New Approach for the Low-Speed Operation of PMSM Drives Without Rotational Position Sensors - Power Electronics, IEEE Transactions on

New Approach for the Low Speed Operation of PMSM Drives Without Rotational Position Sensors Power Electronics, IEEE Transactions on 512 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL 11, NO 3, MAY 1996 N[.]

New Approach for the Low-Speed Operation of PMSM Drives Without Rotational Position Sensors

Joohn-Sheok IOm, Student, IEEE

Abstract- A new approach to the position sensor elimination

of permanent magnet synchronous machine (PMSM) drives for

low-speed operation is presented in this paper Using the position

sensing characteristics of PMSM itself, the actual rotor position,

as well as the machine speed, can be obtained even in the transient

state Since the essential back-EMF information is obtained from

direct measurement during a special test cycle called MCDI, the

operating speed range of the sensorless drives can be extended

to 10 rpm Moreover, the chronic starting problem of the PMSM

drives can be simply settled by the proposed algorithm In order

to verify the feasibility of the proposed scheme, experimental

results in the low-speed range of about 10 - 100 rpm are also

presented

I INTRODUCTION

P to now, machine drive systems without rotational

position sensors, the so-called sensorless drives, have

gained increased popularity in industrial applications because

of the inherent drawbacks of rotational sensors In general,

the rotational encoder-type sensors are used to obtain speed

or position information of the machine The major drawback

of these sensors is the performance degradation caused by

vibration or humidity Furthermore, these extemal sensors

will result in added cost and increase the size of the drives

[ 11-[6] By these reasons, there has been considerable interest

in developing techniques to achieve the position and speed

information for the sinusoidal back-EMF-type permanent mag-

net synchronous machine (PMSM) without extemal position

sensors The PMSM of this type has widely found its applica-

tion fields on the high-performance machine drive because of

the ripple-free torque characteristics and simple control rule

However, since none of the parameters of the PMSM with

sinusoidal back-EMF varies as a function of the rotor position,

it is very difficult to get the rotor position information without

rotational position sensors in this drives

Basically, most previous studies about the sensorless drives

for the PMSM with sinusoidal back-EMF-type start on the

foundation of the voltage equation of the machine and the

information of the machine terminal quantities, such as line

voltage and phase current Using this information with ma-

chine model, the rotor angle and speed were estimated di-

rectly or indirectly based on the back-EMF information by

various sensorless strategies [ 11-[5] In all cases, however,

since the machine parameters should be well-known for the

proper sensorless operation, the sensitivity to the machine

Manuscript received July 13, 199.5; revised November 30, 199.5

The authors are with the School of Electrical Engineering, Seoul National

Publisher Item Identifier S 088S-S993(96)03SS2-1

University, Seoul 15 1-742, Korea

and Seung-Ki s u l , Member, IEEE

parameter is a major drawback Another important demerit

of the previous studies is the limitation of the controllable speed range, especially in low speed range In the previous studies, the essential back-EMF information was estimated using the terminal quantities, such as currents and voltages But these quantities are very noisy because of the PWM operation of the power stage Especially in the low-speed range, the actual voltage information on the machine terminal can be hardly detected because of the small back-EMF of

the machine and the system noise For instance, when a typical machine with 2000 rpm-rated speed is operated at

50 rpm, the back-EMF is about 1.5 V only Considering the precision of the terminal voltage in PWM fashion and the system noise produced by the nonlinear characteristics of the switching devices [lo], the controllable lower speed range of the conventional sensorless drives is generally limited to the value of around 100 rpm Furthermore, another problem to be considered in the sensorless drives for PMSM is the starting capability At standstill state of the machine, the essential back-EMF information cannot be achieved in the conventional drive algorithm for itself In the literature [6]-[8], some special starting algorithm or initial position detection algorithm were studied for just machine starting However, in the case of sinusoidal back-EMF-type PMSM with symmetrical machine parameters, it is quite difficult to apply these methods to machine drive system

In this paper, new drive technique for the sinusoidal back- EMF-type PMSM drives without rotational position sensor

is presented As described above, the rotor position of the PMSM is related only to the phase of the back-EMF So, the information about the back-EMF of the machine plays the important role of the sensorless drives The back-EMF has a sinusoidal waveform of which magnitude is proportional to the rotor angular speed This means that the PMSM itself has a position sensor characteristic like the resolver-type rotational sensors In the proposed method, the back-EMF information

is obtained by direct measuring at the machine terminal not

by indirect estimation For the purpose of direct back-EMF measurement, a special test cycle, called maximum current decaying interval (MCDI), is introduced Since the back-EMF information is obtained from the direct measurement, the controllable speed range in sensorless drives can be extended

to 10 rpm without parameter dependency Additionally, the chronic starting problem of the PMSM drives without rota- tional sensors or with sensors like the incremental encoder can be solved using the proposed direct-voltage measuring strategy For the successful starting of the general PMSM

0885-8993/96$0.5.00 0 1996 IEEE

q r

P

se

(-

vbs

Ld

Fig 1 Space vector diagram for the PMSM

drives, the standstill rotor angle should be known even if a

rotational sensor, such as incremental type encoder, is adopted

for angle detection By detecting the small back-EMF quantity

produced by an initial test current, the initial rotor angle can be

easily determined within several ones [“I in mechanical degree

As a result, the proposed direct back-EMF measuring strat-

egy can be applied to the general sensorless drive system

to increase the controllable lower speed range and to ensure

the initial starting capability To verify the feasibility of the

proposed method, some experimental tests are conducted for

the low speed range (10 N 100rpm)

11 SYSTEM DESCRIPTION

A System Modeling

The general equivalent space vector diagram for the PMSM

is illustrated in Fig 1 It is well-known that none of the

parameters of the PMSM with sinusoidal back-EMF are varied

according to the rotor position [l], [2] Therefore, the ma-

chine model based on the d-q reference frame theory can be

described in simple form as

R, i!pL,]

(Super script s denotes the stationary reference frame.)

As shown in (l), the terminal voltage of the machine con-

sists of the voltage drop terms related to the stator resistance

R, and stator inductance L,, and the back-EMF term including

the rotor speed w, and the back-EMF constant Kp? It is

generally assumed that the back-EMF has the true sinusoidal

waveform related to the rotor position, and its magnitude

is proportional to the rotor speed Therefore, in this aspect,

the PMSM with sinusoidal back-EMF has the positiori sensor

characteristics in itself like the resolver, and the essential in-

formation about rotor angle and speed can be simply obtained

from the back-EMF in the stationary reference frame So the

back-EMF quantity plays an important roll in the machine

drive system without rotational sensor Generally speaking,

this quantity can be obtained using the terminal voltage and

the information about the voltage drop terms In most previous

studies about sensorless drives, the rotor angle and speed are achieved using a specialized estimation algorithm about the back-EMF quantity However, the value of the back-EMF constant is just several 10’s V/krpm in the general commercial PMSM for servo applications Considering the value of K E ,

it is easily surmised that the back-EMF quantity should have

a value about several 1’s V when the machine is operated with 100 rpm mechanical speed Therefore, on the low-speed operating condition, it is quite difficult to obtain the back-EMF information by the conventional estimation method using the machine terminal quantities

B Back-EMF Detection Strategy for Low-Speed Operation

The back-EMF may be detected at the machine terminal through removing the corresponding voltage drops But, since each machine terminal is connected to the center point of the inverter arms in which the pole voltages have two voltage levels, such as zero or dc-link voltage, it is very difficult

to achieve the actual terminal voltage when this voltage is very small In this instance, if the machine terminals are disconnected from the inverter stage, there is one method to detect the small machine voltage at the machine terminal

In this paper, the back-EMF information is obtained by direct measurement of the terminal voltage As previously

mentioned, the back-EMF will appear at the machine terminal when the voltage drop associated with stator resistance and stator inductance is removed A simple way of removing the voltage drops is to hold the machine currents to zero When the machine terminal currents remain zero, the terminal voltage matches the back-EMF And to measure the terminal voltage in analog fashion, the machine terminal should be disconnected from the inverter arms However, it is not easy to actually disconnect the machine and power stage For the purpose of the implementation of disconnection effect, a third state of the inverter pole voltage is introduced in this paper

The states of the inverter pole voltages are always deter- mined by the gating signals Thus, each pole voltage has zero value or DC-link value according to the gating signals When the pole voltages are different from each other, some active voltage is applied to the machine Let’s define this state as the

‘first state’ of inverter In the other case, all the pole voltages have the same value to apply zero voltage to the machine and this condition is defined as the ‘second state’ of inverter In the conventional inverter operation strategy, the machine voltages are generated using these two pole voltage states However, if all the gating signals are removed from the switching devices, the third switching state can be implemented on the inverter side

As shown in Fig 2, when the machine currents flow through

the machine terminals, the phase current would hold its current state because of the inherent characteristics of the stator winding inductance In the normal operation state, the gating signals for upper or lower switching device are always given

by the on state alternately, except during dead time interval

As shown in Fig 2(a), if the gating signal for the lower device

is on, the normal current named, I,,,, flows through the lower switching device, and the current will flow through the upper

Fig 2 Current flow at the inverter state of all gating off

diode attached to each switching device in the other case If

the current flows from machine to power stage, the opposite

state takes place, as shown in Fig 2(b) However, if all the

gating signals for switching devices are in the off state, the

machine current named io^ will unconditionally flow through

the upper or lower diode according to its direction, as shown

in Fig 2 Therefore, in this state, the pole voltages of the

inverter are just determined by the machine current direction

This phenomenon has been utilized for establishing the basic

idea of the dead-time compensation strategy also During this

third switching state, the stored energy in the stator inductance

is recovered on the dc-link side effectively For example, as

shown in Fig 2, the machine current is injected to the dc-

link voltage through upper diode or derived from the zero

voltage of the inverter side through the lower diode This

means that the negative largest output voltage is applied to the

machine terminal, and the terminal currents should be decayed

to zero with the maximum rate When the stored energy is

fully recovered and all machine currents reach zero value, the

terminal voltages are determined by the back-EMF because

of the absence of the voltage drop Therefore, the machine

terminals keep the electrically open state as the disconnection

state In this paper, the third state of inverter action is utilized

in the MCDI test cycle to measure the machine back-EMF

Considering the voltage equation of the MCDI test interval

leads that the terminal currents can be reduced to zero within

a few tenths of a microsecond For instance, when a PMSM

with 5-mH winding inductance is operated under 310-V dc-

link voltage condition, 5-A stator current will be reduced to

zero within 13 ps

5[A1 = 13[psec] (2)

ai

?; v d c $ .310 [VI

At the end of this MCDI test cycle, the back-EMF that

appears at the machine terminal can be directly measured

with analog fashion Since the neutral point of the PMSM

is not generally available, the back-EMF should be obtained

from the line-to-line voltage, as shown in Fig 3 The voltage

measurement circuit can be constructed with a laser-trimmed

differential-type isolation amplifier and some additional com-

ponents The maximum measurable line-to-line back-EMF is

clamped to 5 V by the back-to-back-connected zener diodes

and 10 k 0 resistance Therefore, the maximum operation speed

is limited to several hundred rpm In the upper speed range,

the back-EMF quantities can be properly estimated using an

, Inverter,

Fig 3 Back-EMF measurement circuit

T,, = 3.6msec

Y

I

I

I

E,', EA' detect & Speed Control

Fig 4 MCDI test cycle

estimation method [9] Another considerable point is the effect

of the parasitic capacitors on the switching devices Because

of this parasitic capacitor and the leakage inductance of the terminal lines, some resonant phenomenon takes place during the MCDI test cycle and the machine current does not diminish rapidly, as expected in (2) To prevent this resonance, some damping resistors are inserted between the machine terminal

in Y -connection form

With the consideration of the imaginary neutral point of the PMSM, the phase back-EMF can be derived as follows

E: + Et + E," = O

Therefore, the d-q components of the back-EMF on the

stationary frame can be obtained simply as,

E" = (2Ez - E{ - E , " ) / 3

(4)

E: =(E," - El)/&

111 SENSORLESS CONTROL STRATEGY

A Rotor-Angle Detection Strategy

The back-EMF measurement is performed by adding an MCDI test cycle to the current control algorithm, as shown

in Fig 4 In this figure, Test means the sampling interval for MCDI test cycle and for the rotor quantity estimation process At the beginning of the MCDI test cycle, all gating signals for the inverter are blocked When the stator currents reach the zero value, the back-EMF is measured 1; is the

Fig

r$

I

I

5 The overall control diagram for low-speed sensorless drive algorithm

torque component current in the synchronous frame fixed to

rotor position After TdlP interval, normal current regulation

process is performed for torque generation Therefore, in the

proposed sensorless algorithm, consequential high-frequency

torque ripple would appear in the case of large viscous friction

load condition However, this ripple can be negligible if a

sufficient inertia is mounted on the machine shaft Considering

the general load pattern that the load quantity would be

increased according to the machine speed, the applicability

of the proposed drive scheme is not damaged Furthermore,

since the ripple frequency determined by the sampling interval

Test is considerably high, this torque ripple is not critical

in the aspect of drive Another choice of the sexisorless

drive for low-speed operation is the usage of the PMSM

with trapezoidal back-EMF In this case, the back-EMF also

can be directly measured at the nonconducting phase [ 2 ]

However, since the torque ripple inherently produced by the

two-phases conducting method is quite severe, it is very

difficult to generate continuously stable electrical torque So

the controllable lower-speed range would be limited about

several hundred rpm in the loaded operating condition

From the measured back-EMF, the actual rotor angle can

be achieved as follows

E: = E -E COS(^,) sin(0,) 1, E = J-.sign(4$)

COS(^,) = E i / E

sin(8,) = - E ; / E

: {

However, when some errors occur at the back-EMF de-

tection instance, the control angle might be influenced by

these errors directly, and this results in the system instability

Therefore, instead of above angle detection method, the rotor

angle is achieved using arc-tangent function and an angle

compensation algorithm is introduced in this paper First, the

rotor angle is achieved as follows

-E: - Kew, sin(0,) sin(0,)

E; Kew, cos(8,) cos(8,)

It is noticeable that the sign of the rotor speed should

be considered at the angle calculation because the speed information is canceled by the division operation of the arc- tangent function When the machine speed has negative value, the rotor angle can be described by

K,w, sin(0,)

= tan(0, - 180')

- sin(0,) - sin(0, - 180')

-

-

-

E; K,w, COS(^,) - COS(^,) COS(^, - 180')

: 8, = t a n 2-'(-E:, E,") + 180' (7)

Therefore, the rotor angle is represented as follows

(8)

The speed direction can be identified using the calculated

0, = 0, + 7r

rotor angle as follows

1 if A&,, > 0

-1 if A0,,,, < 0

To avoid the conflict between (8) and (9), a hysteresis with some bands is added to the rotor speed direction determination process In Fig 5 , the overall rotor angle estimation process

is depicted The lower area enclosed by a dashed box corre- sponds to the speed direction detection process with hysteresis function And the upper area enclosed by another dashed box represents the angle compensation algorithm In this algorithm, the deviation of the calculated rotor angle is limited to some value according to the rotor speed to stabilize the control angle

in the transient state The basic operation concept is that the rotor angle deviation during one estimation step has the same value of the time integral value of the rotor speed Thus, the maximum deviation will be limited by following value

AO,,, = 1.5 x ILF(&)I Test

516 MAY 1996

(The L F ( ) function means the low-pass filter.)

If the angle deviation is smaller than these maximum

values, the control angle Bc always follows the rotor angle

6 , calculated from the measured back-EMF However, if the

deviation of the control angle exceeds the maximum value, this

deviation is restricted to the maximum value and the control

angle is more slowly changed than the calculated angle With

the help of this angle compensation algorithm, the control

angle is always maintained to the actual angle with sufficiently

negligible error, even if some measurement errors occur

At the end of the MCDI test cycle, the gating signals are

returned to their original state, and the normal current control

sequence is restarted In this study, the current sampling

interval is chosen to be 60 ps, and the current control algorithm

based on space voltage vector is used for high-performance

current regulation

machine would be slightly rotated to arbitrary direction The rotor angle deviation between the initial point and moved point

is very small, but the angular speed would be sufficiently great to detect the rotor angle Whenever the machine speed

is higher than 10 rpm, the essential back-EMF can be directly measured, and the initial angle detection process can be accomplished through the proposed angle detection strategy Therefore, within several ones ["I deviation of the rotor shaft

in mechanical degree, the initial rotor angle can be easily detected Just after the initial angle determination cycle, a proper starting torque is generated with the information about the electrical initial rotor angle, and the rotor is smoothly rotated according to the direction of the speed command This important feature also can be applied to the general PMSM drives with an incremental encoder type position sensor for smooth machine starting When the incremental encoder is used, the actual rotor position can be achieved

B Speed-Detection Algorithm

From the basic machine equation (I), the back-EMF infor-

mation can be utilized for the accurate rotor angular speed w,

as follows

w, = sign( $) ' d m / K E

when the first zero pulse is encountered Therefore, in most drives, a special type encoder that provides U, V, W pulses

is used for the initial starting However, the rotor position information with just 60" precision is available using this special encoder at the initial starting instance Moreover, in order to measure these pulses, an additional six lines should

be connected between machine and controller for only the initial starting Using the proposed simple voltage measuring minimum area and smallest wiring, the initial position angle

Every MCD1 test cyc1e, the speed can be equipment that can be mounted on the control board with from the back-EMF and speed control algorithm is performed

operating condition, the following remark should be made on

the back-EMF constant

For an accurate determination of w,, it is necessary to

Although the above equation is guaranteed in can be easily detected within reasonable range

IV EXPERIMENTAL RESULTS know the correct back-EMF constant K E This parameter

may be uncertain or may be varying according to the thermal

environment Changing this parameter will cause steady-state

error of the estimated speed This uncertainty can be overcome

by the back-EMF compensation algorithm [9] The basic idea

of this algorithm is that the back-EMF constant slowly varies

according to thermal condition and the long-term average of

the time differentiation value of rotor angle has the same value

of the actual rotor speed With the help of this algorithm,

not only the parameter variation but also the measurement

error of the back-EMF can be compensated easily Because

of the direct rotor position measuring strategy, the back-EMF

constant can be modified to the suitable value to remove the

steady-state error of the estimated speed

C Starting Strategy

Generally, in the PMSM drives, the information about the

standstill rotor angle is required at the initial start-up instant to

produce a proper electrical torque for starting In the case of

the drive system without position sensors, a specially designed

starting algorithm should be included in the drive algorithm

because the back-EMF is not available at the start-up [6]

However, if the proposed direct back-EMF detection strategy

is adopted for the machine drive system, the starting capability

can be given to the drive system in itself

At the start-up instant, a test current with short pulse form

is injected to the machine current By this test current, the

Experimental tests are conducted to evaluate the perfor- mance of the proposed back-EMF detection strategy and the feasibility of the new position sensor elimination technique for low speed operation The overall PMSM drive system for the laboratory prototype test is illustrated in Fig 6 The tested drive system consists of a six-pole Y-connected 0.47 [kW] PMSM and an electric dynamometer load The rating and parameters of the PMSM are presented in Table I The power stage comprises the IGBT inverter with 8.33-kHz switching frequency and 200-V dc link And, high-performance digital signal processor (DSP), TMS320C30 controller specially de- signed for machine drive is used to implement the overall proposed sensorless algorithm, based on C language The whole experimental results including current, machine speed, and angle traces are illustrated by the multichannel oscillo- scope through D/A converter equipped on the main controller Moreover, for monitoring purposes only, an additional incre- mental encoder with 2000 ppr resolution is attached to the machine shaft The actual machine speed and rotor angle are taken from this sensor

The experimental waveforms of the proposed MCDI test cycle are shown in Fig 7 The PMSM is operated with 50 rpm constant speed by the proposed sensorless control algorithm In Fig 7(a), the actual rotor angle in electrical angular degrees is displayed The estimated rotor angle from the proposed angle- estimation strategy is shown in Fig 7 (b), and the detected back-EMF quantity on the stationary q-axis is shown in the

i

KE (DC equivalent model)

Iyo+a(DC equivalent model)

._

- -

.- 6.5 [mH]

0.03 [ V/rpm]

.-

l"32OC30

DSP Controller

- > > -

L L : 1

I

(d) v m l ] ? d T L

0

Fig 6 Laboratory experimental setup

TABLE I PMSM RATING AND PARAMETERS

3-phase 6pole Sinusoidal back-EMF PMSM -

Rate Power I 0.47 [kWl ,_

R" I 1.26 r ni

.-

0 rate I 2000.0 [rpm]

T

Fig 7 Experiment waveforms of the MCDI test cycle (a) Actual rotor angle

(elec angle) (b) Estimated rotor angle (elec angle) (c) Detected stationary

q-axis back-EMF (d) Synchronous q-axis current (enlarged time base)

next figure As shown in these figures, using the proposed

MCDI test cycle, the essential back-EMF quantities can be

clearly detected and the new sensor elimination control :scheme

for the low-speed operation has performed very well In the last

figure, Fig 7(d), the torque component current of the machine

is presented in the sampled data fashion To show the MCDI

test interval precisely, the time duration between two anrows in

the other figures is enlarged in this figure At the initial points

of each test cycle, the machine currents are quickly diminished

to zero value to directly measure the back-EMF When the

back-EMF is completely detected, the machine currents are

rapidly stabilized to produce the required electrical torque

Of course, the whole electrical torque is affected by the test

cycle for back-EMF detection, and the maximum conlhuous

output power of the proposed control strategy is limited to

Fig 8 Speed control performance (10 rpm) (a) and (b) Detected syn- chronous q-axis and d-axis back-EMF (c) and (d) Actual and estimated rotor angle (elec angle) (e) and (f) Actual and estimated rotor speed

some value However, this sensor elimination technique can be effectually utilized to some applications in which a continuous full torque operation is not needed and an effective low speed operation is required without position sensors

Fig 8 shows the speed-control performance of the proposed algorithm with 10 rpm speed reference In Fig 8(a) and (b), the traces of the measured back-EMF are shown As shown

in these figures, some low order harmonics are included in the back-EMF waveform These harmonic components appear actually in the machine itself because of the slightly small misalignment of the stator winding or inherent flux distribution

of the permanent magnet However, this harmonic component would be negligible in the higher speed range because the magnitudes of this harmonics are nearly constant irrespective

of the magnitude of the fundamental back-EMF In Fig 8(c)

-1001 I t

4

4

Fig 10 50 [rpm]) (a) Initial angle difference E 70°, No load (b) Initial angle difference

5Z -120°, No load (c) Initial angle difference 5Z -40' 1/4 load (d) Initial angle difference -150°, 1/4 load (In each figure, first traces: actual rotor angle (elec.); second traces: estimated rotor angle (elec.); third traces: actual machine speed; last traces: torque component current command on the reference frame fixed to 8est.)

Starting performance of the proposed algorithm (0 i

and (d), the actual rotor angle and the rotor angle calculated

from the back-EMF are shown And, the actual rotor speed

and calculated speed are displayed in Fig 8(e) and (0 The

rotor speed quantities are fluctuated owing to the harmonic

component of the back-EMF If a higher control band width

is selected for speed controller, this speed fluctuation can be

decreased to some degree

In Fig 9, the step response of the proposed sensorless

PMSM control algorithm is depicted when the speed reference

is changed from 20 to 100 rpm and back with 40% rating

load From these experimental results, it can be seen that the

measured back-EMF can be successfully used to obtain the

necessary position and speed information for replacing a shaft

sensor at low speed range (10 - 150 rpm) However, this

proposed sensorless method cannot be applied to the system

that has large viscous friction load at low speed operation

range, because the current decaying operation in the MCDI

test circuit may result in the mechanical resonance when high

current flows to machine In the upper speed range, over 150

rpm, the basic machine-model-based method, such as [9], may

be adopted for the sensorless drives

In Fig 10, the starting performance of the proposed scheme

is presented As described above, since the initial rotor position

can be detected through the injected test current and voltage measurement system, the direct back-EMF measuring strategy gives good starting ability to the control system As shown in the lowest trace of Fig 10(a), the test current for producing arbitrary torque is injected with the initial control angle, which

is always maintained at zero on the stationary reference frame

at the power-up instance During the test current injection cycles, the whole rotor angle estimation process previously mentioned is performed And within a sufficiently short time, the information about the initial rotor angle can be obtained Therefore, at the end of the test cycle, the machine gets

started by the proposed sensor elimination control scheme As shown in Fig 1O(c) and (d), the starting process is successfully applied, even in the same loaded case Considering that the outer load is generally given by the pattern that is proportional

to the machine speed or the square value of the machine speed, the machine starting under a general loaded condition is no more a critical problem

V CONCLUSION

A new approach to the position sensor elimination of the PMSM drives for low-speed operation is presented in this

paper It is shown that the actual rotor position, as well as

the machine speed, can be achieved strictly using the: MCDI

test cyc1e and direct measuring the back-EMF with a simp1e

measurement circuit Additionally, a relatively smooth start

[9] J S Kim and S K SUI, “New approach for high performance PMSM

drives without rotational position sensors,’’ APEC Con5 Rec., pp

381-386, 1995

[lo] J S Kim, J W Choi, and S K SUI, “Analysis and compensation

of voltage distortion by zero current clamping in voltage-fed PWM

inverter,”Yokohama ZPEC Conf Rec., pp 265-270, 1995

can be achieved by the injection of a simple test current to

the machine To avoid the steady state-speed estimation error

caused by the parameter variations, the back-EMF constant

compensation algorithm is also proposed The experimental

results show the important features of the new methold in the

low speed region (10-100 [rpm]) with acceptable accuracy

REFERENCES

[ I ] R Wu and G R Slemon, “A permanent magnet motor drive without a

shaft sensor,” ZEEE Trans Znd Applicat., vol 27, no 5, pp 1005-1011,

1991

[2] K Iizuka, H Uzuhashi, M Kano, T Endo, and K Mohri, “Microcom-

puter control for sensorless brushless motor,” ZEEE Trans Znd 4ppZicat.,

vol IA-21, no 4, pp 5 9 5 4 0 1 , 1985

[3] N Matsui and M Shigyo, “Brushless dc motor control without position

and speed sensors,” ZEEE Trans Znd Applicat., vol 28, no 1, pp

Joohn-Sheok Kim (S’92) was born in Korea in

1965 He received the B.S and M.S degrees in electrical engineering from Seoul National Univer- sity, Korea, in 1989 and 1992, respectively, where

he is working toward the Ph.D degree in the area

of power electronics

His major interests are adjustable speed ac drives and static power converter

_ _

120-127, 1992

[4] R Dhaouadi, N Mohan, and L Norum, “Design and implementation

of an extended Kalman filter for the state estimation of a permanent

magnet synchronous motor,” IEEE Trans Power Electron., vol 6, no

3, pp 491497, 1991

[5] S Bolognani, R Oboe, and M Zigliotto, “DSP-based extended Kalman

filter estimation of speed and rotor position of a pm synchronous motor,”

ZECON Conf R e c , pp 2097-2102, 1994

[6] R Krishnan and R Ghosh “Starting algorithm and performance of a pm

dc brushless motor drive system with no position sensor,” PLSC Conf

R e c , pp 815-821, 1986

[7] M Matsui and T Takeshita, “A novel starting method of censorless

salient-pole brushless motor,” ZAS Conf Rec , pp 386-392, 1994

[8] S Kondo, A Takahashi, and T Nishida, “Armature current locus based

estimation method of rotor position of permanent magnet synchronous

motor without mechanical sensor,” ZAS Conf Rec , pp 55-60 1995

Seung-Ki SUI (S’78-M’87) received the B S , M S ,

and P h D degrees in electrical engineering from Seoul National University, Korea, in 1980, 1983, and 1986, respectively

He was with the Department of Electncal and Computer Engineering, University of Wisconsin, Madison, as a Research Associate from 1986 to

1988 He joined Gold-Star Industnal Systems Com- pany as Principal Research Engineer in 1988, where

he remained until 1990 Since 1991 he has been with the Department of Electrical Engineenng, Seoul National University His present research interests are in high-performance electnc machine control using power electronics He is performing vanous research projects for industnal systems and some of the results are applied to the fields of industrial high-power electric machine control