~

274 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 2 5 , NO. 2, MARCHIAPRIL 1989

Modeling, Simulation, and Analysis of

Permanent-Magnet Motor Drives,
Part 11: The Brushless DC
Motor Drive

Abstract-The brushless dc motor has a permanent-magnetrotor, and While this approach has already been proposed, the back EMF
the stator windings are wound such that the back electromotive force is not represented as a Fourier series as is done in [4].Instead,
(EMF) is trapezoidal. It therefore requires rectangular-shaped stator the back EMF is generated according to the position of the
phase currents to produce constant torque. The trapezoidal back EMF
implies that the mutual inductance between the stator and rotor is rotor using piecewise linear curves. This technique avoids the
nonsinusoidal. Theref ore, no particular advantage exists in transformingso-called Gibbs phenomenon that occurs due to the truncation
the machine equations into the well-known two-axis equations, which is of the higher order harmonics necessary when using the
done in the case of machines with sinusoidal back EMF’S. This second Fourier series approach.
part of the two-part paper develops a phase variable model of the BDCM Using this model of the BDCM, a detailed simulation and
and uses it to examine the performance of a BDCM speed servo drive
system when fed by hysteresis and pulsewidth-modulated (PWM) current
analysis of a BDCM speed servo drive is given. The
controllers. Transients similar to those applied to the permanent-magnet simulation includes the state variable model of the motor and
synchronous motor system of Part I are applied to this drive system to speed controller and a real-time model of the inverter
allow a comparative evaluation. Particular attention is paid to the motorswitches. Although the switches are assumed to be ideal
torque pulsations. Some experimental verification is given. devices, the software developed is flexible enough to incorpo-
rate their turn-on and turn-off times. Every instance of a
I. INTRODUCTION
power switch opening or closing is simulated to determine the

T HE ac servo has established itself as a serious competitor current oscillations and consequent torque pulsations. The
to the brush-type dc servo for industrial applications. In effects of the hysteresis window size on the motor torque
the fractional-to-30-hp range, the available ac servos include pulsations is investigated, and the effects of hysteresis and
the induction, permanent-magnet synchronous, and brushless PWM current controllers on the drive system performance are
dc motors (BDCM) [l]. The BDCM has a trapezoidal back also examined. Similar transients that were applied to the
EMF, and rectangular stator currents are needed to produce a permanent-magnet synchronous motor (PMSM) drive in Part I
constant electric torque, as shown in Fig. 1. Typically, [7] are applied here for comparative evaluation. In addition,
hysteresis or pulsewidth-modulated (PWM) current control- both the small and large signal performances are investigated.
lers are used to maintain the actual currents flowing into the Experimental verification is provided.
motor as close as possible to the rectangular reference values. The paper is organized as follows. The mathematical model
Although some steady-state analysis has been done [2], [3], of the BDCM is developed in Section 11. The operation of the
the modeling, detailed simulation, and experimental verifica- hysteresis and PWM current controllers and the structure of
tion of this servo drive has been neglected in the literature. the drive system are presented in Section 111. The results and
The purpose of this paper is to fill this void. conclusions are in Sections IV and V, respectively.
It is shown that, because of the trapezoidal back EMF and
the consequent nonsinusoidal variation of the motor induc- 11. MATHEMATICAL MODELOF THE BRUSHLESS DC MOTOR
tances with rotor angle, a transformation of the machine The BDCM has three stator windings and permanent
equations to the well-known d, q model is not necessarily the magnets on the rotor. Since both the magnet and the stainless-
best approach for modeling and simulation. Instead, the steel retaining sleeves have high resistivity, rotor-induced
natural or phase variable approach offers many advantages. currents can be neglected and no damper windings are
modeled. Hence the circuit equations of the three windings in
Paper IPCSD 88-23, approved by the Industrial Drives Committee of the phase variables are
IEEE Industry Applications Society for presentation at the 1987 Industry
Applications Society Annual Meeting, Atlanta, GA, October 19-23. Manu- R O O
script released for publication July 12, 1988.
P. Pillay was with the Electrical Engineering Department, Virginia
Polytechnic Institute and State University, Blacksburg, VA. He is now with O O R
the Department of Electrical and Electronic Engineering, The University of
Newcastle-upon-Tyne, Merz Court, Newcastle-upon-Tyne, England NE1
7RU.
R. Krishnan is with the Electrical Engineering Department, Virginia
Polytechnic Institute and State University, Blacksburg, VA 24061.
IEEE Log Number 8825303.

0093-9994/89/0300-0274$01.OO 0 1989 IEEE

PILLAY AND KRISHNAN: PERMANENT-MAGNET MOTOR DRIVES, PART 11

where it has been assumed that the stator resistances of all the
windings are equal. (Symbols are defined in the Nomenclature
at the end of the paper.) The back EMF's e,, eb, and e, have
trapezoidal shapes as shown in Fig. 1 . Assuming further that
there is no change in the rotor reluctances with angle, then

Hence

but

Therefore.

Hence

P
+[

1,
O O R

R O O

O O R
L-M
0

*
0
0
La = Lb = L,= L
Lob = Lra = Lcb = Ma

L -0M

and the electromagnetic torque is

M

0
L

i, + ib + i, = 0.

Mib + Mi, = - Mi,.

L -0M ] P [ ; ] + [ $ ] .

In state-space form the equations are arranged as follows:

[;I1-[ 1 / ( L- M )

O O R
0
1 / ( L- M )

[[;I-[: : :I[;]-[$]]
0 1 / ( L- M )
(3)

(5)

(6)
- n
)h-$' 150'
n

The power circuit and switching logic of the hysteresis

current controller that drives the BDCM is exactly the same as
that for the PMSM drive in Part I [7], and the reader is
275

Fig. 1. Back EMF and current waveforms of brushless dc motor.

soidal, hence transformation to a d, q reference frame cannot

be easily accomplished. A possibility is to find a Fourier series
of the back EMF, in which case the back EMF in the d , q
(4) reference frame would also consist of many terms. This is
considered too cumbersome, hence the a, 6 , c phase variable
model already developed will be used without further transfor-
mation.
111. CURRENT
CONTROLLERS
AND DRIVE
SYSTEM

referred to [7, section IV] for the details. The only difference
is in the shape of the reference current, which is sinusoidal for
the PMSM but rectangular for the BDCM, as shown in Fig. 2.
Because of the nonzero inductance of the stator phase
windings, the actual phase currents are unable to assume the
desired rectangular form. Instead, the currents are trapezoidal
due to the finite rise time. This has consequences on the torque
production and the drive performance. This will be elaborated
upon in Section IV.
A second method used to generate the required stator
currents is to use a PWM current controller. The logic for this
is exactly the same as that for the PMSM drive, and the reader
T, = (e,i, + ebib+ ecic)/ur. (7) is referred to [7, section VI for the relevant description.
The structure of the RDCM drive system is somewhat
The equation of motion is similar to that of the PMSM system in [7, fig. 51 with some
differences that have been explained in [5]. The BDCM would
probably not be used for extended speed operation because of
The currents i,, ib, and i, needed to produce a steady torque its limited flux weakening capabilities, and hence the blocks
without torque pulsations are shown in Fig. 1. With ac associated with F W in [7, fig. 51 would not be necessary. In
machines that have sinusoidal back EMF's, a transformation addition, since d , q modeling is not used in the BDCM drive,
can be made from the phase variables to d , q coordinates the torque reference divided by the torque constant Kt would
either in the stationary, rotor, or synchronously rotating give a reference stator current i: instead of the i; shown in
reference frames. Inductances that vary sinusoidally in the a, [7]. The rest of the drive system is essentially the same except
b, c frame become constants in the d , q reference frame. The that a resolver is not absolutely necessary for a BDCM speed
back EMF being nonsinusoidal in the BDCM means that the servo. Hall-effect position sensors located every 60" (electri-
mutual inductance between the stator and rotor is nonsinu- cal) would suffice.
276 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 25, NO. 2, MARCHiAPRlL 1989

1 actual
-
0
p?
I
SPEED
I I

iref U L s t e r e s i s Band

Fig. 2. Hysteresis current controller. U) TORQUE

" k x REF'
IV. RESULTS
Digital computer simulations of the entire BDCM drive
system are presented in this section. The state space model of
the BDCM and speed controller and switching logic of the
current controllers are included in the simulation. Every
instance of a power device switching on or off is modeled.
Although the turn-on and turn-off times of the power devices
are neglected, the simulation program developed is flexible
enough to incorporate this. A proportional-integral (P-I)
speed controller is used. Large and small signal transients are
considered, and in addition, a comparison between the PWM
and hysteresis current controllers on the drive performance is
made. The BDCM drive is run through transients similar to - VOLT RGE
those of the PWSM drive in Part I [7] so that a comparative
evaluation can be made.
Fig. 3 shows key results when a BDCM (see Table I for the
parameters) is started up from standstill to a speed of 1250 r/
min using a PWM current controller. The speed is slightly
underdamped in the design used here and increases in a linear
manner during the initial startup period when the torque is 1 1
0.00 ' o'.ozs ' d. 05
constant and equal to the maximum capability of the motor. TIME(S1
This ensures that the machine runs up in the shortest time Fig. 3. Transients when fed from P W M current-source inverter (CSI).
possible. Every time the stator current is commutated from TABLE I
one phase to another, a torque pulsation is generated. This MOTOR PARAMETERS

torque pulsation may be troublesome at low speeds as it can

R 0.29 fl
affect the accuracy and repeatability of position servo per- L - A4 0.365 mH
formance. The magnitude of the pulsation depends on the J 0.0002265 kgm2
operating current level. At 2-pu current, the pulsation is about eJw, 0.185V/rad/s
twice that at 1-pu current. When the operating current is zero,
so is the commutation induced torque pulsation. In addition to close tracking of the commanded current by the actual is
these pulsations, high-frequency torque pulsations also occur evident except for the initial rise time due to the stator time
due to the current oscillations produced by the PWM current constant. During the 60" period when the phase does not
controller. These are of sufficiently high frequency that they conduct current, no voltage is applied and the voltage visible
are effectively filtered out by the rotor inertia. These high- across the motor terminals is the back EMF as shown in Fig.
frequency pulsations depend only on the PWM switching 3. This is unlike the PMSM drive results presented in Part I [7],
frequency, unlike the commutation induced pulsations which where the voltage in any one phase is applied continuously to
depend on the operating speed. The higher the number of poles generate the sinusoidal currents needed for constant torque in
of the machine, the higher is the frequency of the commutation that drive system.
induced pulsations for a given operating speed. This is an At 0.03 s, a load of 1 pu is applied to the motor. This causes
advantage for speed servo performance since the higher the a small decrease in the speed as shown in Fig. 3. This decrease
frequency of the pulsation, the lower is the effect on the speed is barely perceptible and is less than the speed overshoot that
because of the low-pass filtering effect of the rotor. However, occurred during the startup. The motor electric torque
the number of pulsations increases with the number of poles, increases to 1 pu to satisfy the load torque requirements.
and it is this factor that can be critical for position servo Fig. 4 shows the corresponding curves when a hysteresis
performance. Care should therefore be taken in using this instead of a PWM current controller is used. Clearly, the large
machine for position servos [6]. signal speed transient is the same as when the PWM current
The phase voltage switches continuously in an effort to controller is used. The voltage is used to force the currents to
force the actual current to equal the commanded value. The remain within the hysteresis bands. Although the oscillations
PILLAY AND KRISHNAN: PERMANENT-MAGNET MOTOR DRIVES. PART 11 277

0
SPEED -
p? SPEED
-
I I 1

/ = = - -
/ x REF

L? TORQUE
N
X REF
I I ? TORQUE
I I
X REF
-8 0 RCTURL
3 I

-
a-
z
F9.F I I
9-1
L? CURRENT

- VOLTRGE
Fig. 5 . Transients when fed from hysteresis CSI

L? TORQUE
I I

x REF

0.025’ 0.05
TIME S I
Fig. 4. Transients when fed from hysteresis CSI.
-.
d
I
‘
I
CUR~ENT
I
’
I
in the current, and consequently torque, are slightly larger with U!
NI I I I 1
the hysteresis current controller, the average value of the
torque is the same with both current controllers, thus

i?K
producing the same large signal dynamics. By increasing the
magnitude of the hysteresis bands, the resulting oscillations in
current and subsequent torque pulsations are increased, as 9 =. RCTURL 0.025
shown in Fig. 5. This also results in a reduction in the
0.00
switching frequency of the inverter. TIME( S I
It is therefore clear that the speed transient is similar Fig. 6. Transients when fed from PWM CSI.
irrespective of whether a PWM or hysteresis current controller
is used. However, if the hysteresis bands are so large as to the speed command is input, a pulse of torque is demanded to
produce large magnitude and low-frequency torque pulsations, increase the actual speed of the motor. This is provided by an
then significant speed pulsations would occur. increase in the rectangular current as shown in Fig. 7. In the
The torque and current response, when a load torque of 0.1 PMSM drive, a pulse of sinusoidal current is demanded for the
pu is applied are shown in Fig. 6. These are scaled-down same purpose. Up to now transient results have been pre-
versions of the curves corresponding to the load torque of 1 sented. Steady-state results are presented next.
pu. This indicates that, provided the phase current is input From the results presented earlier, the motor torque
according to the timing strategy in Fig. 1, the transfer function pulsations clearly increase as a function of the hysteresis
between the electric torque and current is linear and the small window size. This trend is plotted in Fig. 8 in pu. The
and large signal responses are similar. If this timing is lost, relationship between the magnitude of the motor torque
then the transfer function becomes nonlinear and the above pulsations and the hysteresis window size is nonlinear, unlike
would no longer be true. In the. case of the PMSM drive in Part the PMSM drive [ 7 ] , where this relationship is linear. This
I [ 7 ] , this transfer function is linear only under vector control. result was obtained by varying the hysteresis window size and
Fig. 7 shows the speed, torque, and current for a 0.1-pu determining the corresponding torque pulsations.
increase in the speed of the machine after it has run up. When From the previous results, it is also clear that the magnitude
278 IEEE TRAN5;ACTIONS ON INDUSTRY APPLICATIONS. VOL. 25. NO. 2, MARCHIAPFUL 1989

m
0.4
0
U

u
2 0.2
3

U! TORQUE a
NI I I 1 I

1 2
Current (pu)
Fig. 9. Commutation torque pulsation versus current

LD CURRENT
1 I I

E?
E?N7 0 RCT UFlL
RCTUFlL
I I I
'0.m 0'.026
0.026 0.05
TIME( S 1
Fig. 7. Transients for 0.1 increase in speed.

Fig. 10. Measured current of BDCM. X axis: 1 ms/div. Y axis: 2.5 A/div.

I
0.1 0.2
Lnl
'0.381
I I
0.4405 '
I I
0.5
Window size (pu) TIME[S) lo-'
N
Fig. 8. Torque pulsations versus window size. Fig. 11. Predicted current of BDCM.

of the commutation-induced torque pulsation depends on the Experimental Vergication

operating current level. In other words, the commutation- Fig. 10 shows the current waveform of the BDCM, thus
induced torque ripple depends on the current being commu- verifying the rectangular shape. The prediction is in Fig. 1 1 .
tated. This relationship is plotted in Fig. 9. The torque The 120" conducting and 60" nonconducting periods can also
pulsation depends linearly on the current being commutated. be seen. To test the model developed, the machine was started
Decreasing the hysteresis window size increases the operat- up on an inertia load and its speed measured as shown in Fig.
ing frequency of the inverter. This trend depends on the motor 12. The theoretical prediction of this speed is given in Fig. 13.
parameters and is somewhat similar to that presented in [7] and The theoretical predictions and the practical measurements
will therefore not be presented here. compare favorably, thus verifying that the model and com-
From Fig. 3 it can be seen that the torque pulsations due to puter simulation program used in this investigation is valid for
the current controller is extremely small, and its effect on the the machine used.
speed is not even noticeable. Just as in the PMSM drive in Part
I, changing the PWM switching frequency does not affect the V. CONCLUSION
torque pulsations as much as varying the window size in the This part of the two-part paper has presented the modeling,
hysteresis current controller. Hence the PWM switching simulation, and analysis of a BDCM drive. Particular attention
frequency should be chosen on the basis of the torque was paid to the motor large- and small-signal dynamics and
bandwidth and inverter switching capability rather than on the motor torque pulsations. The simulation included the state space
resulting torque pulsations, which is the same result as for the model of the motor and speed controller and real-time
PMSM drive. model of the inverter switches. Every instance of a power
PILLAY AND KRISHNAN: PERMANENT-MAGNET MOTOR DRIVES, PART I1 279

The frequency of the commutation-induced torque pulsa-

tions increase as the number of poles of the machine is in-
creased, thus reducing their effects on the speed. A high pole
number is therefore advantageous in a speed servo. However,
since the number of pulsations increase with an increase in the
number of poles for a given mechanical rotation, a very high
pole number may be undesirable for position servo perform-
ance.
NOMENCLATURE
e,, eb, e, a, b, and c phase back-EMF’S. V
EP Peak value of back-EMF, V.
Kt = 2e0/or,torque constant.
L a , Lb, L, Self-inductance of a, 6 , and c phases, H.
Fig 12 Measured speed of BDCM. X axis: 10 ms/div. Y axis: 214 r/min/
div . Lob Mutual inductance between phase a and 6 , H.
ua, V b , U, a, b, and c phase voltages, V.
Other symbols used are listed in [ 7 ] .
ACKNOWLEDGMENT
The authors would like to thank Inland Motor, Specialty
Products Division, Kollmorgan Corporation, of Radford
6 FlCTUFlL Virginia for the loan of one of their BDCM drives.
‘e
a
I
0
I

d.05 0.I REFERENCES

TIME(S1
Fig. 13. Predicted speed of BDCM. 111 R. Krishnan, “Selection criteria for servo motor drives,” in Proc.
IEEE IAS Annu. Meeting, 1986, pp. 301-308.
[21 T. M. Jahns, “Torque production in permanent-magnet synchronous
device turning on or off was simulated to calculate the current motor drives with rectangular current excitation,” IEEE Trans. Ind.
oscillations and resulting torque pulsations. Appl., vol. IA-20, no. 4,pp. 803-813, July/Aug. 1984.
131 S. Funabiki and T. Himei, “Estimation of torque pulsation due to the
The results indicate that the small and large signal responses behavior of a converter and an inverter in a brushless dc-drive system,”
are very similar. This result is only true when the timing of the Proc. Inst. Elec. Eng., vol. 132, pt. B, no. 4, pp. 215-222, July 1985.
input phase currents with the back EMF is correct. In the case 141 N. A. Demerdash and T. W. Nehl, “Dynamic modeling of brushless dc
motors for aerospace actuation,” IEEE Trans. Aerosp. Electron.
of the PMSM drive presented in Part I [ 7 ] , this result is true Syst., vol. AES-16, no. 6, pp. 811-821, Nov. 1980.
under vector control. The large- and small-signal speed P I P. Pillay and R. Krishnan, “Application characteristics of permanent
response is the same whether PWM or hysteresis current magnet synchronous and brushless dc motors for servo drives,” in
Proc. 1987IEEEIASAnnualMeeting, Atlanta, GA, Oct. 19-23, pp.
controllers are used. This is because, even though the torque 380-390.
pulsations may be different due to the use of different current 161 G. Pfaff, A. Weschta, and A. Wick, “Design and experimental results
controllers, the average value which determines the overall of a brushless ac servo-drive,” in Proc. IEEE IAS Annu. Meeting,
1982, pp. 692-697.
speed response is the same. 171 P. Pillay and R. Krishnan, “Modeling, simulation and analysis of
The relationship between the motor torque pulsations permanent magnet motor drives-Part I: The permanent magnet
synchronous motor drive,” this issue, pp. 265-273.
produced as a result of the hysteresis current controller and the
window size is nonlinear, unlike the PMSM drive, where it is
linear. The relationship between the commutation-induced Pragasen Pillay (S’84-M’87), for a photograph and biography please see
page 273 of this TRANSACTIONS.
torque pulsation and the current being commutated, however,
is linear. This has implications in position applications. The
accuracy and repeatability of position servo performance can Ramu Krishnan (S’81-M’82), for a photograph and biography please see
be affected, particularly if high currents are commanded. page 273 of this TRANSACTIONS.