You are on page 1of 141

Introduction to Vector Control of Induction

Machines
1
R.E. Betz
School of Electrical Engineering and Computer Science
University of Newcastle, Australia
email: Robert.Betz@newcastle.edu.au
April, 2012
1
First created: July, 1997.
Revised on 2012-04-29 15:37:10 +1000 (Sun, 29 Apr 2012) by R.E. Betz
SVN Version: 505
Contents
List of Figures iii
List of Figures iv
List of Tables v
List of Tables vi
Preface vii
1 Fundamentals 1-1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1
1.2 The Sinusoidal Assumption . . . . . . . . . . . . . . . . . . . . . 1-3
1.2.1 Winding Interaction with Spatial Flux Density Distribution 1-4
1.2.2 Winding Interaction with Temporal Flux Density Variation 1-7
1.3 Flux Linkage to Sinusoidally Distributed Windings . . . . . . . . 1-10
1.4 Other Important Assumptions . . . . . . . . . . . . . . . . . . . . 1-11
1.4.1 Properties of Three Phase Sinusoidal Windings . . . . . . 1-12
1.5 Torque Expressions . . . . . . . . . . . . . . . . . . . . . . . . . . 1-13
1.5.1 Torque of a Simple Reluctance Machine . . . . . . . . . . 1-14
1.5.1.1 Slow Rotor Movement . . . . . . . . . . . . . . . 1-16
1.5.1.2 Instantaneous Rotor Movement . . . . . . . . . . 1-19
1.5.1.3 Real Movement . . . . . . . . . . . . . . . . . . 1-20
1.5.2 Linear Torque Model . . . . . . . . . . . . . . . . . . . . . 1-23
1.5.2.1 Singly Excited System . . . . . . . . . . . . . . . 1-23
1.5.2.2 Doubly Excited System . . . . . . . . . . . . . . 1-25
1.5.3 The Ellipse Diagram for co-energy . . . . . . . . . . . . . 1-29
2 The Kron Primitive Machine 2-1
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
2.2 Model for the Doubly Fed Machine . . . . . . . . . . . . . . . . . 2-1
2.2.1 Zero saliency case . . . . . . . . . . . . . . . . . . . . . . 2-3
2.2.2 One degree of saliency case . . . . . . . . . . . . . . . . . 2-3
2.2.3 Torque expression . . . . . . . . . . . . . . . . . . . . . . 2-3
2.3 Commutator Machines . . . . . . . . . . . . . . . . . . . . . . . . 2-4
2.4 The Primitive Machine Concept . . . . . . . . . . . . . . . . . . . 2-7
2.4.1 Use of the Primitive Machine - the DC Machine . . . . . 2-11
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-13
CONTENTS ii
3 Frame transformations, DQ and Space Vector Models 3-1
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
3.2 dq Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
3.2.1 Stationary Frame Transformations . . . . . . . . . . . . . 3-2
3.2.1.1 MMF transformations . . . . . . . . . . . . . . . 3-2
3.2.1.2 Current Transformations . . . . . . . . . . . . . 3-5
3.2.1.3 Voltage Transformations . . . . . . . . . . . . . 3-6
3.2.1.4 Impedance Transformations . . . . . . . . . . . . 3-7
3.2.1.5 Flux Linkage Transformations . . . . . . . . . . 3-8
3.2.2 Rotating Frame Transformations . . . . . . . . . . . . . . 3-8
3.2.3 Example: SYNCREL Linear dq Model . . . . . . . . . . . 3-11
3.3 Space Vector Model . . . . . . . . . . . . . . . . . . . . . . . . . 3-16
3.3.1 Current Space Vectors . . . . . . . . . . . . . . . . . . . . 3-16
3.3.1.1 Stationary Frame Current Vectors . . . . . . . . 3-16
3.3.1.2 Rotating Frame Current Vectors . . . . . . . . . 3-20
3.3.2 Flux Linkage Space Vector . . . . . . . . . . . . . . . . . 3-21
3.3.3 Voltage Space Vector . . . . . . . . . . . . . . . . . . . . . 3-23
3.3.4 Example: SYNCREL Space Vector Model . . . . . . . . . 3-24
3.3.5 Space Vector Power Expression . . . . . . . . . . . . . . . 3-25
3.3.6 Example: Space Vector Expression for SYNCREL Torque 3-26
3.3.7 Relationship Between Space Vectors and dq Models . . . 3-28
3.4 Steady State Model . . . . . . . . . . . . . . . . . . . . . . . . . . 3-29
4 Vector Control of Induction Machines 4-1
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
4.2 Vector Models for Induction Machines . . . . . . . . . . . . . . . 4-1
4.2.1 Flux Linkage Expression . . . . . . . . . . . . . . . . . . . 4-1
4.2.2 Magnetising Current . . . . . . . . . . . . . . . . . . . . . 4-4
4.2.3 Power and Torque Expressions . . . . . . . . . . . . . . . 4-4
4.2.4 The Space Vector Model of the Induction Machine . . . . 4-8
4.3 A Heuristic Explanation of Vector Control . . . . . . . . . . . . . 4-11
4.4 Special Reference Frames . . . . . . . . . . . . . . . . . . . . . . 4-13
4.4.1 The Magnetising Flux Linkage Reference Frame . . . . . 4-13
4.4.2 The Rotor Flux Linkage Reference Frame . . . . . . . . . 4-15
4.4.3 Stator Flux Linkage Reference Frame . . . . . . . . . . . 4-17
4.5 Derivation of Rotor Flux Oriented Vector Control . . . . . . . . . 4-18
4.6 Structure of a Rotor Oriented Vector Drive . . . . . . . . . . . . 4-22
4.6.1 Indirect Rotor ux Oriented Controller . . . . . . . . . . 4-22
4.6.2 Direct Rotor Flux Oriented Controller . . . . . . . . . . . 4-23
4.7 Magnetising Flux Orientation . . . . . . . . . . . . . . . . . . . . 4-24
A Calculation of Inductances for Salient Pole Machines A-1
A.1 Calculation of Inductances . . . . . . . . . . . . . . . . . . . . . . A-1
A.1.1 Self Inductances . . . . . . . . . . . . . . . . . . . . . . . A-3
A.1.2 Mutual Inductances . . . . . . . . . . . . . . . . . . . . . A-10
A.1.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . A-13
CONTENTS iii
B Winding Functions B-1
B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-1
B.2 Ideal Sinusoidal Winding . . . . . . . . . . . . . . . . . . . . . . . B-2
B.2.1 Conventional inductance calculation . . . . . . . . . . . . B-2
B.2.2 Alternative inductance calculation . . . . . . . . . . . . . B-5
B.3 Non-sinusoidal winding . . . . . . . . . . . . . . . . . . . . . . . . B-8
B.3.1 Inductance Using Basic Principles . . . . . . . . . . . . . B-8
B.3.2 Inductance Using Winding Functions . . . . . . . . . . . . B-11
B.4 Flux Linkage Expression . . . . . . . . . . . . . . . . . . . . . . . B-12
B.5 A Note on Winding Functions for Multi-pole Machines . . . . . . B-13
B.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-14
Bibliography B-15
List of Figures
1.1 MMF calculation integration path. . . . . . . . . . . . . . . . . . 1-4
1.2 Dimensions of a single coil. . . . . . . . . . . . . . . . . . . . . . 1-8
1.3 Conceptual diagram of sinusoidally distributed three phase wind-
ings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-12
1.4 Simple singly excited reluctance machine. . . . . . . . . . . . . . 1-15
1.5 Flux plots for static movement. . . . . . . . . . . . . . . . . . . . 1-17
1.6 Incremental energy change with small movement of the rotor. . 1-18
1.7 Area representing mechanical output energy. . . . . . . . . . . . 1-19
1.8 Energy with instantaneous movement. . . . . . . . . . . . . . . . 1-20
1.9 Flux versus current trajectory for typical real movement. . . . . 1-21
1.10 Flux versus current for linear magnetic material. . . . . . . . . . 1-24
1.11 Doubly excited reluctance machine. . . . . . . . . . . . . . . . . 1-26
1.12 Self and mutual inductance variation with rotor angle. . . . . . . 1-30
1.13 Flux linkage for the a-phase . . . . . . . . . . . . . . . . . . . . . 1-31
1.14 Total, self and mutual ux versus current loci. . . . . . . . . . . . 1-32
1.15 Segments used for ellipse area. . . . . . . . . . . . . . . . . . . . 1-33
1.16 Co-energy ellipses for a saturated SYNCREL . . . . . . . . . . 1-35
2.1 Magnetic circuit conceptual diagram of a double fed machine . . 2-2
2.2 Torque plot for the double fed machine with DC rotor and stator
currents. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
2.3 A two pole commutator machine . . . . . . . . . . . . . . . . . . 2-5
2.4 Equivalent circuit of a DC machine in motoring mode . . . . . . 2-7
2.5 Primitive dq machine . . . . . . . . . . . . . . . . . . . . . . . . . 2-8
2.6 Separately excited DC machine with a compensating winding . . 2-12
3.1 Three phase to two phase transformation . . . . . . . . . . . . . 3-2
3.2 Two phase stationary to two phase rotating frame transformation 3-10
3.3 Conceptual diagram of a three phase SYNCREL . . . . . . . . . 3-12
3.4 Model for the ideal dq equations . . . . . . . . . . . . . . . . . . 3-14
3.5 Resolving the current space vector onto the abc axes . . . . . . . 3-18
3.6 Relationship between the dq axes and the current space vector. . 3-19
3.7 Space vector rotating frame transformations . . . . . . . . . . . . 3-20
3.8 Phasor diagram for a steady state SYNCREL . . . . . . . . . . . 3-31
4.1 Conceptual diagram of an induction machine. . . . . . . . . . . . 4-2
4.2 Relationship between stationary and rotating frames and the gen-
eral reference frame for the induction machine . . . . . . . . . . . 4-9
LIST OF FIGURES v
4.3 Conceptual diagram of an induction machine with quadrature-
phase stator windings. . . . . . . . . . . . . . . . . . . . . . . . . 4-11
4.4 Space vectors in quadrature induction machine at time t
+
o
. . . . . 4-12
4.5 Space vectors in quadrature induction machine at t > t
o
. . . . . . 4-13
4.6 Position of the space vectors after the stator has been rotated. . 4-14
4.7 Relationship between the dq frame and the special xy frame. . . 4-14
4.8 Relationship between various space phasors in the stator and ro-
tor ux linkage reference frames . . . . . . . . . . . . . . . . . . . 4-16
4.9 Relationship between the stationary reference frame and the spe-
cial reference frame xed to the stator ux linkage space phasor. 4-17
4.10 Block diagram of a indirect rotor ux vector oriented control scheme4-23
4.11 Block diagram of a direct rotor ux eld oriented vector controller.4-24
4.12 Flux model in a rotor ux reference frame . . . . . . . . . . . . . 25
A.1 Two pole three phase SYNCREL conceptual diagram . . . . . . A-2
A.2 Developed diagram of a SYNCREL. . . . . . . . . . . . . . . . . A-5
A.3 d axis developed diagram for SYNCREL . . . . . . . . . . . . . . A-6
A.4 a phase inductance plot. . . . . . . . . . . . . . . . . . . . . . . A-11
B.1 Two pole sinusoidal winding layout . . . . . . . . . . . . . . . . . B-2
B.2 Calculation of the MMF for a sinusoidal winding. . . . . . . . . . B-4
B.3 Dierent methods of calculating the ux linking a coil. . . . . . . B-6
B.4 Cumulative number of turns for a sinusoidally distributed winding.B-6
B.5 Turns function and mmf distribution for two fractional pitch
windings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-9
B.6 Physical layout of the non-sinusoidal winding. . . . . . . . . . . . B-9
List of Tables
3.1 Summary of Stationary Frame Transformations . . . . . . . . . . 3-8
3.2 Summary of Rotating Frame Transformations . . . . . . . . . . . 3-11
Preface
These notes contain material for the rst half of Elec414 - Energy Flow Control.
Their objective is to give students a basic understanding of vector controlled
induction machines. The approach taken is to start o from rst principles and
develop the basic expressions for a machine. Next the primitive dq machine
is developed and this is applied to the DC machine and the synchronous re-
luctance machine. The next stage is to develop stationary and rotating frame
transformations for AC machines, and then show that these machines can also
be developed as dq machines.
The last part of the notes introduces space vector representations for ma-
chines, and shows how they are related to dq representations. These models are
then used to develop various forms of vector control for the induction machine.
Robert E. Betz Newcastle, Australia, 1997
Chapter 1
Fundamentals
1.1 Introduction
This subject provides an introduction to Generalised Machine Theory and the
dynamics of electrical machines. This area has become increasingly important
due to the prevalence of variable speed drive systems in industry. As industry
continues to modernise the penetration of AC variable speed drives into ap-
plications which were previously considered to be the exclusive domain of DC
machines is continuing. Consequently an understanding of how these drive sys-
tems work and what their capabilities are is becoming increasingly important.
Variable speed drive systems rst began to appear in the late 1960s early
1970s. During the early years many of these drive systems were based on chop-
per fed DC machines, this being especially true for applications that required
good transient performance. AC machine drive systems were also available
around this time, but generally these were limited to low performance applica-
tions where the transient performance was of little importance. The other issue
that tended to limit the application of AC variable speed drives during this era
was the reliability of the systems. The dreaded shoot through problem in the
power electronics was a frequent occurrence.
Whilst reasonably high performance drives were developed based on DC
machine technology they still had the problem that the machine was a DC ma-
chine. These machines are inherently costly, they require a lot of maintenance,
and they are less reliable than other machine types (mainly due to the presence
of the commutator and brushes). Therefore there was a motivation to develop
drive systems based on AC machines, and in particular the induction machine.
This machine is extremely reliable and low cost. If has often been said that the
induction machine will continue to work even if an axe has been sunk into the
rotor.
Two problems initially held back the development of AC induction machine
variable speed drive systems:
the previously mentioned lack of reliability in the power electronics.
the lack of control capable of giving good performance from the induction
machine.
Now let us examine these in more detail. The early development of variable
1.1 Introduction 1-2
speed drive systems was dogged by the poor reliability of the power electronics.
This was largely to do with the fragility and diculty of use of the early power
devices used. Consequently variable speed drives got a bad reputation amongst
many industrial people, which took more than 10 years to overcome. Virtually
all of these initial problems have been overcome with the latest power electronic
devices. Also one can not underestimate the inuence that the microprocessor
has had on the development of drives systems. The availability of low cost
computation has allowed cost eective implementation of sophisticated control
algorithms for drive systems.
The other factor that limited the development of AC drives was the control.
The control for high performance DC drives was easy since this machine struc-
ture was very easy to control. In fact with a separately excited DC machine the
system dynamics were a rst order linear dierential equation. However, the AC
machine was a very complex fth order non-linear system. For this reason early
AC drives were designed to operate at steady state for most of the time, the
variable speed capability only being used to move to a dierent set point. The
development of the concept of vector control and the early to 1970s overcame
these limitations, and similar performance to the DC machine can be obtained
from the induction machine. Research into various aspects of vector control has
been active ever since this time.
This set of notes will gradually work its way towards the development of
vector control. The material to presented will follow the outline below:
1. Examination of the fundamental assumptions underlying the development
of the dynamic models.
2. Development of the rotating eld concept.
3. Introduction to the concept of the Generalised Machine
4. Detailed development of the Generalised Machine.
5. Development of the conversions between the three phase machine and the
generalised machine.
6. Detailed development of the concept of space vectors.
7. Development of DQ and space vector models for several dierent types of
machines.
8. Concepts of eld orientation.
9. Vector control of the synchronous and induction machine.
10. Introduction of load dynamics.
11. Hardware and software structure for variable speed drive systems.
Modelling of machines is very complex due to their highly non-linear nature
and the diculty in obtaining analytical models that reect the underlying
physics. Consequently all the modelling approaches used involve assumptions.
It is important to understand what these assumptions are, and how they aect
the validity of the models produced. We will not con-
sider all the as-
sumptions here, but
will concentrate on
the main ones.
The next section will examine one of the fundamental assumptions of ma-
chine modelling, the sinusoidal winding distribution assumption, and consider
its implications.
1.2 The Sinusoidal Assumption 1-3
1.2 The Sinusoidal Assumption
One of the main assumptions that is used in the modelling of many types of AC
machines is the sinusoidal assumption. Essentially the assumption is that the
windings in the machine are arranged so that the resultant mmf has a spatially
sinusoidal distribution. A normal AC machine usually has three windings spaced
at 120

electrical, each producing a spatially sinusoidal mmf when fed with a


current. An amazing property of this arrangement is that if it is fed with three
temporal sinusoidal currents, separated temporally by 120

, then the resultant


mmf is a spatially moving sine wave around the machine (this will be shown
mathematically in later sections in this Chapter).
Why is this assumption so important? From a modelling point of view
the sinusoidal functions have a rich set of mathematical properties which make
the modelling of machines analytically tractable. One of the key properties of
sinusoidal functions is their connection with vectors, and the consequent ability
to take orthogonal components of them.
In reality the mmf produced by real windings are not pure sinusoids. Most
windings for real machines are conned to slots in the stator. This leads to an
mmf that has step changes in it, and consequent higher order spatial harmonics.
However, the winding conguration is designed to minimize these harmonics. As
we shall see below, the signicance of these winding harmonics on the perfor-
mance of the machine also depends on the harmonics in the ux waveforms that
interact with the windings.
The sinusoidal assumption is not only applied to the mmf produced by the
windings, but it is also applied to the resultant uxes produced by the action of
the mmf on the iron circuit of the machine. In the case of the SYNCREL the
iron circuit reluctance varies in a complex fashion due to the rotor saliency. This
means that the ux density produced by the mmf is in general not spatially si-
nusoidal. However, the harmonics in these waveforms are usually neglected, and
only the fundamental component is considered from an analysis point of view.
This may seem to be a gross approximation, but models developed using this
approach have been shown to give reasonable representations of the behaviour
of real machines.
In the remainder of this section we shall look at some of the properties
of a sinusoidally distributed winding. Specically, the characteristics of a non-
sinusoidal ux density interacting with a sinusoidal winding shall be considered.
This is of particular relevance to the SYNCREL and other salient pole machines
as their ux density distributions in general are not sinusoidal.
The remainder of this section discusses the foundations of the sinusoidal
assumption, and why it can be used successfully to simplify the modelling of
machines, with special emphasis on the SYNCREL. Specic issues addressed
are:
Consideration of some of the general properties of sinusoidally distributed
windings (e.g. only link with elds of the same pole number).
Detailed analysis of the variation of inductance with rotor position for a
two pole axially laminated rotor would be benecial.
1.2 The Sinusoidal Assumption 1-4
Figure 1.1: MMF calculation integration path.
1.2.1 Winding Interaction with Spatial Flux Density Dis-
tribution
In this sub-section we shall consider the interaction of a spatially non-sinusoidal
ux density distribution with an ideal sinusoidal winding. Such an ideal wind-
ing will produce a temporal sinusoidally varying current density around the
machine. The following equation can be written for the conductor density as a
function of the angle
p
around the periphery of the machine:
n(
p
) = n
a
sin
p
(1.1)
This waveform has an amplitude of n
a
conductors, and goes positive and neg-
ative. How can one have positive and negative numbers of conductors? The
sign convention is based on the direction of the current in the conductor. The
positive part of this conductor distribution carry currents in one direction, and
the negative part carry the return currents [4].
Given this winding distribution, the mmf spatial distribution readily follows.
If the a-phase is carrying i
a
amps, then the mmf can be calculated by imple-
menting Amperes Law. This is achieved by carrying out a closed path integral
1.2 The Sinusoidal Assumption 1-5
over the full coil span (see Figure 1.1 for the path of integration):
F
aT
(
p
) =
_

p
+

p
n
a
i
a
sin
p
d
p
= 2n
a
i
a
cos
p
= 2

F
a
cos
p
F
a
(
p
) =

F
a
cos
p
where

F
a
= n
a
i
a
(1.2)
The 2 factor in the front of the right hand side of the above expression is
there because the total mmf is expended across two air gaps, and the F
a
(
p
)
expression represents the mmf expended per air gap.
The total number of coils in the winding is simply the sum of the number
of coils at each
p
position. Due to the continuous nature of the proposed
distribution this sum becomes an integral:
N
a
=
_

0
n
a
sin
p
d
p
= 2n
a
(1.3)
Therefore the peak mmf for the winding may be written as:

F
a
=
N
a
i
a
2
=

N
a
i
a
= n
a
i
a
Now let us consider some general ux density waveform that varies in the
following way spatially with respect to
p
around the machine, and also has a
time varying spatial phase angle (t):
B(
p
) =

B
n
sin n(
p
(t)) (1.4)
This ux waveform is a non-sinusoidal waveform as it contains a number of
harmonics denoted by the integer value of n.
Furthermore assume that the winding is on a machine with the following
physical dimensions:
l the length of the machine.
v
B
the linear velocity of the B eld.
r the radius of the stator of the machine.
Therefore the B(
p
) eld phase is changing in the following fashion:
(t) =
v
B
t
r
(1.5)
This expression implies that the B(
p
) eld is spatially moving with respect to
time.
From basic physics we can say the following the voltage induced in a length
of conductor l, moving with a velocity of v
B
perpendicular to a magnetic ux
density of B is:
e = Blv
B
(1.6)
1.2 The Sinusoidal Assumption 1-6
In the case of a sinusoidally distributed coil the length of conductor for one
side of the coil at some position
p
is:
l
T
= n
a
l sin
p
(1.7)
therefore the induced voltage in the conductors at this angular position is:
e(
p
) = Bv
B
n
a
l sin
p
(1.8)
The ux density at this position at a particular instant of time can be determined
from (1.4), and consequently (1.8) becomes:
e(
p
) = v
B
n
a
l

B
n
sin
p
sin n(
p
(t)) (1.9)
To simplify the following manipulations let K
n
v
B
n
a
l

B
n
. In order to cal-
culate the total voltage produced by these conductors we have to add up the
contributions of all the conductors in the coil. This involves integrating the
voltage at each position
p
for the circumference of the machine. Therefore
assuming a single pole pair machine we have:
e
T
=
_
2
0
K
n
sin
p
sin n(
p

v
B
t
r
)d
p
(1.10)
Using the trigonometric relation sin xsin y = 1/2[sin(x+y) +sin(xy)] one can
write:
e
T
=
_
2
0
K
n
2
_
sin
_
(n + 1)
p

nv
B
t
r
_
+ sin
_
(n 1)
p

nv
B
t
r
__
d
p
(1.11)
For the specic case of n = 1 (i.e. only the fundamental harmonic present)
then (1.11) can be integrated and becomes:
e
T
= K
1
sin
_
v
B
t
r
_
= K
1
sin(
B
t) (1.12)
i.e. the voltage induced by the winding is a temporal sinusoidal voltage (as
expected).
Now consider what happens to the higher order harmonics in the ux density
waveform. If we carry out the integration of (1.11) for the case of n > 1 we
have:
e
T
=
K
n
2
_
cos
n + 1
_
2(n + 1)
nv
B
t
r
_

cos
n 1
_
2(n 1)
nv
B
t
r
_
+
_
1
n + 1
+
1
n 1
_
cos
_
nv
B
t
r
__
(1.13)
If we consider the various terms in (1.13) using the trigonometric relation:
cos(x y) = cos xcos y + sin xsin y (1.14)
1.2 The Sinusoidal Assumption 1-7
we get the following:

cos
n + 1
_
2(n + 1)
nv
B
t
r
_
=
1
n + 1
_
cos 2(n + 1) cos
nv
B
t
r
+ sin 2(n + 1) sin
nv
B
t
r
_
=
1
n + 1
cos
nv
B
t
r
(1.15)
and similarly:

cos
n 1
_
2(n 1)
nv
B
t
r
_
=
1
n 1
cos
nv
B
t
r
(1.16)
Therefore (1.13) can be written as:
e
T
=
K
n
2
_

_
1
n + 1
+
1
n 1
_
cos
nv
B
t
r
+
_
1
n + 1
+
1
n 1
_
cos
nv
B
t
r
_
= 0; n > 1 (1.17)
Remark 1.1 The implications of the above expression are that the higher order
harmonics in the ux density spatial waveform do not link to the sinusoidally
distributed winding. In other words the pole number of the ux density waveform
has to be the same as that of the winding. This is a very important property
of sinusoidal windings. One can then consider the ux density harmonics to be
contributing to the leakage ux.
Remark 1.2 Real machine windings are not exactly sinusoidally distributed as
in the ideal case above. Therefore there are spatial harmonics in the winding
distribution itself. Consequently it is possible for higher order harmonics in the
ux density waveform to link with same pole number harmonic in the winding
distribution, resulting in a harmonic voltage. For example, most winding con-
gurations contain a signicant third harmonic spatial component, therefore the
third harmonic in the ux density waveform (introduced by saturation eects)
can link with the individual windings. Consequently third harmonic voltages can
be seen in the phase voltages.
1.2.2 Winding Interaction with Temporal Flux Density
Variation
In this section we consider a non-sinusoidal, spatially stationary ux density
distribution which has a sinusoidal temporal variation, interacting with a sinu-
soidal winding distribution. For the sake of the following argument consider the
ux density to have the following form:
B(
p
) =

B
n
cos n
p
(1.18)
Let us rstly consider the n = 1 case. Consider a single coil which has the
dimensions shown in Figure 1.2. One can calculate the ux linking a coil at any
position using the general expression:
=
_
B.dS (1.19)
1.2 The Sinusoidal Assumption 1-8
Figure 1.2: Dimensions of a single coil.
Consider the situation where there is only the fundamental ux density dis-
tribution. The above surface integral can be written as follows (using Figure 1.2)
for the ux at angle coil position
p
1
:
(
p
) =

B
1
r
_

p
+

p
_
l
0
cos dl d
=

B
1
r
_

p
+

p
l cos d
= 2

B
1
rl sin
p
(1.20)
The dot product is eliminated in this situation as the ux density is perpendic-
ular to the integration surface.
In order to get the voltage induced in the coils at a particular position around
the machine the following calculation has to be carried out:
e(
p
) = n(
p
)
d(
p
)
dt
(1.21)
= n
a
sin
p
d
dt
_
2

B
1
rl sin
p
_
=
2n
a
rlB
1
sin
2

cos t (1.22)
Remark 1.3 Equation (1.22) is obtained by realising that we are dealing with
a sinusoidal temporal variation of a sinusoidal spatial distribution. Therefore
the amplitude of the ux density is varying with respect to time in a sinusoidal
manner. Therefore:

B
1
= B
1
sin t (1.23)
1
Note that the
p
in the following expression is the angle of the most clockwise side of the
coil.
1.2 The Sinusoidal Assumption 1-9
where is the frequency of the temporal variation. It is important to realise
that the
p
angle in (1.22) is constant with respect to time in this case.
To nd the total voltage for the whole winding the individual contributions for
the number of turns at each position
p
have to be added:
e
T
=
_

0
e(
p
) d
p
=
2n
a
rlB
1
cos t

_

0
sin
2

p
d
p
=
n
a
rlB
1

cos t (1.24)
i.e. a temporal sinusoidal voltage is produced from the winding as one would
expect.
The more interesting case is when the ux density spatial distribution is
non-sinusoidal as in (1.18). In this case the ux for a single coil is:
(
p
)
n
= B
n
r
_

p
+

p
_
l
0
cos n dl d
=
B
n
rl
n
[sin n(
p
+ ) sin n
p
] (1.25)
Clearly (
p
)
n
= 0 for n even. Therefore the even harmonics do not link to a
single coil.
For the n odd case it can be seen that the expression for the ux becomes:
(
p
)
n
=
2B
n
rl
n
sin n
p
(1.26)
To calculate the voltage in a single coil at some position
p
we again apply
(1.21). Carrying out the dierentiation on (1.26) we get:
e(
p
)
n
=
2n
a
B
n
rl
n
sin
p
sin n
p
cos t :n is odd (1.27)
To get the total voltage due to the winding the individual contributions are
integrated as in the previous case:
e
T
=
2n
a
B
n
rl cos t
n
_

0
sin
p
sin n
p
d
p
(1.28)
It can be shown that
_

0
sin
p
sin n
p
d
p
= 0, therefore the total voltage due to
the odd harmonics is zero. Therefore, as with the spatially moving ux density
case, only the component of the ux density that has the same pole number as
the winding links with the winding, even if the harmonics are space stationary
and have a time varying amplitude..
Remark 1.4 The main implications of the above analysis is that the ux den-
sity component with the same pole number as the winding links with the winding.
Therefore, for a pure sinusoidally distributed winding, the harmonics in the ux
density only contribute to the leakage ux, and do not have a role in determining
the performance of the machine. However, in reality a pure sinusoidal winding
1.3 Flux Linkage to Sinusoidally Distributed Windings 1-10
cannot be produced, and there are spatial harmonics in the winding distribution.
Therefore, harmonics in the ux density waveform can link with similar pole
number harmonics in the winding distribution resulting in higher order voltage
harmonics being produced in the winding. These harmonics will also have an
inuence on machine performance.
Remark 1.5 The fact that even non-sinusoidally distributed windings react pri-
marily to uxes of the same pole number as the spatial fundamental of the wind-
ing means, from a machine modelling perspective that the models can produce
performance results that are relevant to real machines.
1.3 Flux Linkage to Sinusoidally Distributed Wind-
ings
The ux linkage properties of sinusoidal windings are very important in the
modelling of machines with sinusoidally distributed windings. In these section
we will evaluate the ux linkage expression for the winding specied by (1.1)
except that it is shifted by the angle . Therefore the winding equation is:
n = n
a
sin(
p
) (1.29)
This winding is subjected to a sinusoidally distributed ux density distribu-
tion of the form:
B(
p
) = B
m
cos
p
(1.30)
We wish to work out what the total ux linkage to this winding is. Applying
the fundamental denition of ux (1.19) to a single turn of the winding at some
angle
p
, similarly to (1.20) we get:
() =
_

p
+

p
B()lr d
= 2B
m
lr sin
p
(1.31)
Therefore for the turns distribution given by (1.29) the total linkage at a par-
ticular position
p
is:
(
p
) = n(
p
)
= 2n
a
B
m
lr sin
p
sin(
p
) (1.32)
To nd the total ux linking the coil integrate over the entire positive half cycle
of the ux density waveform:

T
=
_
2

2
(
p
) d
p
= 2n
a
B
m
lr
_
2

2
sin
p
sin(
p
) d
p
= n
a
B
m
lr cos (1.33)
1.4 Other Important Assumptions 1-11
Remark 1.6 The above expression for
T
says that the ux linking a sinusoidal
winding varies co-sinusoidally with the angle between the axis of the winding and
the vector of the sinusoidally distributed ux density distribution. Another way
of interpreting this is to say that a ux linkage vector of amplitude n
a
B
m
lr
lies along the ux density vector, and the component of the ux linkage vector
that lines coincident with the axis of the winding is the ux linking the winding.
Therefore, we are creating the view that there is a ux linkage that is sinusoidally
distributed in space. This concept is important in space vector modelling of
machines.
1.4 Other Important Assumptions
Whilst the sinusoidal assumption is very important, other assumptions are also
made in order to make the modelling of the machine tractable. These assump-
tions are:
1. The stator windings are assumed to be sinusoidally distributed. When
excited with current a sinusoidal spatial distribution of mmf is produced.
2. The machine does not exhibit any stator or rotor slotting eects.
3. The machine iron is a linear material, i.e. it is not subject to magnetic sat-
uration eects. The permeability of the material is very large in compar-
ison to air. Therefore the permeance of the magnetic paths is dominated
by the air gaps.
4. The airgap ux density waveforms can be adequately represented by the
fundamental component.
Let us examine some of these assumptions. The rst assumption says that we
have ideal sinusoidally distributed windings. As mentioned previously this never
occurs in practice because the machine windings have to be made up from nite
diameter wires, and generally speaking these wires are placed for mechanical
and magnetic reasons in slots in the stator or rotor of a machine. However, be-
cause of the properties of sinusoidally distributed windings (even approximately
sinusoidally distributed windings) the eects of slotting only produces secondary
eects in terms of induced voltages. The second assumption is clearly related
to the rst assumption.
The third assumption would seem to be rather restrictive at rst glance.
Almost all practical machines are constructed using iron which exhibits a non-
linear ux density versus magnetising force characteristic . the iron saturates.
Most real machines exhibit saturation in their normal operation regimes. This
assumption under the circumstance of high saturation makes a model generate
erroneous results. However, in many situations the saturation is such that the
linear approximation still gives reasonable results. In any case, the essential
characteristics of the dynamic performance of the machine is preserved under
the linear material assumption. This is fortunate, as the models can be analysed
using standard linear analysis because of this assumption.
The nal assumption essentially says that the modelling will assumed that
the uxes in the machine are spatially sinusoidal. This assumption does not
cause too many problems for reasons stated previously, and is important because
it allows powerful analysis techniques to be applied to machines.
1.4 Other Important Assumptions 1-12
120
q
i
a
i
b
i
c
v
a
v
b
v
c
Figure 1.3: Conceptual diagram of sinusoidally distributed three phase windings
1.4.1 Properties of Three Phase Sinusoidal Windings
Most practical AC machines use three phase windings on their stator, and some
also have these types of winding on the rotor (e.g. wound rotor induction ma-
chine). If one is seeks to understand AC machines it is important to understand
the basic properties of these windings.
Figure 1.3 shows a conceptual layout of a set of three phase windings. Note
that in this diagram the concentrated coils drawn represent sinusoidally dis-
tributed windings. The axes of each of the coils are 120

apart in space. If each


of these coils was fed with a DC current then each winding would give a mmf
distribution that is sinusoidal in space. The three mmf waveforms can be added
to give a single sinusoidal resultant mmf waveform. By controlling the currents
in these three winding the resultant mmf waveform can be made to move in
space. This phenomena becomes interesting if the windings are fed with three
phase currents.
Given that the space distribution of the mmfs for windings a, b and c can
be modelled similarly to (1.2) then the following expressions can be written for
1.5 Torque Expressions 1-13
the mmfs:
F
a
(
p
) =

F
a
cos
p
(1.34)
F
b
(
p
) =

F
b
cos
_

2
3
_
(1.35)
F
c
(
p
) =

F
c
cos
_

p
+
2
3
_
(1.36)
where
p
is as dened in Figure 1.3 and

F
a
,

F
b
, and

F
c
are the peak mmfs.
The resultant mmf distribution at any point
p
around the three phase machine
periphery is:
F
T
=

F
a
cos
p
+

F
b
cos
_

2
3
_
+

F
c
cos
_

p
+
2
3
_
(1.37)
Assuming that the three phase windings have identical turns, and they are being
driven by three phase currents of the form:
i
a
= I
pk
cos t (1.38)
i
b
= I
pk
cos
_
t
2
3
_
(1.39)
i
c
= I
pk
cos
_
t +
2
3
_
(1.40)
then (1.37) can be written as:
F
T
=

NI
pk
_
cos t cos + cos
_
t
2
3
_
cos
_

2
3
_
+ cos
_
t +
2
3
_
cos
_
+
2
3
_
_
(1.41)
=
3
2

NI
pk
cos(t ) (1.42)
where

N = N/2. As mentioned previously the division by 2 is to make the
mmf expression the mmf per airgap (i.e. the total mmf is expended across two
airgaps).
Equation (1.42) means that the resultant mmf has a spatial sinusoidal distri-
bution which is rotating around the machine at t electrical radians per second. This rotating mmf
property is funda-
mental to the oper-
ation of all AC ma-
chines
If the mmf of (1.42) is acting on a uniform air gap between two iron structures
(such as occurs in an electrical machine with a round rotor) then the ux pro-
duced by the mmf will also be spatially sinusoidal. If the mmf is acting on a
non-uniform air gap between two iron structures (such as occurs in a salient
pole machine) then the resultant ux will most probably not be sinusoidal.
2
1.5 Torque Expressions
One of the most fundamental quantities of any machines performance is the
torque. After all one is usually using a machine to produce torque and power
2
Note the the amplitude of the mmf waveform is
3
4
NI
pk
because half the mmf is expended
across each of the airgaps in the machine.
1.5 Torque Expressions 1-14
from electrical energy. Consequently it is essential to have techniques in any
machine model to allow the torque that will be produced to be accurately esti-
mated.
The approach taken in this section is to rstly consider a very general tech-
nique for estimating torque. This technique, based on the concept of co-energy
is capable of accounting for the saturation non-linearities in the machine. Its
generality extends further, since it can be successfully applied to a variety of
dierent machines. The following development will place special emphasis on
the reluctance machine.
The classic way of calculating the power or torque produced by an electrical
machine is to write the equation for the energy balance in the machine. The
following discussion is based on that in [2]. The expressions developed are for
electrical to mechanical energy conversion. Mechanical to electrical conversion
is the dual of this, and will not be discussed separately. The energy balance
equation is based on the conservation of energy principle i.e. the input energy
must be balanced by the losses (both electrical and mechanical), any energy
transiently stored in the system (both mechanical and electrical), and the output
energy in the form of mechanical work. Therefore the energy balance can be
written as:
_
Electrical
energy input
_
=
_
Electrical
losses
_
+
_
Stored energy
in elds
_
+
_
Mechanical
energy
_
(1.43)
In symbol form this may be written as:
E
e
= E
le
+ E
fe
+ E
me
(1.44)
The mechanical energy component may not appear as mechanical work, but
some of it may be stored in forms such as kinetic energy and various forms of
potential energy. Therefore the actual output mechanical energy is:
E
mo
= E
me
E
lm
E
sm
(1.45)
where:
E
lm
the mechanical losses
E
sm
the stored mechanical energy
Therefore (1.44) may be written as:
E
e
= E
le
+ E
lm
. .
Losses
+E
fe
+ E
sm
. .
Stored
+E
mo
(1.46)
1.5.1 Torque of a Simple Reluctance Machine
The following discussion is with reference to Figure 1.4, which is a schematic
diagram of a simple reluctance machine. The following assumptions are made
in the analysis:
1. The iron circuit exhibits saturation, i.e. it has a nonlinear ux vs current
relationship.
1.5 Torque Expressions 1-15
Figure 1.4: Simple singly excited reluctance machine.
2. There is no leakage ux.
3. Hysteresis and eddy currents are ignored.
4. Mechanical energy storage and losses are ignored.
Using standard circuit analysis on circuit in Figure 1.4 gives:
v = Ri +
d
dt
(1.47)
Therefore the power being input into the circuit is:
vi = Ri
2
+ i
d
dt
(1.48)
To get the total energy input into the circuit over a time t integrate (1.48)
assuming that the initial energy storage is zero:
_
t
0
vi dt =
_
t
0
_
Ri
2
+ i
d
dt
_
dt (1.49)
Equation (1.49) can be broken up into two sections the Ri
2
term is clearly
related to the resistive losses in the coil. Therefore the remaining term must be
related to any stored eld energy and mechanical energy. Considering this term
in detail, it can be seen that it is in a form that enables it to be transformed by
a change of variable. Since i = F(), where F denotes a non-linear function.
Therefore the second term in (1.49) can be written as:
i(F())
d
dt
dt = i()d (1.50)
1.5 Torque Expressions 1-16
Therefore (1.49) can be written as:
_
t
0
vi dt =
_
t
0
Ri
2
dt +
_

0
i() d (1.51)
The second term therefore can be written in terms of the energy equation as:
_

0
i() d = E
fe
+ E
mo
(1.52)
Equation (1.52) contains the output energy term, and therefore can be used
to calculate the output power and torque if there is mechanical movement in
the system. However, before this can be achieved the eld energy term must be
separated out. If there isnt mechanical movement then this term would be the
stored eld energy in the coils.
Consider the rotor in Figure 1.4. If it is held stationary, and a voltage and
current are applied then there can be no mechanical energy output. Therefore
ignoring the resistive losses all the input energy must be stored in the magnetic
eld of the stator coils. Consider two dierent rotor positions; one where the
rotor is aligned with the axis of the windings, and the other where it is at some
angle to the windings. In the rst situation there will be more ux for a given
value of current as compared to the second case. A sketch of the ux versus
current plots are shown in Figure 1.5. Notice that the unaligned ux plot is
more linear than the aligned ux plot due to the fact that the ux path is
dominated by air in the former case, and consequently the ux density does not
get high enough to cause saturation.
For the aligned position the integration of (1.52) is represented by the shaded
area in Figure 1.4. Therefore this area represents the eld energy stored in the
system, and is the useful electrical energy applied to the system. Similarly
the area represented by P
A
, P
B
, P
E
, and P
A
, is the eld energy stored in the
unaligned position. The alternative integration
_
i
2
0
di, is called the co-energy
and, as we shall see, is also important in the determination of the mechanical
energy output of the system when the rotor is allowed to move.
In order to calculate the mechanical energy the rotor is allowed to move. In
order to nd the mechanical energy a thought experiment involving two dierent
types of movement is carried out. The movements are:
1. Very slow movement of the rotor which does not produce any voltage
across the stator coil due to d/dt. This means that the current owing
in the coil is determined by its resistance, and is therefore constant.
2. Very fast movement of the rotor. Since the ux linkage cannot change
instantaneously then the ux is considered to be constant throughout this
process.
Ii should be noted that the above two movements are idealised and it is impos-
sible to carry out these experiments with a high degree of accuracy. However it
is possible to approximate them.
1.5.1.1 Slow Rotor Movement
Assume that the rotor is moved from the unaligned position to the aligned
position very slowly. The point P
B
will move along the vertical line to the
1.5 Torque Expressions 1-17
Figure 1.5: Flux plots for static movement.
position P
C
as the rotor moves. The initial energy in the system is that energy
stored in the unaligned position (represented by the area P
A
, P
B
, P
E
, and P
A
in
Figure 1.5). The nal stored energy in the system is represented by the shaded
area of Figure 1.5. The dierence between these two areas is the change in the
stored energy of the system. If the change in the stored energy is subtracted
from the energy input to the system during this movement then the dierence
must be the mechanical energy output during the movement. The energy input
can be ascertained in the following manner. Assume an incremental movement
of the rotor is made as shown in Figure 1.6.
If the current is constant then the shaded area is the incremental electrical
energy added to the system. If these areas are integrated for a movement from
P
B
to P
C
then the energy applied to the system is:
_

2

1
i
2
d = i
2
_

2

1
d = i
2
(
2

1
) (1.53)
This is represented by an area that is a square region of height (
2

1
) and
width i
2
. Therefore the energy which is transferred into the mechanical move-
ment is:
E
mo
= i
2
(
2

1
)
. .
Elec input energy

_
_

2
0
i d
_

1
0
i d
_
. .
Change in stored energy
(1.54)
The sign convention of positive power for energy owing out of the machine
shall be taken. This convention is consistent with the normal convention used
for motoring machines. Power owing into of the system is negative and power
owing out of the system is positive. In the particular case above electrical
energy is owing from the supply into the rotor, and mechanical energy is owing
out of the rotor into an external load via the rotor shaft. The calculation in
1.5 Torque Expressions 1-18
- incremental energy change
Unaligned
Aligned
Figure 1.6: Incremental energy change with small movement of the rotor.
(1.54) can be seen in Figure 1.7. The output energy area is the dierence
between the co-energy at the aligned position and the co-energy at the unaligned
position. Therefore the mechanical output energy can be written as:
E
mo
= E

fe
2
E

fe
1
= E

fe
(1.55)
where:
E

fe
1
=
_
i
2
0

unalign
di the co-energy in the unaligned position.
E

fe
2
=
_
i
2
0

align
di the co-energy in the aligned position.
If we adopt the convention that the energy changes are calculated by subtracting
the initial energy from the nal energy then the co-energy calculation naturally
gives the correct sign for the mechanical output energy using the energy sign
convention that we have dened above.
If the energy for a mechanical movement is known then the torque can be
found using the expression:
T
ave
=
mechanical energy
angular movement
=
E
mo

pd
(1.56)
If one considers a very small movement of
pd
, and take the limit as
pd
0,
then one can nd the expression for the instantaneous torque under the slow
movement condition:
T
e
= lim

p
0
_
E

fe

pd
_
i constant
=
E

fe

pd

i constant
(1.57)
1.5 Torque Expressions 1-19
Figure 1.7: Area representing mechanical output energy.
1.5.1.2 Instantaneous Rotor Movement
One can think of
the instantaneous
movement occur-
ring because the
rotor has virtually
no inertia. The
rotor will tend to
align in the position
to maximise the
ux.
The second thought experiment is to imagine that the rotor is moved virtually
instantaneously from the unaligned to the aligned position. During this move-
ment the ux linkage cannot change. The reason for this is that Lenzs Law
states that the induced voltage during a rate of change of ux linkage is such as
to oppose the rate of change of ux linkage. The constant ux linkage during
these movements is known as the law of constant ux linkage. In this particular
case, the ux linkage would be attempting to increase as the rotor aligns with
the stator pole. Consequently the induced voltage is such so as to reduce the
current from i
2
to i
1
to keep the ux constant at
1
. Once the rotor stops at the
aligned position the induced voltage due to rotor movement disappears and the
current can increase back to the steady state value of i
2
(which is determined
by the resistance of the stator winding). The path followed by the ux linkages
is shown in Figure 1.8.
During the movement from the unaligned to aligned position the ux link-
ages remain constant (from points P
B
to P
F
in Figure 1.8). At position P
F
the
rotor has reached to aligned position and the movement stops. The ux then
follows the P
F
to P
C
path. During the movement phase there is no electrical
energy owing into the system (except to supply the resistive losses). This can
be concluded because
_
i d = 0. However the stored energy in the eld has
changed considerably. Using conservation of energy arguments the conclusion
is that the stored eld energy has been converted to mechanical energy. This
energy is presented by the area P
A
, P
B
, and P
F
in Figure 1.8. Once the rotor is
stationary, the current will increase from the i
1
value at P
F
to i
2
, at a rate de-
termined by the time constant of the stator winding. During this time electrical
energy is owing into the system and being stored in the magnetic eld. This
statement can be deduced from the fact that the rotor is not moving, therefore
there can be no mechanical energy.
1.5 Torque Expressions 1-20
Figure 1.8: Energy with instantaneous movement.
The expression for the mechanical energy can be written as:
E
mo
= [area (P
A
P
F
P
E
P
A
) - area (P
A
P
B
P
E
P
A
)]
=
_
_

1
0
i
align
d
_

1
0
i
unalign
d
_
(1.58)
The mechanical output energy is therefore the change in the eld energy E
fe
,
during this movement. The negative sign results so that the mechanical energy
from this expression is consistent with that calculated using co-energy. Remem-
ber that the convention for the sign of the energy is that energy owing into the
system is negative, and out is positive. In the case of both the slow and fact
rotor movements the mechanical energy ows out of the system (i.e. the energy
is being applied to the shaft load on the rotor).
The torque expression can be found for this case as it was for the case using
slow movement. Assume a small increment,
p
, of the movement shown in
Figure 1.8. The instantaneous torque over such a movement can be found by
letting the angular movement approach zero:
T
e
= lim
0

_
E
fe

pd
_
constant
=
_
E
fe

pd
_
constant
(1.59)
1.5.1.3 Real Movement
Real movements usually dont consist of either a very slow rotor movement or
an instantaneous movement. They are usually somewhere in between. However,
1.5 Torque Expressions 1-21
Figure 1.9: Flux versus current trajectory for typical real movement.
as will be seen in a moment, real movements can be analyzed by using the two
idealised movements in combination. The following discussion is with reference
to Figure 1.9, which shows the versus i trajectory for a typical real movement.
Consider a small elemental movement of the rotor. The area P
A
bdP
A
repre-
sents the actual amount of mechanical output energy. With a slow movement
the mechanical output energy is P
A
bcP
A
, and with instantaneous movement
P
A
baP
A
. Heuristically one can see from Figure 1.9 that as the movement an-
gle becomes less, the slow movement and instantaneous movement areas will
approach each other, and in the limit they will be equal. In other words the
areas bcdb and bdab tend to zero as
pd
0. This means that the areas for
the slow movement and instantaneous movement converge to the shaded area,
which is the area representing the mechanical energy for the actual movement.
Therefore the following expression can be written:
T
e
=
E

fe

pd

i constant
=
E
fe

pd

constant
(1.60)
In other words the same value of torque is obtained if the rate of change of
co-energy is calculated with a constant current, or if the rate of change of eld
energy is calculated with constant ux linkage.
The same result may be obtained via a much more formal route. Both the
eld energy and the eld co-energy can be expressed as functions of the following
form:
E
fe
= G(, i) (1.61)
E

fe
= H(, i) (1.62)
1.5 Torque Expressions 1-22
where G(, i) and H(, i) represent non-linear functions of the variables and
i.
Similarly the current, ux and rotor angle can be expressed as the following
functional relationships:
i = f(,
pd
) (1.63)
= g(i,
pd
) (1.64)
= h(, i) (1.65)
where f, g and h are non-linear functions of their respective variables.
If the current and ux functional relationships are substituted into (1.61)
and (1.62) then it can be seen that E
fe
and E

fe
can be expressed as functions
of (i,
pd
) or (,
pd
), i.e.:
E
fe
= I(,
pd
) (1.66)
E

fe
= J(i,
pd
) (1.67)
where I and J represent the new non-linear functions after the substitution.
Let us consider the co-energy expression. Since i is constant for this expres-
sion we can take the partial derivative with respect to
pd
:
T
e
=
J(i,
pd
)

pd
=
J(i,
pd
)

pd

pd

pd
=
E

fe

pd

i constant
(1.68)
In a similar fashion it can be shown that:
T
e
=
E
fe

pd

constant
(1.69)
Example 1.1 As an example of the use of the above expressions assume that
the current ux relationship is as follows:
= (L
1
+ L
2
cos 2
pd
)i
0.9
(1.70)
This expression is based on the self inductance expression for a single winding
as obtained in (A.28), with the addition of a non-linearity to approximate the
saturation of the steel in the stator of the machine. Note that this is not an
attempt to accurately model saturation, but is simply an artice for this example.
Using the expression for the co-energy we have:
E

fe
=
_
i
1
0
di
=
_
i
1
0
(L
1
+ L
2
cos 2
pd
)i
0.9
di
=
1
1.9
(L
1
+ L
2
cos 2
pd
) i
1.9
1
(1.71)
Using (1.57) on (1.71) one gets:
T
e
=
2L
2
i
1.9
1
1.9
sin 2
pd
(1.72)
1.5 Torque Expressions 1-23
If one now uses (1.52) and rearranges (1.70) with i as the subject of the expres-
sion, then the eld energy can be calculated as follows:
E
fe
=
_

1
0
i d
=
_

1
0

1
0.9
(L
1
+ L
2
cos 2
pd
)
1
0.9
d
=
0.9
1.9
_

1.9
0.9
1
(L
1
+ L
2
cos 2
pd
)
1
0.9
_
(1.73)
Using (1.59) and (1.73) one obtains:
T
e
=
_
1
1.9
2L
2

1.9
0.9
1
sin 2
pd
(L
1
+ L
2
cos 2
pd
)
1.9
0.9
_
(1.74)
Using (1.70) in (1.74) the following expression can be derived:
T
e
=
_
2L
2
i
1.9
1
1.9
sin 2
pd
_
(1.75)
which is the same as (1.72) as expected.
As a nal check on this result, the eld energy and the co-energy should add up
to be
1
i
1
. Substituting (1.70) into (1.71) gives:
E

fe
=
1
1.9

1
i
1
(1.76)
Similarly substituting (1.70) into (1.73) gives:
E
fe
=
0.9
1.9

1
i
1
(1.77)
Adding (1.76) and (1.77) one gets
1
i
1
as expected.
1.5.2 Linear Torque Model
1.5.2.1 Singly Excited System
An important special case of the above co-energy based analysis is for linear
models. Linear models are no particularly accurate, but they are useful because
they can be analysed mathematically and therefore aid our qualitative under-
standing of a machine. We shall rstly analyse the simplest possible reluctance
machine the singly excited reluctance machine.
Consider a linear magnetic system. It can be characterized by the following
expression:
= Li (1.78)
where L the inductance of the system. The ux versus current plot for such
a system is shown in Figure 1.10. If the eld energy of the system is calculated
1.5 Torque Expressions 1-24
Figure 1.10: Flux versus current for linear magnetic material.
one has:
E
fe
=
_

0
i d
=
_

0

L
d
=
1
2

2
L
(1.79)
Given that L = /i then (1.79) can be written as:
E
fe
=
1
2
i =
1
2
Li
2
(1.80)
Clearly from Figure 1.10 it can be seen that the co-energy and the eld
energy are equal, i.e.:
E
fe
= E

fe
(1.81)
If the linear system is of the type shown in Figure 1.4 then as the rotor of the
machine aligns with the stator poles the inductance of the system will increase.
This is indicated in Figure 1.10 by the steeper ux line. For such a movement
we can calculate the torque using a formal method based on (1.57):
T
e
=
E

fe
(i,
pd
)

pd

i constant
=

pd
_
1
2
Li
2
_

i constant
T
e
=
1
2
i
2
dL
d
pd
(1.82)
1.5 Torque Expressions 1-25
A less formal and heuristic approach is to realise that the co-energy under the
less aligned ux line is:
E

fe
1
=
1
2

1
i (1.83)
and under the more aligned line after a rotor movement is:
E

fe
2
=
1
2

2
i (1.84)
Since:

1
= L
1
i and (1.85)

2
= L
2
i (1.86)
we can write:
E

fe
= E

fe
2
E

fe
1
=
1
2
(L
2
L
1
)i
2
=
1
2
Li
2
(1.87)
Therefore:
E

fe

pd
=
1
2
i
2
L

pd
(1.88)
As
pd
0 we end up with (1.82).
1.5.2.2 Doubly Excited System
Now let us consider a machine system with two windings, instead of one. We
shall consider the situation where we have a stator winding as in the previous
singly excited case, and in addition we also have a winding on the rotor itself.
This is a similar situation to a synchronous machine. One might be tempted to
ask Why are we considering a winding on the rotor, after all we are supposed to
be concentrating on reluctance machines without such a winding?. The reason
for considering this case is that the expressions for the eld energy obtained are
also applicable to the two phase reluctance machine i.e. a reluctance machine
with two orthogonal stator windings.
Consider a machine of the form shown in Figure 1.11. The expressions for
the ux linking the windings is:

1
= L
1
i
1
+ Mi
2
(1.89)

2
= L
2
i
2
+ Mi
1
(1.90)
where:
L
1
the self inductance of the stator winding
L
2
the self inductance of the rotor winding
M the mutual inductance between the stator
and the rotor.
1.5 Torque Expressions 1-26
Figure 1.11: Doubly excited reluctance machine.
The instantaneous voltage equations for the coils are:
v
1
= R
1
i
1
+
d
1
dt
(1.91)
v
2
= R
2
i
2
+
d
2
dt
(1.92)
Substituting (1.89) and (1.90) into the above gives:
v
1
= R
1
i
1
+
d
dt
(L
1
i
1
) +
d
dt
(Mi
2
) (1.93)
v
2
= R
2
i
2
+
d
dt
(L
2
i
2
) +
d
dt
(Mi
1
) (1.94)
Because this system has saliency, then the inductances, L
1
and L
2
are func-
tions of . However, because we are considering the machine to be linear, the
inductances are not functions of current. Taking the appropriate derivatives
then the voltage equations become:
v
1
= R
1
i
1
+ L
1
di
1
dt
+ i
1
dL
1
dt
+ M
di
2
dt
+ i
2
dM
dt
(1.95)
v
2
= R
2
i
2
+ L
2
di
2
dt
+ i
2
dL
2
dt
+ M
di
1
dt
+ i
1
dM
dt
(1.96)
If (1.95) and (1.96) are respectively multiplied by i
1
and i
2
, then one ob-
tains the total instantaneous power for each of the two windings. If these two
expressions are added together and integrated from time zero to time t, then
the total energy input to the system can be calculated. This expression can be
shown to have the form:
E
e
=
_
t
0
(v
1
i
1
+ v
2
i
2
) dt
=
_
t
0
(R
1
i
2
1
+ R
2
i
2
2
) dt +
_
(L
1
i
1
di
1
+ L
2
i
2
di
2
+ i
1
M di
2
+ i
2
M di
1
+ 2i
1
i
2
dM + i
2
1
dL
1
+ i
2
2
dL
2
) (1.97)
1.5 Torque Expressions 1-27
Comparing (1.97) to (1.44) allows the following relationships to be written:
_
Useful electrical
energy input
_
=
_
t
0
(v
1
i
1
+ v
2
i
2
) dt
_
t
0
(R
1
i
2
1
+ R
2
i
2
2
) dt (1.98)
_
_
Energy to
eld
storage
_
_
+
_

_
Electrical
energy
to mechanical
energy
_

_
=
_
(L
1
i
1
di
1
+ L
2
i
2
di
2
+ i
1
M di
2
+ i
2
M di
1
+ 2i
1
i
2
dM + i
2
1
dL
1
+ i
2
2
dL
2
) (1.99)
Because the system
is linear we can in-
terchange eld en-
ergy and eld co-
energy.
In order to get a value for the torque produced by this device one needs
to separate the eld energy from the mechanical energy in (1.99). This can
be achieved by locking the rotor of the transducer at some position, and then
energizing the coils from zero current to some required value. Because the rotor
is locked there can be no mechanical energy, therefore all the energy must be
eld energy. Note that this condition also implies that the dL
1
, dL
2
and dM
terms in (1.99) are zero. Therefore the stored eld energy equation becomes:
E
fe
=
_
i
1
0
L
1
i
1
di
1
+
_
i
2
0
L
2
i
2
di
2
+
_
i
2
0
i
1
M di
2
=
1
2
L
1
i
2
1
+
1
2
L
2
i
2
2
+ i
1
i
2
M (1.100)
The integration in (1.100) implicitly relies on the fact that the energy stored
in the eld of an inductor depends on the instantaneous value of the current,
and not on the time history of how the current attained this value. The reason
that only the i
1
M di
2
term is integrated in (1.100) is not immediately obvious.
However it can be reasoned out using the following thought experiment. Assume
that i
2
is initially held at zero by an open circuit. At the same time i
1
is increased
from zero to some nal value of I
1
. During this time the energy stored in the self
inductance is
1
2
L
1
i
2
1
. With the current in winding 1 held constant, the current
in winding 2 is then increased to I
2
. Similarly the energy stored in the winding
2 self inductance is
1
2
L
2
i
2
2
. However, due to the mutual coupling to winding
1 during this current increase there is an induced voltage in winding 1 which
attempts to produce a current to oppose the changing ux. The current source
supplying the I
1
current produces a voltage to oppose this induced voltage so
that the current remains constant at I
1
. Therefore energy is being supplied to
winding 1, and this energy must be going into the stored eld since it cannot
go into mechanical energy. The energy supplied to winding 1 is:
E
M
=
_
T
0
M
di
2
dt
. .
Induced voltage
I
1
dt
=
_
I
2
0
MI
1
di
2
= MI
1
I
2
(1.101)
Therefore the total energy is the self inductance energy terms plus (1.101). A
more formal proof can be obtained if the currents in both winding 1 and 2 are
1.5 Torque Expressions 1-28
assumed to increase linearly from zero to their nal values; i.e.:
i
1
(t) = K
1
t
i
2
(t) = K
2
t
_
for 0 t T (1.102)
If we resort to a form of the stored energy equation based on (1.97), we can
write, using (1.102) the following for the stored eld energy in the mutual ux:
E
M
=
_
T
0
_
i
2
M
di
1
dt
+ i
1
M
di
2
dt
_
dt
=
_
T
0
2MK
1
K
2
t dt
= MK
1
K
2
T
2
= Mi
1
(T)i
2
(T) (1.103)
which is the same expression as (1.101).
It is a simple matter to now calculate the torque produced by this system.
Because the system is linear we know that E
fe
= E

fe
, therefore using (1.57) on
(1.100) one can write the following:
T
e
=
E
fe

i constant
=
1
2
i
2
1
dL
1
d
+
1
2
i
2
2
dL
2
d
+ i
1
i
2
dM
d
(1.104)
If the rotor is moving what happens with the stored eld energy. Intuition
tells one that the eld energy must be changing because the inductances are
changing. Clearly the change of the inductance with the angular movement is
the key to the production of torque. Therefore if torque is being produced then
mechanical energy is also being produced if there is movement. The question
to resolve is how much of the energy being input to the system during this
movement phase is going into mechanical energy and how much is going into
stored eld energy. In order to analyze this situation dierentiate (1.100) with
respect to time:
dE
fe
dt
= L
1
i
1
di
1
dt
+
1
2
i
1
dL
1
dt
+ L
2
i
2
di
2
dt
+
1
2
L
2
i
2
dL
2
dt
+ Mi
2
di
1
dt
+ Mi
1
di
2
dt
(1.105)
If (1.105) is integrated with respect to time then the total eld energy can be
obtained:
E
fe
=
_
dE
fe
=
_
(L
1
i
1
di
1
+
1
2
i
2
1
dL
1
+ L
2
i
2
di
2
+
1
2
i
2
2
dL
2
+ i
1
i
2
dM + i
1
M di
2
+ i
2
M di
1
)
(1.106)
The eld energy related terms in this expression are obviously those used in
(1.100), therefore the other terms are related to the mechanical energy, i.e. the
mechanical energy is:
E
mo
=
_
(
1
2
i
2
1
dL
1
+
1
2
i
2
2
dL
2
+ i
1
i
2
dM) (1.107)
1.5 Torque Expressions 1-29
If this expression is compared with (1.99), and eliminating the stored eld energy
terms one ends up with the following terms only:
_
(2i
1
i
2
dM + i
2
1
dL
1
+ i
2
2
dL
2
) (1.108)
Clearly this is twice (1.107), therefore when the rotor is moved half the electrical
input energy associated with the movement is going into mechanical output
energy, and the other half is being stored in eld energy.
1.5.3 The Ellipse Diagram for co-energy
A classic case study for the use of co-energy is the synchronous reluctance ma-
chine (SYNCREL). The development in the previous section of the basic torque
expressions in terms of eld energy and co-energy may appear to be very low
level. However, an understanding of this basic way of calculating torque is very
relevant to the machines such as the SYNCREL as it exhibits a high degree
of saturation in normal operation. Co-energy torque calculation techniques are
also very useful for the switched reluctance machine for the same reason.
In this section we shall concentrate on the application of co-energy to calcu-
late the torque of the SYNCREL. The advantage of using the co-energy approach
is that the non-linearities caused by saturation and slotting eects can be ac-
counted for, whereas with the conventional linearised models using sinusoidal
approximations these eects are ignored. Using co-energy in combination with
nite element analysis it is possible to get very accurate estimates of the torque
production and ripple for these machines at the design stage [5].
To demonstrate the application of co-energy to the reluctance machine we
shall rstly consider a linear machine i.e. no saturation or slotting eects.
Therefore the inductance expressions for the machine obey (A.55) and (A.56)
derived previously. It will be shown in a future section that the dq-axis induc-
tances can be related to the phase expressions in the following manner:
L
d
= L
sl
+
3
2
(L
1
+ L
2
) (1.109)
L
q
= L
sl
+
3
2
(L
1
L
2
) (1.110)
where L
sl
the self leakage of the phase windings. By manipulating these two
expressions L
1
and L
2
can be obtained in terms of the dq-axes inductances:
L
1
=
L
d
+ L
q
3
+
2
3
L
sl
(1.111)
L
2
=
L
d
L
q
3
(1.112)
It is assumed that this machine is a three phase, four pole machine, and is
excited by currents of the form:
i
a
= I
pk
cos(
pd
+ ) (1.113)
i
b
= I
pk
cos(
pd
+
2
3
) (1.114)
i
c
= I
pk
cos(
pd
+ +
2
3
) (1.115)
1.5 Torque Expressions 1-30
Figure 1.12: Self and mutual inductance variation with rotor angle.
Notice that the currents are synchronized with the rotor position, and that the
angle between the resultant mmf of the stator and the rotor d-axis is always
radians. For the purposes of this example we shall only consider = /4
radians.
From nite element analysis the dq parameters for an axially laminated
motor with the phase current at 15 amps rms, and the current angle at /4
have been determined as [5]:
L
d
= 77.3mH
L
q
= 10.0mH
L
sl
= 2.3mH
_
_
_
(1.116)
If (1.116) are substituted into (1.111) and (1.112) and these are substituted
into (A.55) and (A.56), then the plots in Figure 1.12 can be obtained for the
inductance variation with rotor angle.
To calculate the total ux linkage to the a-phase under these conditions then
the following expressions have to be used:

aa
= L
aa
i
a
(1.117)

ab
= L
ab
i
b
(1.118)

ac
= L
ac
i
c
(1.119)
and the total ux linkages with the a-phase are:

af
=
aa
+
ab
+
ac
(1.120)
Clearly the expressions in (1.117), (1.118) and (1.119) involve multiplicative
sinusoidal functions when the current expressions (1.113), (1.114) and (1.115)
are substituted in. For example, (1.117) is an expression of the form:

aa
= I
pk
[L
1
cos(
pd
+

4
) + L
2
cos 2
pd
cos(
pd
+

4
)] (1.121)
1.5 Torque Expressions 1-31
Figure 1.13: Flux linkage for the a-phase
and similarly for
ab
and
ac
. If (1.117), (1.118) and (1.119) are plotted then
Figure 1.13 results. Even though the self and mutual ux curves are nonsinu-
soidal, the resultant total ux curve is sinusoidal. In [5] nite element (FE)
based a-phase ux results have been compared to these ideal curves, and very
close agreement is achieved [5].
In order to apply the co-energy concept to calculate the torque for the ma-
chine a vs i plot is required. This can be achieved by plotting the expression
(1.120) versus the instantaneous a-phase current shown in (1.113). Figure 1.14
shows this plot. Notice that the total ux linkage curve is an ellipse. The self
and mutual ux linkage terms and the aligned and unaligned linear magnetisa-
tion lines are also shown. Note that the SYNCREL is not bound by these lines,
as is the case for the switched reluctance machine. This is due to the eect
of the mutual ux on the ux linking a single phase. Another major point of
dierence is that the vs i plot traverses two quadrants, whereas the plot for
the switched reluctance machine stays in one quadrant. This occurs because of
the bidirectional nature of the currents in the SYNCREL.
One can interpret this diagram as follows. If at t = 0 we have
pd
= 0,
then i
a
= I
pk
/

2. For the particular case in the plot I


pk
= 22 Amp. The
starting point is marked on Figure 1.14. As
pd
increases in a positive direction
the i
a
current will decrease in value. Therefore anticlockwise rotation of the
rotor corresponds to movement in an anticlockwise direction on the diagram (as
shown by the arrow on the diagram).
Figure 1.14 shows a small movement in the rotor angle, from
pd
= 0 to
pd
=
. The shaded area is the change in the co-energy for this movement, therefore
the instantaneous torque can be calculated using (1.57). Areas enclosed in
a counter-clockwise direction correspond to positive (motoring) torque, while
areas in the clockwise direction give rise to negative (generating) torque.
Remark 1.7 Note that Figure 1.14 has been drawn for a linear system. The
segment corresponding to the movement of the rotor is a triangle because of
1.5 Torque Expressions 1-32
T
Figure 1.14: Total, self and mutual ux versus current loci.
this linear assumption. In a real machine this area would have curved sides
due to saturation at the higher currents when the rotor is aligned with a phase.
Indeed in this situation the outside boundary of the diagram is no longer an
ellipse (this only occurs if the magnetic material is linear). In order to get
accurate instantaneous torques under this condition one would have to calculate
the area taking into account the non-linear shape of the segment [5].
One complete traverse of the ellipse corresponds to 360

electrical, and conse-


quently to the passage of two rotor poles past the axis of the phase winding.
Therefore, to conform with the normal convention applied to the switched re-
luctance machine, the basic unit of energy conversion is half the area of the
ellipse, E
a
stroke
= E
2
/2, where E
2
is the area for one complete revolution of
the ellipse. The ellipse shown in Figure 1.14 is only for one phase, similar ellipses
can be drawn for the other two phases of a three phase machine. Therefore the
total energy per stroke for the machine is:
E
stroke
= 3E
a
stroke
=
3
2
E
2
(1.122)
Assume that the machine in question has p
p
pole pairs; i.e. 2p
p
poles.
Therefore in one mechanical revolution of the machine rotor 2p
p
poles will pass
each phase axis. Clearly the total energy is the number of strokes multiplied by
the energy of a single stroke, i.e.:
E
T
stroke
= 6pE
a
stroke
= 2p
p
E
stroke
(1.123)
The expression for the average torque for a revolution of the rotor is:
T
ave
=
E
fe

=
E
T
stroke
2
=
p
p

E
stroke
(1.124)
1.5 Torque Expressions 1-33
0
180
Figure 1.15: Segments used for ellipse area.
For a general m phase, p
p
pole pair machine the average torque expression
can easily be shown to be:
T
ave
=
mp
p

E
a
stroke
=
p
p

E
stroke
(1.125)
where E
stroke
= mE
a
stroke
.
In order to calculate the torque using this technique one needs to calculate
the area of the ellipse. In general a numerical technique needs to be used to do
this (since the data for this curve in a real situation would come from a nite
element analysis of the motor). One simple technique is the break the ellipse into
a number of sectors, each corresponding to the same small angular movement
of the rotor. The area of each sector is then computed and summed to get the
total area. Figure 1.15 shows the ellipse divided into a number of approximately
triangular sectors. The area of each of these areas can be approximated as a
triangle, especially if the number of sectors is large. A general formula for the
area of a triangle is:
A
triag
=
1
2
[x
3
(y
1
y
2
) + x
2
(y
3
y
1
) + x
1
(y
2
y
3
)] (1.126)
where the x
i
and y
j
(i, j = 1, 2, 3) are the vertices of the triangle in the xy
plane.
If the triangles in Figure 1.15 are summed using (1.126) then the result for
this particular case is 45.4183 Nm. If the torque is calculated via the normal dq
expression for the torque of a three phase machine, i.e.:
T
ave
=
3
2
p
p
(L
d
L
q
)i
d
i
q
(1.127)
1.5 Torque Expressions 1-34
where p
p
the pole pairs, then the result is 45.4275 Nm. Clearly as the number
of triangular segments is increased then the ellipse calculated torque approaches
the dq even more closely.
Remark 1.8 The above calculation is comparing the average torque using the
dq expression and using the ellipse area assuming that the system is linear.
Therefore any inaccuracy is due to the numerical inaccuracy of the triangles.
It should be understood that the co-energy diagram for a real machine is not
an ellipse, and one cannot use triangular segments to work out the accurate
instantaneous torques (as noted above). However even in the case of a saturated
machine, triangles are quite satisfactory for the calculation of the average torque
since it is related to the total co-energy area, and not the area of the individual
segments.
The main benet of the ellipse is realised if one is dealing with data for a
real machine i.e. a machine with saturation and slotting eects. Saturation
characteristics can be accounted for using the more traditional dq based formulae
by including saturation functions for the inductances, but slotting eects do not
t into the fundamental sinusoidal assumptions used to derived these types of
models. However, if a machine is designed using nite element (FE) modelling
techniques then the versus i data is available from the FE package. Since
FE modelling uses the exact geometric dimensions of the machine stator and
rotor, as well as the characteristics of the magnetic materials, this data then will
contain the eects of the saturation and slots. Consequently, application of the
ellipse torque calculation technique to this data allows very precise prediction of
not only the average torque from the machine, but also the torque ripple. This
is an invaluable aid in designing a machine.
Reference [5] compares measured values of torque with torque data calcu-
lated using the ellipse technique on the FE design data. The correspondence
between the two sets of results is very good. One interesting point that is made
in [5] is that the saturated ellipse shapes obtained from FE data are almost
exactly the same as the ideal ones. The torque ripple present in a machine
with slotting manifests itself in irregular separation of the curved radial lines
emanating from the centre of the ellipse. These lines each correspond to an
equal angular movement of the rotor, and the co-energy enclosed in each small
triangle is related to the instantaneous torque. In a slotted machine these areas
become very irregular. These eects are shown schematically in Figure 1.16.
Notice the curvature of the ellipses as the current level increases, and the dier-
ent co-energy areas for the same angular movement of the rotor indicating that
the machine has signicant slotting ripple.
1.5 Torque Expressions 1-35
Figure 1.16: Co-energy ellipses for a saturated SYNCREL
Chapter 2
The Kron Primitive Machine
2.1 Introduction
This chapter will develop some basic tools required for the transient analysis of
machines. the modelling approach will be based on the Kron primitive machine.
The Kron primitive is a generalisation of the DC commutator machine. It is
useful because almost all machines can be transformed into a Kron primitive
machine, allowing similar analysis of most machines within a common frame-
work.
The develop the Kron machine we shall develop a model for the doubly
fed machine, then convert this to a DC machine with a commutator, and then
nally develop the basic primitive machine. The basic primitive model will then
be used to analyse the DC machine, with particular emphasis on the separately
excited DC machine.
2.2 Model for the Doubly Fed Machine
The following development is with reference to Figure 1.11. This machine can be
represented schematically as in Figure 2.1. Note that even though this machine
is shown with a salient rotor and stator, we shall be considering the situation
where the either the rotor or the stator is cylindrical. The following voltage
equations can be written for this system:
v
1
= R
1
i
1
+ p
1
(2.1)
v
2
= R
2
i
2
+ p
2
(2.2)
where p denotes
d
dt
.
These expressions can be expanded in terms of self and mutual inductances
to give:
v
1
= R
1
i
1
+ p(L
1
i
1
+ M
12
i
2
) (2.3)
v
2
= R
2
i
2
+ p(L
2
i
2
+ M
21
i
1
) (2.4)
This in turn can be further expanded. Note that we shall assume that L
12
and
M
12
are functions of (which they would be in the case of this doubly salient
2.2 Model for the Doubly Fed Machine 2-2
q
v
2
i
2
i
1
v
1
F
1
F
2
w
( axis) d
Figure 2.1: Magnetic circuit conceptual diagram of a double fed machine
machine), and in turn that is a function of time (i.e. the rotor is moving):
v
1
= R
1
i
1
+ L
1
pi
1
+ i
1
pL
1
+ M
12
pi
2
+ i
2
pM
12
(2.5)
v
2
= R
2
i
2
+ L
2
pi
2
+ i
2
pL
2
+ M
21
pi
1
+ i
1
pM
21
(2.6)
Now considering the following term:
i
1
pL
1
= i
1
dL
1
d
d
dt
= i
1

dL
1
d
(2.7)
where = d/dt and is the angular velocity of the rotor.
Therefore (2.5) and (2.6) can be written as follows:
v
1
= R
1
i
1
+ L
1
pi
1
+ M
12
pi
2
+ i
1
dL
1
d
+ i
2
dM
12
d
(2.8)
v
2
= R
2
i
2
+ L
2
pi
2
+ M
21
pi
1
+ i
2
dL
2
d
+ i
1
dM
21
d
(2.9)
which can be written in matrix form as follows:
v = [R+Lp + G] i (2.10)
where:
v =
_
v
1
v
2
_
i =
_
i
1
i
2
_
R =
_
R
1
0
0 R
2
_
L =
_
L
1
M
12
M
21
L
2
_
2.2 Model for the Doubly Fed Machine 2-3
G =
d
d
_
L
1
M
12
M
21
L
2
_
=
dL
d
Consider these equations for two separate cases: namely a cylindrical stator
and rotor, and a salient pole rotor and a cylindrical rotor.
2.2.1 Zero saliency case
Because both the stator and the rotor are cylindrical then we have the following
situation for the inductances of the machine: Note that the mu-
tual inductance
expression is ap-
proximately true
even for a concen-
trated or uniform
distributed winding
in a DC machine.
L
1
= constant
L
2
= constant
M
12
= M
21
= M
m
cos
Note that the above relationship for the mutual inductance will be assumed
here. Therefore under these conditions the voltage expression for the machine
becomes:
_
v
1
v
2
_
=
__
R
1
0
0 R
2
_
+
_
L
1
M
m
cos
M
m
cos L
2
_
p +
_
0 M
m
sin
M
m
sin 0
___
i
1
i
2
_
(2.11)
This expression describes the complete electrical dynamics for this machine
system. We have not considered the torque produced by the machine as yet,
but when this is included then the mechanical dynamics can also be included
to give a full system description.
2.2.2 One degree of saliency case
In this case we consider the expressions for (2.10) with a salient stator and a
cylindrical rotor. It can be shown that the inductances for this situation are:
L
1
= constant
L
2
= L

2
+ L

2
cos 2
M
12
= M
21
= M
m
cos
Therefore the voltage expressions become:
_
v
1
v
2
_
=
__
R
1
0
0 R
2
_
+
_
L
1
M
m
cos
M
m
cos (L

1
+ L

1
cos 2)
_
p +
_
0 M
m
cos
M
m
cos 2L

1
sin 2
___
i
1
i
2
_
(2.12)
2.2.3 Torque expression
The basic torque expression for this machine was found as (1.104), and is re-
peated here for convenience:
T
e
=
E
fe

i constant
=
1
2
i
2
1
dL
1
d
+
1
2
i
2
2
dL
2
d
+ i
1
i
2
dM
12
d
(2.13)
2.3 Commutator Machines 2-4
q
p
2
-
p
2
p
-p
Statorcoilaxis
Generation
Motoring
-i L
2
2
2
2
''
sin q
-i i M
m 1 2
sinq
Figure 2.2: Torque plot for the double fed machine with DC rotor and stator
currents.
By inspection this expression can be written in matrix form as follows:
T
e
=
1
2
[ i
1
i
2
]
_
dL
1
d
dM
12
d
dM
12
d
dL
2
d
_ _
i
1
i
2
_
(2.14)
which can be written in matrix form as:
T
e
=
1
2
i
T
Gi (2.15)
This equation turns out to be a general expression for virtually all machines. For
example the torque for both the machine types presented above can be found
for any particular stator and rotor currents by substituting for the inductance
terms in the G matrix.
Example: Let us consider a couple of examples of the use of the torque
expression. Substitute into the G matrix the inductance expressions for the one
degree of saliency case, and we get:
T
e
=
1
2
_
i
1
i
2

_
0 M
m
sin
M
m
sin 2L

2
sin 2
_ _
i
1
i
2
_
(2.16)
Consider that both the stator and the rotor are fed with DC currents. If the
torque expression components and their total are plotted then Figure 2.2 results.
Note that the torque is pulsating, and that the average torque over a complete
cycle of rotation is zero. Another observation that one can make from this Note that torque is
positive in the anti-
clockwise direction
diagram is that a net torque can be produced if the current in the either of the
windings (but not both) is reversed at the = 0 or = angles.
2.3 Commutator Machines
An important group of AC and DC machines are those employing a commuta-
tor. Whilst these machines are of considerable practical importance, they are
2.3 Commutator Machines 2-5
+ +
+
+
+ +
+
+
+
+
+
+
+
+
+
F
1
F
1
F
2
F
2
daxis
daxis
qaxis qaxis
q
w
i
2
w
q
i
1
i
2
v
1
v
2
pseudo-stationary
Figure 2.3: A two pole commutator machine
also important from a machine theory point of view as they form the basis of
the generalised primitive machine. The concept of a pseudo stationary coil is
fundamental to this, and this is the main reason that we are revisiting these
machines. At the end of this section we shall have a full dynamic model of the
DC motor.
Figure 2.3 is a physical and conceptual diagram of a two pole commutator
machine. The coil on the rotor is said to be pseudo stationary because the
conductors in the winding are moving (and therefore cutting ux) but the mmf
produced by the winding remains stationary in space. Of course it is the brushes
that allow this magic to occur. The conceptual diagram to the right of Figure 2.3
shows the machine coils represented by concentrated coil equivalents. The coil
on the rotor is now a pseudo-stationary coil.
The equations for this machine can now be written from (2.8) and (2.9)
by realising that the d/d terms in (2.8) will be zero as the rotor coil from a
magnetic point of view is space stationary with respect to the stator coil and
the rotor is round. Therefore the voltage equations expressions are:
v
1
= R
1
i
1
+ L
1
pi
1
+ M
12
pi
2
(2.17)
v
2
= R
2
i
2
+ L
2
pi
2
+ M
21
pi
1
+ i
2
dL
2
d
+ i
1
dM
21
d
(2.18)
Remark 2.1 Note that the expression for the rotor equation is the same as for
the non-commutator machine as the windings in the rotor are still moving at
an angular velocity with respect to the stator eld and consequently have the
same voltages induced in them due to the mutual coupling from the stator and
the variation in rotor magnetic circuit permeance due to the stator saliency. It
is an interesting property of pseudo stationary coils that they have rotational
voltages induced in them from movement, but magnetically they are stationary
(hence the name). Note that based expressions for mutual and self ux do not
change with time and are only dependent on the brush angle.
The torque expression for this machine is a little dierent to that of (2.15) be-
cause of the disappearance of all the d/d terms in the rst voltage equation.
2.3 Commutator Machines 2-6
The torque expression can be calculated by looking at the total dierential en-
ergy input to the system and then assigning various components in the resultant
expression to losses, stored eld energy and mechanical output energy. Now the
incremental energy supplied in time dt is:
dE
e
= (v
1
i
1
+ v
2
i
2
)dt (2.19)
which can be expanded to:
dE
e
= (i
2
1
R
1
+ i
2
2
R
2
)dt
. .
incremental resistive loss
+ [
1
2
L
1
p(i
2
1
) +
1
2
L
2
p(i
2
2
) + M
12
p(i
1
i
2
)]dt
. .
incremental eld storage
+
_
i
1
i
2
dM
12
d
+ i
2
2
dL
2
d
_
dt
. .
incremental mechanical output energy
(2.20)
Using the relationship between torque and power P = T, then one can
write the following expression for the torque of the commutator machine with
stator saliency:
T
e
= i
1
i
2
dM
12
d
+ i
2
2
dL
2
d
T
e
= i
T
Gi (2.21)
where:
i =
_
i
1
i
2
_
G =
_
0 0
dM
12
d
dL
2
d
_
Note that (2.21) is exactly twice the previous general expression (2.15) where
the magnetic axis of the rotor winding was moving with respect to the stator.
Now let us write down the full expression for these equations for the stator
saliency commutator machine. Therefore the inductance values to be used are:
L
1
= constant
L
2
= L

2
+ L

2
cos 2
dL
2
d
= 2L

2
sin 2
M
12
= M
m
cos
dM
12
d
= M
m
sin
and the voltage and torque expressions for a general brush displacement are:
T
e
=
_
i
1
i
2

_
0 0
M
m
sin 2L

2
sin 2
_ _
i
1
i
2
_
(2.22)
_
v
1
v
2
_
=
_
_
_
_
R
1
0
0 R
2
_
+
_
L
1
M
m
cos
M
m
cos L
2
_
p
+
_
0 0
M
m
sin 2L

2
sin 2
_
_
_
_
_
i
1
i
2
_
(2.23)
2.4 The Primitive Machine Concept 2-7
rotor
coil
R
a
L
r
wM I
m f
+ -
I
a
V
a
Figure 2.4: Equivalent circuit of a DC machine in motoring mode
If one considers the situation where the brushes are at = /2 then the
expressions above become:
T
e
= i
1
i
2
M
m
(2.24)
v
1
= i
1
R
1
+ L
1
pi
1
(2.25)
v
2
= i
2
R
2
+ L
2
pi
2
+ M
m
i
1
(2.26)
If one considers the steady state situation, then the normal steady state DC
machine equation can be obtained:
T
e
= I
f
I
a
M
m
(2.27)
V
f
= I
f
R
f
(2.28)
V
a
= I
a
R
a
+ M
m
I
f
(2.29)
where f refers to eld quantities, and a refers to the armature quantities. The
resultant equivalent circuit is shown in Figure 2.4.
We shall see in the next section that the DC commutator machine is the
basis of the Kron primitive machine
2.4 The Primitive Machine Concept
It will be shown in the subsequent chapters that most of the common machine
types can be transformed into a form called the primitive machine or Kron
primitive machine form. The transformed machine has identical performance
characteristics to the original machine. The reasons for carrying out these trans-
formations are:
1. A variety of totally dierent machines can be analysed using the same
primitive machine. The specics of a machine are taken care of by the
transformations to and from the primitive machine.
2. There are a number of ways of doing the transformation, and some of
the techniques result in signicant simplications in the complexity of the
machine models.
2.4 The Primitive Machine Concept 2-8
i
dr
i
qr
i
qs
i
ds
v
ds
v
dr
v
qr
v
qs
qaxis
daxis
w
qs
qr
dr
ds
pseudostationary
Figure 2.5: Primitive dq machine
3. It is often much easier to understand how a machine works in its primitive
form, as much of the complexity present in many machines become hidden
by the transformation process.
The primitive machine is essentially a generalization of the DC commutator ma-
chine. Figure 2.5 shows a diagram of the general layout of a primitive machine.
Note that these machines are sometimes called dq machines as they position the
coils on dq axes. This machine could contain an arbitrary number of coils, but
only two are shown in this case to keep the derivation simple. The nomenclature
of the coil suxes indicates the axis the coil lies on and the supporting magnetic
structure (s for stator and r for rotor).
By analogy with the commutator machine of the previous section one can
proceed to write down the L and G matrices for a stator salient machine of
the form of Figure 2.5. The coil pairs relative to the d axis coils are at 0 and
/2 radians, and relative to the q axis coils are at 0 and /2 radians. Let us
consider the matrices for two coils as in (2.23):
L =
_
L
1
M
m
cos
M
m
cos L
2
_
Because the windings are separated by /2 then the mutual inductance elements
of this matrix are non-zero only on the same magnetic axis (since cos = 0 for
= /2).
Now consider the G matrix:
G =
_
0 0
M
m
sin 2L

2
sin 2
_
2.4 The Primitive Machine Concept 2-9
The top row of this matrix is zero because the rotor coil is pseudo stationary
with respect to the stator coil, and consequently their is no relative movement of
the rotor ux relative to the stator ux, and the stator circuit does no experience
any change in permeance as the rotor rotates. For = 0 and /2, sin 2 = 0
and the only G term of concern is M
m
sin . Therefore the G matrix for the
orthogonal coils becomes:
G =
_
0 0
M
m
0
_
Note that the sign can be deduced by applying Lenzs law to the system. Terms in the G
matrix are present
only when coils are
in space quadrature
and at least one is
pseudo stationary.
We shall now consider the four coil primitive machine case. Writing down
the voltage equations for this system in the most general form we get:
v
ds
= R
ds
i
ds
+
d
dt
(L
ds
i
ds
) +
d
dt
(M
dsdr
i
dr
) +
d
dt
(M
dsqr
i
qr
)
+
d
dt
(M
dsqs
i
qs
) (2.30)
v
dr
= R
dr
i
dr
+
d
dt
(L
dr
i
dr
) +
d
dt
(M
drds
i
ds
) +
d
dt
(M
drqr
i
qr
)
+
d
dt
(M
drqs
i
qs
) (2.31)
v
qs
= R
qs
i
qs
+
d
dt
(L
qs
i
qs
) +
d
dt
(M
qsqr
i
qr
) +
d
dt
(M
qsdr
i
dr
)
+
d
dt
(M
qsds
i
ds
) (2.32)
v
qr
= R
qr
i
qr
+
d
dt
(L
qr
i
qr
) +
d
dt
(M
qrqs
i
qs
) +
d
dt
(M
qrdr
i
dr
)
+
d
dt
(M
qrds
i
ds
) (2.33)
The above expressions can be signicantly simplied using the observations
noted above in relation to the terms that are relevant when the windings angles
are 0 or /2 radians. Applying these principles the equations become:
v
ds
= R
ds
i
ds
+ L
ds
di
ds
dt
+ M
dsdr
di
dr
dt
(2.34)
v
dr
= R
dr
i
dr
+ L
dr
di
dr
dt
+ M
drds
di
ds
dt
+
_
i
qr
dM
drqr
d
+ i
qs
dM
drqs
d
_
(2.35)
v
qs
= R
qs
i
qs
+ L
qs
di
qs
dt
+ M
qsqr
di
qr
dt
(2.36)
v
qr
= R
qr
i
qr
+ L
qr
di
qr
dt
+ M
qrqs
di
qs
dt
+
_
i
dr
dM
qrdr
d
+ i
ds
dM
qrds
d
_
(2.37)
Note that the winding that is the source of the ux is taken as the reference
when evaluating the angles between windings.
2.4 The Primitive Machine Concept 2-10
The above expression can be written in matrix form as follows:
_

_
v
ds
v
dr
v
qs
v
qr
_

_
=
_

_
_

_
R
ds
0 0 0
0 R
dr
0 0
0 0 R
qs
0
0 0 0 R
qr
_

_
+
_

_
L
ds
M
dsdr
0 0
M
drds
L
dr
0 0
0 0 L
qs
M
qsqr
0 0 M
qrqs
L
qr
_

_
. .
L matrix
p
+
_

_
0 0 0 0
0 0
dM
drqs
()
d
dM
drqr
()
d
0 0 0 0
dM
qrds
()
d
dM
qrdr
()
d
0 0
_

_
. .
G matrix
_

_
_

_
i
ds
i
dr
i
qs
i
qr
_

_
(2.38)
Substituting for the mutual inductance expressions one can write the above
equation in matrix form as follows:
v = {R+Lp + G}i (2.39)
where:
R =
_

_
R
ds
0 0 0
0 R
dr
0 0
0 0 R
qs
0
0 0 0 R
qr
_

_
L =
_

_
L
ds
M
dsdr
0 0
M
drds
L
dr
0 0
0 0 L
qs
M
qsqr
0 0 M
qrqs
L
qr
_

_
G =
_

_
0 0 0 0
0 0 M
drqs
L
qr
0 0 0 0
M
qrds
L
dr
0 0
_

_
v =
_

_
v
ds
v
dr
v
qs
v
qr
_

_
i =
_

_
i
ds
i
dr
i
qs
i
qr
_

_
Let us consider the G matrix coecients for a moment. Consider the
dM
drqs
()
d
term. Now M
drds
() = M
drds
cos is the mutual inductance between
the dr and qs windings. In this part of the equation it is being used to give
the voltage induced in the dr winding due to ux generated in the qs winding.
Therefore the angle between the winding is measured relative to the qs winding.
Expanding:
dM
drqs
()
d
= M
drqs
sin (2.40)
where M
drqs
is the maximum mutual inductance between the d axis rotor wind-
ing and the q axis stator winding. In this case the angle is /2 radians,
therefore:
dM
drqs
()
d
= M
drqs
(2.41)
The other term in the G matrix is the
dM
drqr
()
d
type terms. These terms
relate to mutual inductance between two windings on the same magnetic struc-
ture. Therefore if the windings are at an angle of = 0 then the dr and qr
windings are coincident, and assuming perfect couplings under theses condi-
tions then the maximum mutual inductance would be equal to each windings
self inductance.
2.4 The Primitive Machine Concept 2-11
The torque expression for the primitive machine can be calculated using the
expression (2.21):
T
e
= [ i
ds
i
dr
i
qs
i
qr
]
_

_
0 0 0 0
0 0 M
drqs
L
qr
0 0 0 0
M
qrds
L
dr
0 0
_

_
_

_
i
ds
i
dr
i
qs
i
qr
_

_
(2.42)
Expanding and collecting terms:
T
e
= (M
drqs
i
qs
i
dr
M
qrds
i
ds
i
qr
) (L
dr
L
qr
)i
dr
i
qr
(2.43)
The second term in this expression is due at saliency (i.e. dierent permeances
in the d and q axis directions). In a non-salient machine this term would
disappear. The internal mechanical power produced by the machine is easily
calculated from this expression using the expression:
P
e
= T
e
(2.44)
Often (2.39) is written in its expanded form as follows:
_

_
v
ds
v
dr
v
qs
v
qr
_

_
=
_

_
R
ds
+ L
ds
p M
dsdr
p 0 0
M
drds
p R
dr
+ L
dr
p M
drqs
L
qr
0 0 R
qs
+ L
qs
p M
qsqr
p
M
qrds
L
dr
M
qrqs
p R
qr
+ L
qr
p
_

_
_

_
i
ds
i
dr
i
qs
i
qr
_

_
(2.45)
2.4.1 Use of the Primitive Machine - the DC Machine
In this section we shall consider an application of the primitive machine model.
To keep things simple we shall look at a DC machine. As one might suspect the
conversion of the DC machine model to the primitive machine is very straight
forward, since, as we have seen in the previous section the primitive machine
model derivation was based on the DC machine.
Figure 2.6 shows a conceptual diagram of a separately excited DC machine
with a compensating winding or interpole.
In order to develop the primitive machine equations we can apply the rules
mentioned in the previous section. The machine is assumed to have a salient
pole stator and a round rotor (which most machines of this type have). Clearly
from Figure 2.6 one can see that one of the windings present in the primitive
machine is not present in this machine, therefore it has one equation less than
the general primitive machine equation. Let us construct the main inductance
matrices. In order to understand where the elements go in these matrices we
shall nominate the current vector to be:
i =
_
_
i
f
i
c
i
a
_
_
Firstly the L matrix. The only locations that elements appear are for mutual
inductance terms for coils on the same axes and for self inductance terms:
L =
_
_
L
f
0 0
0 L
c
M
ca
0 M
ac
L
a
_
_
2.4 The Primitive Machine Concept 2-12
f
a
c
Compensating
winding
daxis
qaxis
w
(b)Primitivecoildiagram
qaxis
daxis
Compensatingwinding
(a)Physicalarrangement
i
f
v
f
i
c
v
c
i
a
v
a
v
arm
Figure 2.6: Separately excited DC machine with a compensating winding
Now consider the G matrix. These terms occur in pseudo stationary coils
that are orthogonal to a stationary coil or another pseudo stationary coil. The
terms take the general form of M sin where the is measured relative to
the coil that is producing the ux (i.e. the seat of the ux). Therefore in this
particular case the relevant coils are the armature coil (coil a) and the eld coil
(coil f) and the maximum mutual ux if the axes were lying on the same axis
would be M
af
. Therefore we will have rotational voltage terms in the armature
and eld equations. Therefore the G matrix becomes:
G =
_
_
0 0 0
0 0 0
M
af
0 0
_
_
Substituting the R, L and G matrices into the general expression (2.39)
gives the following expression:
_
_
v
f
v
c
v
a
_
_
=
_

_
0.3cm
_
_
R
f
0 0
0 R
c
0
0 0 R
a
_
_
+
_
_
L
f
0 0
0 L
c
M
ca
0 M
ac
L
a
_
_
p
+
_
_
0 0 0
0 0 0
M
af
0 0
_
_
0.3cm
_

_
_
_
i
f
i
c
i
a
_
_
(2.46)
which can be expanded giving:
_
_
v
f
v
c
v
a
_
_
=
_
_
R
f
+ L
f
p 0 0
0 R
c
+ L
c
p M
ca
p
M
af
M
ac
p R
a
+ L
a
p
_
_
_
_
i
f
i
c
i
a
_
_
(2.47)
Now considering the currents we have in the connection shown in Figure 2.6:
i
a
= i
c
and v
arm
= v
a
v
c
(2.48)
2.5 Summary 2-13
Using this expression one can then further simplify (2.47) to give:
_
v
f
v
arm
_
=
_
R
f
+ L
f
p 0
M
af
(R
a
+ R
c
) + (L
a
+ L
c
2M
ac
)p
_ _
i
f
i
a
_
(2.49)
which is the same expression as that for a separately excited DC machine with
an armature inductance of (L
a
+ L
c
2M
ac
). This inductance is small due to
the eect of the compensating winding since L
a
L
c
M
ac
.
The torque expression for this machine can be obtained by using the general
expression for the primitive machine and substituting for the specic G matrix
for this case:
T
e
= i
T
Gi
= [ i
f
i
c
i
a
]
_
_
0 0 0
0 0 0
M
af
0 0
_
_
_
_
i
f
i
c
i
a
_
_
= M
af
i
a
i
f
(2.50)
Note that the torque
is negative since if
i
a
and i
f
are posi-
tive then the torque
is clockwise, which
is negative using
our convention.
The only remaining equation required to obtain a full dynamic model of a
DC machine drive system is the load model. The standard model for a rotating
load is:
J + f() + T
F
= T
e
(2.51)
where:
J rotational moment of inertia
f() linear or non-linear friction coecient
T
F
xed load torque
Equations (2.49), (2.50) and (2.51) constitute a complete dynamic model of the
compensated, separately excited DC machine. These equations can be used to
calculate the response of the system under various conditions short circuited
generator, full voltage start etc. In addition they can be used for the design of
controllers for this machine.
Remark 2.2 If the i
f
current is held constant, then the torque expression be-
comes T
e
= Ki
a
, where i
a
is governed by a rst order linear dierential equa-
tion. One can see why the separately excited DC machine is simple to use in
high performance control systems.
2.5 Summary
This chapter has shown how the primitive machine is derived from the basic
DC commutator machine. Rules have been derived to determine the values of
the various inductances used in the primitive machine model.
The primitive machine forms the basis for the analysis of a whole range of
machines whose winding distributions can be modelled as a sinusoidal funda-
mental without too much error. This assumption applies to common machines
such as the induction machine, the synchronous machine, the synchronous re-
luctance machine, and many brushless DC machines. However this modelling
concept cannot be applied to machines such as the switched reluctance machine.
Chapter 3
Frame transformations, DQ
and Space Vector Models
3.1 Introduction
This chapter develops the basic theory behind transformation of AC machine
models into dq and space vector machine models. In the previous chapter we
developed the primitive machine model from the DC commutator machine. The
transformations presented in this chapter allow an AC machine to be converted
into a primitive machine form. This has the benet of allowing all these machine
types to be analysed using the same techniques. Another advantage is that the
transformed models are often considerably simpler than the non-transformed
model, allowing simpler analysis and a more intuitive understanding of how the
machine works.
The remainder of the chapter is organised as follows. The next major section
will present the basics of dq modelling, for both stationary frame and rotating
frame transformations. An example of how the transformations are applied is
given for the synchronous reluctance machine. The nal major section presents
the other form of dynamic machine equations space vector models, and shows
the relationship between this modelling technique and dq modelling. Once again
the SYNCREL is used as an example.
3.2 dq Models
Most electrical machines with sinusoidally distributed windings are modelled
mathematically using a technique called dq modelling. It is not the purpose of
this chapter to give an exhaustive derivation of dq modelling of machines, as
this could ll a whole text book in its own right. However, a brief overview
of the principles of dq modelling will be presented, and as an example the dq
model for the SYNCREL will be derived.
The fundamental assumption used as the basis of dq modelling is that the
winding distribution in a machine is sinusoidal. In addition a number of other
secondary assumptions are made, which are similar to the assumptions used in
Appendix A, namely:
3.2 dq Models 3-2
120
Figure 3.1: Three phase to two phase transformation
1. The machine does not exhibit stator or rotor slotting eects.
2. The machine iron is linear material, i.e. there is no saturation eects.
The combined eect of these assumption is that linear traditional circuit analysis
techniques can be used to analysis the electrical circuit of a machine. The
sinusoidal assumption means that various spacial quantities in the machine can
be broken into orthogonal components. It is this that is used to carry out
coordinate transformations from the three phase axes of a machine to the two
phase dq axes.
3.2.1 Stationary Frame Transformations
The general idea of a dq type of transformation can be obtained by considering
the transformation of the three phase currents to their two phase equivalents.
Consider Figure 3.1. This shows a conceptual diagram of a three phase ma-
chine. The windings represented by the concentrated coils are actually spacially
sinusoidally distributed windings similar to that shown in Figure 1.1. The lines
through the centre of the coils are the axes of the associated winding mmfs, and
therefore can be thought of as a vector that represents the sinusoidal quantities.
3.2.1.1 MMF transformations
Given that the space distribution of the mmfs for windings a, b and c can be
modelled similarly to (1.2) then the following expressions can be written for the
mmfs:
F
a
(
p
) =

F
a
cos
p
(3.1)
F
b
(
p
) =

F
b
cos
_

2
3
_
(3.2)
F
c
(
p
) =

F
c
cos
_

p
+
2
3
_
(3.3)
3.2 dq Models 3-3
where
p
is as dened in Figure 3.1. The resultant mmf distribution for the
three phase machine is:
F
T
=

F
a
cos
p
+

F
b
cos
_

2
3
_
+

F
c
cos
_

p
+
2
3
_
(3.4)
Assuming that the three phase windings have identical turns, and they are being
driven by three phase currents of the form:
i
a
= I
pk
cos t (3.5)
i
b
= I
pk
cos
_
t
2
3
_
(3.6)
i
c
= I
pk
cos
_
t +
2
3
_
(3.7)
then the peak mmfs for each of the phases are:
F
a
=

NI
pk
cos t (3.8)
F
b
=

NI
pk
cos
_
t
2
3
_
(3.9)
F
c
=

NI
pk
cos
_
t +
2
3
_
(3.10)
where
1

N = N/2.
Therefore (3.4) can be written as:
F
T
=

NI
pk
_
cos t cos + cos
_
t
2
3
_
cos
_

2
3
_
+ cos
_
t +
2
3
_
cos
_
+
2
3
_
_
(3.11)
=
3
2

NI
pk
cos(t ) (3.12)
i.e. the resultant mmf is has a spacial sinusoidal distribution which is rotating
around the machine at t electrical radians per second.
If the vectors associated with (3.1), (3.2) and (3.3) are resolved along two
orthogonal axes called the dq axes then the following expressions can be written
for the resultant dq axes mmfs:
F
s
dq
= TF
abc
(3.13)
i.e.
_
F
s
d
F
s
q
_
=
_
1
1
2

1
2
0

3
2

3
2
_
_
_
F
a
F
b
F
c
_
_
(3.14)
The s superscript on these variables means that the dq axes are in a stationary
frame. The meaning of this will become clearer when we look at rotating frame
transformations later. The s subscript next to the d says that the d axis is
for the stator. An r subscript is used to refer to rotor quantities.
1
Note that this denition for

N results from the fact that the total mmf of the winding is
expended across two airgaps. Therefore the mmf per airgap is half the total mmf.
3.2 dq Models 3-4
Now let us consider what each of the mmfs for the axes are. Firstly expanding
the d axis expression we have:
F
s
ds
=

NI
pk
_
cos t
1
2
cos
_
t
2
3
_

1
2
cos
_
t +
2
3
__
(3.15)
=
3
2

NI
pk
cos t (3.16)
Similarly for the q axis we have:
F
s
qs
=

3
2

NI
pk
_
cos(t
2
3
) cos(t +
2
3
)
_
(3.17)
=
3
2

NI
pk
sin t (3.18)
Notice that the converted machine has time carrying sinusoidal mmfs on each
axis that are 90

out of phase. Note that both the d and q axis windings are also
sinusoidally distributed as well ie. F
s
ds
() = F
s
ds
cos and F
s
qs
() = F
s
qs
sin .
In order to get the resultant space distribution from both these windings we
add together F
s
ds
() and F
s
qs
() similarly to the three phase case. Therefore we
get:
F
T
= F
s
ds
() + F
s
qs
()
=
3
2

NI
pk
(cos t cos + sin t sin )
=
3
2

NI
pk
cos(t ) (3.19)
Therefore the mmf distribution for the two phase machine is exactly the same
as the distribution for the three phase machine.
We now have a technique for going from a three phase machine mmf to an
equivalent two phase machine mmf. However, in order for these transformations
to be very useful we have to have a technique to do the reverse. If we have an
inverse transformation then we can relate values calculated in the two phase
machine back to the three phase machine. Therefore we need to make the
transformation of (3.13) invertible. In order to do this the T matrix and F
s
dq
vector are augmented as follows:
_
_
F
s
d
F
s
q
F
s
_
_
=
_

_
1
1
2

1
2
0

3
2

3
2
1

2
1

2
1

2
_

_
_
_
F
a
F
b
F
c
_
_
(3.20)
i.e. F
s
dq
= SF
abc
(3.21)
and F
abc
= S
1
F
s
dq
(3.22)
3.2 dq Models 3-5
where:
S
1
=
_

_
2
3
0

2
3

1
3
1

2
3

1
3

1

2
3
_

_
=
2
3
_

_
1 0
1

1
2

3
2
1

1
2

3
2
1

2
_

_
i.e. S
1
=
2
3
S
T
(3.23)
The choice of the 1/

2 augmentation of T was made so that the property


in (3.23) was obtained. Note that the F term is zero if the three phase mmfs
contain no zero sequence components, else this term is not zero. Therefore, for
a star connected machine F always equals zero, since one cannot have zero
sequence currents with this conguration.
Remark 3.1 Although the above analysis has been carried out assuming bal-
anced sinusoidal currents the transformation expressions are valid for arbitrary
current waveforms including DC.
Remark 3.2 The above dq mmf is mmf invariant with the three phase mmf.
As we shall see this is not the preferred transformation.
3.2.1.2 Current Transformations
Given the mmf transformation in the previous section, it is a simple matter to
construct the transformation for the three phase currents to their equivalent two
phase currents. This transformation can be handled in two sensible ways. The
transformation could be carried out in such a way that the transformed machine
produces the same total power as the original three phase machine. Such trans-
formations are called power invariant transformations. Another transformation
can be implemented such that the transformed machine produces 2/3rds the
power of the three phase machine. This is one particular example of a power
variant transformation. Usually the power variant transformation is used, since
it turns out that in steady state the two phase currents and voltages have ex-
actly the same amplitude as the phase voltages and currents of the three phase
machine. If the magnitude of the two phase quantity is taken, and then pro-
jected onto the relevant three phase axis, then the instantaneous value can be
found for that phase. This transformation is commonly used in the literature
because of this property. Another advantage of this transformation is that the
per phase inductance values found by the normal testing procedures can be ap-
plied to each of the windings of the two phase machine. As we shall see shortly
this implies that end winding of the two phase machine has the same number
of turns as each individual winding of the three phase machine.
Consider the situation where we desire a power variant transformation the
two phase machine in this situation has 2/3rds the resultant mmf of the three
phase machine. It can be seen from (3.19) that this means that the right hand
3.2 dq Models 3-6
side of (3.21) has to be multiplied by 2/3. Therefore (3.21) and (3.22) can be
written as:
F
s
dqy
=
2
3
SF
abc
(3.24)
F
abc
=
3
2
S
1
F
s
dq
= S
T
F
s
dq
(3.25)
Now consider the mmf expressions expressed in terms of currents and winding
turns:
F
s
dq
=

N
2
i
s
dq
(3.26)
F
abc
=

N
3
i
abc
(3.27)
Using (3.24) one can write:

N
2
i
s
dq
=
2
3
S

N
3
i
abc
(3.28)
where
2
:

N
2
= N
2
/2 the half the number of turns for a winding
of the two phase dq machine.

N
3
= N
3
/2 the half the number of turns for each winding
of the three phase machine.
Since the two phase dq machine is an articial machine of our creation, we
are free to choose the number of turns for each of the windings. Clearly if

N
2
=

N
3
, i.e. the two phase machine has the same number of turns on its windings
as the three phase machine, and the i
s
relationship has the same form as the
mmf relationship above. Consequently the i
s
dq
vector has 2/3rds the magnitude
of the i
abc
resultant current vector. Therefore the current relationships between
the two machines is:
i
s
dq
=
2
3
Si
abc
(3.29)
i
abc
= S
T
i
s
dq
(3.30)
3.2.1.3 Voltage Transformations
Similarly, one can derive the relationship between the three phase and two phase
voltages. Consider the power relationships for the two machines:
P
3
= v
s
T
abc
i
abc
(3.31)
P
2
= v
s
T
dq
i
dq
(3.32)
We want P
2
= 2/3P
3
. Therefore substituting (3.31) and (3.32) into this
expression and using (3.29) one can obtain:
v
s
dq
=
2
3
Sv
abc
(3.33)
v
abc
= S
T
v
s
dq
(3.34)
2
Note again that this denition arise from the mmf per airgap condition which is half the
total mmf.
3.2 dq Models 3-7
Notice that this expression is in the same form as that for the current. Therefore
it has the same property that the magnitude of the voltage vector is 2/3rds that
of the voltage vector for the three phase machine. If one considers the case
where the windings are excited by three phase currents of the form in (3.5),
(3.6) and (3.7), then it is easy to show that:

i
s
dq

= I
pk
(3.35)
i.e. the magnitude of the resultant dq vector is equal to the peak current in a
phase in steady state. Similarly then we can write:

v
s
dq

= V
pk
(3.36)
where V
pk
the peak of three phase sinusoidal voltages supplying the abc
windings. Therefore the use of the 2/3rds power relationship has allowed one to
easily correlate the dq voltages and currents to the abc voltages and currents.
3.2.1.4 Impedance Transformations
Next we need to consider the transformation of the machine parameters be-
tween the three phase and two phase machines. Consider the following general
expressions for the two machines:
v
abc
= Z
abc
i
abc
(3.37)
v
s
dq
= Z
s
dq
i
s
dq
(3.38)
Using (3.37) together with (3.30) and (3.34) one can write:
v
s
dq
=
2
3
SZ
abc
S
T
i
s
dq
(3.39)
Comparing this expression with (3.38) one can see that:
Z
s
dq
=
2
3
SZ
abc
S
T
(3.40)
and Z
abc
=
2
3
S
T
Z
s
dq
S (3.41)
These general impedance transformations can be used to generate specic
transformations for the inductances and resistances for a three phase winding.
For a three phase winding the impedance matrix can be written as:
Z
abc
=
_
_
R
a
+ L
aa
p L
ab
p L
ac
p
L
ba
p R
b
+ L
bb
p L
bc
p
L
ca
p L
cb
p R
c
+ L
cc
p
_
_
(3.42)
where p d/dt.
By inspection it can be seen that the resistive and inductive transformations
become:
R
s
dq
=
2
3
SR
abc
S
T
(3.43)
R
abc
=
2
3
S
T
R
s
dq
S (3.44)
L
s
dq
=
2
3
SL
abc
S
T
(3.45)
L
abc
=
2
3
S
T
L
s
dq
S (3.46)
3.2 dq Models 3-8
To dq
s
To abc
F
s
dqy
=
2
3
SF
abc
F
abc
= S
T
F
s
dq
i
s
dq
=
2
3
Si
abc
i
abc
= S
T
i
s
dq
v
s
dq
=
2
3
Sv
abc
v
abc
= S
T
v
s
dq

s
dq
=
2
3
S
abc

abc
= S
T

s
dq
L
s
dq
=
2
3
SL
abc
S
T
L
abc
=
2
3
S
T
L
s
dq
S
R
s
dq
=
2
3
SR
abc
S
T
R
abc
=
2
3
S
T
R
s
dq
S
Z
s
dq
=
2
3
SZ
abc
S
T
Z
abc
=
2
3
S
T
Z
s
dq
S
Table 3.1: Summary of Stationary Frame Transformations
where:
R
abc
=
_
_
R
a
0 0
0 R
b
0
0 0 R
c
_
_
L
abc
=
_
_
L
aa
L
ab
L
ac
L
ba
L
bb
L
bc
L
ca
L
cb
L
cc
_
_
3.2.1.5 Flux Linkage Transformations
Now that we have the inductance and current transformations it is possible to
develop the transformations for the ux linkages. The ux linkage expressions
for the three and two phase machines are:

abc
= L
abc
i
abc
(3.47)

s
dq
= L
s
dq
i
s
dq
(3.48)
If (3.45) and (3.30) are substituted into (3.48) then one gets:

s
dq
=
2
3
SL
abc
2
3
S
T
Si
abc
=
2
3
SL
abc
i
abc

s
dq
=
2
3
S
s
abc
(3.49)
and
abc
= S
T

s
dq
(3.50)
The stationary frame transformations are summarized in Table 3.1.
3.2.2 Rotating Frame Transformations
The transformation in (3.21) allows the three phase windings to be represented
by an equivalent set of two phase windings. These winding are stationary with
respect to the original three phase winding. It is then possible to project the
stationary two phase winding onto two phase windings that are at some angle
to the stationary winding axes and moving with respect to these axes.
The following discussion is with respect to Figure 3.2. This diagram shows a
rotating dq axes with respect to the stationary dq axes derived in the previous
3.2 dq Models 3-9
section. The angle
sr
is dened with reference to the rotating axis (the subscript
being read as the angle of the stator axis (s) with respect to the rotating axis
(r)) as this makes it easier to see the projections of the stationary quantities
onto this axis. Using the normal convention for angle sign (anticlockwise is
positive angle), one can write the following expressions:
F
r
d
1
= F
s
d
cos
sr
(3.51)
F
r
d
2
= F
s
q
cos(
sr
+

2
) = F
s
q
sin
sr
(3.52)
F
r
q
1
= F
s
q
cos
sr
(3.53)
F
r
q
2
= F
s
d
sin
sr
(3.54)
Clearly the total mmf then on each of the rotating axes is:
F
r
d
= F
r
d
1
+ F
r
d
2
= F
s
d
cos
sr
F
s
q
sin
sr
(3.55)
F
r
q
= F
r
q
1
+ F
r
q
2
= F
s
q
cos
sr
+ F
s
d
sin
sr
(3.56)
This expression can be written more succinctly in matrix form:
_
F
r
d
F
r
q
_
=
_
cos
sr
sin
sr
sin
sr
cos
sr
_ _
F
s
d
F
s
q
_
(3.57)
The zero sequence component can be included by ensuring that it makes no
contribution to the projected vectors as follows:
_
_
F
r
d
F
r
q
F
r

_
_
=
_
_
cos
sr
sin
sr
0
sin
sr
cos
sr
0
0 0 1
_
_
_
_
F
s
d
F
s
q
F
s

_
_
(3.58)
and
_
_
F
s
d
F
s
q
F
s

_
_
=
_
_
cos
sr
sin
sr
0
sin
sr
cos
sr
0
0 0 1
_
_
_
_
F
r
d
F
r
q
F
r

_
_
(3.59)
To make the denition consistent with the angle denition used to dene the
inductance expressions, use
sr
=
rs
, where
rs
is the angle of the rotating
axis with respect to the stationary axis. Therefore the above can be written as:
_
_
F
r
d
F
r
q
F
r

_
_
=
_
_
cos
rs
sin
rs
0
sin
rs
cos
rs
0
0 0 1
_
_
_
_
F
s
d
F
s
q
F
s

_
_
(3.60)
and
_
_
F
s
d
F
s
q
F
s

_
_
=
_
_
cos
rs
sin
rs
0
sin
rs
cos
rs
0
0 0 1
_
_
_
_
F
r
d
F
r
q
F
r

_
_
(3.61)
These relationships can be written in short form as:
F
r
dq
= BF
s
dq
(3.62)
F
s
dq
= B
T
F
r
dq
(3.63)
3.2 dq Models 3-10
Figure 3.2: Two phase stationary to two phase rotating frame transformation
The stationary to rotating frame transformation can be combined with the
three phase to stationary two phase transformation to give the transformation
from a three phase stationary frame to an arbitrary rotating frame. Clearly the
transformations for the mmf are (using (3.24) and (3.25)):
F
r
dq
=
2
3
CF
abc
(3.64)
F
abc
= C
T
F
r
dq
(3.65)
where:
C = BS =
_
_
cos
rs
cos(
rs

2
3
) cos(
rs
+
2
3
)
sin
rs
sin(
rs

2
3
) sin(
rs
+
2
3
)
1

2
1

2
1

2
_
_
(3.66)
C
T
= S
T
B
T
=
_

_
cos
rs
sin
rs
1

2
cos(
rs

2
3
) sin(
rs

2
3
)
1

2
cos(
rs
+
2
3
) sin(
rs
+
2
3
)
1

2
_

_ (3.67)
It can be shown that the all the transformations from the abc frame to
the dq
r
frame have the same form as the stationary frame transformations of
Table 3.1, except that C and C
T
are substituted for S and S
T
respectively, and
the superscript on the variables becomes
r
.
From Faradays law it is possible to express the voltages in terms of rate
of change of ux linkage. In the case of the rotating transformations, this rate
of change can be from two causes; (a) the time rate of change of ux linkage
3.2 dq Models 3-11
To dq
r
To abc
F
r
dqy
=
2
3
CF
abc
F
abc
= C
T
F
r
dq
i
r
dq
=
2
3
Ci
abc
i
abc
= C
T
i
r
dq
v
r
dq
=
2
3
Cv
abc
v
abc
= C
T
v
r
dq

r
dq
=
2
3
C
abc

abc
= C
T

r
dq
L
r
dq
=
2
3
CL
abc
C
T
L
abc
=
2
3
C
T
L
r
dq
C
R
r
dq
=
2
3
CR
abc
C
T
R
abc
=
2
3
C
T
R
r
dq
C
Z
r
dq
=
2
3
CZ
abc
C
T
Z
abc
=
2
3
C
T
Z
r
dq
C
Table 3.2: Summary of Rotating Frame Transformations
caused by the time rate of change of currents, and (b) the rate of change due to
the relative movement of the frames. The general Faraday relationship is:
v
abc
= p
abc
(3.68)
and
abc
= C
T

r
dq
(3.69)
therefore v
abc
= pC
T

r
dq
(3.70)
As can be seen from (3.66), the C matrix is in general a time dependent
matrix, since
rs
could be changing with respect to time. Therefore expanding
(3.70) using the chain rule one gets:
v
abc
= {pC
T
}
r
dq
+C
T
{p
r
dq
} (3.71)
If one expands (3.71) by taking the appropriate derivatives, and then rearranges
the result the following expression can be obtained:
v
abc
= C
T
_
_
_
p
_
_

r
d

r
q

_
_
+
rs
_
_

r
q

r
d
0
_
_
_
_
_
= C
T
v
r
dq
(3.72)
v
r
dq
= p
_
_

r
d

r
q

_
_
+
rs
_
_

r
q

r
d
0
_
_
(3.73)
As we shall see in the next section, (3.73) is the form of the reluctance machine
dq equations.
A summary of the transformations from a stationary frame to a rotating
frame are summarized in Table 3.2.
3.2.3 Example: SYNCREL Linear dq Model
As an example of the use of the above transformations we shall consider the
synchronous reluctance machine (SYNCREL). This machine was chosen partly
out of convenience (I already had the model developed), and also because it
is a relatively simple machine that demonstrates saliency. The transformation
process will be carried out in a two stage process. The reason for this is that
the nature of the two phase stationary frame machine will be exposed, whereas
if the direct transformation to the rotating frame is carried out then this model
is stepped over. The rst step in process is to convert the three phase model
3.2 dq Models 3-12
Figure 3.3: Conceptual diagram of a three phase SYNCREL
of the machine to the two phase model of the machine. The following discus-
sion is with reference to Figure 3.3. This diagram shows a three phase, two
pole SYNCREL. The stationary dq frame is aligned with the d-axis along the
a-phase mmf axis. The rotating d-axis is located along the high permeance axis
of the rotor. Because the SYNCREL is a synchronous machine, the rotor has
to be synchronized with the rotating eld in steady state to produce any useful
torque. Hence this frame is also synchronized with this eld, and is known as
a synchronously rotating reference frame. The synchronously rotating reference
frame has some very important properties that make it the frame that is most
useful for control purposes. It will be seen in this frame that the angle depen-
dence of the inductances disappears, and the currents and voltages become DC
values in steady state.
The most complicated part of the three phase machine to two phase machine
conversion is the inductance transformation, so we shall look at this in detail.
The inductances for this model are calculated in Appendix A and appear in
(A.55) and (A.56). These inductance expressions have to be transformed using
the transformations in Table 3.1. Applying these transformations the inductance
matrix in the stationary frame becomes:
L
s
dq
=
3
2
_
_
2
3
L
l
+ L
1
+ L
2
cos 2 L
2
sin 2 0
L
2
sin 2
2
3
L
l
+ L
1
L
2
cos 2 0
0 0
2
3
L
l
_
_
(3.74)
Equation (3.74) can now be converted to the rotating frame by carrying out
the BL
s
dq
B
T
transformation. After considerable manipulation one arrives at
3.2 dq Models 3-13
the following expression for the dq inductance matrix:
L
r
dq
=
_
_
L
l
+
3
2
(L
1
+ L
2
) 0 0
0 L
l
+
3
2
(L
1
L
2
) 0
0 0 L
l
_
_
(3.75)
If one assumes that the system has no zero sequence currents owing (i.e.
the machine has its winding Y connected for example) then the last column and
row can be deleted from the above matrices. Therefore the relevant matrix for
the dq inductances is:
L
r
dq
=
_
L
l
+
3
2
(L
1
+ L
2
) 0
0 L
l
+
3
2
(L
1
L
2
)
_
(3.76)
Notice in (3.76) that the dependent inductance values of the original three
phase model have been converted to time invariant and independent induc-
tances in the dq frame. This results from the fact that the dq reference frame is
tied to the rotor. If one were measuring the inductance whilst xed to the rotor,
the inductance will not change as the rotor is rotated (assuming a non-salient
stator). In addition, the transformed windings that are xed to this frame do
not see any movement of the rotor from the moving d-axis, and therefore the
mutual inductance term to the orthogonal winding is zero. A consequence of
this simplication of the inductances is that the dq frame dynamic equations
are much simpler than the three phase equations.
In (3.73) we calculated the generic form of the dq dynamic equations taking
into account only the voltage terms due to the ux linkages. If the three phase
conversion process is carried out for the resistance it can be shown that the dq
values are identical to the three phase values. Therefore, the generic dq equation
can be rewritten in the following form if we include the resistive drop term and
use the fact that the dq inductances are time invariant:
v
r
dq
=
_
R 0
0 R
_
i
r
dq
+
_
L
r
d
0
0 L
r
q
_
pi
r
dq
+
rs
_
L
r
q
0
0 L
r
d
_
i
r
qd
(3.77)
which can be written in scalar form as:
v
r
d
= Ri
r
d
+ L
r
d
di
r
d
dt

rs
L
r
q
i
r
q
v
r
q
= Ri
r
q
+ L
r
q
di
r
q
dt
+
rs
L
r
d
i
r
d
_
(3.78)
where:
L
r
d
= L
l
+
3
2
(L
1
+ L
2
)
L
r
q
= L
l
+
3
2
(L
1
L
2
)
Equation (3.78) is shown in diagram form in Figure 3.4.
It should be noted that the magnitude of the total ux linkage for the SYN-
CREL can be written in terms of the d and q-axis inductances as follows:
=
_
(L
r
d
i
r
d
)
2
+ (L
r
q
i
r
q
)
2
(3.79)
The other relevant part of the machine model is the torque. The torque
expression (1.104) can be used as the basis for the development of the torque
3.2 dq Models 3-14
R
R
pd q
r
q
r
L i
pd d
r
d
r
L i
+
+
-
-
L
d
r
L
q
r
v
d
r
v
q
r
d-axis
q-axis
i
d
r
i
q
r
Figure 3.4: Model for the ideal dq equations
for the linear reluctance machine. The expression developed in (1.104) was for
a system where the rotor had a single excitation winding and the stator a single
winding. However, the location of the second coil does not have to be on the
rotor, and it can be the q-axis coil instead. Obviously the self and mutual
coupling terms for the coils will now be the same as those in the stationary
frame dq model derived above.
The expression for the torque is (using (1.104)):
T
e
=
1
2
i
s
2
d
dL
s
d
d
rs
+
1
2
i
s
2
q
dL
s
q
d
rs
+ i
s
d
i
s
q
dL
s
dq
d
rs
(3.80)
where the inductance terms are dened as in (3.74). Taking the derivatives in
this expression, and introducing the 3/2 factor to account for three phases, we
get the following expression for the torque in terms of the stationary frame dq
currents:
T
s
e
=
3
2
_
(i
s
2
q
i
s
2
d
)L
2
sin 2
rs
+ 2L
2
i
s
d
i
s
q
cos 2
rs
_
(3.81)
Using the relationship:
i
s
dq
= B
T
i
r
dq
(3.82)
one can substitute for i
s
d
and i
s
q
in (3.81) in terms of i
r
d
and i
r
q
, and obtain:
T
r
e
=
3
2
2L
2
i
r
d
i
r
q
=
3
2
(L
r
d
L
r
q
)i
r
d
i
r
q
(3.83)
3.2 dq Models 3-15
All of the analysis thus far has been for a single pole pair machine. A multiple
pole machine only requires a slight modication to the torque expression, and the

rs
term is in electrical radians per second in the dynamic equation. Therefore
for a p
p
pole pair machine the torque expression is:
T
r
e
=
3
2
p
p
(L
r
d
L
r
q
)i
r
d
i
r
q
(3.84)
The torque expression could also be found by applying the general torque
expression derived for the primitive machine (2.21), repeated here for conve-
nience:
T
e
= i
T
Gi (3.85)
where:
i =
_
i
r
d
i
r
q
_
and G =
_
L
r
q
0
0 L
r
d
_
Substituting for these vectors into the general torque expression:
T
e
= [ i
r
d
i
r
q
]
_
L
r
q
0
0 L
r
d
_ _
i
r
d
i
r
q
_
= [ i
r
d
i
r
q
]
_
L
r
q
i
r
d
L
r
d
i
r
q
_
= (L
r
d
L
r
q
)i
r
d
i
r
q
(3.86)
as was previously obtained. We can introduce the
3
2
p
p
factor to give the same
torque as a p
p
pole pair three phase machine.
The only other transformation of immediate interest that has not been ex-
plicitly carried out is the current transformation. It was eluded to in Sec-
tion 3.2.1.3 that one property of the rotating transformations was that the
magnitude of the current and voltage vectors was equal to that of a single phase
of the three phase machine in steady state. Another property that occurs is
that in steady state is that i
r
d
and i
r
q
have DC values if the dq-axes are rotating
synchronously with the rotor. To formally show these properties consider the
abc machine is being driven by currents of the form:
_
_
i
a
i
b
i
c
_
_
=
_
_
I
pk
cos(
rs
+ )
I
pk
cos(
rs
+
2
3
)
I
pk
cos(
rs
+ +
2
3
)
_
_
(3.87)
These currents are synchronized to the rotation of the rotor, and consequently
so is the resultant current vector. Carrying out the transformation from the abc
frame to the dq stationary frame we get:
i
s
dq
=
_
_
I
pk
cos(
rs
+ )
I
pk
sin(
rs
+ )
0
_
_
(3.88)
and the further transformation to the dq rotating frame gives:
i
r
dq
=
_
_
I
pk
cos
I
pk
sin
0
_
_
(3.89)
3.3 Space Vector Model 3-16
Notice that if the phase angle is zero then the q-axis current is zero, and
all the current lies in the d-axisi.e. along the high permeance axis of the rotor.
If the peak value of the abc currents are constant then we have a constant
amplitude D.C. value equal to the abc phase amplitude, in the dq-axes.
3.3 Space Vector Model
An alternative method for modelling machines that has become popular is the
space vector technique. This method of modelling is very similar to the dq
modelling technique, and in fact it is very simple to convert between the two
dierent types of models. The main reason for the popularity of the technique
has been the growth in vector based control techniques, since this modelling
method naturally ts this view of the machine. Its main advantage is that a
simpler notation can be used for machine equations. For example, the electrical
dynamics of an induction machine can represented by two equations (instead of
four with a dq model). The form of the equations also evokes a resultant vector
way of thinking about the machines operation, as opposed to a component
vector approach with the dq modelling technique. A full discussion of space
vectors applied to the control of machines can be found in [6]. The application
of space vectors to the reluctance machines has not been as pervasive as it has
with induction machines because the reluctance machine more naturally relates
to a component viewpoint. This is due to the presence two dierent permeance
axes in the machine. However, in some situations space vectors are a useful
tool for viewing this machines operation, and it is therefore justied to have
a brief view of the space vector concepts applicable to reluctance machines. It
should be emphasized that because the reluctance machine does not have any
rotor winding we have no need to develop rotor expressions, as is the case with
the induction machine.
Space vector modelling is based on the concept that the mmf of a three
phase machine can be represented by a resultant vector that has a physical
location in space. This stems from the fact that the individual windings of the
phases are sinusoidally distributed, and the vector for each of the windings can
be considered to lie on the axis of the phases. It should be noted that the dq
modelling developed in the previous sections used similar assumptions, but the
modelling approach was dierent.
3.3.1 Current Space Vectors
3.3.1.1 Stationary Frame Current Vectors
In a manner similar to (3.4) we can write the following expression for the resul-
tant mmf in a three phase machine:
F
s
T
= N
3
[i
a
(t)cos + i
b
(t)cos(
2
3
) + i
c
(t)cos(
4
3
)] (3.90)
where is the angle from the axis of the a-phase winding as previously.
The notational simplicity of the space vector formulation is obtained by
introducing complex notation. In the following equations the _ is used to
3.3 Space Vector Model 3-17
denote vectors in the complex form. Equation (3.90) can be written as:
F
s
T
= N
3
Re[i
a
(t)e
j
+ i
b
(t)e
j(2/3)
+ i
c
(t)e
j(4/3)
]
=
3
2
N
3
2
3
Re
_
[i
a
(t) + ai
b
(t) + a
2
i
c
(t)]e
j
_
(3.91)
where a = e
j2/3
, and is a vector of unit length lying spatially along the axis of
the b-phase. Similarly a
2
= e
j4/3
, and lies along the c-phase axis. Notice that
this complex notation implicitly means that we have a set of pseudo dq axes,
which now correspond to the real and imaginary axes.
Now consider the central part of the above mmf expression, namely:
i
a
(t) + ai
b
(t) + a
2
i
c
(t) =
_
i
a
(t)
1
2
i
b
(t)
1
2
i
c
(t)
_
+ j

3
2
[i
b
(t) i
c
(t)] (3.92)
which is the same expression that is obtained for the three phase to two phase
transformations for the dq model, where the imaginary axis expression corre-
sponds to the q axis expression in the dq model.
Therefore let us dene the current vector as follows:
i
s
=
2
3
[i
a
(t) + ai
b
(t) + a
2
i
c
(t)]
= |i
s
| e
j
s
(3.93)
Using this denition of the current vector we can write the expression for
the mmf in the machine as:
F
s
T
=
3
2
N
3

_
i
s
e
j
_
(3.94)
Figure 3.5 shows pictorially (3.93) means. The |i
s
| vector is the magnitude of
the resultant current vector. Notice that the direction of this vector is spatially
the same direction as the original mmf vector (since the two are related by a
scalar). The
s
angle is the angle of this vector with respect to the reference
a-phase axis. If one were to add together the i
a
, i
b
, and i
c
current vectors
graphically on this diagram, the resultant current vector would have the angle

s
but be 3/2 times the magnitude. The 2/3rd term was introduced into (3.91),
and then carried into (3.93), since the resultant current vector has the property
that the vector can be directly projected back onto the three phase axes. This
is the same situation as with dq modelling in that the 2/3 factor allows the
currents in the two phase space vector representation to be directly correlated
with the phase currents of the three phase machine. It should be noted that
implicit in this projection is that there are not zero sequence currents owing
(i.e. i
a
+ i
b
+ i
c
= 0). It can also be shown that under this restriction that
the space vector to three phase projections can be represented by the following
relationships:
(i
s
) = i
a
(3.95)
(a
2
i
s
) = i
b
(3.96)
(ai
s
) = i
c
(3.97)
3.3 Space Vector Model 3-18
Figure 3.5: Resolving the current space vector onto the abc axes
In the particular case where the currents are of the form (3.53.7) then the
space vector can be written as follows:
i
s
=
2
3
I
pk
[cos t + (cos
2
3
+ j sin
2
3
) cos(t
2
3
)
+ (cos
4
3
+ j sin
4
3
) cos(t +
2
3
)]
=
2
3
I
pk
[cos t
1
2
(
1
2
cos t +

3
2
sin t)
+ j

3
2
(
1
2
cos t +

3
2
sin t)
1
2
(
1
2
cos t

3
2
sin t)
j

3
2
(
1
2
cos t

3
2
sin t)
= I
pk
(cos t + j sin t)
i
s
= I
pk
e
jt
(3.98)
Therefore the resultant current vector has a constant magnitude and the angle

s
is changing at the constant rate of , i.e. the vector is rotating around the
machine at a constant angular frequency.
The space vector representation can be simply related back to the dq repre-
3.3 Space Vector Model 3-19
Figure 3.6: Relationship between the dq axes and the current space vector.
sentation. From (3.29) is can be seen that:
i
s
d
=
2
3
_
i
a

1
2
i
b

1
2
i
c
_
(3.99)
i
s
q
=
1

3
[i
b
i
c
] (3.100)
If one takes the real and imaginary components of (3.93) then one can write the
following:
(i
s
) =
_
2
3
(i
a
+ ai
b
+ a
2
i
c
)
_
=
_
2
3
(i
a
+ (cos
2
3
+ j sin
2
3
)i
b
+ (cos
4
3
+ j sin
4
3
)i
c
)
_
=
2
3
_
i
a

1
2
i
b

1
2
i
c
_
= i
s
d
(3.101)
Similarly:
(i
s
) =
1

3
[i
b
i
c
] = i
s
q
(3.102)
These projections can be seen in Figure 3.6. Note that the dq projections are
not as restrictive as the projections onto the abc axes. For example, if there
are zero sequence currents then (3.101) does not equal (3.95). Zero sequence
currents require the presence of an additional space vector equation, as was
the case with the dq equations. However, the discussion in this book shall be
focussed on balanced (and usually Y connected) machines that do not have zero
sequence current components.
3.3 Space Vector Model 3-20
Figure 3.7: Space vector rotating frame transformations
3.3.1.2 Rotating Frame Current Vectors
Similar space vector expressions can be derived for frames that are not stationary
to the rotor. Consider the situation shown in Figure 3.7. Here we have the
original i
s
vector as in Figure 3.6, as well as the same vector projected onto
another frame which is possibly rotating.
The current vector can be written with reference to the rotating frame as:
i
r
= |i
r
| e
j
r
= |i
r
| e
j(
s

rs
)
Now i
s
= |i
s
| e
j
s
= |i
r
| e
j(
rs
+
r
)
(as |i
s
| = |i
r
| )
= |i
r
| e
j
r
e
j
rs
i
s
= i
r
e
j
rs
= i
r
e
j
sr
(3.103)
The sign of the angle in (3.103) is dependent on the reference axis for the an-
gle dierence between the two reference frames. The normal convention adopted
is that the old frame is taken as the reference, therefore the sign convention is:
x
new
= x
old
e
j
newold
(3.104)
where:

newold
the angle between the new and old axes
with reference to the old axis.
This relationship is general and can be applied to all space vector axis trans-
formations.
In the case of the transformation in (3.103) we are transforming from the
rotating axis to the stationary axis. Applying the rule above, the angle from
3.3 Space Vector Model 3-21
the old axis to the new axis is
rs
(or
sr
). Therefore applying (3.103):
i
s
= i
r
e
j
sr
(3.105)
3.3.2 Flux Linkage Space Vector
The total ux linking the phases in a three phase machine are:

a
= L
aa
i
a
+ L
ab
i
b
+ L
ac
i
c
(3.106)

b
= L
bb
i
b
+ L
ba
i
a
+ L
bc
i
c
(3.107)

c
= L
cc
i
c
+ L
ca
i
a
+ L
cb
i
b
(3.108)
Dene the ux linkage space vector as follows:

s
=
2
3
(
a
+ a
b
+ a
2

c
) (3.109)
The justication for the denition of the space ux vector is that the funda-
mental of the ux linkage to a single phase varies as a sinusoidal function of the
current angle to the axis of any particular phase. This was shown in Chapter 1.
Therefore the ux linkage has similar sinusoidal properties to the mmf of the
machine, and the same techniques can therefore be applied.
Substituting (3.1063.108) into (3.109)and assuming that:
L
ab
= L
ba
L
ac
= L
ca
L
bc
= L
cb
after a small amount of manipulation gives:

s
=
2
3
_
_

1
2

1
2

c
_
+ j

3
2
(
b

c
)
_
(3.110)
which after further manipulation gives:

s
=
2
3
_
1
2
_
(2L
aa
L
ab
L
ac
) i
a
+ (2L
ab
L
bb
L
bc
) i
b

+ (2L
ac
L
cc
L
bc
) i
c
_

+ j

3
2
_
(L
bb
L
bc
)i
b
+ (L
ab
L
ac
)i
a
+ (L
bc
L
cc
)i
c
__
(3.111)
This expression can be simplied greatly for a cylindrical rotor machine (i.e.
the self inductances are equal, and the mutual inductances are equal) . Dening:
L
aa
= L
bb
= L
cc
= L
s
and
L
ab
= L
ba
= L
ac
= L
ca
= L
bc
= L
cb
= M
ss
where M
ss
is the mutual inductance between 120

separated phases, then we


can write (3.111) as:

s
= (L
s
M
ss
)
_
2
3
_
(i
a

1
2
i
b

1
2
i
c
) + j

3
2
(i
b
i
c
)
__
=

L
s
i
s
(3.112)
3.3 Space Vector Model 3-22
where

L
s
= L
s
M
ss
. In the case of a normal three phase machine M
ss
=
M
m
cos 120

=
1
2
M
m
. Note that ignoring leakage M
m
= L
s
. Therefore

L
s
=
3
2
L
s
. This is the three phase equivalent inductance. We shall look at the
induction machine
in detail in a fol-
lowing chapter
Remark 3.3 Note that even though the inductance term is
3
2
L
s
in (3.112), the
ux linkage
s
is still 2/3rds that produced by the three phase machine as the
current vector is still 2/3rds the magnitude of the three phase current vector.
For the reluctance machine the expression is much more complicated. Substitut-
ing the inductance expressions (A.55) and (A.56) into (3.111) letting
rs
=
d
,
and after considerable manipulation one obtains the following expression for the
ux space vector in a stationary reference frame:

s
= (L
1
+ L
2
cos 2
rs
)i
a
+ (
L
1
2
+ L
2
cos 2(
rs


3
))i
b

+ (
L
1
2
+ L
2
cos 2(
rs
+

3
))i
c

+ j
1

3
_
(

3L
2
sin 2
rs
)i
a
+ (
3
2
L
1
L
2
(
3
2
cos 2
rs
+

3
2
sin 2
rs
))i
b
(3.113)
+ (
3
2
L
1
+ L
2
(
3
2
cos 2
rs

3
2
sin 2
rs
))i
c
_
(3.114)
The validity of this expression can be checked as follows. If (3.114) is calculated
for
rs
= 0 and i
a
= I
pk
, i
b
= I
pk
/2, i
c
= I
pk
/2, (i.e. the mmf vector lies
coincident with the a-phase) then the real part of the inductance is 3/2(L
1
+L
2
)
as expected from the dq analysis. A similar result can be found for the imaginary
component for
rs
= 90

, i
a
= 0, i
b
=

3/2I
pk
, i
c
=

3/2I
pk
(in this case the
mmf is at 90

and the rotor d-axis is also in this position).


If the currents are in the form of (3.87) then (3.114) can be simplied to the
following expression:

3 currents
=
3
2
I
pk
_
(L
1
cos(
rs
+ ) + L
2
cos (
rs
))
+ j (L
1
sin(
rs
+ ) + L
2
sin(
rs
))
_
=
3
2
I
pk
_
L
1
e
j(
rs
+)
+ L
2
e
j(
rs
)
_
(3.115)
This special case for the currents has been chosen because it is the form of the
currents that are applied to the machine when it is being vector controlled. It
can be seen from that the resultant current is synchronized to the rotor position
such that the resultant current vector has an angle of radians with the rotor
high permeance axis. Note that similar to the cylindrical rotor case the current
amplitude only appears as I
pk
in (3.115), the 3/2 term being from the inductance
part of the expression. Therefore the magnitude of the ux linkage is 2/3rds of
the ux linkage for the three phase machine, as was indicated from the denition
of the ux linkage expression (3.109).
3.3 Space Vector Model 3-23
3.3.3 Voltage Space Vector
In a manner analogous to the vector denitions of the current vector and ux
vector one can dene the voltage vector:
v
s
=
2
3
(v
a
+ av
b
+ a
2
v
c
) (3.116)
where v
a
,v
b
,v
c
are the individual phase voltages. As with the current vectors,
if there are no zero sequence voltages then the individual phase voltages can
be simply obtained from the voltage space vector using a similar expression to
(3.95), (3.96) and (3.97):
Re(v) = v
a
(3.117)
Re(a
2
v) = v
b
(3.118)
Re(av) = v
c
(3.119)
The concept of the voltage space vector is quite abstract. However, its
existence can be justied from the vectors already dened. The voltage in
a machine is made up of two components; the resistive drop, and the induced
voltage from changing ux linkages. We have already dened the current vector,
and the resistive drop is simply this vector multiplied by the resistance (which is
a scalar). The ux linkages have also been dened as a vector, and taking their
derivative in a vector sense also results in a vector. Therefore both components
of the voltage are vectors, and consequently the voltage can be considered to be a
vector. It is easy to demonstrate that if one takes components of a voltage vector
for a set of three phase windings one does get the individual abc components
of the voltages. Note that this process requires that there are no zero sequence
voltages present.
Voltage in across a coil is usually related to two main things in an electrical
circuit current through the coil resistance, or a changing linkage through the
coil (due either to time or spatial variations). For the concept of a voltage vector
to be useful it should give the correct results under these two conditions. Now
consider the situation where we have the three coils of a three phase circuit
carrying DC currents. Therefore the individual currents in the windings are
i
a
, i
b
, and i
c
. Now if the voltage vector concept is to useful then:
v = iR (3.120)
Now writing down the voltage vector using its denition and the individual
phase voltages under this condition:
v =
2
3
(i
a
R + ai
b
R + a
2
i
c
R)
= R
2
3
(i
a
+ ai
b
+ a
2
i
c
)
= iR (3.121)
Note that under this excitation the voltage vector and the current vector are in
the same spatial direction.
The other case to consider is when the coil is excited by a change in ux
linkage. As mentioned previously there are two situations that occur the coil
3.3 Space Vector Model 3-24
is subjected to a time varying ux linkage which is stationary in space, and
the coil is subjected to a moving sinusoidal ux density waveform. We shall
consider each of these. Firstly the time varying ux linkage. Now, realising
that is constant we can write:
d
s
dt
=
d
dt
_

e
j
_
=
d

dt
e
j
= |v
s
| e
j
= v
s
(3.122)
Therefore the voltage vector is in the same direction as the ux vector under
this condition.
The other situation to consider is when we have a spatial sinusoidally dis-
tributed ux distribution moving with respect to the coils. The best way to
approach this case is to again use the complex form of the ux linkage expres-
sion:

s
=

e
j(t)
where (t) is the time varying vector angle
=

e
jt
(3.123)
Therefore we can write:
v
s
=
d
s
dt
=

d
dt
e
jt
= j

e
jt
= j
s
(3.124)
Therefore the voltage vector is 90

out of phase with the ux linkage vector. As


can be seen from the above expression the voltage vector leads the ux linkage
vector by 90

, which is the same result that is obtained in the temporal domain


for sinusoidal variation of ux linkages in an inductor.
3.3.4 Example: SYNCREL Space Vector Model
We have assembled enough of the space vector model machinery to construct
the space vector electrical model for the SYNCREL. This model is very simple,
in fact about as simple a model as one can get for a machine. The simplicity
results from the space vector notation. A SYNCREL has only one set of three
phase windings on the stator, therefore the expression for the stator voltage in
space vector notation using stationary frame variables is:
v
s
= Ri
s
+
d
s
dt
(3.125)
It is a straight forward process to verify this expression from the denitions
already presented for the various space vectors.
Evaluation of the voltage from (3.125) is complex due to the nature of the
ux linkage term in a stationary reference frame. A great simplication can
3.3 Space Vector Model 3-25
be achieved by converting this expression into a rotating frame synchronized
with the rotor (as was done with the dq equations). Applying (3.104) to the
voltage, current and ux linkage vectors, one can write the following relationship
between the stationary and rotating reference frame vectors:
v
s
= v
r
e
j
rs
i
s
= i
r
e
j
rs

s
=
r
e
j
rs
_
_
_
(3.126)
Substituting (3.126) into (3.125) gives:
v
r
e
j
rs
= Ri
r
e
j
rs
+
d
dt
_

r
e
j
rs
_
= Ri
r
e
j
rs
+
d
r
dt
e
j
rs
+
r
e
j
rs
j
d
rs
dt
v
r
= Ri
r
+
d
r
dt
+ j
rs

r
(3.127)
where
rs
=
d
rs
dt

rs
angle of rotating frame wrt stationary frame (3.128)
Assuming that the currents being applied to the machine are of the form
(3.87) then it is not dicult to show that:
i
s
= I
pk
e
j(
rs
+)
(3.129)
Therefore the rotating frame current space vector is, using (3.103):
i
r
= I
pk
e
j
(3.130)
Applying a similar transformation to (3.115), one obtains:

r
=
3
2
I
pk
[(L
1
+ L
2
) cos + j(L
1
L
2
) sin ]
=
3
2
I
pk
_
L
1
e
j
+ L
2
e
j

(3.131)
3.3.5 Space Vector Power Expression
Now that we have the space vector representations for the dynamic equations
we are in a position to calculate the input stator power for the machine in terms
of space vectors. Assuming that there are no zero sequence components the
following expression can be written for the three phase power of the machine:
P
3
= v
a
i
a
+ v
b
i
b
+ v
c
i
c
= (v
s
)(i
s
) +(a
2
v
s
)(a
2
i
s
) +(av
s
)(ai
s
) (3.132)
This equation can also be written in a more compact form:
P
3
=
3
2
(v
s
i

s
) (3.133)
where the means complex conjugate (ie. i

s
=
2
3
(i
a
+a

i
b
+a
2

i
c
) ). Because
the space vectors are closely related to the time domain phasors in steady state
3.3 Space Vector Model 3-26
the similarity of this expression with the time domain complex power expression
should not be surprising. This expression can be conrmed by the following
expansion:
(v
s
i

s
) =
_
_
2
3
_
2
_
v
a
+ av
b
+ a
2
v
c
_
_
i
a
+ a

i
b
+ a
2

i
c
_
_
(3.134)
=
_
_
2
3
_
2 _
3
2
v
a
i
a
+
3
2
v
b
i
b
+
3
2
v
c
i
c

j

3
2
(v
a
i
a
+ v
b
i
b
+ v
c
i
c
)
__
=
2
3
(v
a
i
a
+ v
b
i
b
+ v
c
i
c
) (3.135)
As can be seen from (3.135) the space vector representation of the machine is
absorbing 2/3rds the power of the three phase machine. Hence the space vector
transformations we have developed are power variant transformations, as was
the case for the dq transformations.
It should also be noted that power expressions are reference frame indepen-
dent (as one would naturally expect if the reference frame concept was to be
useful). This can be shown as follows. Let use convert the voltage and current
vectors in the previous power expression to an arbitrary reference frame at some
angle , with respect to the ie.
v

s
= v
s
e
j
i

s
= i

s
e
j
Substituting these expressions into (3.133) we get:
(v

s
i

s
) =
_
_
2
3
_
2
_
v
a
+ av
b
+ a
2
v
c
_
e
j
_
i
a
+ a

i
b
+ a
2

i
c
_
e
j
_
=
_
_
2
3
_
2
_
v
a
+ av
b
+ a
2
v
c
_
_
i
a
+ a

i
b
+ a
2

i
c
_
_
which is the same as (3.134). Therefore the power is invariant in dierent
reference frames.
3.3.6 Example: Space Vector Expression for SYNCREL
Torque
The power expression developed above can be used as a means to calculate the
torque produced by the machine. A general expression for torque is:
T = P (3.136)
This expression can be used to develop the electro-magnetic torque for the space
vector model of the machine by utilizing the energy balance expressed in (1.44).
If one can identify the loss and eld storage terms then they can be subtracted
from the total input energy to give the mechanical output energy. This can then
be substituted into (3.136) to give the electromagnetic torque.
3.3 Space Vector Model 3-27
Consider the expression (3.125). The power expression for the machine can
be written using the relationship (3.133) as follows:
P
3
=
3
2
(v
s
i

s
) =
3
2

__
Ri
s
i

s
+
d
s
dt
i

s
__
(3.137)
Clearly the Ri
s
i

s
term is related to the power losses in the machine, therefore
the
d
s
dt
i

s
term must be related to stored eld energy and mechanical output
power. Considering the last term for the special case of currents in the form
(3.87), with I
pk
constant with respect to time, and using (3.115) we can write:
d
s
dt
=
d
dt
_
3
2
I
pk
_
L
1
e
j(
rs
+)
+ L
2
e
j(
rs
)
_
_
= j
rs
_
3
2
I
pk
_
L
1
e
j(
rs
+)
+ L
2
e
j(
rs
)
_
_
= j
rs

s
(3.138)
where
rs
=
d
rs
dt
Therefore the power expression under this steady state condition becomes:
P
3
=
3
2

__
Ri
s
i

s
+ j
rs

s
i

s
__
(3.139)
Clearly there is only one term related to the rotational power and that is
j
rs

s
i

s
. Expanding this using i

s
= I
pk
e
j(
rs
+)
and (3.115) one gets:
P
3
=
3
2

_
3
2
j
rs
I
2
pk
_
L
1
+ L
2
e
j2
_
_
=
9
4

rs
I
2
pk
L
2
sin 2 (3.140)
T
e
=
9
4
I
2
pk
L
2
sin 2 (3.141)
If we remove the restriction that I
pk
has to be constant, then we would end
up with Ldi/dt type terms in (3.138). These terms are not related to in any
way, and result in change of stored eld energy terms in (3.137). Therefore
(3.140) is valid for transient conditions as well as for steady state.
The same expression can be obtained if the torque is calculated using the
rotating reference frame expression of (3.127). In this case the rotational power
term is even more easily identied. Consider the power expression in this frame:
P
3
=
3
2

_
j
rs

r
i

r
_
=
3
2

_
j
rs
3
2
I
pk
_
L
1
e
j
+ L
2
e
j

I
pk
e
j
_
=
9
4
I
2
pk

_
j
rs
L
1
+ j
rs
L
2
e
j2
_
=
9
4

rs
I
2
pk
L
2
sin 2 (3.142)
3.3 Space Vector Model 3-28
which is the same as the power in the stationary frame case. Therefore the power
and torque produced is reference frame independent (as one should expect).
It is possible to develop a more general form of the mechanical power/torque
expression. In the two examples above the key expression for the rotational
power has the form:
P
rot
=
3
2

_
ji

_
(3.143)
Regardless of whether the variables are in a rotating frame or a stationary frame
the current and ux have the following form:
=

e
j
and i = |i| e
j
therefore:
P
rot
=
3
2

_
j

|i| e
j()
_
=
3
2

_
j

|i| (cos( ) + j sin( ))


_
=
3
2
(j

|i| cos( )

|i| sin( ))
=
3
2

|i| sin( )
P
rot
=
3
2
i (3.144)
and T
rot
=
3
2
i (3.145)
Remark 3.4 The above general expressions for power and torque are the same
expressions as rotational power and torque for the DC machine. In this machine
the torque produced is of the form:
T
rot
= K
f
i
a
(3.146)
where i
a
is the armature current and
f
is the ux linkage due to the eld. In
a DC machine the physical arrangement with the commutator is such that the
spatial angle between these two values is 90

. Therefore the DC machine torque


expression is the same at that for an AC machine with a right angle between
the current and ux vectors. As we shall see in the next chapter this torque
expression is general for all AC machines satisfying sinusoidal assumptions.
Remark 3.5 In the power/torque expressions above the
d
s
dt
has a contribution
to the rotational power because of the saliency of the rotor. The rotational terms
result because
s
has a spatial component due to the inductance variation with
rotor position. This situation does not occur with round rotor machines such as
the induction machine.
3.3.7 Relationship Between Space Vectors and dq Models
Clearly the space vector model and the dq model of a machine are very closely
related. The Real and Imaginary axes of the space vector model can be con-
sidered to be the same as the dq axes. Therefore, by taking the components
of the space vectors (i.e. taking the Re and Im parts) onto these axes one can
3.4 Steady State Model 3-29
obtain the dq representation of the variable or equation. For example, consider
the (3.115) representation for the ux linkage. Taking the Re and Im parts we
obtain:

3 currents
_
=
3
2
I
pk
[L
1
cos( + ) + L
2
cos( )] (3.147)

3 currents
_
=
3
2
I
pk
[L
1
sin( + ) + L
2
sin( )] (3.148)
Calculating the ux linkage using (3.74) and (3.88) one gets the following:
_

s
d

s
q
_
=
3
2
I
pk
_
L
1
cos( + ) + L
2
(cos 2 cos( + ) + sin 2 sin( + ))
L
1
sin( + ) + L
2
(sin 2 cos( + ) cos 2 sin( + ))
_
(3.149)
Since:
cos 2 cos( + ) + sin 2 sin( + ) = cos( )
sin 2 cos( + ) cos 2 sin( + ) = sin( )
then the Space Vector and dq expressions are equivalent. This equivalence is
more easily veried if the rotating versions of the two models are compared.
Consider (3.131). If Re and Im parts are taken we have:

3 currents
_
=
3
2
I
pk
[(L
1
+ L
2
) cos ] = L
d
i
d
(3.150)

3 currents
_
=
3
2
I
pk
[(L
1
L
2
) sin ] = L
q
i
q
(3.151)
since we know that L
d
= 3/2(L
1
+ L
2
) and L
q
= 3/2(L
1
L
2
), and i
d
=
I
pk
cos , i
q
= I
pk
sin from the dq model theory.
Finally it can be shown that the Space Vector and dq model theory give the
same torque and power expressions. Consider the following relationships:
I
2
pk
sin 2 = 2(I
pk
cos )(I
pk
sin )
= 2i
d
i
q
(3.152)
2L
2
=
2
3
_
3
2
(L
1
+ L
2
)
3
2
(L
1
L
2
)
_
=
2
3
(L
d
L
q
) (3.153)
Substituting these into (3.141) gives the normal dq torque expression (3.83).
3.4 Steady State Model
Thus far this Chapter has been mainly considering the dynamic models. The
steady state model of machines can be derived from the dq models, and as
an example this section will develop the steady state voltages and currents for
the SYNCREL using the dynamic model as a starting point. The approach is
similar to that in [2].
3.4 Steady State Model 3-30
Consider the ideal dq equation in a rotating reference frame as shown in
(3.78). If the machine is in steady state then the derivative terms in this ex-
pression will be zero. Hence the steady state form of (3.78) is:
v
r
d
= Ri
r
d

rs
L
r
q
i
r
q
v
r
q
= Ri
r
d
+
rs
L
r
d
i
r
d
_
(3.154)
Using the standard transformation shown in Table 3.2 one can write:
v
abc
= C
T
v
r
dq
(3.155)
Considering only phase a in this transformation it can be seen that (3.155) leads
to the following expression for the a phase voltage:
v
a
= [
v
r
d
..
(Ri
r
d

rs
L
r
q
i
r
q
) cos
rs

v
r
q
..
(Ri
r
q
+
rs
L
r
d
i
r
d
) sin
rs
] (3.156)
This equation can be written in the form:
v
a
= Re
_
(v
r
d
+ jv
r
q
)e
j
rs
t
_
(3.157)
We wish to express the voltage in terms of the phase voltage. Consider the
following form for the time domain expression for the voltage on phase a:
v
a
= V
m
cos(
rs
t + ) (3.158)
which may be written in complex notation as:
v
a
= Re
_

2
V
m

2
e
j
e
j
rs
t
_
(3.159)
The phasor voltage for v
a
may be written by inspection of (3.159) as:
V
a
=
V
m

2
e
j
(3.160)
therefore (3.159) may be written as:
v
a
= Re
_

2V
a
e
j
rs
t
_
(3.161)
Comparing (3.161) with (3.157) one can write the following:
V
a
=
1

2
_
v
r
d
+ jv
r
q
_
= V
r
d
+ jV
r
q
(3.162)
Similarly the phase current can be written in terms of the steady state dq cur-
rents:
I
a
=
1

2
_
i
r
d
+ i
r
q
_
= I
r
d
+ jI
r
q
(3.163)
Using the expressions for V
a
and I
a
above one can write (3.162) as:
V
a
= R
r
d
I
r
d

rs
L
r
q
I
r
q
+ j(R
r
q
I
r
q
+
rs
L
r
d
I
r
d
) (3.164)
3.4 Steady State Model 3-31
Figure 3.8: Phasor diagram for a steady state SYNCREL
which allows the steady state phasor diagram of Figure 3.8 to be drawn. Notice
from this Figure that the power factor for this machine can never be leading. If
the (3.164) is written as:
V
a
= R
r
d
I
r
d

rs
L
r
d

I
r
q
+ j(R
r
q
I
r
q
+
rs
L
r
d
I
r
d
) (3.165)
where = L
d
/L
q
then it can be seen that as and the d-axis ux in the
machine stays constant, then the voltage and current vectors in Figure 3.8 will
approach each other, and the power factor will be unity.
Chapter 4
Vector Control of Induction
Machines
4.1 Introduction
In the previous chapter we assembled much of the basic theory to tackle the
modelling of the induction machine, and the derivation of vector control of the
induction machine. In this chapter we will extend the basic theory where neces-
sary for the induction machine situation, and then apply this to the development
of a number of dierent vector control strategies for this machine.
4.2 Vector Models for Induction Machines
The purpose of this section is to extend the theory developed for the single
winding AC machine in the previous chapter to that of the induction machine.
This involves considering modelling for a machine where one has windings on
both the rotor and the stator. Although the general principles are the same as
for the stator winding case the detailed expressions are dierent.
Figure 4.1 shows a conceptual diagram of a machine with three phase wind-
ings on the rotor. Note that it this machine is an induction machine then the
rotor windings are usually short circuited. However for the moment the equa-
tions will be developed as thought the windings are not shorted.
Similar expressions to those developed in the previous chapter for the stator
winding can be developed for the rotor winding. These will not be repeated here.
These expressions are developed assuming that the rotor winding is stationary
in other words we are developing the expressions in a reference frame aligned
with the rotor.
4.2.1 Flux Linkage Expression
Some of the expressions developed previously have to be altered slightly in
order to account for the presence of the rotor. One of these is the ux linkage
expression as expressed in (3.111). With the rotor present the various total ux
4.2 Vector Models for Induction Machines 4-2
w
rs
a
s
b
s
c
s
a
r
b
r
c
r
q
rs
Figure 4.1: Conceptual diagram of an induction machine.
linkage expressions for the stator phases become:

as
= L
s
i
as
+ M
ss
i
bs
+ M
ss
i
cs
+ M
sr
i
ar
cos
rs
+
M
sr
i
br
cos(
rs
+
2
3
) + M
sr
i
cr
cos(
rs
+
4
3
) (4.1)

bs
= L
s
i
bs
+ M
ss
i
as
+ M
ss
i
cs
+ M
sr
i
ar
cos(
rs
+
4
3
)+
M
sr
i
br
cos
rs
+ M
sr
i
cr
cos(
rs
+
2
3
) (4.2)

cs
= L
s
i
cs
+ M
ss
i
as
+ M
ss
i
bs
+ M
sr
i
ar
cos(
rs
+
2
3
)+
M
sr
i
br
cos(
rs
+
4
3
) + M
sr
i
cr
cos
rs
(4.3)
where:
L
s
the self inductance of the phase
M
ss
the mutual inductance between stator phases
M
sr
the maximum mutual inductance between a stator and rotor phase
Note the self inductance of a single stator winding can be written in terms
of a leakage inductance and a magnetising inductance:
L
s
= L
sl
+ L
sm
(4.4)
where L
sl
is a stator winding leakage inductance, and L
sm
is a stator winding
magnetising inductance. If the stator has an eective number of turns equal
to N
se
and the rotor has an eective number of turns N
re
then the stator
4.2 Vector Models for Induction Machines 4-3
magnetising inductance is related to the mutual inductance between the rotor
and the stator as follows: This is a standard
result from trans-
former theory
L
sm
=
N
se
N
re
M
sr
(4.5)
If these expressions are substituted into the denition of the ux linkage
vector (3.109), after considerable manipulation one can obtain the following
expression:

s
= L
s
i
s
+ L
m
i
r
e
j
rs
= L
s
i
s
+ L
m
i

r
(4.6)
where the

means that the rotor current vector has been referenced to the
stator reference frame. Notice that this expression has a marked resemblance
to the previous expression (3.112) there is an additional term related to the
rotor. The various inductances in this expression are:
L
s
= L
s
L
ss
= L
sl
+ L
sm
M
ss
and M
ss
=
1
2
M
ss
m
=
1
2
L
sm
L
s
= L
sl
+
3
2
L
sm
(three phase stator inductance)
L
m
=
3
2
M
sr
Note that this expression can be further manipulated as follows:

s
=
_
L
sl
+
3
2
L
sm
_
i
s
+
3
2
N
re
N
se
L
sm
i

r
(4.7)
Note that
N
re
N
se
i

r
is the rotor current referred to the stator. Therefore the ex-
pression can be written in terms of the stator magnetising current (which is the
normal form for a transformer) as follows: Note that referring
currents to the sta-
tor using turns ra-
tios is dierent to
referencing current
vectors to dierent
reference frames.

s
= L
s
i
s
+

L
m
i

r
s
(4.8)
where:

L
m
=
3
2
L
sm
i

r
s
the rotor current referenced and referred to the stator.
Remark 4.1 Note that the above expression is exactly analogous to an expres-
sion for a transformer where the secondary current has been referred to the
stator.
Similar analysis can be applied to the rotor giving the expression:

r
=

L
r
i
r
+ L
m
i

s
(4.9)
where:

L
r
= L
r
M
rr
= L
rl
+ L
rm
and M
rr
=
1
2
M
rr
m
=
1
2
L
rm


L
r
= L
rl
+
3
2
L
rm
(three phase rotor inductance)
L
m
=
3
2
M
rs
=
3
2
M
sr
L
rm
the rotor magnetising inductance =
N
re
N
se
M
rs
4.2 Vector Models for Induction Machines 4-4
Remark 4.2 Note that if the eective rotor and stator turns are the same (i.e.
N
se
= N
re
) then L
rm
= L
sm
= M
sr
. Therefore L
m
=
3
2
L
sm
=
3
2
L
rm
and
both (4.6) and (4.9) are in the form of the conventional transformer equation.
Referring currents to the stator or the rotor has no eect because the turns ratio
is unity.
4.2.2 Magnetising Current
In a machine such as the induction machine there are two mmfs contributing
to the total mmf in the machine the stator mmf and the rotor mmf. The
total mmf in the machine can be calculated by adding together these two mmfs.
Referring to (3.94) we can write:
F
T
= F
s
T
+ F
r
T
=
3
2
N
se

_
i
s
e
j
_
+
3
2
N
re

_
i

r
e
j
_
=
3
2
N
se
_

_
i
s
e
j
_
+
N
re
N
se

_
i

r
e
j
_
_
=
3
2
N
se

_
(i
s
+
N
re
N
se
i

r
)e
j
_
(4.10)
Notice that the
N
re
N
se
i

r
term is the rotor current in a referenced to a stationary
frame referred to the stator (in the same way the secondary currents can be
referred to the primary in a transformer). The current expression in (4.10) is
the magnetising current as this is the current that produces the magnetising
current in the machine:
i
m
= i
s
+
N
re
N
se
i

r
= i
s
+ i

r
s
(4.11)
4.2.3 Power and Torque Expressions
Other expressions that dier when we have a rotor with windings are those for
power and torque. Realising that the torque and power expressions are closely
related we will concentrate on the torque expression in this discussion. The
main dierence between the evaluation of the torque expression here and that
carried out in the previous chapter is that we must account for the power in
the rotor, since the equations for the induction machine allow for the double
fed machine case. Furthermore, if we dont account for the rotor power then
we would not have a torque expression at all, since it is the power in the rotor
that is contributing to the torque. As noted in the evaluation of the torque for
the SYNCREL we only needed to consider the stator in that case because the
eects of the rotor were reected into the stator via the spatial variation of the
stator inductances (which was caused by movement of the rotor). The rotor
had no electrical circuit, therefore was electrically passive.
Similarly to the expression (3.133) we can write the following expression for
the total power for the induction machine:
P
3
=
3
2
(u
s
i

s
+ u

r
i

r
) (4.12)
4.2 Vector Models for Induction Machines 4-5
1
We also know from conservation of energy arguments that equation (1.44)
holds. Power is related to energy via the relation:
P
3
=
dE
e
dt
dE
e
= P
3
dt (4.13)
Therefore the power expression (4.12) can be broken into various components
based on (1.44). The losses section of this expression is normally broken in
various losses such as friction, iron losses etc., but we for simplicity shall only
consider the resistive loss component. Let us write the expression for the stator
resistance loss:
P
R
s
le
=
3
2
(u
R
s
le
i

s
) (4.14)
=
3
2
(i
s
R
s
i

s
)
=
3
2
(|i
s
|
2
R
s
) (4.15)
where u
R
s
le
is the voltage across the stator resistance R
s
.
Similarly the losses in the rotor can be written as:
P
R
r
le
=
3
2
(|i

r
|
2
R
r
) (4.16)
Using these two expressions we can write the increment energy loss due to
resistance losses as:
dE
le
= P
R
le
dt =
3
2

_
|i
s
|
2
R
s
+|i

r
|
2
R
r
_
dt (4.17)
Now let us consider the energy that is being put into the eld. The energy
being put into the eld is the voltage across the inductive elements the current
through the inductive elements. The voltage across the inductive elements is
the rate of change of ux linkage. The general expression for the eld power is:
P
fe
=
3
2

_
d
s
dt
i

s
+
d
r
dt
i

r
_
=
3
2

_
d
s
dt
i

s
+
d

r
dt
i

r
_
(4.18)
Note that this expression can also be written with the rotor variables referred
to the stator reference frame and vice-versa since power expressions are invariant
under reference frame transformations. If (4.18) is integrate with respect to time
then the following expected expression for the eld energy results:
E
fe
=
3
2
(
s
i

s
+
r
i

r
) =
3
2
(
s
i

s
+

r
i

r
) (4.19)
Remark 4.3 Note that it is implicit in our derivation of the eld energy that the
rotor of the machine is stationary. If the rotor is not stationary then the d
r
/dt
term will contain a rotationally related term. Such a term is not connected with
eld energy but is connected with mechanical output power.
1
Note that the rotor variables are expressed in the stator reference frame.
4.2 Vector Models for Induction Machines 4-6
Rearranging (1.44) so that the mechanical energy term is the subject and writing
in terms of dierentials gives:
dE
me
= dE
e
dE
le
dE
fe
(4.20)
Substituting for the terms in this expression, assuming that there is movement
of the rotor, gives:
dE
me
=
3
2
(u
s
i

s
+ u

r
i

r
)dt
3
2

_
|i
s
|
2
R
s
+|i

r
|
2
R
r
_
dt

3
2

_
d
s
dt
i

s
+ i

r
_
d

r
dt

rs
const
+ j
rs

r
_
. .
The
d

r
dt
term
_

_
dt (4.21)
Terms in this expression can be collected together so that it consists of compo-
nents due to the stator and components due to the rotor:
dE
me
=
3
2

_
u
s
i

s
|i
s
|
2
R
s

d
s
dt
i

s
_
dt
. .
Stator mechanical energy components
+
3
2

_
u

r
i

r
|i

r
|
2
R
r
i

r
_
d

r
dt

rs
const
+ j
rs

r
__
dt
. .
Rotor mechanical energy components
(4.22)
The stator section of this equation cannot contribute to mechanical output
power, therefore all the terms in this section of the equation must add to be
zero. Let us consider the rotor section of the equation the terms:
u

r
i

r
|i

r
|
2
R
r
i

r
d

r
dt

rs
const
= 0
for the same reason that the stator section terms equal zero these terms do not
involve anything related to mechanical motion and therefore cannot contribute
to the mechanical output power. The last term on the other hand involves

rs
and therefore must have something to do with mechanical output power.
Therefore the mechanical output power is:
P
rot
= P
me
=
dE
me
dt
=
3
2
(j
rs

r
i

r
) (4.23)
T
rot
= T
me
=
P
me

rs
=
3
2
(j

r
i

r
) (4.24)
Note the similarity of these expressions with (3.143). Therefore in a manner
similar to (3.144) and (3.145) we can write that the power and torque for the
induction machine is:
P
rot
=
3
2

rs

r
i

r
(4.25)
T
rot
=
3
2

r
i

r
(4.26)
4.2 Vector Models for Induction Machines 4-7
Remark 4.4 As can be seen from the torque and power expressions equations
(3.144) and (3.145) are very general.
It is possible to derive a number of alternative expressions for the torque of an
induction machine using the ux expressions (4.6) and (4.9). Converting (4.9)
to be relative to the stator reference frame and dening the magnetising ux
vector: Note that we have
assumed that the
stator to rotor
eective turns are
equal to get the
expression for
m
.

r
=

L
r
i

r
+ L
m
i
s
(4.27)

m
= L
m
i

r
+ L
m
i
s
= L
m
i
m
(4.28)
where

m
refers to the magnetising ux referred to the stator reference frame
(from its natural magnetising ux reference frame). Rearranging (4.28):
L
m
i
s
=

m
L
m
i

r
and substituting into (4.27) gives:

r
=

L
r
i

r
+

m
L
m
i

r
Substituting this into (4.26) gives: Note that the
torque is related
to the interaction
of the magnetising
ux and the rotor
current. The rotor
leakage ux does
not contribute to
torque production.
T
rot
=
3
2
(

L
r
i

r
+

m
L
m
i

r
) i

r
=
3
2

m
i

r
(4.29)
This expression can be further manipulated by using (4.28):
T
rot
=
3
2
(L
m
(i

r
+ i
s
) i

r
(4.30)
=
3
2
L
m
i
s
i

r
(4.31)
i.e. the torque is a function of the cross product of the stator and rotor current Assuming N
se
=
N
re
space vectors.
A dierent expression can be obtained from (4.26) as follows:
T
rot
=
3
2

r
i

r
=
3
2
(

L
r
i

r
+ L
m
i
s
) i

r
=
3
2
L
m
i
s
i

r
(as above)
Now realising that i

r
i

r
= 0 allows one to write:
T
rot
=
3
2
L
m

L
s
(

L
s
i
s
+ L
m
i

r
) i

r
=
3
2
L
m

L
s

s
i

r
(4.32)
The torque can also be expressed totally in terms of stator quantities. Con-
sider (4.31), by reversing the order of the cross product we can write:
T
rot
=
3
2
L
m
i

r
i
s
(4.33)
4.2 Vector Models for Induction Machines 4-8
Again using the fact that a vector crossed with itself is equal to zero we can
write this expression as:
T
rot
=
3
2
(

L
s
i
s
+ L
m
i

r
) i
s
=
3
2

s
i
s
(4.34)
This expression can be further developed. Assuming that the stator and
rotor eective turns are equal we can write:

s
=
sl
+

m
(4.35)
where:

sl
= L
sl
i
s

m
= L
m
i
m
= L
m
(i
s
+ i

r
)
Substituting for
s
in (4.35) we can write:
T
rot
=
3
2
(L
sl
i
s
+ L
m
i
m
) i
s
=
3
2
L
m
i
m
i
s
=
3
2

m
i
s
(4.36)
Note that the sta-
tor leakage ux does
not contribute to
the torque.
4.2.4 The Space Vector Model of the Induction Machine
In this section we develop the space vector model of the induction machine in
a general reference frame. From this general reference frame it is then easy
to generate models of the machine in the natural reference frames such as the
stationary reference frame or the rotor reference frame.
Let us begin with the basic equations for the induction machine. These are
simple to deduce from the work we have already carried out on space vector
modelling. We simply write expressions of the form of (3.125) for both the
stator and the rotor. Note that the equation for each of these has been written
in its nature reference frame (i.e. stationary for the stator and rotating with
the rotor for the rotor):
v
s
= R
s
i
s
+
d
s
dt
(4.37)
v
r
= R
r
i
r
+
d
r
dt
(4.38)
We want to express these equations in a general reference frame rotating at an
angular velocity of
g
rad/sec. Figure 4.2 shows the relationship between the
stationary frame, the rotating frame and the general reference frame.
We can use the frame conversion factors to reference the stator expression
and the rotor expression to the general reference frame. Using (3.104) we can
4.2 Vector Models for Induction Machines 4-9
x
s
y
s
x
g
y
g
x
r
y
r
q
rs
q
gs
q q
gs rs
-
w
rs
w
gs
Figure 4.2: Relationship between stationary and rotating frames and the general
reference frame for the induction machine
write for a generic vector in each of the natural reference frames:
x
sg
= x
s
e
j
gs
(4.39)
x
rg
= x
r
e
j(
gs

rs
)
= x
r
e
j(
rs

gs
)
(4.40)
Applying these conversion factors we can write the following:
v
s
= v
sg
e
j
gs
; i
s
= i
sg
e
j
gs
;
s
=
sg
e
j
gs
(4.41)
Substituting these expressions into (4.37) we can write:
v
sg
e
j
gs
= R
s
i
sg
e
j
gs
+
d
dt
_

sg
e
j
gs
_
= R
s
i
sg
e
j
gs
+
sg
e
j
gs
. j
d
gs
dt
+ e
j
gs
d
sg
dt
= R
s
i
sg
e
j
gs
+
_
j
g

sg
+
d
sg
dt
_
e
j
gs
v
sg
= R
s
i
sg
+
d
sg
dt
+ j
gs

sg
(4.42)
In a similar manner we can do the same with the rotor equation for the
machine. In this particular case we have:
v
r
= v
rg
e
j(
gs

rs
)
; i
r
= i
rg
e
j(
gs

rs
)
;
r
=
rg
e
j(
gs

rs
)
(4.43)
4.2 Vector Models for Induction Machines 4-10
Substituting these into (4.38) gives:
v
rg
e
j(
gs

rs
)
= R
r
i
rg
e
j(
gs

rs
)
+
d
dt
_

rg
e
j(
gs

rs
)
_
= R
r
i
rg
e
j(
gs

rs
)
+
rg
e
j(
gs

rs
)
. j
d(
gs

rs
)
dt
+ e
j(
gs

rs
)
d
rg
dt
= R
r
i
rg
e
j(
gs

rs
)
+
_
j(
gs

rs
)
rg
+
d
rg
dt
_
e
j(
gs

rs
)
v
rg
= R
r
i
rg
+
d
rg
dt
+ j(
gs

rs
)
rg
(4.44)
Summarising, the induction machine equations in a general reference frame
are
2
:
v
sg
= R
s
i
sg
+
d
sg
dt
+ j
g

sg
v
rg
= R
r
i
rg
+
d
rg
dt
+ j(
g

r
)
rg
(4.45)
Remark 4.5 With equation (4.45) one can easily generate the induction ma-
chine equation in an arbitrary reference frame. This will be very handy when we
consider the various forms of vector control which result from these expressions
in particular reference frames.
Using reasoning similar to that in the previous section we can write that the
rotational power is due to the true rotational motion in (4.45). Therefore the
expression for the rotational power is:
P
rot
=
3
2

_
j
r

rg
i

rg
_
=
3
2

rg
i
rg
(4.46)
Therefore it is obvious that the rotational torque produced is:
T
rot
=
3
2

rg
i
rg
(4.47)
This has the same form as the expression derived in the previous section, there-
fore it can be seen that the various torque expressions derived in the previous
section carry over to the general reference frame except that the uxes and
currents have to be replaced by the general reference frame equivalents.. The
expressions for the uxes are:

sg
=

L
s
i
sg
+ L
m
i
rg
(4.48)

rg
=

L
r
i
rg
+ L
m
i
sg
(4.49)
2
Using the simple notation of
g
for
gs
and
r
for
rs
4.3 A Heuristic Explanation of Vector Control 4-11
sD
sQ
y
m
'
D
Q
Figure 4.3: Conceptual diagram of an induction machine with quadrature-phase
stator windings.
One particular version of the torque expression that will be useful later is
derived as follows. From (4.33) we know that we can write the following in the
general reference frame:
T
rot
=
3
2
L
m
i
rg
i
sg
(4.50)
This expression can be expanded as follows:
T
rot
=
3
2
L
m

L
r
(

L
r
i
rg
+ L
m
i
sg
) i
sg
since i
sg
i
sg
= 0
=
3
2
L
m

L
r

rg
i
sg
(4.51)
4.3 A Heuristic Explanation of Vector Control
In the previous section we developed some general space vector models of the
induction machine. These models are very useful in quantitative denitions of
the fundamental equations for vector control. However, in this section we shall
consider vector control from a heuristic point of view.
The following discussion will be with reference to Figure 4.3. This diagram
shows a squirrel cage induction machine with quadrature stator windings. One
could consider the stator to be that of a DC machine where the q axis winding
corresponds to the compensation winding, and the d axis winding is the eld
winding. For the sake of the following argument we will assume that the stator
and rotor windings have the same number of eective turns. Let us now conduct
a thought experiment on this conguration.
4.3 A Heuristic Explanation of Vector Control 4-12
D
Q
i
s
i i
m ds
' =
i
qs
i
r
'
y
m
'
Figure 4.4: Space vectors in quadrature induction machine at time t
+
o
.
Assume that a current i
qs
is suddenly injected into the sQ winding. By
Lenzs Law a current will ow in the rotor in such a way as to oppose the
change in ux caused by the increasing current in the sQ winding. The current
directions are those marked in Figure 4.3. Therefore the current owing at some
time t
+
o
(where t
o
is the time of application of the q axis current) is:
i

r
= i
qs
(4.52)
The eect of the injection of the current can be better scene from a space vector
diagram of the machine see Figure 4.4. The resultant stator current is simply
the vector addition of the two stator currents. The induced rotor current directly
opposes the q axis stator current, and therefore there is no ux produced in the
q axis. Therefore at time t
+
o
the magnetising current is the d axis current, as it
would be in a compensated DC machine.
Assume that the rotor is held stationary. After the initial current induced
in the rotor the current will die away with the time constant being that of the
rotor circuit. Therefore at some time t > t
o
the situation could be that in
Figure 4.5. The length of the i

r
vector will decrease in length until it is zero. It
is clear that under this condition that the magnetising current vector and ux
no longer coincide with the d axis of the machine. Therefore this situation is
now dierent from that of the DC machine.
What happens if we can move the stator so that the d axis remains aligned
with the magnetising current vector? That is, as the magnetising current vector
moves by
m
toward i
s
we move the stator by the same amount. This move-
ment of the stator can also be viewed as a movement of the rotor in a clockwise
direction if the stator is taken as the reference frame. If we have relative mo-
tion between the stator and the rotor then we have the rotor bars cutting ux
and consequently there will be voltage produced in the bars. If one used the
F = qv B expression the voltage induced for the relative motion in this exam-
ple is such that the decreasing rotor current will be increased. Note that the B
4.4 Special Reference Frames 4-13
D
Q
i
s
i
ds
i
qs
i
r
'
i
m
'
m
m
Movementof i
m
'
Figure 4.5: Space vectors in quadrature induction machine at t > t
o
.
eld would be slightly oset from the d axis because the i
qs
current is no longer
fully compensated. This in turn changes the orientation of the induced voltage
in the rotor and a consequent shift in the resultant rotor current so that com-
pensation is reestablished. Figure 4.6 shows the situation where the stator has
been rotated so that the d axis still lies along the magnetising current vector.
Note that this also implies that the rotor current, i

r
is still orthogonal to the
magnetising ux.
Remark 4.6 If one keeps the d axis winding aligned with the magnetising cur-
rent vector, then the i
s
vector will move by the same angular displacement as the
d axis. Therefore implicitly the i

r
vector must be orthogonal to the magnetising
current vector.
In a real situation one would not move the stator winding, but the same
eect could be achieved by controlling the stator d and q axis currents so that
the resultant mmf (and hence current vector i
s
) produced by the winding moves
around the machine by
m
.
The movement of the stator current vector in relation to the magnetising
ux linkage phasor has important implications on the torque production of the
machine. This will be investigated in the next section.
4.4 Special Reference Frames
4.4.1 The Magnetising Flux Linkage Reference Frame
In the previous section we heuristically discussed the basic concepts behind
vector control. This discussion was with respect to the magnetising ux linkage.
Now let us consider the implications of the magnetising ux reference frame on
the torque production of the machine.
4.4 Special Reference Frames 4-14
D
Q
i
s
i i
r q
' =
i i
m d
' =
m
m
sD
sQ
Figure 4.6: Position of the space vectors after the stator has been rotated.
i i
s sm
,
m
m
sD
sQ
x
y
w
m
i
sx
i
sy
i
ds
i
qs
y y y
m mm
mx
' = =
Figure 4.7: Relationship between the dq frame and the special xy frame.
4.4 Special Reference Frames 4-15
The following discussion is with reference to Figure 4.7. As can be seen we
have dened a reference axes (the xy axes) that are rotating with the magnetis-
ing current vector. Therefore the xy axes are rotating at:

m
=
d
m
dt
(4.53)
One can also see that the relationship between the stator current in the mag-
netising reference frame and the stator current in the stator frame is (using
(3.104)):
i
sm
= i
s
e
j
m
= i
sx
+ ji
ys
(4.54)
From (4.36) we can deduce that the torque in a magnetising ux reference
frame is:
T
rot
=
3
2

m
i
sm
=
3
2

i
sy
(4.55)
since there is no orthogonal component of the magnetising ux with this par-
ticular orientation of the reference axes. Clearly the:

m
= L
m
(i
sm
+ i
rm
)
= L
m
i
sx
(4.56)
=
sm
(4.57)
therefore:
T
rot
=
3
2

sm
i
sy
=
3
2
L
m
i
sx
i
sy
The basic principle of alignment with the magnetising axis forms the basis
of magnetising ux vector orientated control. This shall be considered in more
detail later.
4.4.2 The Rotor Flux Linkage Reference Frame
Another reference frame that is very commonly used is the rotor ux reference
frame. In this reference frame the special axes are xed to the rotor ux vector.
Figure 4.8 is used in the following discussion.
We know from previous work that;
T
rot
=
3
2
L
m

L
r

rg
i
sg
(4.58)
Therefore if a reference frame, xy, is chosen so that it is coincident with the
rotor ux linkage vector then there by denition cannot be an orthogonal com-
ponent of the rotor ux linkage in this frame of reference. Therefore the torque
expression becomes:
T
rot
=
3
2
L
m

L
r

rr
i
sr
(4.59)
4.4 Special Reference Frames 4-16
i i
s s r
,
y
r
r
sD
sQ
x
y
w
mr
i
sx
i
sy
i
ds
i
qs
w
r
q
r
a
s
i i
mr mr
=
y
y r r
Figure 4.8: Relationship between various space phasors in the stator and rotor
ux linkage reference frames
Now:
i
sr
= i
s
e
j
r
= i
sx
+ ji
sy
(4.60)
If
r
is the rotor ux linkage in a rotating frame, then:

rr
=
r
e
j(
r

r
)
=
r
e
j
r
e
j
r
=

r
e
j
r
=

e
j
r
e
j
r
=

=
rx
(4.61)
Substituting this into the torque expression (4.59) gives:
T
rot
=
3
2
L
m

L
r

rx
(i
sx
+ ji
sy
)
=
3
2
L
m

L
r

rx
i
sy
(4.62)
This expression can be further manipulated into a form that is more familiar in
the vector control literature. This form of the expression is based in deriving
a single current value related to the rotor ux linkage (the normal denition of
the ux linkage involves the rotor and the stator uxes):
i
mr
=

rr
L
m
=
(

L
r
i
rr
+ L
m
i
sr
)
L
m
= i
sr
+

L
r
L
m
i
rr
= i
sr
+ (1 +
r
)i
rr
;
r
=
L
rl
L
m
(4.63)
This current vector lies along the x axis of the reference frame (since it is
related to the rotor ux linkage vector via a scalar) as shown in Figure 4.8.
4.4 Special Reference Frames 4-17
i i
s s s
,
y
i i
ms ms
=
r
s
sD
sQ
x
y
a
s
i
sx
i
sy
i
qs
i
ds
y y y
y s s s
sx
, =
w
ms
Figure 4.9: Relationship between the stationary reference frame and the special
reference frame xed to the stator ux linkage space phasor.
Since
rr
=
rx
then
rx
= L
m
|i
mr
| . Hence (4.62) becomes:
T
rot
=
3
2
L
2
m

L
r
|i
mr
| i
sy
(4.64)
=
3
2
L
m
(1 +
r
)
|i
mr
| i
sy
using

L
r
= (1 +
r
)L
m
(4.65)
As with the magnetising ux alignment case the torque can be controlled
separately from the ux in the machine if one can keep the |i
mr
| value constant
and use the i
sy
current for the torque control.
4.4.3 Stator Flux Linkage Reference Frame
In a manner similar to rotor ux linkage vector alignment we can also align the
special reference frame to the stator ux linkage vector. In a manner similar to
(4.34), and realising that the reference frame is aligned with the stator ux, we
can write the torque as:
T
rot
=
3
2

sx
i
sy
(4.66)
Figure 4.9 shows the relationship between the special frame and the stator
current vector.
It is possible to put the equation for the torque in terms of currents. The
stator ux vector can be written as follows:

ss
=

L
s
i
ss
+ L
m
i
rs
(4.67)
where i
rs
is the rotor current referenced to the special stator ux reference
frame. Let us consider this current vector further:
i
rs
= i
rx
+ ji
ry
= i
r
e
j(
s

r
)
= i
r
e
j
r
= i

r
e
j
s
(4.68)
4.5 Derivation of Rotor Flux Oriented Vector Control 4-18
Now let us dene the stator magnetising current in the stator ux linkage
reference frame:
i
ms
=

ss
L
m
=

L
s
L
m
i
ss
+ i
rs
= (1 +
s
)i
ss
+ i
rs
(4.69)
where
s
=
L
sl
L
m
(4.70)
Since the
ss
is coincident with the x axis of the reference frame then so is
i
ms
. Therefore
sx
= L
m
i
ms
. Substituting for
sx
in (4.66) we can write:
T
rot
=
3
2
L
m
|i
ms
| i
sy
(4.71)
As with the previous special reference frames the torque is related to the
interaction of a current and a ux linkage vector that are in space quadrature.
Remark 4.7 In all these special reference frames one can see that the form of
the torque expressions is identical to those for a separately excited DC machine.
Furthermore the currents in these expressions are DC values in steady state.
Therefore the choice of the reference frame is very important in simplifying a
complex model to that of a DC machine. The fact that the rotating reference
frames chosen create steady state DC currents has important implications on
the control strategies employed.
4.5 Derivation of Rotor Flux Oriented Vector Con-
trol
In the previous sections of this chapter we have derived a number of supporting
expressions that allow us to dene the relevant expressions for vector control
in a number of dierent reference frames. In this section we shall derive the
expressions required for control in the rotor reference frame. This frame was
chosen as it is one of the most popular control strategies used in practice. The
derivation of controllers in the other reference frames will be left as an exercise.
We begin with the equations for the induction machine in the general refer-
ence frame (4.45), repeated here for convenience:
v
sg
= R
s
i
sg
+
d
sg
dt
+ j
g

sg
(4.72)
v
rg
= R
r
i
rg
+
d
rg
dt
+ j(
g

r
)
rg
(4.73)
In a frame rotating with the rotor ux the speed of the frame is
mr
therefore
these expressions are modied as follows:
v
sr
= R
s
i
sr
+
d
sr
dt
+ j
mr

sr
(4.74)
v
rr
= R
r
i
rr
+
d
rr
dt
+ j(
mr

r
)
rr
(4.75)
4.5 Derivation of Rotor Flux Oriented Vector Control 4-19
Let us rstly consider the stator voltage equation. We know that the stator
ux can be expresses as follows:

sr
=

L
s
i
sr
+ L
m
i
rr
(4.76)
and substituting this into (4.74) we get:
v
sr
= R
s
i
sr
+
d
dt
(

L
s
i
sr
+ L
m
i
rr
) + j
mr
(

L
s
i
sr
+ L
m
i
rr
)
= R
s
i
sr
+

L
s
di
sr
dt
+ L
m
di
rr
dt
+ j
mr

L
s
i
sr
+ j
mr
L
m
i
rr
(4.77)
From the expression for the rotor magnetising current (4.63) we can write:
i
rr
=
i
mr
i
sr
(1 +
r
)
(4.78)
Because of the choice of the frame, we know that the frame lies along the i
mr
vector and consequently there is no orthogonal component to i
mr
. Therefore
the expression for the rotor current can be rewritten as:
i
rr
=
|i
mr
| i
sr
(1 +
r
)
(4.79)
Substituting this for i
rr
in (4.77) we get:
v
sr
= R
s
i
sr
+

L
s
di
sr
dt
+ L
m
d
dt
_
|i
mr
| i
sr
(1 +
r
)
_
+ j
mr

L
s
i
sr
+ j
mr
L
m
_
|i
mr
| i
sr
(1 +
r
)
_
(4.80)
Now manipulating this equation to make the derivative of the stator current the
subject of the expression we get:
di
sr
dt
=
v
sr

L
s

R
s
i
sr

L
s

L
m

L
s
d
dt
_
|i
mr
| i
sr
(1 +
r
)
_
j
mr
i
sr
j
mr
L
m

L
s
_
|i
mr
| i
sr
(1 +
r
)
_
(4.81)
Collecting terms we can write this expression as follows:
_
1
L
m

L
s
(1 +
r
)
_
di
sr
dt
=
v
sr

L
s

R
s
i
sr

L
s

L
m

L
s
(1 +
r
)
d |i
mr
|
dt
j
mr
i
sr
_
1
L
m

L
s
(1 +
r
)
_
j
mr
L
m
|i
mr
|

L
s
(1 +
r
)
(4.82)
Realising that:
1
L
m

L
s
(1 +
r
)
= 1
L
m

L
s

L
r
L
m
= 1
L
2
m

L
s

L
r
4.5 Derivation of Rotor Flux Oriented Vector Control 4-20
and, the leakage coecient is dened as:
= 1 k
2
r
k is the coupling coecient
L
m
= k
_

L
s

L
r
k =
L
m
_

L
s

L
r
= 1
L
2
m

L
s

L
r
then we can write:

di
sr
dt
=
v
sr

L
s

R
s
i
sr

L
s
(1)
d |i
mr
|
dt
j
mr
i
sr
j
mr
(1) |i
mr
| (4.83)
which can be simplied to:
di
sr
dt
=
v
sr

L
s

R
s
i
sr

L
s

(1 )

d |i
mr
|
dt
j
mr
i
sr
j
mr
(1
)

|i
mr
| (4.84)
Breaking this into real and imaginary parts we get two coupled dierential
equations for the stator circuit when rotor ux orientation is implemented:
di
sx
dt
=
v
sx

L
s

R
s
i
sx

L
s

(1 )

d |i
mr
|
dt
+
mr
i
sy
(4.85)
di
sy
dt
=
v
sy

L
s

R
s
i
sy

L
s

mr
i
sx

mr
(1 )

|i
mr
| (4.86)
Remark 4.8 Notice that the above expressions for i
sx
and i
sy
are cross coupled
if we are attempting to control them through the voltages v
sx
and v
sy
. In addition
the rotor magnetising current has an inuence on the expressions.
Remark 4.9 If we are using a voltage source inverter to control the above stator
equations, we need a parameter dependent decoupler in order to gain decoupled
control of i
sx
and i
sy
.
Remark 4.10 If the stator is being fed from an ideal current source then the
values of i
sx
and i
sy
are impressed on the stator. In this situation the current
sources are producing whatever voltage is required to make the currents in the x
and y axes the current source value.
Following on from the last remark, if we do not have an ideal current source
then we can still achieve most of the advantages of the ideal current source. We
can generate an current source by applying current feedback around a voltage
source inverter. Clearly this cannot be an ideal current source as it has a certain
bandwidth in relation to reacting to changes in the load, and there is a nite
voltage that can be applied to the machine. However, even with these limitations
a reasonable current source can be obtained.
The main advantage of using current source feed to the inverter is that
the complex dynamics of the stator are removed from the system. Let us for
the moment assume that the stator dynamics can be ignored due to an ideal
current source, and then consider the rotor circuit. Realising that the rotor of
an induction machine is short circuited then we can write (4.75) as:
0 = R
r
i
rr
+
d
rr
dt
+ j(
mr

r
)
rr
(4.87)
4.5 Derivation of Rotor Flux Oriented Vector Control 4-21
Now for rotor orientation, then by denition we have:

rr
= L
m
i
mr
= L
m
|i
mr
| (4.88)
therefore we can write:
0 = R
r
i
rr
+ L
m
d |i
mr
|
dt
+ j(
mr

r
)L
m
|i
mr
| (4.89)
The rotor current can be expressed in terms of the rotor magnetising current
and the stator current using (4.79):
i
rr
=
|i
mr
| i
sr
(1 +
r
)
=
L
m

L
r
_
|i
mr
| i
sr
_
(4.90)
Substituting into (4.89) gives:
0 =
R
r
L
m

L
r
|i
mr
|
R
r
L
m

L
r
i
sr
+ L
m
d |i
mr
|
dt
+ j(
mr

r
)L
m
|i
mr
| (4.91)
Multiplying this by

L
r
/R
r
L
m
and rearranging we get:
T
r
d |i
mr
|
dt
+ j(
mr

r
)T
r
|i
mr
| +|i
mr
| i
sr
= 0 (4.92)
where:
where T
r
=

L
r
R
r
the rotor time constant
Breaking this expression into real and imaginary parts:
T
r
d |i
mr
|
dt
+|i
mr
| = i
sx
(4.93)
(
mr

r
)T
r
|i
mr
| i
sy
= 0 (4.94)
Rearranging the second of these equations and adding the torque expression
(4.64) we end up with the conventional electrical rotor ux orientated equations
for the induction machine:
T
r
d|i
mr
|
dt
+|i
mr
| = i
sx

mr
=
r
+
i
sy
T
r|i
mr
|
T
rot
=
3
2
L
2
m

L
r
|i
mr
| i
sy
(4.95)
Remark 4.11 The rst equation in (4.95) is the ux equation since i
mr
is
directly related to
rr
in the machine.
Remark 4.12 The second equation can be rearranged so that the last term is
the subject of the expression:
i
sy
T
r
|i
mr
|
=
mr

r
(4.96)
This is clearly in the form
sl
=
e

r
where
sl
is the slip frequency. There-
fore
i
sy
T
r|i
mr
|
is the slip frequency of the machine.
4.6 Structure of a Rotor Oriented Vector Drive 4-22
Remark 4.13 This expression is the same form as the dynamic equation of
a separately excited DC machine. The rst equation has no coupling from the
second equation. The second equation is related to the torque produced by the
machine. This can be seen easily by rearranging the torque expression in terms
of i
sy
and substituting into the
mr
equation:

mr
=
r
+
2
3
T
rot
L
m
|i
mr
|
2
(4.97)
Remark 4.14 One of the main control techniques used with these equations
is that |i
mr
| is kept constant at the i
sx
value, and the torque can be controlled
separately by the use of i
sy
.
Remark 4.15 The above equations are fundamentally based on the fact that
we are assuming that the position of the rotor ux is known. Therefore the i
sx
and i
sy
currents can be accurately applied at the correct position spatially in the
machine.
Remark 4.16 If we have an ideal current source then the torque is algebraically
related to the i
sy
current . In practice a delay will occur in the rise of the torque,
this being related to the leakage inductance of the machine.
Remark 4.17 There is no break away torque limit in rotor ux eld oriented
control the torque increases linearly with i
sy
.
Torque control based on the equations in (4.95) are generally used in two control
philosophies direct eld oriented control, and indirect eld oriented control. In
direct eld oriented control the position of the rotor ux is determined using
ux feedback, the ux magnitude and position being obtained from measure-
ments of the ux via Hall Eect sensors or search coils, or via a ux model of
the machine. Indirect eld orientation on the other hand uses a feed-forward
technique to calculate the position of the ux using the current rotor position
and the reference position of the slip frequency. The latter technique has proved
to be very popular as it does not involve sensors or a ux model. Note that
this technique implicitly has parameter sensitivity since the reference position
is calculated using an expression that involves T
r
.
4.6 Structure of a Rotor Oriented Vector Drive
4.6.1 Indirect Rotor ux Oriented Controller
The block diagram of Figure 4.10 shows the basic structure of a indirect vector
controlled drive based on the expressions developed in the previous section.
Notice that all the control calculations are carried out in the rotating reference
frame.
The rotor ux reference angle is calculated on-line by integrating the refer-
ence slip frequency:

r
=
_

slref
dt +
r
=
_
i
syref
T
r
i
sxref
dt +
r
(4.98)
4.6 Structure of a Rotor Oriented Vector Drive 4-23
IM

PWM
modulator
andcurrent
controller
Inverter
i
a
i
b
i
c
Currentcontrolledinverter
S
S
S
S
S

dt
d
T
r
+ 1
Field
Weakening
Generator

r
T
dt
d
w
rref
r j
e
r
3 2
Rotating2phaseto
stationary3phaseX'formation
r
m
L
L
2
3
2
w
r
+
-
+
+
-
+
- +
-
r
r
r
r
sxref r
i T
sxref r
syref
slref
i T
i
= w
mrref
i
sxref
i
syref
i
dsref
i
qsref
i
Speed
Controller
rotref
T
r
q
Mains
aref
i
bref
i
cref
i
slref
q
+
Figure 4.10: Block diagram of a indirect rotor ux vector oriented control scheme
This integration clearly has to be implemented accurately in order for the rotor
ux angle to be correct. In modern implementations this is carried out as a
digital integration.
In the block diagram the current controller is implemented in the stationary
reference frame, however this is implementation dependent. For example, if the
current controllers are implemented using conventional PI controllers, then the
current control is carried out in the synchronously rotating rotor ux reference
frame. The reason for this is that PI controllers cannot provide accurate tracking
of a sinusoidal reference, and the references in the stationary frame are sinusoidal
in steady state. However in the rotating frame the steady state values of the
currents are DC values, and PI controllers can track DC values with zero error.
The eld weakening block in the diagram generates the i
mrref
value above
the base speed. Therefore, this contains a function generator that drops the
rotor ux in a prescribed way so that the machine will operate in a constant
power mode above base speed. Below base speed the rotor ux is usually kept
constant at the maximum desired ux so that the transient performance of the
drive will be at its maximum.
The speed control block is usually implemented as a PI controller. The
output of the speed control block can be interpreted as the desired torque. If
one wishes to implement position control then a further loop can be added
outside the speed control loop. The control type is again usually a PI controller
(although other more sophisticated control strategies can be used). The output
of this block is interpreted by the speed control loop as a desired speed.
4.6.2 Direct Rotor Flux Oriented Controller
This form of the rotor ux controller is probably a more obvious way of imple-
menting a rotor ux oriented vector controller since it uses classical feedback
4.7 Magnetising Flux Orientation 4-24
IM
PWM
modulator
andcurrent
controller
Inverter
i
a
i
b
i
c
Currentcontrolledinverter
S
S
S
S
S
S
Field
Weakening
Generator
dt
d
w
rref
r j
e
r
3 2
Rotating2phaseto
stationary3phaseX'formation
r
m
L
L
2
3
2
w
r
w
r
+
-
+
-
+
- +
- r
r
r
r
mrref
i
sxref
i
syref
i
dsref
i
qsref
i
Speed
Controller
rotref
T
r
q
Mains
aref
i
bref
i
cref
i
mr
i
mr
i
-
+
-
+
T
rot
Flux
Model
i
sy
Flux
controller
Torque
Controller
Figure 4.11: Block diagram of a direct rotor ux eld oriented vector controller.
that most engineers are familiar with. Figure 4.11 shows a block diagram of
the controller. In this controller the rotor ux position is determined by a ux
model that processes current measurements from the machine. The torque and
ux controllers would normally be conventional PI regulators.
The ux model is generated directly from the equations in (4.95). Figure 4.12
is a block diagram of the ux model in a rotor ux reference frame. It is also
possible to develop the ux model in a stationary reference frame.
4.7 Magnetising Flux Orientation
We shall not derive the equations for magnetising ux orientation in these notes.
However, it is useful to look at the rotor equations in this frame and compare
them with those for the rotor ux oriented frame:
di
sx
dt
+
i
sx
T
rl

sl
i
sy
=
_
|i
mm
| + T
r
d|i
mm
|
dt
_
T
rl
(4.99)
di
sy
dt
+
i
sy
T
rl
=
sl
_
T
r
T
rl
|i
mm
| i
sx
_
(4.100)
where:
T
rl
=
L
rl
R
r

sl
=
m

r
i
mm
= i
rm
+ i
sm
where the m subscript means magnetising frame
Notice that the cross coupling in these equations is much more complex than
in the case of rotor ux orientation. Therefore to gain decoupled control of the
torque and ux using this frame we require decoupling equations.
4.7 Magnetising Flux Orientation 25
S
L
m

r
T
r
y
r
r
+
+
r r
q - r
r mr
w - w
i
sy
i
sx
r
j
e
r -
2 3
dt
d
T
r
+ 1
1
r
T
i
a
i
b
i
c
q
r
mr
i
Figure 4.12: Flux model in a rotor ux reference frame
Remark 4.18 Clearly rotor ux orientation gives the classical DC machine
equations for the control without any decoupling equations when the system is
being current fed.
Appendix A
Calculation of Inductances for
Salient Pole Machines
A.1 Calculation of Inductances
One of the fundamental parameters of any machine model is the inductance of
the armature windings of the machine under all operating conditions. Later in
this chapter we shall that the variation of inductances with respect to the me-
chanical position of the rotor is directly connected with electromagnetic energy
conversion in all machines, and hence with the production of torque. Therefore
the calculation of the SYNCREL inductances is fundamental to understanding
the machines operation.
In the case of the SYNCREL, the armature is on the stator, since the rotor
does not have any windings. We will nd that the inductance of a particular
winding varies depending on the position of the rotor in relation to the winding,
and the degree of magnetic saturation of the stator and the rotor iron.
This section will determine the self and mutual inductances for the stator
windings of the SYNCREL. The derivation of these inductances will be carried
out in a detailed and formal manner using a traditional approach [2]. A dierent
approach, and in many ways a simpler and more elegant one using the concept
of winding functions is shown in Appendix B
Remember!! Modify the winding function stu for the SYNCREL.
The following standard assumptions are made in the following analysis:
1. The stator windings are sinusoidally distributed. When excited with cur-
rent a sinusoidal spatial distribution of mmf is produced.
2. The machine does not exhibit any stator or rotor slotting eects.
3. The machine iron is a linear material, i.e. it is not subject to magnetic sat-
uration eects. The permeability of the material is very large in compar-
ison to air. Therefore the permeance of the magnetic paths is dominated
by the air gaps.
4. The air gap ux density waveforms can be adequately represented by their
fundamental component.
A.1 Calculation of Inductances A-2
Figure A.1: Two pole three phase SYNCREL conceptual diagram
5. The stator turns are all full pitched (i.e. they cover electrical radians).
6. There is no leakage ux i.e. there is perfect coupling between the wind-
ings.
Figure A.1 is a conceptual schematic of a two pole, three phase SYNCREL.
Note that the rotor shape does not represent a realistic rotor, but is drawn in
this manner to accentuate the variable reluctance in the d and q axes. The axis
of the rotor which oers the minimum reluctance to the passage of ux across
the air gap from the stator to the rotor is called the d-axis. The maximum
reluctance path is denoted as the q-axis. Note that following development will
use the concept of dq axes before the concept has been rigorously developed
in a more general framework. In the following development the dq axes are
closely associated with the physical conguration of the rotor, therefore the
general development can be left to later without having too many problems
understanding this material.
A few preliminary conjectures, based on heuristics, can be made about the
variation of the winding inductance with respect to the angular rotor position:
Conjecture A.1 The winding self inductance will be a maximum when the d-
axis of the rotor is aligned with the axis of the winding.
Remark A.1 This conjecture concurs with ones intuitive understanding of ux
interacting with iron. The presence of iron in a coil will result in more ux per
unit of current. When the d-axis is aligned with the axis of a coil then there
will be more iron in the ux path for the coil.
A.1 Calculation of Inductances A-3
Conjecture A.2 The winding self inductance will be a minimum when the q-
axis of the rotor is aligned with the axis of the winding.
Remark A.2 If there is less iron in the coils ux path then it is harder to
produce ux for a given amount of current in the coil. Clearly if the q-axis is
aligned with the axis of the coil then there is a larger air path and less iron for
the ux to travel through.
Conjecture A.3 As the rotor is rotated between these two positions the self
inductance varies. The period of the phase inductance variation is half the
period of the mmf variation for the phase winding.
Remark A.3 This is fairly obvious since the phase inductance is a maximum
when a d-axis rotor pole aligns with the phase axis, and this occurs when the
rotor has rotated through electrical radians.
Conjecture A.4 There is mutual inductance between the three phase stator
windings that is a function of the rotor position.
Remark A.4 Clearly as the rotor is rotated the amount of iron in the paths
that would be taken by the mutual ux will vary, and hence the amount of ux
linking the windings will vary.
A complete analysis of this situation involves computing of all the harmonics
of the ux density waveform and then calculating the total ux linkage with
the winding. One then obtains inductance expressions containing a number
of harmonic terms [2]. The harmonic term amplitudes decrease rapidly with
increasing harmonic number, allowing the approximation of considering only the
fundamental to be made. The constant reluctance path approximation made
in the following analysis is essentially the same approximation. If the winding
function technique is use to calculate the inductances then the harmonic eects
are sometimes more readily included. However, the accuracy of this technique
is critically dependent on the accuracy of the inverse air gap function.
A.1.1 Self Inductances
Firstly consider the self inductance of the a-phase sinusoidally distributed wind-
ing. A useful technique to calculate inductances in situations like this is to
consider that the stator mmf can be broken into two sinusoidally distributed
components which can be considered to be acting along the d-axis and the q-
axis of the rotor.(this is possible because of the assumed sinusoidal nature of
the mmf, which implicitly allows components to be taken). Let us consider a
few simple cases of the application of this concept. Figure A.1 can be used as
an aid to visualise the situation. If, for example. the rotor d-axis is aligned with
the axis of the a-phase winding then the total a-phase mmf acts on the d-axis
permeance, and there is no component acting on the q-axis permeance. Since
the stator mmf is spatially sinusoidally distributed, then this means that the
air gap ux density waveform would be sinusoidally distributed. Similarly if the
rotor q-axis is aligned with the a-phase axis, then the total a-phase mmf acts
on the q-axis permeance. Between these two rotor positions the permeance seen
by the winding is, in general, a complex function of the rotor angular position.
A.1 Calculation of Inductances A-4
Consequently, the air gap ux density distribution is also a complex function of
the rotor angle.
The sinusoidally distributed mmf on the stator of the machine can be broken
into two sinusoidal components which are centred on the d and q-axes respec-
tively, regardless of the position of the rotor. These component mmfs are then
acting on the d and q-axis permeances, P
d
and P
q
. Since these permeances are
constant, this is equivalent to saying that the component mmfs are acting on two
constant air gaps, g
d
and g
q
, for the d and q-axes [2]. Therefore the resultant
component air gap ux densities should be spatially sinusoidal, and consequently
the resultant total air gap ux density should also be sinusoidal. This contra-
dicts the statements made in the previous paragraph about the complex nature
of the air gap ux density. However, the fundamental of the actual air gap ux
density is, in practice, very close to that obtained using this approximation, and
measured inductances for real machines are in reasonable agreement with the
calculated values based on the approximation. The reason for this is that sinu-
soidally distributed windings will only link to the components on a ux density
waveform that have the same pole number as the winding, as was previously
shown in Section 1.2. Therefore, for an ideal sinusoidally distributed winding
only the fundamental component of the ux density can link to the winding,
and consequently harmonic ux densities only contribute to leakages.
Remark A.5 An ideal sinusoidally distributed winding cannot be constructed
all true windings have winding space harmonics. These winding harmonics can
therefore link to harmonic ux densities of the same poll number. This can lead
to the generation of harmonic voltages, and more complex inductance variations
with rotor position.
Addition Could add a section here examining the assumption that the d and
q-axes air gaps can be modelled as constant air gaps. Could consider an
ideal 2 pole axially laminated machine, looking at the eective air gap
seen by the mmf in both the axes.
The following is with reference to Figure A.2, which is a laid out diagram of
Figure A.1. This diagram shows the two ctitious air gaps, with the component
mmfs acting on the d and q-axes respectively. The resultant air gap ux density
distributions are shown for the two axis waveforms. Notice that the resultant
air gap ux density waveform is distorted away from the d-axis of the rotor by
the q-axis ux waveform, the degree of distortion being related to the dierence
between the air gap lengths and the mmf applied in the axes.
In order to calculate the self inductance of the a-phase winding the total
self ux linkage must be calculated for the winding. This self ux linkage has
separate components contributed by both the d and q-axis uxes.
Using the approach in [2] we calculate the ux due to one of the component
mmfs acting on one of the air gaps by proceeding in the following manner:
1. Calculate the ux in an incremental area at some angular position in the
machine accounting for the spatial distribution of the mmf.
2. One then integrates up these incremental uxes for a total span of a single
coil. This gives the total ux linking one coil.
A.1 Calculation of Inductances A-5
Figure A.2: Developed diagram of a SYNCREL.
3. Calculate the ux linking all the coils that have their axes at some angular
position around the machine. This is achieved by multiplying the value
obtained in point 2 by the number of turns that lie in the same position
as the single coil.
4. Finally integrate up the previous value over the coil span accounting for
the change in the number of turns with angular spatial variation.
5. Once the ux linkage for each air gap is found then the total ux linkage
to the a-phase is found by adding together the linkages due to the d and
q axes.
Consider the d-axis, as shown in Figure A.3. The expression for the ux over
a 180

electrical span of the d-axis mmf can be found as follows. Consider the
incremental permeance over an angle of d:
dP
d
=

o
dA
g
d
(A.1)
A.1 Calculation of Inductances A-6
Figure A.3: d axis developed diagram for SYNCREL
where:
dA the incremental area.
= l
m
r d
machine periphery angle relative to the d axis.
l
m
the length of the machine.
r the radius of the machine at the centre of the air gap.

o
the permeability of free space.
Therefore the incremental ux can be written as:
d
d
= dP
d
_

F
d
cos
_
=

F
d

o
l
m
r
g
d
cos d (A.2)
where

F
d
cos is the d-axis component mmf.
To nd the total ux linking a single coil whose most clockwise coil side starts
at radians relative to the d-axis position, we integrate the d-axis incremental
A.1 Calculation of Inductances A-7
uxes d
d
for the dA elements using the following integration:

d
=
_
+

d
d
=
_
+

_

F
d

o
l
m
r
g
d
cos
_
d
=
2

F
d

o
l
m
r
g
d
sin (A.3)
where:

F
d
=

F
a
cos
pd
the component mmf at
pd
, and (A.4)

F
a
the peak mmf of the a-phase.

pd
the angle of the d-axis around the machine periphery (elec-rad) (A.5)
Remark A.6 Note that the above denition of the ux linkage per turn implies
that the normal vector for the coil area is at the angle +/2 radians. Realising
this is important in getting the correct sign for the total ux linkage of the coil.
Clearly the maximum ux of 2

F
d

o
l
m
r/g
d
is obtained when the coil side =
/2 this means that the coil axis is a 0 radians and hence aligns with the
component mmf axis. Equation (A.3) can be written in terms of the total d-axis
permeance by utilising the following expression:
P
d
=
_
2

2
dP
d
=
_
2

o
l
m
r
g
d
d
=

o
l
m
r
g
d
(A.6)
therefore (A.3) can be written as:

d
() =
2

F
d
P
d

sin (A.7)
If a coil side starts at some angle with respect to the d-axis then the coil
axis is at + /2. Dene:

a
angle of the coil axis relative to the d-axis
and hence:

a
= +

2
(A.8)
and consequently:
=
a


2
(A.9)
Substituting this into (A.7) we can write the ux for a single turn whose axis is
at
a
with respect to the d-axis as:

d
(
a
) =
2

F
d
P
d

sin(
a


2
)
=
2

F
d
P
d

cos
a
(A.10)
A.1 Calculation of Inductances A-8
This expression can be further manipulated so that the ux is a function of the
angle of the d-axis and the coil axis with respect to the axis of the a-phase. Let:

a
the angle of the coil axis with respect to the a-phase
therefore:

a
=
a

pd
(A.11)
Substituting this into (A.10) we can write the following:

d
(
a
,
pd
) =
2

F
d
P
d

cos(
a

pd
) (A.12)
We are now in a position to calculate the ux linkage to the turns of a-phase
at some particular coil axis angle
a
for some constant d-axis. The number of
turns that have their coil axis at angle
a
can be deduced from the turns density
function (1.1) as:
n
ta
(
a
) = n
a
cos
a
(A.13)
Remark A.7 Clearly n
ta
(
a
) can be negative. The concept of a negative num-
ber of turns/radian at a particular coil axis angle is related to the concept of a
negative number of conductors around the periphery of the machine (the sign
in this case arising from the direction of current in the conductors at that
point).The turns density function expressed in
a
is essentially the mmf/ampere
for the winding at a particular position. This is also known as a winding func-
tion. Therefore the negative sign indicates that the ux produced is in the op-
posite direction across the air gap (i.e. from the stator to the rotor instead of
from the rotor to the stator).
Therefore the total ux linkage for the number of turns at
a
is:

d
(
a
) =
2

F
d
P
d
n
a

cos
a
cos(
a

pd
) (A.14)
We are now in the position to calculate the total ux linkage of the d-axis
ux to the a-phase by integrating the ux linkage
d
(
a
) at each position
a
for
the coil span of the winding. Therefore the total ux linkage is:

ad
(
pd
) =
2

F
d
P
d
n
a

_
+

cos
a
cos(
a

pd
)d
a
(A.15)
Carrying out this integration and simplifying the result we obtain:

ad
(
pd
) =

F
d
P
d
n
a
cos
pd
(A.16)
In a similar fashion, the ux linkage of the q-axis ux with the a-phase can
be found. The procedure is identical to the above so it will not be presented
in detail. Instead we will simply state the results of the intermediate steps and
then present the nal ux linkage result.
The incremental permeance for the q-axis is:
dP
q
=

o
l
m
r d
g
q
(A.17)
A.1 Calculation of Inductances A-9
and therefore the total permeance of over a coil span is:
P
q
=
_

0
dP
q
=

o
l
m
r
g
q
(A.18)
The q-axis is at an angle of /2 radians with respect to the d-axis. Therefore
the variation of the q-axis mmf is:
F
q
=

F
q
cos(

2
) (A.19)
Therefore the q-axis incremental ux linkage is:
d
q
= dP
q

F
q
cos(

2
) (A.20)
Since cos(

2
) = sin , and substituting for dP
q
in (A.20) gives:
d
q
=

o

F
q
l
m
r
g
q
sin d =

F
q
P
q

sin d (A.21)
Consequently the expression for the ux linkage for a single coil can be written
as:

q
() =
_
+

d
q
=

F
q
P
q

_
+

sin d
=
2

F
q
P
q

cos (A.22)
where an angle relative to the d-axis.
Carrying out the angle conversion to the coil axes relative to the a-phase as
was done in the d-axis case we can write:

q
(
a
) =
2

F
q
P
q

sin(
a

pd
) (A.23)
The total ux linkage of the q-axis ux to the a-phase can therefore be
written as:

aq
(
pd
) =
_
+

n
ta
(
a
)
q
(
a
) d
a
=
2

F
q
P
q
n
a

_
+

cos
a
sin(
a

pd
) d
a

aq
(
pd
) =

F
q
P
q
n
a
sin
pd
(A.24)
In the above expressions the peak values of the d and q-axes mmfs are found
by taking components of the a-phase mmf onto the d and q-axes respectively
as follows:

F
d
=

F
a
cos(
pd
) =

F
a
cos
pd
(A.25)

F
q
=

F
a
sin(
pd
) =

F
a
sin
pd
(A.26)
A.1 Calculation of Inductances A-10
where

F
a
the peak of the a-phase mmf = n
a
i
a
(from (1.2)). Note that the
negative sign in front of the
pd
terms results from the fact that the angle is
measured relative to the d-axis, and not the a-phase axis,since we are projecting
the a-phase mmf onto the d and q axes.
The total ux linkage to the a-phase can now be calculated by using super-
position and adding the components linking to it from the d and q-axes. Using
(A.16) and (A.24) we get:

aa
(
d
) =
ad
(
d
) +
aq
(
d
)
= n
a

F
a
(P
d
cos
2

pd
+ P
q
sin
2

pd
)
= n
2
a
i
a
(P
d
cos
2

pd
+ P
q
sin
2

pd
)
=
n
2
a
i
a
2
[(P
d
+ P
q
) + (P
d
P
q
) cos 2
pd
] (A.27)
The rotor self inductance can therefore be calculated as a function of the
d-axis position as:
L
aa
=

aa
i
a
= L
1
+ L
2
cos 2
pd
(A.28)
where:
L
1
=
N
2
8
(P
d
+ P
q
)
L
2
=
N
2
8
(P
d
P
q
)
N total number of turns in sinusoidal winding
= 2n
a
Figure A.4 shows a plot of (A.28). Notice that the inductance varies as a
function of cos 2
pd
with a constant oset as mentioned in conjecture A.3.
The self inductances for the other two phases can be found similarly as:
L
bb
= L
1
+ L
2
cos 2
_

pd

2
3
_
(A.29)
L
cc
= L
1
+ L
2
cos 2
_

pd
+
2
3
_
(A.30)
Addition Perhaps a remark about the fact that this analysis gives accurate in-
ductance results since only the fundamental components of the ux density
distribution link to the sinusoidal winding, as proved in a previous section.
A.1.2 Mutual Inductances
In addition to the self inductance of the winding there is also mutual inductance
between the a, b, and c-phases. These inductances are also a function of the
position of the rotor, since its position clearly changes the reluctance of the ux
paths between the windings. The process of calculating the general expressions
for these inductances is very similar to that for the self inductances. We shall
work out in detail the mutual inductance between two windings and then simply
state the relationships between the other windings.
A.1 Calculation of Inductances A-11
Figure A.4: a phase inductance plot.
Let us consider the mutual inductance between the a-phase and the b-phase.
The spatial sequence of the phases is as shown in Figure 3.3. The winding
conductor density distribution for the b-phase is:
n
b
(
p
) = n
b
sin(
p

2
3
) (A.31)
Therefore the number of coils with their axes at some angle
a
with respect to
the a-phase (i.e. the winding function) is:
n
tb
() = n
b
cos(
a

2
3
) (A.32)
As with the self inductance we shall work out the ux linkage for the d and
q axes separately, and then use superposition to calculate the total ux linkage.
We can write the expression for the ux linkage for a single turn using the
expression (A.10) calculated for the self inductance case:

d
(
a
) =
2

F
d
P
d

cos
a
(A.33)
Again we can say:

a
=
a

pd
(A.34)
allowing us to again write the ux expression as:

d
(
a
,
pd
) =
2

F
d
P
d

cos(
a

pd
) (A.35)
A.1 Calculation of Inductances A-12
We can now write the ux expression for the coils that have their axis at
a
as:

d
(
a
,
pd
) =
2

F
d
P
d
n
b

cos(
a

2
3
) cos(
a

pd
) (A.36)
Finally we now nd the total linkage of the d-axis ux by integrating over a coil
span of the b-phase:

dba
=
2

F
d
P
d
n
b

_
+

cos(
a

2
3
) cos(
a

pd
)d
a
(A.37)
After considerable manipulation this expression can be written as:

dba
=

F
d
P
d
n
b
cos(
pd

2
3
) (A.38)
Using (A.25) the expression becomes:

dba
=

F
a
P
d
n
b
cos
pd
cos(
pd

2
3
) (A.39)
Now let us consider the q-axis contribution to the b-phase ux. Using (A.10)
we can again write an expression for the q-axis ux linking a single turn centred
at the angle
aq
relative to the q-axis:

q
(
aq
) =
2

F
q
P
q

cos
aq
(A.40)
The
aq
angle con be converted to angle relative to the a-phase:

a
=
pd
+
aq
+

2
(A.41)
and therefore:

aq
=
a
(
pd
+

2
) (A.42)
Hence
q
can be written as:

q
(
a
,
pd
) =
2

F
q
P
q

cos(
a

pd


2
) (A.43)
=
2

F
q
P
q

sin(
a

pd
) (A.44)
Now using the winding function we can write:

q
(
a
,
pd
) =
2

F
q
P
q
n
b

cos(
a

2
3
) sin(
a

pd
) (A.45)
Integrating over the coil span:

qba
=
2

F
q
P
q
n
b

_
+

cos(
a

2
3
) sin(
a

pd
)d
a
(A.46)
After considerable manipulation we arrive at the expression for the ux
linkage from the q-axis to the b-phase:

qba
=

F
q
P
q
n
b
cos(
pd


6
) (A.47)
A.1 Calculation of Inductances A-13
Using (A.26) this expression can be written as:

qba
=

F
a
P
q
n
b
sin
pd
cos(
pd


6
) (A.48)
We are now in a position to calculate the total mutual ux linkage to the
b-phase from the a-phase as follows:

ba
=
dba
+
qba
=

F
a
n
b
[P
d
cos
pd
cos(
pd

2
3
) P
q
sin
pd
cos(
pd


6
)]
=

F
a
n
b
_
P
d
_

1
4
(1 + cos 2
pd
) +

3
4
sin 2
pd
_
+
P
q
_

3
4
sin 2
pd
+
1
4
(1 cos 2
pd
)
_
_
(A.49)
After manipulation we get the following expression for this mutual inductance:

ba
=
n
a
n
b
i
a
2
_
(P
d
+ P
q
)
2
+ (P
d
P
q
) cos(2
pd

2
3
)
_
(A.50)
For a balanced machine n
a
= n
b
, therefore the term in front of this expression
is n
2
a
i
a
/2. Therefore this expression is the same as that for the self inductances
and hence we can write the mutual inductance in the same form as that for the
self inductances:

ba
=
N
2
i
a
8
_
(P
d
+ P
q
)
2
+ (P
d
P
q
) cos(2
pd

2
3
)
_
(A.51)
Dividing (A.51) by i
a
gives the inductance expression:
L
ba
= L
ab
=
L
1
2
+ L
2
cos 2(
pd


3
) (A.52)
where L
1
and L
2
are as dened in (A.28).
By a similar process it can be shown that the other mutual inductances are:
L
ca
= L
ac
=
L
1
2
+ L
2
cos 2(
pd
+

3
) (A.53)
L
cb
= L
bc
=
L
1
2
+ L
2
cos 2
pd
(A.54)
A.1.3 Summary
Assuming that the mmf for each phase varies sinusoidally around the machine,
and that the resultant mmf in the machine acts on two dierent air gaps for the
low and high permeance axes, then the self and mutual inductances of a phase
winding vary as follows with
pd
(the angle of the d-axis with the a-phase). In
the above derivations we did not take into account the leakage inductance term
in each of the self inductances. If we assume that the leakage does not change
with rotor position (which may not be a valid assumption) then the leakage can
be included by the addition of the term L
l
as shown below:
A.1 Calculation of Inductances A-14
Self Inductances
L
aa
= L
l
+ L
1
+ L
2
cos 2
pd
L
bb
= L
l
+ L
1
+ L
2
cos 2
_

pd

2
3
_
L
cc
= L
l
+ L
1
+ L
2
cos 2
_

pd
+
2
3
_
_
_
_
(A.55)
Mutual Inductances
L
ba
= L
ab
=
L
1
2
+ L
2
cos 2(
pd


3
)
L
cb
= L
bc
=
L
1
2
+ L
2
cos 2
pd
L
ca
= L
ac
=
L
1
2
+ L
2
cos 2(
pd
+

3
)
_
_
_
(A.56)
where:
L
1
=
N
2
8
(P
d
+ P
q
)
L
2
=
N
2
8
(P
d
P
q
)
N total number of turns in sinusoidal winding
= 2n
a
L
l
the leakage inductance of each phase
Appendix B
Winding Functions
B.1 Introduction
The computation of self and mutual inductances for machines is crucial when
one is trying to develop a set of dynamic equations for the performance of a
machine. A technique using the concept of winding functions has been shown
to be useful for developing inductance expressions for complex machine and
winding congurations [1, 7]. Because of the diculty of obtaining the original
source for this technique [3], this appendix will attempt to develop the basis of
the technique for some very simple winding congurations, and then show that
the basic idea can be extended to more complex geometries.
The winding function based expression for the mutual inductance between
two arbitrary windings , i and j, in a machine is:
L
ij
=
0
rl
_
2
0
g
1
(,
rm
)N
i
(,
rm
)N
j
(,
rm
) d (B.1)
where:
r radius of the circle of integration
l length of the stack of the machine
g(,
rm
) airgap function
N
i
(,
rm
) i phase winding function
N
j
(,
rm
) j phase winding function
mechanical angle around the machine

rm
mechanical angle of the rotor

0
permeability of free space
Remark B.1
rm
is the angle of the rotor with respect to a reference. This
angle is only relevant in relation to the calculation of inductance if the rotor has
saliency that is the air gap function is not a constant with respect to .
Remark B.2 The radius r is not obvious when we are dealing with singly or
doubly salient structures. The value must be chosen in such a way as to ensure
B.2 Ideal Sinusoidal Winding B-2
n n
a a
( ) sin
n n
b b
( ) sin( )
(Rad)
Figure B.1: Two pole sinusoidal winding layout
that the ux linking into the winding is correct, since the radius is a crucial
component in determining the dA areas when computing incremental uxes.
The following sections will, by examples, develop an understanding of the terms
in this equation and how they are derived for a particular machine.
B.2 Ideal Sinusoidal Winding
We shall begin by considering the standard sinusoidally distributed windings.
The windings we will consider will be idealized ones, but the same technique can
be used for real windings that are distributed in slots. The situation that we
are considering is the mutual inductance between two sinusoidally distributed
windings with the same pole number that are at some phase angle to each other.
For simplicity we shall assume that the machine has a uniform air gap g and the
iron of the machine has innite relative permeability. This in turn implies that
one can consider that the mmf is all expended in driving ux across the machine
air gaps. Figure B.1 shows the layout of the windings. We are assuming that
we have a two pole conguration.
B.2.1 Conventional inductance calculation
The plan of attack is to rstly look at calculating the mutual inductance between
the windings using a conventional technique. Then we shall use a slightly dier-
ent way of calculating the inductance, which will lead to the winding function
formulation.
In order to calculate the mutual inductance between the two windings we
have to calculate the ux that links the two windings. In this example we shall
B.2 Ideal Sinusoidal Winding B-3
assume that winding a has a current of i
a
amps and we are interested in the
ux that links to winding b.
Figure B.2 shows the a phase winding sitting on the stator of a machine.
The mmf for winding a can be computed by carrying out the Amperes law
line integral:
i =
_
H.dl (B.2)
around the path shown in Figure B.2, and realising that H = F/l, where F is
the mmf.
For some small increment d the amount of enclosed current is:
di = n
a
i
a
sin d (B.3)
where is an angle around the machine circumference.
If we carry out integrate over the coil span then we are enclosing the current
inside the path of integration . To calculate this current we need to add together
the incremental currents around the periphery of the machine for the length of
the enclosed path. This gives the total enclosed ampere turns of the winding at
some position . Therefore the expression is:
i
T
() =
_
+

di
=
_
+

n
a
i
a
sin d
= 2n
a
i
a
cos
= 2

F
a
cos (B.4)
For a typical machine we assume that the magnetic materials have innite
permeability. This in turn means that no magnetic eld intensity is required
to force ux through the iron. Consequently all of the H, and therefore F, is
expended to force ux across the airgap.
The mmf for the a phase can be found from i
T
() by realising that the
integration path in Figure B.2 crossing two air gaps, therefore half the mmf
calculated is used to cross each air gap. Therefore the mmf expression for the
a phase becomes:
F
a
() =

F
a
cos (B.5)
where

F
a
() = n
a
i
a
.
Now that we have an expression for the mmf of the a phase we can compute
the ux density at some arbitrary angle around the periphery of the machine
using the basic fact that:
B =
o
H (B.6)
and H = F/l we can write:
B
a
() =

o
g
F
a
()
=

o

F
a
g
cos (B.7)
B.2 Ideal Sinusoidal Winding B-4
Figure B.2: Calculation of the MMF for a sinusoidal winding.
In order to nd the total ux linkage to winding b we need to add up all
the incremental uxes over a complete coil span of winding b.
The incremental ux linking any incremental area of the machine is at some
angle is:
d = B
a
()dA (B.8)
Now if the radius of the area in question is r then:
dA = rl d (B.9)
where d is an incremental angle. Therefore:
d = B
a
()rl d (B.10)
To get the total ux linking a single turn of the b phase whose coil side starts
at we need to integrate up the incremental uxes for the total coil span of the
coil. Stated mathematically this is:

ab
() =
_
+

d
=

o
rl

F
a
g
_
+

cos d
=
2
o
rl

F
a
g
sin (B.11)
In order to nd the total ux linkage for the whole phase we need to integrate
up the ux for each coil of the phase.
B.2 Ideal Sinusoidal Winding B-5
The number of turns whose coil sides start in the b phase at some particular
angle around the machine is:
n
b
() d = n
b
sin( ) d (B.12)
Therefore the ux linking the coils starting at is:
n
b
()
ab
() d (B.13)
Finally in order to get the total ux linkage for the entire phase we need to
integrate up the ux linking the coils for each value of for the entire phase
span. Therefore the expression for the ux linking from the a phase to the b
phase is:

ab
=
_
+

n
b
()() d
=
2
o
rl

F
a
n
b
g
_
+

sin( ) sin d

ab
=

o
rl

F
a
n
b

g
cos (B.14)
B.2.2 Alternative inductance calculation
In this section we shall carry out the same inductance calculation as above, but
in this case using the turns function for the b phase. One of the main dier-
ences that arise from this approach is that the integration is carried out over
2 mechanical radians. The only real dierence between these two techniques
is that we calculate the total ux linkage to winding b for an incremental
area dA at some angle of , and then add up all the areas for the whole wind-
ing. At any particular there are in general a number of turns that are being
linked. In the previous method we calculated the ux linking a whole single
coil of phase b with its coil side starting at some angle , and then added
up the ux for all the coils. The dierence between the two techniques can be
seen in Figure B.3, where the rst technique nds the ux linking a group of
coils whose coil sides start at
1,
and the second technique nds the incremen-
tal ux linking all the coils at position
2
. This is represented in the diagram
as B(
2
)
_
_

2
o
nsin d
_
rl d where the
_

2
o
nsin d term corresponds to the
cumulative number of turns to the
2
point along the winding.
Therefore general cumulative coil number expression for any arbitrary angle
is:
C() =
_

nsin( ) d
= n[1 cos( )] (B.15)
This function is plotted in Figure B.4. Notice that the maximum number of
cumulative turns occurs at + (i.e. after the pole pitch of the winding) as
intuition would tell us.
In order to calculate the total mutual ux between the a and b phases
using this approach we have to add up the incremental mutual ux linkages
B.2 Ideal Sinusoidal Winding B-6
Figure B.3: Dierent methods of calculating the ux linking a coil.
2
2n
Figure B.4: Cumulative number of turns for a sinusoidally distributed winding.
B.2 Ideal Sinusoidal Winding B-7
for the cumulative number of b phase coils over the 2 radian span of the b
phase. Using (B.15) the b phase cumulative number of coils at any position
is:
C
b
() = n
b
[1 cos( )] (B.16)
Clearly the total ux linking the b winding is:

ab
=
_
2
0
d
=
_
2
0
C
b
() d (B.17)
Using (B.7) and (B.8) we can expand the d expression so that we get:

ab
=

0
rl
g
_
2
0
C
b
()F
a
() d (B.18)
This can be further expanded by substituting for C
b
() and F
a
() to give:

ab
=

0
rl

F
a
n
b
g
_
2
0
[1 cos( )] cos d
=

0
rl

F
a
n
b
g
__
2
0
cos d
_
2
0
cos( ) cos d
_
=

0
rl

F
a
n
b
g
_
2
0
cos( ) cos d
=

0
rli
a
g
_
2
0
n
b
cos( )n
a
cos d (B.19)
Notice in (B.19) that the terms inside the integral are F
b
()/i
b
and F
a
()/i
a
.
These terms are called the winding functions for the winding and are usually
given the notation below:
N
a
() F
a
()/i
a
(B.20)
N
b
() F
b
()/i
a
(B.21)
allowing the above expression to be written as:

ab
=

0
rli
a
g
_
2
0
N
b
()N
a
() d (B.22)
As a test we can substitute the appropriate expressions for our example
system into the above and we get:

ab
=

0
rli
a
g
_
2
0
n
b
cos( )n
a
cos d
=

0
rli
a
n
a
n
b
g
cos
=

0
rl

F
a
n
b

g
cos (B.23)
B.3 Non-sinusoidal winding B-8
which is the same as (B.14).
From (B.22) expression it is simple to see that the mutual inductance be-
tween these two windings is:
L
ab
=
|
ab
|
i
a
=
0
rl
_
2
0
g
1
N
b
()N
a
() d (B.24)
The situation used for the above development is very simple. Consequently
(B.24) is a simpler function form compared to what can happen in more complex
machines. For example, one of the windings can be located on a rotor, and
therefore the winding function for this winding can become a function of the
rotor position as well as . Also in the above development the air gap g has been
assumed to be constant. However, if saliency is present in a machine structure
then g will be a function of . The situation is even more complex if the saliency
is on the rotor, as g then also becomes a function of the rotor angle as well.
Therefore, the general form of the mutual inductance expression becomes:
L
ab
=
0
rl
_
2
0
g
1
(,
rm
)N
a
(,
rm
)N
b
(,
rm
) d (B.25)
B.3 Non-sinusoidal winding
As an example of the utility of this technique for determining the mutual induc-
tances we shall calculate the mutual inductance between two windings that have
dierent pole pitches. Furthermore the distribution of the windings around the
mechanical periphery of the machine is not symmetric. The case is simplied by
considering the two windings to be concentrated windings and to be mounted on
cylindrical magnetic structures. The usual innite iron permeability assumption
is used, and the air gap is assumed to be constant.
The approach taken will be to rstly calculate the mutual inductance us-
ing basic principles, and then to calculate the inductances using the winding
function technique.
B.3.1 Inductance Using Basic Principles
Figure B.5 is a developed diagram of the winding arrangement for the machine.
The plots show the turns function for the b winding and the mmf as a function
of for the a winding. Figure B.6 shows the physical arrangement of the
windings.
Remark B.3 The span of the coil dened in Figure B.5 is arbitrary. There are
two possible denitions for coil span for both coils, and it does not matter which
is chosen.
The mmf diagram for the a winding can be determined by using the mag-
netic circuit concept. The current in the winding produces a ux across an air
gap of length g an area rl
a
, where r is the circumference of the machine, and
l is the axial length as in previous derivations. This same ux has to cross the
return air gap, which is the same length but has an area of rl(2
a
). Clearly
the rst air gap and the second air gaps have dierent reluctances due to the
B.3 Non-sinusoidal winding B-9
b
2
2
n
b
F
a
( )
a
2
2
a
a a
n i
a
a a
n i
2
Figure B.5: Turns function and mmf distribution for two fractional pitch wind-
ings.
a
b
Figure B.6: Physical layout of the non-sinusoidal winding.
B.3 Non-sinusoidal winding B-10
dierent areas of the air gaps. This can be clearly seen from the denition of
reluctance:
R =
g

0
A
(B.26)
where A is the area orthogonal to the ux.
Using these ideas we can write the following expression for the ux produced
by the a winding:
=
n
a
i
a
R

a
+ R

a
(B.27)
where R

a
and R

a
correspond to the reluctance of the
a
and 2
a
air gaps
that the ux has to cross. Since magnetic circuits obey the analogous relation-
ships as conventional electrical circuits, we can write the following expressions
for the mmf required for the ux to bridge each of the air gaps:
F

a
a
=
R

a
R

a
+ R

a
n
a
i
a
(B.28)
F

a
a
=
R

a
R

a
+ R

a
n
a
i
a
(B.29)
Now:
R

a
=
g

0
rl
a
(B.30)
R

a
=
g

0
rl(2
a
)
(B.31)
therefore:
F

a
a
=
g

0
rl
a
g

0
rl
a
+
g

0
rl(2
a
)
n
a
i
a
=
_
2
a
2
_
n
a
i
a
(B.32)
and
F

a
a
=
g

0
rl(2
a
)
g

0
rl
a
+
g

0
rl(2
a
)
n
a
i
a
=

a
2
n
a
i
a
(B.33)
Clearly F
a
= F

a
a
+ F

a
a
= n
a
i
a
. The negative sign results in Figure B.5 due to
the convention that ux owing out of the rotor is due to positive mmf.
Given this information we can now write down the ux density produced by
the a winding in the air gap corresponding to
a
:
B

a
a
=
0
F

a
a
g
=

0
g
_
2
a
2
_
n
a
i
a
(B.34)
B.3 Non-sinusoidal winding B-11
This ux density links the b phase winding from to
a
, therefore the ux
through the winding due to this ux density is:

a
ab
=
_

a

0
g
_
2
a
2
_
n
a
i
a
rl d
=

0
g
_
2
a
2
_
n
a
i
a
rl(
a
) (B.35)
Similarly the remaining section of the b phase is linked by the return ux
from the a winding. The return ux density is:
B

a
a
=
0
F

a
a
g
=

0
g
_

a
2
_
n
a
i
a
(B.36)
Therefore the ux component linking to the b winding is:


a
ab
=
_

b
+

0
g
_

a
2
_
n
a
i
a
rl d
=

0
g
_

a
2
_
n
a
i
a
rl(
b
+
a
) (B.37)
The total ux linkage from winding a to b is therefore:

ab
= n
b
(

a
ab
+

a
ab
)
=

0
n
b
n
a
i
a
rl
g
__
2
a
2
_
(
a
)
_

a
2
_
(
b
+
a
)
_
=

0
n
b
n
a
i
a
rl
g
_
2(
a
)
a

b
2
_
(B.38)
and consequently the mutual inductance is:
L
ab
=

0
n
b
n
a
rl
g
_
2(
a
)
a

b
2
_
(B.39)
B.3.2 Inductance Using Winding Functions
We shall now compute the inductance using (B.24). The rst step is to nd
the winding functions for the windings. This is very simple as we have already
calculated the mmf waveform for the a winding. Its winding function is sim-
ply this divided by i
a
. The winding function for the b phase can be written
by inspection because of the similarity to the a winding. Therefore the two
winding functions are:
N
a
() =
_ _
2
a
2
_
n
a
for 0
a

a
2
_
n
a
for 0
a
(B.40)
N
b
() =
_ _
2
b
2
_
n
b
for
b
+

b
2
_
n
b
for
b
+
(B.41)
B.4 Flux Linkage Expression B-12
Breaking (B.24) into a piecewise continuous integral and substituting the
appropriate values for N
a
() and N
b
() in each of the integrals we can write:

ab
=

0
n
a
n
b
rl
g
__

0
_
2
a
2
_
_

b
2
_
d +
_

a

_
2
a
2
__
2
b
2
_
d
+
_

b
+

a
_

a
2
_
_
2
b
2
_
d +
_
2

b
+
_

a
2
__

b
2
_
d
_
(B.42)

ab
=

0
n
a
n
b
rl
g
__
2
a
2
_
_

b
2
_
+
_
2
a
2
__
2
b
2
_
(
a
)
+
_

a
2
_
_
2
b
2
_
(
b
+
a
) +
_

a
2
__

b
2
_
(2
b

a
)
_
(B.43)
After some simplication and dividing by i
a
we obtain the following expres-
sion for the mutual inductance:
L
ab
=

0
n
b
n
a
rl
g
_
2(
a
)
a

b
2
_
(B.44)
which is exactly the same as that derived using the basic principles approach.
B.4 Flux Linkage Expression
There is another useful expression that is related to the general inductance
expression, and that is the general ux linkage expression. This expression can
be simply derived from (B.1) by using the following relations:
B(,
rm
) =
0
H and (B.45)
F(,
rm
) =
_
H.dl = Hg(,
rm
) (B.46)
therefore (B.46) can be written as:
F(,
rm
) =
g(,
rm
)B(,
rm
)

0
(B.47)
If (B.1) is multiplied by i
i
(or i
j
) then we and up with an expression of the
form:
L
ij
i
i
=
ij
=
0
rl
_
2
0
g
1
(,
rm
)F
i
(,
rm
)N
j
(,
rm
) d (B.48)
and substituting (B.47) for F
i
(,
rm
) in this we can then write the following
for the general mutual ux linkage expression:

ij
= rl
_
2
0
B
i
(,
rm
)N
j
(,
rm
) d (B.49)
B.5 A Note on Winding Functions for Multi-pole Machines B-13
B.5 A Note on Winding Functions for Multi-pole
Machines
Consider a generic winding with the following conductor distribution:
n() = n
pk
sin p
p
(B.50)
where:
n
pk
the peak conductor density [conductors/rad]
periphery angle around machine [mech-rad]
p
p
pole pairs of the winding.
Assuming that the poles are series connected, the total number of turns in
the winding is equal to the number of turns in a pole pair multiplied by the
number of pole pairs. We can nd the number of turns in a pole pair by adding
up the number of conductors in electrical radians of the winding,which is /p
mechanical radians. We can therefore write the total turns for a phase of the
winding as follows:
N = p
_
p
p
0
n
pk
sin p
p
d
= p
_

n
pk
p
p
cos p
p

_
p
0
= 2n
pk
(B.51)
Now let use consider the mmf for a winding with the distribution as in (B.50).
Assume that the winding is carrying a current of i Amp in each conductor.
Therefore we can see that for an angle d we have the following total current in
an element of the winding at some angle :
di
T
= n
pk
i sin p
p
d (B.52)
To nd the mmf produced by the winding we employ Amperes Law and inte-
grate to get the total current enclosed by a coil span at some angle :
F
T
=
_
+

n
pk
i sin p
p
d
=
2n
pk
i
p
p
cos p
p

=
Ni
p
p
cos p
= N

i cos p (B.53)
where:
N

the turns/phase/pole pair = N/p


p
B.6 Conclusion B-14
F
T
is the total mmf, which is expended across two air gaps in a machine.
Therefore the mmf/air gap is:
F =
F
T
2
=
N

i
2
cos p
p
(B.54)
=
Ni
2p
p
cos p
p
=
2n
pk
i
2p
p
cos p
p

=
n
pk
i
p
p
cos p
p
(B.55)
The winding function is dened as the mmf/amp, therefore the winding
function for this multi-pole winding becomes:
N
A
() =
N

2
cos p
p
(B.56)
=
n
pk
p
p
cos p
p
(B.57)
Therefore we can express the amplitude of the winding function in terms of the
turns/phase/pole pair, or alternatively in the peak conductors/rad/pole pair.
B.6 Conclusion
This appendix has attempted to show the basis for the use of winding functions
for the calculation of the mutual inductance between windings. Two examples
have been used to achieve this, one involving idealised sinusoidal windings, and
the other an unusual set of windings. In a real machine one does not have
pure sinusoidally distributed windings. In these situations one can compute
the winding functions by simply evaluating the mmf waveforms that are pro-
duced by the real windings. These mmf waveforms will contain all the winding
harmonics, therefore the inductances calculated will be accurate subject to the
innite permeability assumption and the modelling of the air gap function.
Bibliography
[1] F. Liang, L. Xu, and T. Lipo. D-q analysis of a variable speed doubly AC
excited reluctance motor. Electric Machines and Power Systems, 19(2):125
138, March 1991.
[2] D. OKelly and S. Simmons. Introduction to Generalized Electrical Machine
Theory. McGraw Hill, England, 1968.
[3] N. Schmidt and D. Novotny. Introductory Electro-Mechanics. Ronald Press:
New York, 1965.
[4] G. Slemon. Electric Machines and Drives. Addison-Wesley, 1992.
[5] D. Staton, W. Soong, and T. Miller. Unied theory of torque production
in switched reluctance and synchronous reluctance motors. IEEE Trans. on
Industry Applications, IA-31(2):329337, 1995.
[6] P. Vas. Vector Control of AC Machines. Oxford University Press, 1990.
[7] L. Xu, F. Liang, and T. Lipo. Transient model of a doubly excited reluctance
motor. IEEE Trans. on Energy Conversion, 6(1):126133, March 1991.

You might also like