Real-Time Simulation and Control of Reluctance

Motor Drives for High Speed Operation with

Reduced Torque Ripple
1

Minh C. Ta, 2Christian Dufour

Hanoi University of Technology, 1 Dai Co Viet Road, Hanoi, Vietnam, e-mail: minhtc-auto@mail.hut.edu.vn
Opal-RT Technologies Inc., 1751 Richardson, Montreal, H3K 1G6 Canada, e-mail: christian.dufour@opal-rt.com

AbstractThis paper presents real-time simulation results

of a switched reluctance motor (SRM) drive with a novel
Torque Distribution Function (TDF) for high-speed
applications, in order to reduce torque ripple. The SRM is
fed by a three-phase unidirectional power converter having
three legs, each of which consist of two IGBTs and two freewheeling diodes. The SRM model incorporates all nonlinearities between excitation currents, rotor position and
flux linkages. For the purpose of control SRM drives, an
improvement of the TDF method is proposed for high-speed
applications, in order to reduce torque ripple. The real-time
simulation of the drive is conducted on the RT-LAB realtime simulation platform. Since the converter is currentcontrolled, simulator latency is critical to achieving good
accuracy and avoiding current overshoot. The paper
demonstrates that this type of drive with simple hysteretic
current control can be simulated in real-time at a time-step
of 15s, with good accuracy. The paper also introduces
FPGA-based simulation technology required to test
advanced algorithms like TDF.
KeywordsSwitched reluctance motor (SRM), Torque
Distribution Function (TDF), real-time simulation.

I. INTRODUCTION
Today, before using a motor controller with a real
motor drive, it is a common industrial engineering practice
to test a controller against a simulated motor model
running in real-time. This has several advantages. For
example, the simulated motor drive can be tested with
borderline conditions that would damage a real motor,
often a costly prototype. The motor itself may be under
development in parallel to the controller and therefore not
available for testing. While testing, a controller is
interfaced with the real-time simulated motor drive
through a set of proper I/Os: this approach is called
hardware-in-the-loop (HIL) simulation. Such motor drive
simulation is required by hybrid vehicle OEMs and other
power electronic motor drive manufacturers to speed up
development and testing time by using real-time
simulation before conducting tests on physical prototypes.
At the production stage, HIL simulation can also be used
to verify code integrity with automated correlation tests.
The switched reluctance motor drive is interesting
because of its low manufacturing costs, rugged
construction and simplicity of controller design. Since the
rotor has no windings and no magnets, it is a good
candidate for drive operation at high speed and in adverse
environments. Position sensor reliability poses a more
important problem in the latter case. A developer or

integrator of such drives will want to test the behavior of

the drive in case of a position sensor malfunction, for
example. A real-time simulator is the ideal tool to safely
test this scenario. Sensorless drive designs pose other
challenges related to sensor accuracy that could be
overcome through the use of a real-time simulator. The
real-time simulation of current-controlled drives poses
other challenges, including simulator latency that can
create a delay in hysteresis-type current controllers. This,
in turn, can cause current overshoots that are usually not
present in real drives.
Several SRM models that are ready for real-time
simulation are discussed in detail in references [6], [7],
[8]. In these models, a key aspect is the flux-linkage curve
approximations by polynomial functions or tables. The
switched reluctance machine has wound field coils of a
DC motor for stator windings and has no coils or magnets
on its rotor. A typical 6/4 pole SRM is shown in Fig. 1.
In an SRM, the rotor is aligned whenever
diametrically opposite stator poles are excited. At the
same time, when two rotor poles are aligned to two stator
poles, another set of stator poles is out of alignment with a
different set of stator poles. Continuous torque production
is therefore obtained by sequential excitation of the stator
phases using a unipolar power converter with 2 IGBTs
and 2 diodes for each leg.

Fig. 1. 6/4 pole Switched Reluctance Motor.

Well-known algorithms for controlling an SRM, such

as advanced rise and commutation angles, adaptive
control, fuzzy logic control, and neural network control
have been proposed in the literature. Torque Distribution
Function may be the most well-known method. There are
many kinds of TDFs; developed by Husain [3] (denoted as

TDF-I), Ilic-Spong [4] (denoted as TDF-II), and Krishnan

[1] (denoted as TDF-III).
In this paper, the authors present the topic of real-time
simulation of switched reluctance motor drives using
TDF-III, and also propose a novel Torque Distribution
Function, based on TDF-III, for high-speed model
applications. Next, a real-time simulation system for
switched reluctance motor drives is demonstrated. It is
implemented using a novel TDF for high-speed
applications, in order to reduce torque ripple.

Flux Linkages (Wb)

Rotor Position (Deg)

0.50

35
30
25
20
15
10
5
0

40 - 45

0.45
0.40
0.35
0.30
0.25
0.20

II. TORQUE PRODUCTION STRATEGY

In SRM control strategies, each phase winding is
excited separately in the rising slope of the inductance for
the positive torque, and in the falling slope for negative
torque. The overall torque of the motor is the sum of the
phase torques. The inductance profile, phase current and
voltage are shown in Fig. 2.

0.05
0

50

100

150 200 250 300

Excitation Current (A)

350

400

450

The phase flux linkage depends on both rotor position

and phase current. The model of an SRM can be described
as follows:

Aligned
phases

Vbus
Phase
Voltage

0.10

Fig. 3. Three-dimensional relationship between excitation current,

rotor position and flux linkages.

Active
control
region
T= (40-70)

Ideal
Phase
Inductance

0.15

Unaligned
phases

Tload
Vabc

-Vbus

Imax

Phase
Current

1/s

i=f

Te=fi

Te

1/(Js+B)

Imin
0

Rs

1/s

Fig. 4. Model of SRM used for simulation.

Fig. 2. Inductance profile, phase current and voltage.

A. Model of the SRM

In the SRM, there is a three-dimensional relationship
between excitation current, rotor position and flux
linkages [2]. Therefore, electromagnetic torque depends
on both excitation current and rotor position. Those
relationships are presented in Fig. 3. The machine under
study also has a stator winding resistance equal to 0.2
and a rotational inertia of 0.1 kg.m2.
This SRM model integrates its flux at each winding,
considered as uncoupled. Its equation is

(V

abc

RI abc )dt abc

(1)

where Iabc is the stator current inside the winding, R is

the stator resistance and Vabc is the voltage across the
stator windings. The current is then derived from the
machine flux characteristics abc ( I , ) and stored in
tables.

The produced electrical torque is also derived from the

flux-current characteristic, pre-computed and stored in
tables prior to the real-time simulation in a normalized
form (dK/d).

( I , )dI

K ( , i) 2Te ( I , )

i2

Te ( I , )

B. Torque Distribution Functions

In many machines, such as ac and dc machines, the airgap torque is directly proportional to the excitation current
(in a DC machine) or the transformed current variable (qaxis current in an AC machine). However, in the case of
an SRM drive system, a three-dimensional non-linear
relationship exists between the flux linkages, excitation
current and rotor position. This means that the air-gap
torque depends on both excitation current and rotor
position. Therefore, in the SRM drive system, it is
necessary to add the torque control to the control loop.
The control configuration for an SRM drive system using
TDF is shown in Fig. 5. The extraction of the excitation
current reference from air-gap torque reference is made
using a three-dimensional relationship. The equation for
extracting the current reference is shown as follows:

i* =

2Te*
K(,i)

(2)

Fig. 5. SRM drive control system diagram

C. Torque Distribution Function Methods

The method of torque control depends on how many
phases are excited at a given time. If, in an SRM drive
system, only one phase is excited at a given time, the
torque ripple produced is very large. So, it is necessary to
excite more than one phase at a given time to reduce
torque ripple. In this paper, the TDF method is used.
When using the TDF method, it is necessary to make
certain assumptions: (1) Neglect the mutual coupling
effect between phases, (2) The current controller is
idealized (without tracking error). Working based on these
assumptions, the torque command will be divided into two
parts, one for the incoming phase (called phase x for
short) and another for outgoing phase (phase y). This
division makes the current commands vary more slowly
so that the actual ones can keep up.
There are many ways to get Tx* and Ty* in the
commutation interval, depending on the inductance profile
of a given machine [1], [3], [4]. The TDF III that was
developed by Krishnan [1] is shown in the following
equations:

gx =

K(,i x )

gy =

In reference [1], Krishnan has also proposed the

method TDF IV, which takes the mutual coupling into
consideration and the TDF V, which includes both
saturation and mutual coupling. The practical
implementation of methods TDF IV and TDF V is
however very difficult, due to the fact that the mutual
coupling between the phases is not easily determined.
Therefore, we prefer to consider only the method TDF III,
and try to improve its drawback (Section III).
D. Simulation Results
The performance of the SRM drive system was
evaluated by using Torque Distribution Function, TDF-III.
The torque command was set to 100 Nm and the motors
speed was set to 1000 rpm and 3000 rpm. In commutation
intervals, the torque command was divided into two parts
so that two phases could reduce torque ripple. Simulation
results illustrate that at 1000 rpm, the torque ripple is very
small. However, at a speed of 3000 rpm, torque ripple is
very large.

K(,i y )

(3)

Te* =Tx* +Ty*

Tx* =

g 2x
g +g 2y
2
x

Ty* =

2
y

2
x

g +g 2y

where g x ,g y are pre-calculated and stored in a table and

extracted by linear interpolation method. The torque
commands during a commutation interval are illustrated in
Fig. 6. In this method (TDF III), the saturation and mutual
coupling between the phases are neglected.

Fig. 6. Torque commands in commutation interval.

TABLE I. EQUATIONS OF PROPOSED TDF

Angle

0-150

150 300

Ta*

g a2
T*
2
2 e
ga gc

Te* (Tc* 12 gcic2 )

Tb*

Tc*

Angle

Ta*

III. PROPOSED TORQUE DISTRIBUTION FUNCTION

Because of the non-idealized current controller, the
actual currents need a specified time to track their
commands. As a result, actual phase torques cannot track
commands that produce torque ripple, especially in highspeed applications; in other words, the higher the speed of
the motor, the larger the torque ripple that will be
produced. For this reason, a new TDF is proposed to
reduce torque ripple, especially in SRM drive high-speed
applications.
In the proposed Torque Distribution Function [14], the
torque command is actively compensated for in each
phase, resulting in a flat total command torque. The
*
following equations describe this method. In the table, Ta
*

, Tb , Tc are the phase torque commands, and

torque command.

300 450

g a2
T * (Tb* 12 gbib2 )
2
2 e
g a gb

450 - 600
0

Tb*

gb2
Te*
g a2 gb2

Te* (Ta* 12 g aia2 )

Tc*

Angle

600 - 750
0

750 - 900

Tb*

g b2
Te* (Tc* 12 g cic2 )
g b2 g c2

Tc*

g c2
T*
2
2 e
gb g c

Te* (Tb* 12 gbib2 )

Ta*

Fig. 7. Phase current, phase torques and total torque at 1000rpm

(upper) and 3000 rpm (lower) and torque reference of 100Nm (red line
real torque and green line command torque).

g c2
T * (Ta* 12 g aia2 )
2
2 e
g a gc

As illustrated in the simulation results, the torque

command is not flat (green curve). However, the actual
torque (red curve) is quite flat, thereby significantly
reducing torque ripple.

Te* is the
Fig. 8. Phase current, phase torques and total torque at speed of 3000
rpm and load torque of 100 Nm (red line real torque and green line
command torque).

IV. REAL-TIME SIMULATION OF SRM DRIVES

A. RT-LAB real-time simulation platform
RT-LAB, from Opal-RT Technologies, is a real-time
simulation platform that enables HIL simulation of
controllers, electric plants or both, through automatic code
generation methods. The entire process occurs without the
need for handwritten C code, enabling very rapid
deployment of prototyped controllers or HIL-simulated
plants. The process is notably very efficient when applied
to I/O code because RT-LAB provides a set of Simulink
blocks that automatically configure common I/O
functions, like analog inputs/outputs and time-stamping
capable digital I/Os, with a 10 nanosecond resolution.
Special interpolating models use this timing information
to greatly increase simulation accuracy [10]. RT-LAB
simulators can also be equipped with a userprogrammable FPGA card. The FPGA card can be
programmed with the Xilinx System Generator blockset
for Simulink enabling implementation of complex sensor
models like resolvers, Resolver-To-Digital and FM
resolvers or even complex motor drives [11], [12].

leave margin available for a more complex model. In this

test, however, the sensors are idealized, and only the
numerical value of the angle is transmitted between cores.

TABLE II. RT-LAB SIMULATOR CHARACTERISTICS

Hardware

PC with dual Quad-Core Intel

processor (2.3 GHz)
QNX or RedHawk Linux

Real-time operating
system
RT-LAB version
Digital I/O
Analog I/O

8.2.5
16 Din-16 Dout (Time Stamped)
16 Analog Inputs and 16 Outputs

B. Test case: constant turn-on and turn-off angles and

10A dead band.
The SRM drive discussed in this paper is simulated in
real-time on the RT-LAB real-time simulation platform,
using 2 cores of a dual quad-core PC. Table II summarizes
the characteristics of the RT-LAB system used in this
paper. A very basic type of SRM control with fixed turnon and turn-off angles will be used as an example.
CPU 1: (Ts=15 us)

Drive

a a'

b b'

Vdc

Switched Reluctance Motor

c'
b'
a

c c'
a'
b
c

IGBT pulses

CPU 2: (Ts= 15 us)

FPU

motor

i motor

Motor controller

Fig. 9. Real-time simulation of an SRM drive: task separation.

The task separation for the experiment is shown in Fig.

9. A sample time of 13 s was reached with an I/O
interface consisting of 6 digital inputs for the IGBT gates,
3 digital outputs (used for quadrature encoder emulation)
and 5 Analog Outputs for motor currents and resolver
signals. The final simulation time of 15 s was chosen to

Fig. 10. Simulation runs at 1 us and 15 us (with zoom in the lower).

The most important difference is the current ripple

amplitude during current regulation. The 15 s simulation
run exhibits higher ripple (as much as 2 times the nominal
10A gap) because of the increased delay in the current
loop. In an SRM, this component of the torque ripple is a
magnitude lower than the overall torque ripple and may
therefore minimally affect the overall testing purpose of
the real-time simulator.
As previously mentioned, the SRM drive was simulated
on an RT-LAB system with Hardware-In-the-Loop
capability. This section shows the model signals obtained
during real-time simulation. A 100 N.m torque is applied
so that the SRM runs in current regulation mode in steadystate.

multiplication by inductance inverse to compute machine

currents. However, in the case of the SRM, the fluxcurrent characteristic is two-dimensional, while in [11] a
3-D table was necessary to store inductance values as a
function of current amplitude and sector, as well as rotor
position. Consequently, the SRM implementation on an
FPGA should be easier. Simulation results similar to Fig.
11. (1 microsecond curves), but with minimal overshoot,
are expected.
The RT-LAB system comes with a user programmable
FPGA card based on the Virtex II-Pro (11K cells, 44
multipliers, PCI interface). An improved FPGA card
based on the SPARTAN-3 chip from Xilinx has recently
become available (77K cells 104 multipliers, PCIe
interface). The support for the SPARTAN-3 chip provides
an undeniable boost to the capability of the system. Logic
cells, the standard measure of FPGA resource availability,
are increased 7 fold with the SPARTAN-3. A more subtle
advantage, the 2.5 times increase in hardware multiplier,
greatly increases FPGA drive design possibilities, since
multiplication operations are very common in drive
models. The increased I/O number also enables the
possible connection of several drives implemented on the
chip. Finally, RT-LAB can also be interfaced with a
Virtex-5 board with 288 multiplier units and PCIe
interface.

Simulink
Subsystems
Rate=10-100 s

RTW

RT-LAB

Fig. 11. Motor currents: all 3 phases (upper); phase A with gate signals
(lower).

Fig. 11. shows the currents for the 3 phases of the

motor, with an analog output scaling of 1/50 V/A. The
lower oscilloscope grab of Fig. 11 shows the current of
phase A, along with the corresponding IGBT gate signals.
The correspondence between the IGBT gate level and
current slope sign is clearly apparent in this figure.
V. REAL-TIME SIMULATION OF AN SRM WITH FPGAS AS
COMPUTATIONAL ENGINE.
Real-time simulation of SRM drives on a CPU-based
real-time simulator can produce accurate results, but can
also have the undesirable effect of causing current
overshoots because of model latency. In this case, this
latency has prevented the testing of the proposed TDF
method in real-time. To remedy this problem, an FPGA
implementation is desirable because it offers a very low
calculation and I/O latency. FPGA implementations of a
standard Park-model PMSM drive [6] and a FiniteElement Analysis (FEA) based PMSM drive [7] have
been demonstrated, and have a total latency just above 1
microsecond, including I/Os. The SRM model has a
calculation structure very similar to an FEA-based
PMSM, as follows: flux integration, followed by the

Xilinx System Generator

Simulink Subsystems
Rate= 10 n s

Xilinx/ ISE

PCIe

DIO
AIO

Single/dual multi-core CPU PC

Simulink
Model

Code
Generation

Distributed
Real-Time
Target

FPGA card with embedded IO

Fig. 12. Design flow in RT-LAB

The interface to the CPU-side of the simulator is also

very important. A typical RT-LAB application
(automotive applications, for example) will have several
motor drives interacting electrically, as well as through a
mechanical system. The latter, for example, is typically
complex and relatively slow, and therefore best
implemented using Simulink on a regular CPU, using a
standard code generation technique. Consequently, a fast
data exchange mechanism is required to conduct mixedtype simulation (CPU-FPGA). The PCI Express (PCIe)
support available on the SPARTAN-3 design greatly
increases the data exchange rate between CPU and FPGA.
Finally, the I/O access time is minimal with FPGA
implementations because the FPGA card is hardwired
with the Digital I/Os and Analog converters.
The most important objective is to make the whole
design process easy and straightforward. With RT-LAB,

this process follows the Model-Based Design paradigm in

which the user develops specifications, and conducts
model design and test simply by interacting with the highlevel model, usually designed in Simulink. When FPGA
design is involved, the same process flow, described in
Fig. 12, is maintained except that Xilinx ISE FPGA code
generation tools are involved in the process.
VI. CONCLUSION
The paper has presented modeling, simulation and
control SRM using Torque Distribution Function,
developed by R.Krishnan, to reduce torque ripple. In this
paper, the authors have also proposed a novel Torque
Distribution Function to reduce torque ripple at high
speed. The simulation results have confirmed excellent
response of the torque by novel TDF. This paper has also
demonstrated the feasibility of real-time simulation of a
SRM drive, based on the RT-LAB simulator. The realtime simulation was conducted on 2 cores of an 8-core
standard PC at a sample time of 13 s, including digital
and analog I/O access time. Acceptable results have been
produced, despite some overshoot when the current
converter hysteresis gap is not too small.
However, for a very demanding control algorithm,
such as Torque Distribution Function method, the
standard CPU-based approach presents input-output
latency issues at the I/O level that will prevent an
accurate simulation from being performed. In this case,
an implementation of the SRM drive on the FPGA should
solve the problem. This FPGA-implementation was not
presented in the paper. It is believed that the FPGA
implementation should not be problematic because a
machine of similar complexity, a FPGA-based PMSM
with FEA-based inductance profile, has already been
designed using RT-LAB.
REFERENCE
[1]

R. Krishnan, Switched Reluctance Motor Drives, CRC Press LLC,

2001.
[2] T.J.E. Miller, Switched Reluctance Motors and Their Control,
Magna Physics, Oxford, 1992.
[3] Husain, I. and M. Ehsani, Torque ripple Minimization in
Switched Reluctance Motor Drives by PWM Control, IEEE
Trans. on Power Electronics, Vol. 11, No. 1, 1996, pp. 83-88.
[4] M. Ilic-Spong, T. J. E. Miller, S. R. MacMinn, and J. S. Thorp,
Instantaneous torque control of electric motor drives, IEEE
Trans. on Power Electronics, vol. 2, no. 1, 1987, pp. 5561.
[5] B.K Bose, Power Electronics and Variable Frequency Drives,
IEEE Press, 1997.
[6] Le-Huy, H.; Brunelle, P. A versatile nonlinear switched
reluctance motor model in Simulink using realistic and analytical
magnetization characteristics, Proceedings of the 31st IEEE
Annual Conference of the Industrial Electronics Society 2005
(IECON-2005), 6-10 Nov. 2005
[7] I. Boldea, S.A. Nasar, Electric Drives, CRC Press, ISBN 0-84932521-8
[8] T.J.E. Miller, et al., Ultra-fast model of the switched reluctance
motor, IEEE Industry Applications Conference Proceedings, Vol.
1, 1998, pp. 319-326.
[9] D.A. Torrey, X.M. Niu, E.J. Unkauf, Analytical modeling of
variable-reluctance machine magnetisation characteristics, IEE
Proceedings - Electric Power Applications, Vol. 142, No. 1,
January 1995, pp. 14-22.
[10] M. Harakawa, H. Yamasaki, T. Nagano, S. Abourida, C. Dufour
and J. Blanger, Real-Time Simulation of a Complete PMSM
Drive at 10 us Time Step, Proceedings of the 2005 International

[11]

[12]

[13]
[14]

Power Electronics Conference (IPEC 2005), April 4-8, 2005,

Niigata, Japan.
C.Dufour, J. Blanger, V. Lapointe, FPGA-Based Ultra-Low
Latency HIL Fault Testing of a Permanent Magnet Motor Drive
using RT-LAB-XSG, Simulation: Transactions of the Society for
Modeling and Simulation International, SAGE Publications, Vol.
84, Issue 2/3, February/March 2008, pp. 161-172.
C. Dufour, J. Blanger, S. Abourida, V. Lapointe, FPGA-Based
Real-Time Simulation of Finite-Element Analysis Permanent
Magnet Synchronous Machine Drives, Proceedings of the 38th
Annual IEEE Power Electronics Specialists Conference (PESC
07), Orlando, Florida, USA, June 17-21, 2007.
M. Ehsani, Y. Gao, S. E. Gay, A. Emadi, Modern Electric, Hybrid
Electric and Fuel Cell Vehicles, CRC Press, ISBN 0-8493-3154-4
L. V. Do and M. C. Ta, Modeling, Simulation and Control of
Reluctance Motor Drives for High-Speed Operations,
Proceedings of the IEEE Energy Conversion Congress and
Exposition (ECCE2009), San Jose, USA, September 20-24, 2009.