Dynamic Motor Parameter Identification for High Speed Flux Weakening Operation

of Brushless Permanent Magnet Synchronous Machines

Uwe Schaible, Student Member, IEEE

Barna Szabados, Senior Member, IEEE

Power Research Laboratory

McMaster University
Hamilton, ON, CANADA

Power Research Laboratory

McMaster University
Hamilton, ON, CANADA

Abstract: An experimental investigation is conducted to determine the

behaviour of brushless PM synchronous machine parameters in the
high speed flux weakening operating range. Real-time operating
characteristics are determined for two commercially available 4 Nm
interior PM synchronous machines. Special computer assisted
measuring techniques are employed using an experimental vector
controlled drive, capable of providing sinusoidal excitation waveforms
at up to 8000 rpm. Perturbation analysis is performed on the measured
data and results are applied to a variation of the classical d-q axis
equivalent circuit model. The experimental results are repeatable and
show a significant variation in the machine parameters as a function of
the torque angle. Magnetic non-linearities are also identified and
accounted for through recomputation of the machine model over a wide
range of operating points. The authors demonstrate that all tests can be
performed with no previous knowledge of manufacturer proprietary
information or results from finite element simulation.

efficiency levels when compared with conventional id= 0

control [1]-[4]. Proper implementation of flux weakening
control requires the knowledge of synchronous machine
parameters. The most common parameters required for the
implementation of such advanced control algorithms are the
classical simplified model parameters: Ld - the direct axis selfinductance, Lq - the quadrature axis self- inductance, and mag the permanent magnet flux linkage. Prior knowledge of the
previously mentioned parameters, and the number of pole pairs
p, allows for the implementation of torque control through the
use of current vector control and the generalized synchronous
machine torque expression, shown in (1).

Keywords: Parameter estimation, permanent magnet machines,

current control, microcomputers, electric variables measurement.

The required parameters determine the linearized representation for the d-axis reactance Xd, q-axis reactance Xq, and magnet
excited voltage Eo. Such representation is used to provide a
linearized machine model from which control rules such as
voltage limit ellipses and maximum torque-per-ampere
trajectories can be calculated [1]. The non-linearity of certain
types of salient-pole synchronous machines has made it difficult
to apply these control rules. Previous work has shown that the
synchronous machine parameters can be highly non-linear and
can vary significantly as the machine is loaded [5]-[9]. Hence,
modelling of Ld, Lq, and mag as fixed values over the machines
entire range of speed and loading results in an inaccurate and less
effective implementation of flux weakening in advanced torque
control applications.
The research presented in this paper is a result of the difficulty
encountered by the authors in developing a high performance
vector controlled drive for the high speed operation of
commercially available brushless PM synchronous machines.
The lack of both manufacturer proprietary information and the
results of finite element simulation is overcome by conducting
special computer assisted measuring techniques on a laboratory
test setup. The conducted tests account for magnetic nonlinearities and demonstrate the ability to perform repeatable
analysis of any brushless PM synchronous machine over its
entire operating range.

I. INTRODUCTION

INUSOIDAL back-EMF brushless PM synchronous

machines are receiving much attention due to their high
speed, power density and efficiency characteristics. New rotor
configurations, and the commercial availability of high field
strength neodymium-iron-boron magnets has reduced the cost
of such machines to a level where they can now provide a
significant, yet affordable, performance improvement in many
variable speed applications. Unfortunately, manufacturers of
such commercially available machines provide very little
information for drive designers wishing to implement high
performance torque control.
Advanced high speed salient-pole synchronous machine
drives use vector control in the synchronously rotating reference
frame to actively vary the d-axis armature current as a function
of loading and speed. Such operation, commonly referred to as
flux weakening, allows for higher speed, torque, and

3
Te = 2 p [ mag iq + id iq ( Ld - Lq) ]

(1)

II. EXPERIMENTAL INVESTIGATION

A. Test Setup
Two commercially available brushless PM synchronous
machines were purchased from the same manufacturer. Both
machines were specified as having a sinusoidal back-EMF,
balanced on all three phases. A summary of the manufacturer
supplied technical specifications is provided in Table I.

TABLE I
MANUFACTURER SUPPLIED TECHNICAL SPECIFICATIONS
Parameter

Mac hine A

Machine B

Max. O perating Speed (rp m)

80 00

6000

Max. C o nt. Po w er (W)

22 90

1490

Max. S tall Torq ue (N @m)

4.1 1

3.95

Line-Line Resista nce ( S )

0.2 7

0.5 7

Line-Line Ind ucta nce (mH )

1.8 6

2.30

Back-E M F (Volts/ krp m )

24 .9

19.9

0.2 3 5

0.2 3 5

Fer riteCeramic

Neo d y miu mIro n- Bo ro n

Torq ue Se nsitivity
(N @ m/A mp)
Per ma nent Mag net Typ e

Machine B was recommended as a suitable replacement for the

now obsolete Machine A. No further control-specific information was available from the manufacturer, either due to a lack of
availability or due to its proprietary nature.
The basic block diagram of the test setup is shown in Fig. 1. A
conventional brushed dc machine is coupled, through a speed
reducer, to the Machine Under Test (MUT). Loading of the
MUT is accomplished by configuring the dc machine as a
separately excited generator and varying its field current as it
generates into a resistive load bank. A microprocessor is used to
interface with the user and allows for manual control of the
MUTs phase current vectors. The MUT operating point is set by
simultaneously varying the machine load and the current vector
command sent by the microprocessor to the vector controlled
drive. The current vector command magnitude I*mag and angle I*
are adjusted through user manipulation. Inputs to the
microprocessor include real-time information about the
machine speed (), the true rms phase voltage (Vrms), true rms
phase current (Irms), and the average power consumed (P). The
real-time information is recorded at different operating points as
the MUT is tested.
Phase current and voltage inputs to the MUT are measured
with respect to a fixed reference frame through use of a data
acquisition system. A real-time snapshot of the MUT excitation
waveforms is acquired to supplement the operating point
information recorded by the microprocessor. A shaft position
encoder is connected to the rotor of the test machine and is used
to provide the marker pulse (Mkr) from which the reference
frame is established. The position encoder is also used to
provide rotor position information to the vector controlled drive.
B. Control Methodology
Sinusoidal back-EMF is a characteristic obtained from brushless permanent magnet machines incorporating a buried or
interior permanent magnet rotor configuration [4]. This type of
rotor configuration can cause the reluctance to be greater along
the d-axis flux paths than along the q-axis flux paths, translating
into a higher q-axis inductance (L q) than d-axis inductance (Ld).
Hence, the second term in (1), also known as the reluctance
torque component, can be exploited to produce positive torque

Microprocessor

Vrms
Irms
P

I*
I* mag

Data
Acquisition
System

V
I

Mkr

Shaft Position
Encoder

,Mkr

Vector
Controlled
Drive

Machine
Under
Test

Field
Control

DC
Machine

Load
Bank

Fig. 1. Test setup for parameter identification

I s1 R s
V s1

V d1
V id1
V i1

I s1
1

1
i q1

V iq1

V q1 q-axis

d-axis
Fig. 2. Fundamental frequency phasor diagram

in addition to the main magnetizing torque, only if there is a

negative d-axis current component (id). Flux weakening
operation is possible by using vector control that produces such
positive reluctance torque through the active manipulation of the
input current phasor in the second quadrant of the synchronously
rotating d-q axis reference frame.
Machines A and B were specified as having a sinusoidal backEMF. It is assumed that the test machines have interior
permanent magnet rotors and also the saliency to cause Lq to
exceed Ld. A phasor diagram for the near rated load flux weakening operation of a brushless PM synchronous machine is
shown in Fig. 2. The angle 1 is used to represent the torque
angle of the machine. Current phasor Is1 is offset by an angle 1
from the q-axis to establish a negative d-axis current component.

4
32

80

Phase
Voltage

-16

-40

Torque (N-m)

40

Phase
Current

Phase Voltage (V)

Phase Current (A)

3
16

0
-32

40

80

120

160

200

240

280

320

-80
360

1000

2000

3000

4000

5000

6000

7000

8000

Speed (rpm)

Mechanical Degrees

Specified Cont. Torque Range

Fig. 3. Machine A measured waveforms at 7500 rpm and rated load

Tested Elec. Torque Range

Fig. 5. Machine A specified and tested torque range

5

20

90

45

Phase
Voltage
0

Phase
Current

-10

-45

Torque (N-m)

10

Phase Voltage (V)

Phase Current (A)

0
-20

40

80

120

160

200

240

280

320

-90
360

Mechanical Degrees

Fig. 4. Machine B measured waveforms at 7500 rpm and rated load

1000

2000

3000

4000

5000

6000

7000

8000

Speed (rpm)
Specified Cont. Torque Range

Tested Elec. Torque Range

Fig. 6. Machine B specified and tested torque range

Hence, a positive reluctance torque would be developed for an

interior permanent magnet machine. It is desired to operate
Machines A and B in this manner.
The phasors shown in Fig. 2 consist of only the fundamental
frequency components, denoted by subscript 1, and can be
derived from measurable input quantities to the MUT. The
effects of higher order harmonics on the MUTs torque
production can be neglected by maintaining near sinusoidal
current and voltage excitation waveforms. Conventional synchronous machine drives employ sinusoidal waveform excitation
at speeds up to 3500 rpm, after which the excitation is switched
to a simpler six step square wave scheme [10]. The authors have
constructed a 7.5 kVA vector controlled drive which is capable
of supplying low harmonic sinusoidal excitation for high speed
machine operation up to 8000 rpm.

Having established the synchronously rotating d-q axis

reference frame, the VSI is reconnected to the MUT and the dc
machine is reconfigured as a separately excited generator. The
reference current phasor magnitude I*mag and angle I* is varied to
place the actual current phasor to the MUT at desired positions
relative to the d- and q-axis. I*mag and I* can be set to produce the
Maximum Torque Conversion Efficiency (MTCE) for a specific
loading by monitoring the ratio between the operating speed and
the product between the input phase current and voltage. Tests
are conducted by establishing the MTCE operating points at
different loadings while maintaining a constant speed. A realtime snapshot of the MUTs excitation waveforms is obtained
from the data acquisition system for each MTCE operating
point. Speed, voltage, current and power information is simultaneously recorded by the microprocessor.

C. Test Procedure
Reconstruction of the MUTs phasor diagram begins with the
establishment of its q-axis. The dc machine in Fig. 1 is reconnected to operate as a motor and drive the MUT over the
desired operating range. The MUT is operated as a generator
and its open circuit terminal voltage is measured by the data
acquisition system. The fundamental component of the terminal
voltage is reconstructed from the measured data, and plotted
with respect to the fixed reference frame obtained from the
marker signal. The angular displacement of the crest of the
reconstructed sinusoid with respect to the marker is used to
establish the q-axis, and thereby the d-axis.

D. Measured Results
The real-time snapshots of the excitation current and voltage
waveforms for 7500 rpm, rated load operation of Machines A
and B are shown in Figs. 3 and 4. In both cases, the total
harmonic distortion factor of the waveforms is less than 10%.
All waveforms are measured with respect to the marker at 0
mechanical degrees. The q-axis is assumed to remain fixed with
respect to the marker, allowing the fundamental frequency
diagram to be reconstructed at any operating point.
Tests were conducted on both machines for constant speed
operation between full load and no load. The tests were repeated
at intervals of 500 rpm, spanning the high speed range of 3500

24

39

Rs

16

26

id
13

Id

-8

-13

Vd

-16

10

20

i od
Rc

Vd
icd

-26

-24
0

Vd, Vq (V)

30

40

50

(a)

-39
60

Rs

V iq

X d iod

Torque Angle (delta)

iq

Fig. 7. Machine A measured data and polynomial fits at 5500 rpm

i oq
Rc

Vq
20

+
Eo

60

icq

Vq

10

30

Iq

15

Id
Vd

-5

-15

-10
0

10

15

45

Vd, Vq (V)

15

Id, Iq (A)

+
X q i oq

Iq
Id, Iq (A)

V id

Vq

20

25

-30
30

Torque Angle (delta)

Fig. 8. Machine B measured data and polynomial fits at 5500 rpm

to 7500 rpm. Figures 5 and 6 show a comparison between the

manufacturer specified continuous torque capabilities, and the
actual electrical torque measured during the tests. The results
show that the continuous torque capabilities of Machine A have
been overstated by the manufacturer.
III. ANALYSIS OF MEASURED RESULTS
A. Phasor Diagram Reconstruction
Phasor diagram reconstruction provides the necessary data
required to identify load dependent trends in the d-q axis
representation of the excitation waveforms. Significant errors
can be developed through inaccurate determination of the Fig. 2
phasor diagram angles and . Accurate analysis is possible by
computing a Fourier analysis of the excitation waveforms over a
complete mechanical revolution. Phasor angles and are
determined by measuring the electrical angle of the peak of the
fundamental and adjusting for the q-axis. Phasor magnitudes are
determined by calculating the measured waveforms distortion
factor and adjusting the microprocessor measured true rms value
for that waveform. Reconstruction of Fig. 2 is completed by
using an empirical relationship to calculate the value of the
frequency and temperature dependent stator resistance Rs.
The d-q axis transformed excitation waveform magnitudes
are plotted against their calculated electrical torque angle . A
relationship between the excitation inputs and the corresponding
electrical torque angle is established by fitting a polynomial
through the measured points. The polynomial fits resulting from
the measured data of a 5500 rpm loading test on each machine
are shown in Figs. 7 and 8. All of the polynomial fits are
repeatable to within the 5% measurement error inherent with

(b)

Fig. 10. (a) d-axis equivalent circuit (b) q-axis equivalent circuit

this investigation.
Machine A has a discontinuity at 35. The discontinuity
is detected through a sudden change in the trend of I* at the
MTCE. The discontinuity characteristic shown in Fig. 7 is
repeatable and is believed to be caused by a change in the
reluctance of the flux path in the machine. As the machine
undergoes increased flux weakening at higher speeds and
loadings, certain critical sections along the flux path come out of
saturation. The torque angle at the discontinuity shifts from a
higher angle of 48 to a lower angle of <20 as the speed
setpoint is increased from 4000 to 7500 rpm. No discontinuities
are observed at 3500, 4000 and 7500 rpm. The lack of discontinuites at these speeds indicates that for the tested load
range, the critical flux path sections remain unsaturated at 7500
rpm, and fully saturated at speeds 4000 rpm. Machine B has no
discontinuities over the entire operating range. The difference
between the two machines is most likely caused by the use of
higher field strength neodymium magnets in Machine B.
Machine Bs critical flux path sections remain unaltered, even
when it is subjected to higher levels of flux weakening.
B. Data Modelling
The general machine model for a brushless PM synchronous
machine consists of only a highly simplified representation of
the actual operation of the machine. The static parameter model
assumes lossless operation without the effects of magnetic
saturation and non-linearities. Acceptable results are possible
when using this model at lower speeds and over a narrow speed
range. The effects of core losses and non-linear behaviour do
become more prevalent at higher speeds and over a wide speed
range.
Core losses must be included when modelling the higher
frequency PM synchronous machine operation. Fig. 10 shows
the d- and q-axis equivalent circuits used to model high speed
operation. The model includes copper losses which are
represented by resistance Rs. Core losses are represented by
including a loss component proportional to the machines
internal voltage Vi [3][6]. A resistance Rc is included for this

24

18

32

12

Xd

Xd ; Xq (ohms)

Eo (V)

Eo

24

16

Xq

Xq

Xd

0
10

20

30

40

50

0
60

10

15

Fig. 11. Machine A computed parameters at 3500 rpm

32

24

48

16

Xq
Xq

Xd

0
20

30

40

50

Xd ; Xq (ohms)

Xd
2

Eo (V)

Xd ; Xq (ohms)

Eo
Eo

10

0
60

36

24

Xq
1

Xd

0
0

Torque Angle (delta)

10

48

24

Xd
1

12

0
40

50

Xd ; Xq (ohms)

Eo (V)

Xd ; Xq (ohms)

36

Xq

30

0
30

64

0
60

48

Xq

32

16

Xd

0
0

Torque Angle (delta)

10

15

20

25

0
30

Torque Angle (delta)

Fig. 13. Machine A computed parameters at 7500 rpm

Fig. 16. Machine B computed parameters at 7500 rpm

3500 rpm
Torque (N-m)

3500 rpm

Torque (N-m)

25

Eo

20

20

Fig. 15. Machine B computed parameters at 5500 rpm

Eo

10

15

12

Torque Angle (delta)

Fig. 12. Machine A computed parameters at 5500 rpm

0
30

Fig. 14. Machine B computed parameters at 3500 rpm

Eo
3

25

Torque Angle (delta)

Torque Angle (delta)

20

7500 rpm
5500 rpm

5500 rpm

7500 rpm
1

0
0

10

20

30

40

Eo (V)

50

60

Torque Angle (delta)

Fig. 17. Machine A calculated electrical torque at MTCE

10

15

20

25

30

Torque Angle (delta)

Fig. 18. Machine B calculated electrical torque at MTCE

Eo (V)

Xd ; Xq (ohms)

Eo
3

Eo (V)

investigation. All vector controllable losses are accounted for in

this model, however, the model does not account for stray load
losses such as those incurred by circulating flux in the machines
back-EMF harmonics.
C. Parameter Identification
Evaluation of the general torque expression, given by (1),
requires the knowledge of the unknown parameters Xd, Xq, and
Eo. Mathematical analysis of the Fig. 10 d- and q- axis
equivalent circuits yields (2) and (3).

D. Dynamic Parameter Utilization

The measured parameters show that the manufacturer
supplied data in Table I is insufficient for high performance flux
weakening operation of the test machines. The variation of the
machine parameters indicates the need to re-write (1) to account
for the actual operating conditions of the brushless PM
synchronous machine. A dynamic parameter equivalent of (1) is
given by (6), as derived from the Fig. 10 equivalent circuits.

iq Xq Rs XqVq
Vd id Rs Xq iq

(2)

Te (, ) =

1
1
Vq ioq Rs + Xd iod
K
K

(3)

where:

Rc =

Eo =

observed for both machines.

ioq =
where:
K = 1+

3 p
[Eo(, ) ioq
2

Rs
Rc

Vq Rs iq

Rc

ioq = iq

Vd Rs id

Rc

iod = id

The desired unknown parameters can be calculated by

assuming that the parameters remain constant for small changes
in the operating point. Polynomial fits, such as those in Figs. 7
and 8, are used to predict a new operating point at 2 by
perturbing the torque angle 1 with a small perturbation angle
0.1 [8][9]. Setting Xd=Xd1 =Xd2, Xq=Xq1=Xq2, and Eo=Eo1=Eo2
over a small change in yields the desired parameters through
(4) and (5), with back substitution into (2) and (3).
Vd 1 A Vd 2 Rs (id 1 A id 2 )
iq1 A iq 2

where:
A=

Xd =

( Xd (,) - Xq (,))] (6)

Rc(, ) iq - Xd (, ) id - Eo(, )
Rc(, ) + Rc (, ) 1 Xd (, ) Xq(, )

iod = id +

Xq =

+ ido iqo

(4)

Vq1 iq 1 Rs
Vq 2 iq 2 Rs

Vq1 Vq 2 + Rs (ioq 2 ioq1)

K (iod 2 iod 1)

(5)

Perturbation analysis results are shown in Figs. 11 thru 16 for

3500, 5500, and 7500 rpm operation. The characteristics
indicate that the q-axis reactance (Xq) remains almost linear and
constant with torque angle for both machines. The magnet
exited voltage (Eo) varies considerably for Machine A, yet
remains approximately linear and constant with torque angle for
Machine B. The difference is mainly due to less fluctuation in
the flux path reluctance of Machine B for variations in speed and
loading. A wide variation in the d-axis reactance (Xd) can be

Rc(, ) iq - Xd (, ) id - Eo(, )
Rc(, ) 2 Xq (, ) 1 + Xd (, )

Equation (6) is used to calculate the dynamic parameter

electrical torque for the two machines, as shown in Figs. 17 and
18. The Te(,) characteristic can be used to form a look-up
table in a high performance MTCE control algorithm.
Simplification of (6) can be made for machines such as Machine
B, where Xq and Eo can be approximated as constant values with
changes in torque angle . The results clearly indicate that this
simplification cannot be extended to Xd.
IV. CONCLUSIONS
The preceding investigation has successfully demonstrated
the ease and repeatability of the proposed parameter identification method. All tests can be performed using only the input
excitation waveforms to the test machine, without results from
finite element simulation, and without previous knowledge of
manufacturer proprietary information. The test results clearly
indicate a significant variation in the control parameters required
for high speed flux weakening operation of brushless permanent
magnet synchronous machines.
Incomplete manufacturer supplied data can be supplemented
with the results from this investigation. The proposed parameter
identification method can be used by machine drive designers to
collect data for use in high performance flux weakening vector
control algorithms. Flux weakening control schemes that
assume a constant Ld, fail to properly utilize the reluctance
torque component of salient pole synchronous machines.
Dynamic characterization of the control parameters enables full
utilization of the reluctance torque. Such utilization provides the
benefits of higher torque and efficiency over the entire specified
speed and torque operating range. It also enables machine
designers to identify previously immeasurable performance
characteristics and optimize their designs for specific operating
torque regions.

V. REFERENCES
[1]

T.M. Jahns, Motion Control with Permanent-Magnet AC Machines,

Proc. IEEE, vol.82, no. 8, August 1994, pp. 1241-1252.
[2] S.R. MacMinn, and T.M. Jahns, Control Techniques for Improved HighSpeed Performance of Interior PM Synchronous Motor Drives, IEEE
Trans. Industry Applications, vol. 27, no. 5, September 1991, pp. 9971004, .
[3] S. Morimoto, Y. Tong, Y. Takeda, and T. Hirasa, Loss Minimization
Control of Permanent Magnet Synchronous Motor Drives, IEEE Trans.
Industry Applications, vol. 41, no. 5, October 1995, pp. 511-517.
[4] T.M. Jahns, G.B. Kliman, and T.W. Neumann, Interior Permanent
Magnet Synchronous Motors for Adjustable-Speed Drives, IEEE Trans.
Industry Applications, vol. IA-22, no. 4, July 1986, pp. 738-747.
[5] V.B. Honsinger, The Fields and Parameters of Interior Type AC
Permanent Magnet Machines, IEEE Trans. Power Apparatus and
Systems, vol. 101, no. 4, July 1986, pp. 867-875.
[6] A. Consoli, and A. Raciti, Analysis of Permanent Magnet Synchronous
Motors, IEEE Trans. Industrial Electronics, vol. 27, no. 2, March 1991,
pp. 350-354.
[7] M.A. Rahman, and P. Zhou, Analysis of Brushless Permanent Magnet
Synchronous Motors, IEEE Trans. Industrial Electronics, vol. 43, no. 2,
April 1995, pp. 256-267.
[8] M.A. Rahman, and P. Zhou, Accurate Determination of Permanent
Magnet Motor Parameters by Digital Torque Angle Measurement,
Journal of Applied Physics, vol. 76, no. 10, November 1994, pp. 68686870.
[9] P. Zhou, M.A. Rahman, and M.A. Jabbar, Field Circuit Analysis of
Permanent Magnet Synchronous Motors, IEEE Trans. on Magnetics,
vol. 30, no. 4, July 1994, pp. 1350-1358.
[10] B.K. Bose, A High Performance Inverter-Fed Drive System of an Interior Permanent Magnet Synchronous Machine, IEEE Trans. Industry
Applications, vol. 24, no. 6, November 1990, pp. 987-997.

VI. BIOGRAPHIES
Uwe Schaible (S91) received the B.Eng. &
Mgt., M.Eng., and Ph.D. degrees from
McMaster University, Hamilton, Ontario,
Canada, in 1992, 1994, and 1997, respectively.
He is currently conducting research at the
McMasters Power Research Laboratory in the
area of solid state variable frequency drives for
high speed machines,