IEI:E TRANSACTIONS ON I'OWER ELECTRONICS. VOL. 5. NO.2.

. .

AI'RIL I'NO
1

151 C1

Control Characteristicsand Speed Controller Design for a High Perfonnance Pennanent Magnet Synchronous Motor Drive
PRAGASEN PILLA Y, MEMBER, IEEE, AND RAMU KRISHNAN, MEMBER, IEEE

Abslracl- The Iheory uf vectur control is applied to Ihe nonlinear modcl of a permancnt magnel synchronous motor to dcvclop a lincar modcl fur cuntruller design. purpuses. The operation and relevant mathematics of a pseudo derivative feedback controller is then presentcd. Three different speed controller designs are then considered and a comparative evaluation made based on their largc and small signal behaviur. In ordcr to test the large signal response, the detailed nonline.ar model of the machine and real time model of the inverter swilches are used.

I.

INTRODUCTION

HE DC servo motor drive has many drawbacks when compared 10ac servo mOlOrdrives. These include lack of robustness and overload capability, lower torque, and speed bandwidths and the regular brush and commutator maintenance required. For these reasons, ac servo drives are preferred in spite of the fact that the controllers required are more complicated and expensive than those for dc servos. The machines used for ac servos in tBe fractional to 30 hp range are the induction, brush less dc and permanent magnet synchronous motors. Vector control [I] (also known as field oriented control) is needed in ac drives to transform the ac machine, performance wise, into an equivalent separately excited dc machine. This gives the ac machine the highly desirable performance capabilities of the separately excited dc machine, while retaining the general advantages of ac machines over dc. Vector control was first applied to the induction motor and a considerable amoun't of literature has been devoted 10 this area. Recent research [2]-[4] has indicated that the permanent magnet synchronous motor (PMSM) could become a serious competitor to the induction motor for high performance servo applications. The PMSM is more efficient and has a larger torque to inertia ratio and power density when compared to the 1M in the fractional to 30-hp range. In addition, for the same output capacity in the above range, the PMSM is smaller in size
Manuscripl received April 6. 1987: re\'ised Ocluber 23. 1989. This paper was presented allhe 1987 IEEE Power Electronics Specialists Conference. Blacksburg. VA. June 21-26. P. Pillay is with the Depanment of Electrical and Electronic Engineer' ing. Merz Coun. University of Newcastle Upon Tyne. Newcastle. NEI 7RU. England. R. Krishnan is with Ihe Depanmem of Electrical Engineering. Virginia Polytechnic and Slate University. Blacksburg. V A 24061. IEEE Log Number 8933771.

and lower in weight that makes it preferable for certain high performance applications like robotics and aerospace actuators. A fair amount of research has been done on PM ac mo- . tor drives [2]-[ 11]. Axial field designs have been examined [2] as well as the possibility of replacing industrial PM motors with line start designs [3]. Extended speed operation through flux weakening [5]-[7] has also been examined and key equations developed for the performance prediction. Design criteria have also been laid down [8]. [9] while a comparison between the sinusoidalfed and rectangular-fed motor drives has been presented [10]. Crucial to the successful implementation of vector control is proper control of the armature currents [1'1]-[ 14] that may be accomplished with a hysteresis, ramp comparison, space vector, or some form of predictive current controller. Efficient use of the PMSM necessitates a knowledge of its characteristics from a control point of view. The model of the PMSM is however nonlinear. This paper applies the concept of vector control that has been extensively applied to induction machines to derive a linear model of the PMSM for controller design purposes. The speed controller is then designed. Although a linear model is used for controller design purposes, the speed controller is requirep to function over the entire speed range when the state variables of the system experience large excursions. Hence the large signal behavior of the drive system is evaluated by simulating the entire nonlinear system. The nonlinear equations of the PMSM, state space speed controller equations and a real time model of the inverter switches are used in this simulation. Every instant of a power switch opening or closing is included. Key theoretical results are verified experimentally. The paper is organized as follows. The operation of the entire drive system is described in Section II. Each of the components of the drive system are described in subsequent sections. Section III presents the nonlinear model of the PMSM. In Section IV, the application of vector control to the PMSM is discussed from which a linear model for controller design purposes is developed. Section V describes the operation of current controllers used in high performance motor drives. In Section VI, the operation and relevant equations of the speed controller are
@ 1990 IEEE

0885-8993/90/0400-0J51$01.00

152

IEEE TRANSACTIONS ON rOWER EI.ECTRONIC!>. VOL. ~. NO.2.

ArRIL 1990

," a
Tun. Corml

It

'"

,a .
vc

Current

Controller & Base Dr! ve Amp1if leu

T I, T2. T3, T!., T5, T6

Fig. I. Schematic or PMSM speed servo drive system.

presented while Sections VII and VIII contain the results and conclusions. II. DRIVESYSTEM Fig. I is a schematic of the overall speed servo drive' system. All reference or commanded values are siJperscripted with a * in the diagram. The error between the referen~e and actual speeds is operated upon by the speed controller to generate the torque reference. In the constant airgap nux mode of operation where i,} = 0, the torque reference is divided by the motor torque constant to give the reference quadrature axis current. This goes through the inverse Park transform to generate the A, B, and C phase stator reference currents. Rotor position feedback is needed in order to generate these currents. The current controller attempts to force the actual motor currents to equal the commanded valu~s at all times. Current feedback is required in order for the hysteresis or PWM cur~ rent controllers to achieve this. Current control is implemented by the appropriate firing of power devices T1 to T6. For a speed control system, speed feedback is also needed; both position and speed feedback can be obtained from a resolver/signal processor combination. When greater than rated speed is commanded. the machine then operates in the constant power or flux weakening mode. Here, the airgap nux is weakened by applying a direct axis current in opposition to the rotor magnet nux. The torque speed profile of the drive is as shown in the block labeled FW. The output of the block is unity up to rated speed and decreases hyperbolically with speed between the rated and maximum speeds to ensure constant output power. When the output of FW is unily then 1..111 = Atifand i,} = O. If the output of FW is less than unily then a lower T: is demanded. In addition Aliiis less than A'!fso

that a negative i,}. is commanded in order to buck the magnet flux. The speed'controller is designed at rated speed but is required to operate prop<>rIy up to the maximum speed. \ ,
Each of the components shown in the overall drive block diagram of Fig. I are described and elaborated upon in subsequent sections. III. MACHINEMODEL The following are assumptions made in the derivation: I) saturation is neglected although it can be taken into. account by parameter changes; 2) the induced EMF is sinusoidal; 3) eddy currents and hysteresis losses are negligible; 4) there are no field current dynamics; 5) there is no cage on the rotor. With these assumptions, the stator d, q equations of the PMSM in the rotor reference frame are [I5]:
Uq

= Riq + PAq +

CJJ"A,I

( I) (2)
(3)

Vel = Riel

pAeI

- CJJ.,Aq

where
1..'1= Lqiq

and A,I = L" it! + A(/I'

(4)

lIeIand Uqand the d, q axis voltages, ieland iq are the d, q axis stator currents, Lei and Lq are the d. q axis inductances, A" and 1..'1 are the d, q axis stator nux linkages, while Rand CJJJ are the stator resistance and inverter frequency respectively. A(/! is the nux linkage due to the rotor magnets linking the stator.

'".0.'''

,"'"

H.."'"

~".L</

,-V.""'11.LI;"

ul;~lui'

nil(

~li'LlI"VN()U::;

MOIUH

DHIVI:

153

The el~ctric torque

Tc

lq .1.

3P[ Aufiq + (L"

Lq)i"iq]/2

(5)

and the equation for the motor dynamics is Tc = TL + BIJJ,+ JplJJ,. ( 6)

id Aaf

P is the number of pole pairs, TL is the load torque, B is the damping coefficient, IJJ,is the rotor speed and J the moment of inertia. The inverter frequency is related to the rotor speed as
IJJ..= PIJJ,.

(7)

phue

The machine model is nonlinear as it contains product terms such as speed with i" and iq. Note. that IJJ"iq and i" are state variables. For dynamic simulation, the equations of the PMSM presented in (1) to (6) must .be expressed in state space form as pit/ = (u" - Ri" + 1JJ.,Lqiq)/Lti piq = (Vq - Riq - 1JJ.,L"i" - 1JJ.,A<if) / Lq (8) (9)'. ( 10)

Fig. 2. Phasor diagram or veclor contniller.

current iq be in quadrature to the rotor flux. Consequently itl has to be along the rotor flux since in the reference used it/lags iq by 90. If i" is in the same direction as the rotor flux, the d axis stator flux adds to the rotor flux and an jncrease in the net airgap flux results. On the other hand if i" is negative then the stator d axis flux is in opposition to that of the rotor resulting in a decreased airgap flux. The permanent magnet machines are designed such that the rotor magnet alone is capable of providing the re-

PIJJ, = (T, - TL - BIJJ,)/J.

The d, q variables are obtained from abc v3riables through the Park transform defined as
llq
lI"

I COS (0) I
: 1/2 L.

cos (0

21r/3)

cos (0
sin

+ h/3);

llll I

= 2/3 :sin I (0) sin (0

1/2'

21r/3)

(0 + 21r/3): ~

lI"
lIt"

( II )

lIO

1/2

The abc variables are obtained from d, q variables through the inverse of the Park transform defined as
Vt;;
I
I

COS

(0)

sin (0)

If lIq 1 . lItl

v,,!= cos (0 - 21r/3) sin(0-21r/3) lie: .cos (0 + 21r / 3) sin (0 + 21r/3)

I f'

qui red airgap flux up to the rated speed. Hence it/ is normally zero in the constant torque mode of operation. The electric torque equation (5) then becomes T, = 3P( A<ifiq)/2.

Ii ;

i (12) I

( 15)

Hence the electric torque depends' only on the quadrature Note that these transformations apply equally well to cur- . axis current and a constant torque is obtainable by ensurrents and flux linkages. The total input power to the ma- ing that iq is constant. The constant airgap flux required chine in terms of abc variables is up to rated speed, which is indicated by Allf is provided power = lIlI:1I + lI"i" + lie it' ( 13) by the magnet. If the airgap flux needs to be boosted (i" in the same direction as Allf) or bucked (i" in opposite while in d, q variables direction to Allf) then t~e required i" can be input into the ( 14) machine. Hence independent torque andairgap flux conpower = 3(v"i" + vqiq)/2 trol is available just like a separately excited dc machine. for a balanced systel\l. The required input phase currents needed to produce the commanded i" and iq are calculated from the inverse Park IV. VECTORCONTROL OF A PMSM Transform in (12). Instantaneous knowledge of the rotor The transformation of a PMSM into an equivalent sepposition is needed in order to apply this transformation. Vector control is therefore only possible when precise arately excited dc motor has been made possible with the knowledge of the instantaneous rotor flux is available. vector control technique. The vector controller for the Hence it is inherently easier in the PMSM than in the inPMSM is derived systematically in this section. Fig. 2 shows a vector diagram of the PMSM. Phase A duction motor because the position of the rotor nux is is assumed to be the reference. The instantaneous position uniquely determined by that of the rotor position in the PMSM. In the induction motor however, the rotor nux is of the rotor (and hence rotor flux) is at angle 8, from phase A. The application of vector control, so as to make it simlocated at some angle away from that of the rotor, deilar to a dc machine, demands that the quadrature axis pending on the value of the slip frequency. If a direct

Ij'Vo

--

154

!HIm TRANSACTIONS ON l'ownR

nI.ECTRONln.

VOl.. :i. NO.2.

APRil. 1')1)(1

method is used in the implementation, i.e.; where the airgap flux is measured, then operation around zero speed is complicated by the necessity of integrating the voltage signal to obtain flux. In an indirect implementation, where the position of the rotor flux is estimated, a model of the machine must be used which depends on parameters that can experience large (excess of 50 %) variations due to temperature, saturation and operating frequency. Parameters lIsed in the model that arc different from those in the actual machine cause a degradation in the 1M drive performance; a corresponding degradation is not experienced in the vector-controlled PMSM. Hence with the application of vector control, independent control of the torque (i,,) and flux (i,,) producing currents are possible. The equations for controller design then are (8), (9), (10), and (IS). These equations busically described a voltage source inverter Fed PMSM. The technological cfevelopment of high switching frequency, high current rating transistors has realized high bandw.idth current sources that has lead to improved servo performance in terms of torque and speed bandwidths. The mode of operation of the current controllers used in PMSM drives is discussed next. V. CURRENTCONTROLLERS Current regulated voltage source inverters arc used widely in the Fractional to 30 hp range to achieve high servo performance. Amongst the most common are the hysteresis, ramp comparison, space vector, and predictive controllers. A. Hysteresis Currellt COlltroller Fig. 3 shows a schematic of a hysteresis current controller. Sinusoidal curtents are needed in order to produce constant torque in the PMSM. Upper and lower hysteresis levels are defined relative to the reference value and the inverter is used to ensure that the actual current is confined between the hysteresis bands. Tight control of the actual current is possible by defining small hysteresis bands; however this demands a higher switching frequency from the inverter. Care must be taken not to exceed the switching capability of the inverter power devices, The actual current therefore contains harmoriics that produce high frequency torque ripples. However the motor inel1ia effectively filters out these torque ripples such that the speed is virtualJy ripple free. This current controller reacts instuntaneously (theoretically) to changes in the current command, hence there is no delay or lag in the model of this current controller. B. Ramp Comparisoll Currellt COlltroller Fig. 4 shows a schematic of a ramp comparison current controller. The actual current is compared to the reference and an error signal generated. This is compared to u triangular wave and if the error is larger than the sawtooth, the phase voltage is switched positively and vice versa. The advantage of the ramp comparison current controller over the hysteresis is that the invener switching fre-

8
N

x
0 0 <0 ~ 'f'

I.'

I.' . AI
I.'

- AI

0 0
0

8
to
I

8
N '0.00

0.05

0.10

0.\5

Fig. 3. Hysteresis

current controller.

\ \

-u '.. ~

~-

Fig. 4. Ranlp comparison current controller,

quency is preset and the power device switching capability can easily be taken into account. Its disadvantage however is that there can be a time delay between the time the reference changes and corrective switching action taken by the current contrpller. However, if the ramp comparison delay is small (Le., 'about 10 times less than the stator time constant) then it has been found in practice that it has~a negligibfe effect on the system behavior. C. Space Vector alld Predictive Currellt Co//trollers

The space vector current controller [13] operates by attempting to control the orientation of stator current vector
using the

voltage vectors available, instead of individual phase currents as discussed in Sections V-A and V-B.

Since the sum of the three phase currents must equal zero, when no neutral is connected, there is :In interaction between the phase currents when the hysteresis or ramp comparison current controllers are used. This means that although the current flowing in anyone phase depends on the stale of the other two phases, no account is taken of this in the current control algorithm. The space vector current controller tries to rectify this at the cost of udditional circuit complexity. The predictive current controller decides on u switching strategy based on an estimate of the probable trajectory of the stator current vector. It works best when the load torque profile is known before-hand. since then the stator

current trajectory can be predetermined. More complexity is req4ired to precalculate the probable, trajectory, which is a disadvantag'e from an industrial application point of , vIew. A detailed comparison of the different current controllers is beyond the scope of this paper. A compromise between the complexity of the space vector and predictive algorithms with the simplicity of the hysteresis with its attendant problems of variable switching frequency with motor speed and parameters is the ramp comparison controller. Its switching frequency can be preset to lie within the capability of the inverter, while maintaining as good a current fidelity as the hysteresis current controller. This probably accounts for its use in several industrial drive products and is also the current controller used in this investigation.
,

. '" .
Fig. 5. Pseud<HJcrivalive feedback conlroller.

I~

OJ.

VI. SPEEDCONTROLLER DESIGN The speed controller may be designed using classical frequency-domain techniques or modern state space methods. The state space method requires an accurate knowledge of the drive parameters for proper placement of the .poles. Frequency domain techniques display a general robustness to parameter changes; indeed several techniques exist to check for the sensitivity of the design to plant parameter changes and indeed to include parameter variations from the outset, using for e.g., the Nichol's chart [16]. This probably accounts for its popularity in the drives industry, with a different configuration from t~e conventional proportional integral (PI) or proportiorpl integral and differential (PID) being considered in this paper. It was shown in Section IV that (8), (9), (10), and (IS) are needed for the speed controller design. The stator time constant of the machine in this paper is such that the pulsewidth modulated (PWM) switching delay can be neglected. The motor parameters of the machine are R 1.411 Ld 5.6 mH L'I 5.61 mH J 0.00176 kgm2 B 0.00038818 Nm/rud/s Auf 0.466 V/rad/s, In addition, (8) and (9) describe the stator current dynamics. If the stator time constant is much smaller than the mechanical time constant of the machine then it can be assumed that the actual current assumes the commanded value in negligible time when compared to the mechanical time constant of the motor. Hence the final equation describing the motor operation
IS

of the control system is shown in Fig. 5. The speed and its derivative are fed back and compared to the reference. The error generated is operated upon by an integrator to create the electric torque reference. By placing the dif1"erentiator in the feedback loop, differentiation of the reference is avoided which 'can be problematic when the commanded speed is a step. Further, differentiation of any practical signal should be avoided because of the amplification of noise on the output of the differentiator. The ,derivative term in the feedback loop is still obtainable however by moving !he feedback loop as shown in Fig. 6. Figs. 5 and 6 are equivalent from a transfer function point of view provided K" in'Fig. 6 is equal Kj in Fig. 5. An analysis of the system in Fig. 6 reveals that the characteristic equation is given by

J s + (Kd + B) s + Kj = O.

. ( 17)

from which the damping ratio and natural frequency are easily determined as Z

= (K" + B)/(2-J(JKi))

j'

T,. = Tl. + BWr + Jpwr

(16)

T" is a controlled input while Tl. is an uncontrolled input. The type of controller that is designed for this speed servo system is known as a pseudo-derivative feedback (PDF) controller as developed in [17]. A block diagram

---

-----

--

---

156

IEEE TRANSACTIONS ON POWER ELECTRONICS. VOL. 5. NO.2. A"'UI. 2'

1')1}(1

'

. "'c

"'c

.;
=,. ",0
ILl", g"! 0-0

..

Fig. 6. PDF controllcr without difi'crcntialion.

i~ 0
i* q

7F
Fig. 7. Operating regionsor a currenl sourcc inverter.

".v

!!

.;
0.06

0.31 0.12 TJ1il~SJ 0.25 .lg~~1

0."

0.&0

Fig. B. Speed or critically

d:lIl1pcd syslem.

lrate these concepts, three designs are considered in this paper. The first is such that none of the control or power level signals are driven into saturation. In addition the system is critically damped so that no speed overshoot occurs. In this design the maximum capability of the inverter and motor is utilized at only one point during the entire startup. This is referred to as the critically damped design in the rest of this paper. This is compared to a second design where the maximum current capability of the inverter is commanded over an extended period of time in order to improve the response. In a practical Implementation this means limiting the control signal outputs by zener diodes to ensure that the inverter is not commanded to produce more than its capability. This frees the control level signals from having to remain within the saturation limits as in the first design. The gains can be chosen to improve the small signal dynamic performance. This second design is done for a fast response application and the
,

~ D
11
='
~

.;
51

'"
0'\

X 0

Commanded Ac tual

M .;

. M

>< :i!
-a =' ... D ...
Q/ .;T! \ \

00 n!. :f: 2

c .;
D '0.00 O.GO . 0.06 0.'2 TJ'H~~s I 0.26 .lg~~1 0.31 0."

Fig. 9. Torque or critically damped syslem. 2

final result ends up being slightly underdamped. Therefore this second design is referred to as the underdamped design. In some applications in industry, it is undesirable to have speed overshoot. Therefore a third design is done that is similar to the second but whose final result is overdamped to prevent speed overslioot. This is referred to as the overdamped design. VII. RESULTS The design criteria discussed in the last section were used to obtain the derivative and integral constants in Fig. 5. A computer program was then written to simulate the entire speed servo drive system including the ramp comparison current controller, PDF speed controller and d, q axis equations of the motor. The response of the motor drive for speed reference changes only are considered in this section. Every instant of a power switch opening or closing is represented. Figs. 8-10 show the speed, electric torque and current of the critically damped speed servo design. From Fig, 8. it is clear that there is no speed overshoot that is the dcI
0

&i

Commanded

'" ~
='

.;

0 Actual

0'\ oq

'; .;
M:
Q/ .; -a
>< N ...

::I u c: f . 000 n! :f:

~
90.00
0.06 0.12

TJ'H~1 5I

O.SO 0.25 .lg~1' 0.37 0."

Fig. 10. Currcnl or critically damped systcm.

fining feature of a critically damped system. The motor takes about 50 ms to reach the commanded speed. The current and hence torque limits arc taken into account in the design process and it is ensured that the system always

.. ..
0

Sm ~
(")0 :>< g~ ...0

~~

......

=' . p-o

'" ';!o

X Commanded 0 Actual

g~ .-0

~ r[;;
0

(3 ~!:: 0

~ 0

~
0.06 0.12

TI~~~SI 0'.25 .lg:~1

0'.31

0."

0.60

~o.oo

-+0.06

v
I
TI~~~ S)

I
0.26

I
.lg:~1

0.12

0.31

+ 0."

0.60

Fig. II. Underdamped speed response.

Fig. 12. Torque or underdamped syslem.

2

operates in the linear region. Although the PWM inverter produces high frequency harmonics, the actual current tracks the reference closely. These current harmonics produce high frequency torque harmonics as shown in Fig. 9. The mechanical system filters these harmonics out quite effectively so that they are not reflected in the speed response shown in Fig. 8. The current and torque limits are reached at only one point in the transient. The speed response can be improved by demanding the maximum current and hence torque capability of the machine over a longer period. This is shown in the next set of results. Figs. 11-13 show the speed, torque and current re-

~
'"

='0;

(")0

C;~
:>< ~~ :J. ...0
13 ~!:!
0

X Commanded 0 Actual

~
'

sponse of the speed servo system when designed for a'fast

response application. The motor attains the reference speed after about 15 ms as opposed to the 50 ms in the previous case. This reduction is accomplished by demanding the maximum torque capability of the machine over a period of time as shown in Fig. 12. The actual and commanded phase currents are shown in Fig. 13. The finite time needed by the actual current to reach the commanded is due to the stator time constant. This produces a corresponding rise time in the torque in Fig. 12. Since the maximum torque is demanded until the reference speed is almost reached, there is a speed overshoot. When this occurs, the commanded torque goes negative to bring the actual speed back to the commanded value as shown in Fig. 12. When speed overshoot is not tolerable, then an overdamped system could be the proper choice. The speed response of such a system is shown in Fig. 14. The torque and current curves of this system look similar to the previous ones and are therefore not shown. During the linear rise in the speed, the control signal output is clamped so that the inverter delivers its maximum output. The operation of this system has been verified experimentally as shown in Fig. 15. The initial linear rise in speed and subsequent overdamped response are clearly evident in both the practical measurement and theoretical prediction. Up to now, large signal responses of the speed control servo system have been considered. In the latter two de-

~ 0 '0.00
0.06 0.\2 TI~t~S) 0.25 .lg:~1 0'.31 0."

0.60

Fig. 13. Current or undcnJamped syslem. f!

... CD 0

~r-

='~ p- gS! .-0

>=:::
0

0.60
0.06 0.12 TIC;;t~SJ 0.25 .1g:~1 0.31 0."

Fig. 14. Speed respunse or uverdamped syslem.

signs presented, the control level signals were clamped to ensure that the inverter operated in the linear mode. During small signal operation, the inverter would not in general be required to perform at its maximum capability.

.-158

IEEF. TRANSACTIONS ON POWER m.ECTRoNICS.

VOL. 5. NO.1.

APRil. 11)'1(1

E! Ii!

~
~

. E!
05! w. 0"1

~8

X D

Amplitude
Phase

B::i

. Ii! .
~
~
1.(.0

Fig. 15. Mc~sllrcd

S(.~rtllp speed.'

x-axis

ms/div.

y-axis

437.5

r (min /iliv.

1.(.1

1.(.2 FREQIRRO/SI

1.(.)

1.(..

Fig. 17. Ootle plul ror untierJampeti

system.

8 ...

~
!:: '"
... ...

~
... ...

~
oS!
w'

"
X 0 Amplitude Phase
w' o~ 0: 0 'm~ O'

o~ 0: o~ m. o~
::: ~

oS! X Amplitude
0
\

Phase

.
1.(.0 LE.' . 1.(.2 FREQIRRO/SJ

'

... ..

~
g
1.(.3 1.("

. '"
E! ~ ,
1.(.0
1.(.1 1.(.2 FREaIRROISJ
1.(')

1.(..

Fil;. 16. Dutlc plot rur critically

tlamped syslel11.

Fig. 18. Bode plol rur overd:lmpeJ syslcm.

The system behavior during this mode of operation can be determined from small signal time responses. However they can ;llso be determined from Bode plots ;IS shown next. Fig. 16 shows the amplitude and ph;lse response for the critically damped system. The bandwidth is only about 10 radls ;lnd this system would give an extremely sluggish response. Hence clamping of the control level signals in this design is unnecessary. Fig. 17 shows the Bode plot for the underdamped system. The bandwidth is increased dramatically to about 800 rad/s when compared to the first design. This system would respond much more quickly than the previous one. Fig. 18 shows the corresponding curves for the overdamped system. Although the bandwidth here is lower than that in Fig. 17, it is higher than that in Fig. 16, which is a critic;llly d;lmped design. Thus this overdampcd system performs beller than the critically damped system. VIII. CONCLUSION This paper has presented various designs for the speed controller of a high performance PMSM speed servo using

PDF control. The control parameters obtained from the design process were used in a detailed system simulation which included the inverter switches, speed controller and d, q axis equations of the motor. The results indicate that a critically damped design done so as to ensure that all control and power signals never saturate gives an extremely poor design. Much better small and large signal responses are achieved by avoiding this constraint and using zener diodes instead to limit the commanded input into the inverter. Two designs were presented using this technique, an underdamped design with low speed overshoot arid an overdamped design with no speed overshoot. The response of the underdamped design was much quicker than that of the overdamped. However the overdamped design has applications where speed overshoot is intolerable.
REFERENCES
III F. Dlaschke. "The principle of fielJ nrient;otion;os IIpplicd to the new transveclor closed-loop conlrol system ror rolnling field IIwchines." SiclII"" '.f RCI"ic''''. PI'. 217-221. 1972.

r'LL..,

n.'"

"'''J'''',,,,,

J'LLU

~U..."U~LL"

.h.,.~"

.,."

~."~..",,."""

121 R. Krishnan and A. J. Deutler. "Performance and design of an axial . field p';nnanent /1'lagnet synchronous motor drive," in COllf ReI'. IEEE IAS AII/I/IlII Mutillg. 1985, pp. 635-640. 131 E. Richter. T. 1. E. Miller and T. W. Neumann. "The ferrite permanent magner AC motor-A technical and economic assessment," in COllf R/'('. IEEE. IAS AI/I/I/al M(.(,tillg. 1984. pp. 1353-1358. 141 P. Enjeti. J. F. Lindsay. and M. H, Rashid, "Parameler estimation and dynamic performance of penmlnent magnet synchronous motors." in COllf R(,c. IEEE IAS A//II/IlII Meetillg. 1985. pp. 627-633. 15) T. M. Jahns. "Flux weakening regime operation of an interior permanent magnet synchronous motor drive." in Calif ReI'. IEEE IAS AI///IIlII Meetill!:, 1986. pp. 814-823. f6f T. M. Jahns, G. D. KHnuII1, and T. W. Neumann, "Interior permanent magnet synchronous motors for adjustable speed drives." in COllf Ree. IEEE IAS AI/III/alMeet;II!:. 1985. pp. 814-823. 171 T. Sebastian and G. R. Siemon. "Operating limits of inverter-driven

permanent magnet motor drives."

in Calif Rec. IEEE

IAS

AlIl/l/lIl

Meetillg. 1987. pp. 800-805. . 18) W. Leonard. COl/trol of Electric Dril'e.f. New York: Springer-Verlag. 1984. 19) D. Pauly. G. Pfaff, and A. Weschta. "Brushless servo drives with permanent magnet motors or squirrel cage induction motors-A comparison." in COl/f Rec. IEEE IAS AI/I/I/al Mertil/g. 1984. pp. 503509. . 1101 P. Pillay and R. 'Krishnan. "Application characteristics of PM synchronous and brushless dc mOlors for servo drives." in Calif ReI'. IEEE IAS Al/lIlIlIl Meetillg. 1987. pp. 380-390. Ill) C. Schauder and R. Caddy. "Current control of voltage-source inverters for fast four-quadrant drive perrormance." IEEE TrclllJ. II/d. 1.1'1'1.. vol. IA-18. no. 2, pp. 163-171. Mar.lApr. 1982. 11'21 G. D. Kliman and A. D. Plunkett. "Development of a modulalion . strategy for a PWM inverter drive." IEEE Tral/s. II/d. 1.1'1'1.. vol. IA-15, no. I. Jan.lFeb. 1978. 113) G. pralT. A. Weschta, and A. Wick. "Design and experimental resulls of a brush less ac servo drive," in COl/f Rec. IEEE IAS AI/IIIUlI Mutil/g. 1982, pp. 692-697. 114] D. M. Drod and D. W. Novotny. "Current control of VSI-PWM inverters." in COl/f Rec. IAS AlI/lIIlIl Mee/il/g. 1984, pp. 418-425. 115) P. C. Krause. AI/alysis oJ Electric Machillery. New York: McGrawJ-lill, 1986. (16] E. Eitelberg. J. C. Dalda. E. S.. Boje. and R. G. Harley. "Stab~izing SSR oscillations with a shunt reactor controller for uncertain levels

interests systems. Dr. Pillay is n past recipient of an IEEE prize paper award. He is u membcr of the Industry Applications. Industrial Electronics and Power Engineering Societies of the IEEE and a member of the Industrial Drives and Electric Machines Commillees of the Industry Applicalions Society. He is also a member of the Greck Honor Society. Phi-Kappa-Phi.

I'ragasen Pillay (S'84-M'87) received Ihe D.Eng degree from the University of Durban-Westville, South Arrica. in 1981. the M.Sc(Eng) degree from the University or Natal. South Africa. in t983. and the Ph.D degree from Virginia Polytechnic Institute & State University. Dhlcksburg. in 1987. The Ph. D degree was funded by a Fulbright scholarship. He is currently with the Department or Eleclrical and Electronic Engineering. University of Newcastle upon Tyne. England. UK. His research are in the modeling. control. and design of electric molor drive

.,
,;

Ramu

Krishnan

(S'81-M'82)

received the B.E..

and the M.E. degrees from the University of Madras, India and the Ph.D. degree from Concordia University. Montreal. Canada. all in electrical engineering. He taught nearly seven years in India. then from April 1982 to September 1985. he was a Slalr Engineer and hller a Principal Investigator or ac servo drives at Gould Research Center. Rolling Meadows. II. Since September 1985 he has been a member of the faculty In the Electrical Enginccring Department. Virginia Polytechnic Institute and Stale Universily (VPI & SU), Blacksburg. His rcsearch interesls are in vector controlled induction molor drives. PM dc brush less and synchronous motor drives. switched reluclance Inotor drives. eleclrical machine design. and stalic power conversion and he has published more lhan 50 papers on thcsc lopics. In addition, he has built a graduale program in motor drives and machine design at (VPI & SU). Dr. Krishnan has been awarded four IEEE prizes for his research papers and he has bcen an Associate Editor or IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICSbctween June 1987and Seplember 1988.

of scries compensation," IEEE TrcIlIJ; POlI'er Sy.H.. vol. PS-3. no. 3,

pp. 936-943, Aug. 1988. (17) R. M. Phelan. AlllolI/o/ic COll/rol Sy.\'/eIl1J. IIhaca. N.Y.: Cornell University Press. 1977.

I:!

:i1
~..,