (IPM) with a switching frequency of 10 kHz. The
phase currents are measured by the Hall effect devices and are converted through
two 12-bit A/D converters. An incremental position encoder of 2500 lines is used
to measure the rotor speed and absolute rotor position.
We compare the system performances under the SMC and the SMC þ ESO schemes.
The control gains of both current loops are selected to be kp ¼ 42 and ki ¼ 2600.
The step responses are shown in Figure 8. It can be seen that the SMC and the
SMC þ ESO (with a smaller control gain k) schemes both show smaller overshoots and
shorter settling times.
According to theoretical analysis, the disturbance rejection performance of system
mainly depends on the switching gain k. So here we spend time to tune k from small to
large values. Here two indexes are introduced for comparison, i.e.
nmax  nmin
speed fluctuation index ¼ ,
2

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 535

(a) 1005 (b) 6

5 i*q
Load
1000 iq
Speed (rpm)

Current (A)
4

995 3

2
990
SMC k =300 1
SMC + ESO k = 300
985 0
0.058 0.06 0.062 0.064 0.066 0.058 0.059 0.06 0.061 0.062 0.063 0.064
t (s) t (s)
(c) 6
Load
5
Current (A)

2
i*q
1
iq
0
0.058 0.06 0.062 0.064
t (s)

Figure 5 Steady-state speed responses with sudden load

(simulation): (a) speed; (b) q-axis current (SMC k ¼ 300); (c) q-axis
current (SMCþESO k ¼ 300)

PMSM
AC Vdc
Incremental
Phase a current

Phase b current

position
encoder

Gate drive
Load

Host DSP
PC TMS320F2808 A/D

Figure 6 Configuration of the experimental system

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

536 Speed-regulation problem for a PMSM servo system

Figure 7 Experimental test setup

(b) 12
(a) 1000
i*q
10 iq
800
Speed (rpm)

8
Current (A)

600
6
400
4
200 SMC k = 800 2
SMC + ESO k = 500
0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
t (s) t (s)

(c) 12
i*q
10 iq
Current (A)

0
0 0.05 0.1 0.15 0.2 0.25 0.3
t (s)

Figure 8 Step responses (experiment): (a) speed; (b) q-axis current

(SMC k ¼ 800); (c) q-axis current (SMCþESO k ¼ 500)

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 537

(a) 1020 (b) 6

Load
Load i*q
5
1010 iq
Speed (rpm)

Current (A)
1000 3

2
990
1

980 0
0.5 1 1.5 0.5 1 1.5
t (s) t (s)

Figure 9 Response with a sudden load (SMC with k ¼ 650):

(a) speed response; (b) q-axis current response

(a) 1020 (b) 6

i*q
Load 5 Load iq
1010
Speed (rpm)

4
Current (A)

1000 3

2
990
1

980 0
0.5 1 1.5 0.5 1 1.5
t (s) t (s)

Figure 10 Response with a sudden load (SMC with k ¼ 800):

(a) speed response; (b) q-axis current response

where nmax, nmin are the maximum and minimum speed value in the steady-state
stage, and recovering time means the time required for the system speed output to
leave from and return to the steady state when a load torque is added or removed
suddenly.
Let us first observe the closed-loop system performance under the SMC method.
After the speed output of the system (without load) is in the steady state, a step load
disturbance torque, i.e. a rated torque of TN ¼ 2.4 N  m, is added suddenly. When k is
tuned to a small value, i.e. k ¼ 650, as shown in Figure 9, the motor speed can recover
to its reference value after the load disturbance is added. However, the speed decrease
amplitude is large, about 17.7 rpm. Usually it is expected that when the switching gain
k is increased, the disturbance rejection capacity of closed-loop system may become
better. The speed response in the case of step load torque is shown in Figure 10 when

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

538 Speed-regulation problem for a PMSM servo system

(a) 1020 (b) 6

Load Load i*q
5
1010 iq

Speed (rpm)
4

Current (A)
1000 3

2
990
1

980 0
0.5 1 1.5 0.5 1 1.5
t (s) t (s)

Figure 11 Response with a sudden load (SMC with k ¼ 900):

(a) speed response; (b) q-axis current response

(b) 6
(a) 1020 Load
Load i*q
5 iq
1010
4
Speed (rpm)

Current (A)

1000 3

2
990
1

980 0
0.5 1 1.5 0.5 1 1.5
t (s) t (s)

Figure 12 Response with a sudden load (SMCþESO with

k ¼ 500): (a) speed response; (b) q-axis current response

k is tuned to a larger value, i.e. k ¼ 800. Indeed, the closed-loop system has a smaller
speed decrease, about 12 rpm. However, it shows a larger chattering for steady state.
When k is tuned to a large value, i.e. k ¼ 900, as shown in Figure 11, the motor speed
can recover quickly in the case of load disturbance with an even smaller decrease,
about 9 rpm. However, it shows an even larger chattering. It seems that it is very
difficult to select have a good balance (control gain selection) between disturbance
rejection and chattering under the standard SMC method. In fact, this phenomenon of
the SMC method has already been mentioned in Chiang and Tseng (2004) for a
synchronous reluctance motor system.
Now let us use the SMC þ ESO control method. The control gain of SMC þ ESO is
selected as k ¼ 500, smaller than that of the SMC scheme. The pole of ESO is selected as
p ¼ 200. Figure 12 shows the speed responses when the PMSM system runs at a
steady state of 1000 rpm without load, then a rated load TL ¼ 2 N  m is added
suddenly. The speed decrease is about 11 rpm. The composite controller, SMC þ ESO,

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 539

Table 1 Speed response comparison

The rated load added suddenly Steady-state with load

Algorithm Decrease (rpm) Speed fluctuation (rpm)
SMC k ¼ 650 17.7 6.750
k ¼ 800 12 7.875
k ¼ 900 9 8.875
SMC þ ESO k ¼ 500 11 3.875

although using a smaller control gain, makes the closed-loop system with obviously
less steady-state fluctuations, while the disturbance rejection property is still kept.
A detailed comparison can also be found in Table 1, where the second column shows
the speed fluctuation performance calculated after recovering from the speed step
load disturbance.

5 Conclusion

The speed-regulation problem for a PMSM servo system has been addressed in this
paper. A second-order model has been built to describe the relationship between the
reference quadrature axis current and the speed output for a PMSM speed-regulation
system. A composite controller based on sliding mode feedback control and
disturbance compensation based on ESO has been developed. Compared with that
of a standard SMC scheme, the composite SMC method has obviously reduced the
chattering of the closed-loop PMSM system. Simulation and experiment results and
comparisons have been given to show the effectiveness of the proposed method.

Acknowledgement

This work was supported by the Natural Science Foundation of Jiangsu Province (Project
BK2008295).

References
Achour, A.Y., Mendil, B., Bacha, S. and integral sliding mode control technique.
Munteanu, I. 2009: Passivity-based current IEEE Transactions on Control Systems and
controller design for a permanent-magnet Technology 8, 47–54.
synchronous motor. ISA Transactions 48, Chen, W.H. 2004: Disturbance observer based
336–46. control for nonlinear systems. IEEE
Baik, I.C., Kim, K.-H. and Youn, M.J. 2000: Transactions on Mechatronics 9, 706–10.
Robust nonlinear speed control of PM syn- Chiang, H.K. and Tseng, C.H. 2004: Integral
chronous motor using boundary layer variable structure controller with grey

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

540 Speed-regulation problem for a PMSM servo system

prediction for synchronous reluctance motor technique. IEEE Transactions on Industrial

drive. IEE Proceedings—Control Theory and Electronics 49, 524–34.
Application 151, 349–58. Krishnan, R. 2001: Electric Motor Drives:
Feng, G., Huang, L.P. and Zhu, D.Q. 2004: Modeling, Analysis, and Control. Prentice-Hall.
A new robust algorithm to improve the Kung, Y.-S. and Tsai, M.-H. 2007: FPGA-based
dynamic performance on the speed control speed control IC for PMSM drive with
of induction motor drive. IEEE Transactions adaptive fuzzy control. IEEE Transactions on
on Power Electronics 19, 1614–27. Power Electronics 22, 2476–86.
Grcar, B., Cafuta, P., Znidaric, M. and Gausch, Lai, C.K. and Shyu, K.K. 2005: A novel motor
F. 1996: Nonlinear control of synchronous drive design for incremental motion system
servo drive. IEEE Transactions on Control via sliding-mode control method. IEEE
Systems and Technology 4, 177–84. Transactions on Industrial Electronics 52,
Guo, L. and Chen, W.H. 2005: Disturbance 499–507.
attenuation and rejection for a class of non- Li, S.H. and Liu, Z.G. 2009: Adaptive speed
linear systems via DOBC approach. control for permanent magnet synchronous
International Journal of Robust and Nonlinear motor system with variations of load inertia.
Control 15, 109–25. IEEE Transactions on Industrial Electronics 56,
Han, J.Q. 1995: The extended state observer of 3050–9.
a class of uncertain systems. Control and Li, S.H., Liu, H.X. and Ding, S.H. 2010: A
Decision 10, 85–8, (in Chinese). speed control for a PMSM using finite-time
Han, J.Q. 2009: From PID to active disturbance feedback control and disturbance compensa-
rejection control. IEEE Transactions on tion. Transactions of the Institute of
Industrial Electronics 56, 900–6. Measurement and Control 32, 170–87.
Ho, E.Y.Y. and Sen, P.C. 1991: Control Lin, F.J., Chiu, S.L. and Shyu, K.K. 1998:
dynamics of speed drive systems using Novel sliding mode controller for synchro-
sliding mode controllers with integral com- nous motor drive. IEEE Transactions on
pensation. IEEE Transactions on Industry Aerospace and Electronic Systems 34, 532–42.
Applications 27, 883–92. Lin, C.K., Liu, T.H. and Yang, S.H. 2008:
Hsien, T.-L., Sun, Y.-Y. and Tsai, M.-C. 1997: Nonlinear position controller design with
H1 control for a sensorless permanent- input–output linearisation technique for an
magnet synchronous drive. IEE interior permanent magnet synchronous
Proceedings—Electric Power Applications 144, motor control system. IET Power Electronics
173–81. 1, 14–26.
Hung, J.Y., Gao, W. and Hung, J.C. 1993: Miklosovic, R. and Gao, Z.Q. 2004: A robust
Variable structure control: a survey. IEEE two-degree-of-freedom control design tech-
Transactions on Industrial Electronics 40, 2–21. nique and its practical application. 39th IAS
Jan, R.-M., Tseng, C.-S. and Liu, R.-J. 2008: Annual Meeting on Industry Application
Robust PID control design for permanent Conference, 1495–502.
magnet synchronous motor: a genetic Mohamed, Y.A.-R.I. 2007: Design and imple-
approach. Electric Power Systems Research 78, mentation of a robust current-control scheme
1161–8. for a PMSM vector drive with a simple
Jin, H.Z. and Lee, J.M. 2009: An RMRAC adaptive disturbance observer. IEEE
current regulator for permanent-magnet syn- Transactions on Industrial Electronics 54,
chronous motor based on statistical model 1981–8.
interpretation. IEEE Transactions on Industrial Mohamed, Y.A.-R.I. and El-Saadany, E.F. 2008:
Electronics 56, 169–77. A current control scheme with an adaptive
Kim, K.H., Baik, I.C., Moon, G.-W. and Youn, internal model for torque ripple minimiza-
M.-J. 1999: A current control for a permanent tion and robust current regulation in PMSM
magnet synchronous motor with a simple drive systems. IEEE Transaction on Energy
disturbance estimation scheme. IEEE Conversion 23, 92–100.
Transactions on Control Systems and Nandam, P.K. and Sen, P.C. 1989: Sliding
Technology 7, 630–3. mode speed control of a self-controlled syn-
Kim, K.-H. and Youn, M.-J. 2002: A nonlinear chronous motor based on state estimation
speed control for a PM synchronous motor and digital differentiation. Proceedings of
using a simple disturbance estimation IPEC 1989 291–5.

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014

 Li et al. 541

Ohnishi, K. 1987: A new servo method in motor. IEEE Transactions on Industry

mechatronics. Transactions of the Japanese Applications 39, 408–15.
Society of Electrical Engineers 107-D, 83–6. Wai, R.J. 2001: Total sliding-mode controller
Pan, J.F., Cheung, N.C. and Yang, J.M. 2006: for PM synchronous servo motor drive using
Auto-disturbance rejection controller for recurrent fuzzy neural network. IEEE
novel planar switched reluctance motor. IEE Transactions on Industrial Electronics 48,
Proceedings—Electric Power Applications 153, 926–44.
307–15. Wang, G.J., Fong, C.T. and Chang, K.J. 2001:
Pillay, P. and Krishnan, R. 1989: Modeling, Neural-network-based self-tuning PI control-
simulation, and analysis of permanent- ler for precise motion control of PMAC
magnet motor drives. IEEE Transactions on motors. IEEE Transactions on Industrial
Industry Applications 25, 265–72. Electronics 48, 408–15.
Slotine, J.J.E. and Li, W. 1991: Applied Wei, X.J. and Guo, L. 2009: Composite distur-
Nonlinear Control. Prentice-Hall. bance-observer-based control and terminal
Su, J., Qiu, W., Ma, H. and Woo, P.-Y. 2004: sliding mode control for non-linear systems
Calibration-free robotic eye-hand coordina- with disturbances. International Journal of
tion based on an auto disturbance-rejection Control 82, 1082–1098.
controller. IEEE Transactions on Robotics 20, Wei, X.J., Zhang, H.F. and Guo, L. 2009a:
899–907. Composite disturbance-observer-based con-
Su, Y.X., Zheng, C.H. and Duan, B.Y. 2005: trol and variable structure control for non-
Automatic disturbances rejection controller linear systems with disturbances.
for precise motion control of permanent- Transactions of the Institute of Measurement
magnet synchronous motors. IEEE and Control 31, 401–23.
Transactions on Industrial Electronics 52, Wei, X.J., Zhang, H.F. and Guo, L. 2009b:
814–23. Composite disturbance-observer-based con-
Sun, B. and Gao, Z. 2005: A DSP-based active trol and terminal sliding mode control for
disturbance rejection control design for a uncertain structural systems. International
1-kW H-bridge DC–DC power converter. Journal of Systems Science 40, 1009–17.
IEEE Transactions on Industrial Electronics 52, Wu, D., Chen, K. and Wang, X. 2007: Tracking
1271–7. control and active disturbance rejection with
Utkin, V.I. 1977: Variable structure systems application to noncircular machining.
with sliding modes. IEEE Transactions on International Journal of Machine Tools and
Automatic Control 22, 212–22. Manufacture 47, 2207–17.
Utkin, V.I. 1993: Sliding mode control design Zhou, J. and Wang, Y. 2002: Adaptive back-
principles and applications to electric dri- stepping speed controller design for a per-
vers. IEEE Transactions on Electronic Systems manent magnet synchronous motor. IEE
14, 23–6. Proceedings—Electric Power Applications 149,
Vilathgamuwa, M., Rahman, M.A., Tseng, K. 165–72.
and Uddin, M.N. 2003: Nonlinear control of
interior permanent magnet synchronous

Downloaded from tim.sagepub.com at FRESNO PACIFIC UNIV on December 23, 2014