Parasitic Effects in PM Machines with
Concentrated Windings
Freddy Magnussen, Heinz Lendenmann
ABB
Corporate Research
Vsters, Sweden
Abstract PM machines using concentrated windings are gaining
in popularity at the expense of distributed windings in various
applications, mainly due to cost savings. The result is an
increased amount of parasitic effects like ripple torque,
alternating magnetic fields in the rotor, unbalanced radial forces
and magnetic noise. This paper describes the reasons for the
parasitic effects, which machine topologies that are especially
sensitive and suggests measures in order to reduce their
importance. Both traditional and modular concentrated windings
are analyzed, as well as double layers and single layer windings.
Measurements on a prototype motor and three commercial
servomotors have demonstrated that modular motors are
favorable regarding ripple torque minimization.
Keywords-permanent magnet machines; parasitic effects;
concentrated windings; ripple torque; unbalanced radial forces;
magnetic noise
I. INTRODUCTION
A. Traditional brushless PM machines
The traditional winding configuration of a PM machine
with concentrated windings has a number of slots per pole per
phase of 0.5 and normally double layers per slot, i.e. one phase
coil around every single tooth. Some small motors use a single
layer per slot, i.e. one phase coil around every second tooth,
because of simpler manufacturing. The relation of the slot and
pole combinations for these three-phase machines is written
3p
Q= , (1)
2
where Q and p is the number of slots and poles,
respectively. In these types of machines the slot pitch is 2/3rd
the length of the pole pitch and consequently the stator
windings are also only linking 2/3rd of the flux produced by the
permanent magnets in the rotor. The result is a fundamental
winding factor of 0.87. The electromotive force and the torque
production capability are proportional to this factor. Compared
to the conventional distributed winding, which is linking
almost all the flux from the rotor and therefore has a
fundamental winding factor equal or close to unity, this is an
inherit drawback with the traditional concentrated windings.
The strengths are cost effective manufacturing, short end-
windings, high copper fill factor in slots, high thermal
IAS 2005 1044
conductivity between winding and core and short axial build.
These machines are therefore gaining popularity in applications
that strongly emphasizes the described qualities, like
servomotors for robots and traction motors for electric or
hybrid electric vehicles. The analysis of concentrated windings
in this paper deals with machines with constant slot widths.
Machines can also be designed with variable slot widths [1].
There are means to increase the fundamental winding factor in
machines with concentrated windings. Machines may have
coils joining the same phase around consecutive teeth.
Brushless PM machines with such a distinction are sometimes
called modular [2]-[3].
B. Modular PM machines
Besides the high fundamental winding factor and excellent
torque capability, the modular PM machines in general also
have a manufacturing advantage, thanks to their smaller
number of slots for a given number of poles, and low cogging
and ripple torques, due to a high least common multiple (LCM)
between the number of poles and slots. The number of feasible
slot and pole combinations for three-phase modular PM
machines increases with the number of poles. The number of
slots must be divisible by three and the phase windings must
fulfill the basic requirement of a three-phase winding, i.e. the
resulting phase windings are displaced an electrical angle of
120 degrees from each other. When the number of slots per
phase is odd, the feasible slot and pole combinations are
Q = p 1 , (2)
while the combinations when the number of slots per phase
is even are given by
Q = p 2. (3)
The coils forming one phase from combinations given by
(2) must all be connected in series, since the EMF which is
induced in each phase coil is not exactly in phase [3]. If they
hypothetically were connected in parallel, this would result in
circulating currents within the phase windings. The
combinations from (3) however can be connected in series or
series-parallel groups if the number of pole pairs is even or in
series-anti-parallel groups if the number of pole pairs is odd
0-7803-9208-6/05/$20.00 2005 IEEE
[3]. The combinations from (3) therefore offer greater
flexibility in production.
Some disadvantages with the modular machine topology
have been identified [4]. Especially the 14 poles and 15 slots
combination and furthermore odd slots per phase combinations
have some parasitic effects like unbalanced radial magnetic
forces and alternating magnetic fields in the rotor. In fact there
are not that many feasible combinations for machines with few
pole numbers that even have high fundamental winding factors
[4]. The most promising candidate with the lowest pole number
is the 10 poles and 12 slots combination. In medium- to high-
speed applications like traction, this seems to be the most
attractive alternative in order to avoid high frequency problems
in the machine core and with the control. This combination is
therefore studied and compared with a traditional brushless PM
machine with the same number of slots, but with 8 poles, in the
next section. Both double layers and single layer windings are
considered. The idea is to give an intuitive understanding of the
typical performance of modular versus traditional brushless
PM machines.
II. ELECTROMAGNETIC ANALYSIS
In this section a comparative study of traditional and
modular brushless PM machines intended for sinusoidal
excitation is made. The traditional winding configuration of a
PM machine with concentrated windings has a number of slots
per pole per phase of 0.5 and double layers per slot, i.e. one
phase coil around each tooth. This topology is the reference in
the study. The studied machines are shown in Figure 1. They
all have the same stator cores with 12 slots. The outer stator
diameter is 105 mm, while the rotor diameter is 60 mm. The
active length is 100 mm. The machines differ in their number
of poles (8 or 10) and winding configurations. The radii of the
permanent magnets are slightly smaller than the air gap radius
and the permanent magnets are diametrically magnetized in
order to achieve more sinusoidal open-circuit voltages. The
studied parameters are electromotive force (EMF),
magnetomotive force (MMF), inductance, cogging and ripple
torque, torque capability, radial magnetic forces and alternating
flux densities in the rotor core. The inductance and torque
computations are done with a current loading of 350 A/cm
(RMS). This value of the current loading is theoretically
regarded as being the rated working point of the machines,
which in practice also is dependent of the cooling system.
A. Electromotive force (EMF)
The analytically calculated fundamental winding factors of
the machines in Figure 1 (a)-(d) are 0.87, 0.87, 0.93 and 0.97,
respectively. The finite element computed open-circuit
voltages of the machines at an angular speed of 1000 rpm are
shown in Figure 2. It is shown that the modular machines have
higher fundamental EMFs than the traditional machines and
their waveforms are also more sinusoidal, resulting in lower
ripple torques.
B. Magnetomotive force (MMF)
The winding layouts and the FEM-computed magnetic flux
distribution caused by the magnetomotive force only (the
IAS 2005 1045
permanent magnets are inactive) for the four machines are
shown in Figure 1. The analytically calculated total and
fundamental three-phase MMF, with peak current in one phase,
versus the pole pitch and harmonic content for the four
machines are shown in Figure 3. The effect of slotting is
disregarded, in order to make the influence of the winding
distribution more transparent. The MMFs are normalised in the
sense that all four machines have the same current loading, i.e.
the windings with double layers have one coil turn around
every tooth with unity current amplitude, while the windings
with single layer have two coils around every second tooth
with unity current amplitude. Some interesting characteristics
of machines with concentrated windings are illustrated.
All machines with concentrated windings have even MMF
harmonics, because the space angle between the coil-sides in
each phase is not constant. The machines with single layer
windings have sub-harmonics that are greater than their
fundamental MMF harmonic and they have overall higher
harmonic contents than machines with double layers windings.
The modular machines with double layers windings have
sub-harmonics, while the traditional machines with double
layers windings do not. The only desired MMF component is
the fundamental, which is required to create torque. The rest of
the components are undesired, but are evidently present with
varying magnitudes in these types of machines. The harmonics
cause higher inductance, earlier saturation of the core, higher
core losses, rotor losses and unbalanced radial magnetic forces
or noise.
The machines with single layer windings have sub-
harmonics that are greater than their fundamental MMF
harmonic and they have overall higher harmonic contents than
machines with double layers windings.
+A
-A
+B
-B
(a) (b) +C
-C
Air
Iron
PM
(c) (d)
Figure 1. The winding layout and magnetic flux distribution caused by the
magnetomotive force only (the permanent magnets are inactive) for the
machines with (a) 8 poles and double layers, (b) 8 poles and single layer, (c)
10 poles and double layers and (d) 10 poles and single layer.
0-7803-9208-6/05/$20.00 2005 IEEE
 70
60
the least common multiple between the pole and slot numbers,
50
10 poles single layer
10 poles double layers
which is 24 compared to 60. The peak-to-peak cogging torque
40 8 poles double layers is approximately 4% of the rated torque for the traditional
30 8 poles s ingle layer
machines, while the cogging torque for the modular machines
20
10 is practically zero. As expected from the study of the
EMF [V]

0 electromotive forces, the traditional machines also have higher

-10 ripple torques, because of their more influential 5th and 7th
-20
-30
harmonic that interacts with the fundamental current
-40 component and produce a 6th torque harmonic component (6
-50 times the fundamental electrical frequency). The average
-60
-70
torque of both the modular machines is higher though, due to
0 30 60 90 120 150 180 210 240 270 300 330 360 their higher fundamental open-circuit voltages.
Electrical angle [DEG]
D. Inductance
Figure 2. (a) FEM-computed open-circuit voltages at an angular speed of
1000 rpm of the four PM machines. (The curves of the 8 pole machines are on
The unsaturated inductances are FEM-computed at a
each other.) current loading of 350 A/cm (RMS) and 100 turns per phase.
The machines are consequently operated at the same phase
current. The end-winding inductance is neglected. The
inductances of the machines with single layer windings are
51% higher, while the inductance for the 10 poles machine
with double layers windings is 8% higher, than for the 8 poles
machine with double layers windings.

E. Alternating magnetic fields in the rotor

(a)
(c)
Figure 3. The analytically calculated total and fundamental three-phase
MMF versus the pole pitch and harmonic content for the machines with (a) 8
poles and double layers, (b) 8 poles and single layer, (c) 10 poles and double
layers and (d) 10 poles and single layer. The effect of slotting is disregarded.
The modular machines with double layers windings have
sub-harmonics, while the traditional machines with double
layers windings do not. The only desired MMF component is
The FEM-computed tangential flux density component at a
fixed point on the rotor core at a radius of 22 mm in between
(b) two permanent magnets, when the rotor is rotated two pole
pitches at a current loading of 350 A/cm (RMS) is shown for
each of the four studied machines in Figure 5.
The traditional machine with double layer windings is not
especially sensitive to alternating fields in the rotor, as seen in
Figure 5. However the traditional machine with single layer
windings has a considerable flux density variation in the rotor.
The modular machine with double layer windings has also got
a moderate alternating field. The modular machine with single
layer windings has a variation that is quite severe. The reason
for the different behaviours can be intuitively understood by
(d) considering the different MMF waveforms in Figure 3. The air
gap flux density caused by the armature reaction has the same
waveform as the total three-phase MMF. The armature reaction
interacts with the permanent magnet produced flux density to
form the resulting air gap flux density. Since the MMF of the
traditional machine with double layer windings is pretty
regular, the air gap flux and the rotor flux is also regular.
16

the fundamental, which is required to create torque. The rest of 14

the components are undesired, but are evidently present with 12

varying magnitudes in these types of machines. The harmonics 10

10 poles single layer
Torque [Nm]

cause higher inductance, earlier saturation of the core, higher 8 10 poles double layers
8 poles double layers
core losses, rotor losses and unbalanced radial magnetic forces 6
8 poles single layer

or noise. 4 Cogging (8 poles)

Cogging (10 poles)
2

0
C. Torque capability and ripple torque
-2
The finite element computed ripple torques at a current 0 30 60 90 120 150 180 210 240 270 300 330 360

loading of 350 A/cm (RMS) and the cogging torques are Electrical rotor position [DEG]

shown in Figure 4. The windings are fed with purely sinusoidal

Figure 4. The finite element computed ripple torques at a current loading of
currents. The cogging torque of the traditional machines is 350 A/cm (RMS) and the cogging torques.
higher than for the modular machines, due to the lower value of

IAS 2005 1046 0-7803-9208-6/05/$20.00 2005 IEEE

 1,5
1,4
1,3
1,2
1,1
1,0
Flux density [T]

0,9
0,8
0,7
0,6
0,5 8 poles single layer
0,4
0,3
10 poles single layer Figure 6. Some radial vibration mode shapes for the stator and rotor.
8 poles double layers
0,2
0,1 10 poles double layers
0,0
0 30 60 90 120 150 180 210 240 270 300 330 360
Electrical angle [DEG]

Figure 5. The FEM-computed tangential flux density component at a fixed

point on the rotor core at a radius of 22 mm in between two permanent
magnets, when the rotor is rotated two pole pitches at a current loading of 350
A/cm (RMS).

However, for the machines with single layer windings in

particular, the MMF waveforms are highly irregular, and
consequently cause the alternating fields in the rotor. In order
to avoid considerable rotor losses in the modular machine with
double layers windings it is perhaps sufficient to laminate the
rotor core and manufacture the permanent magnets in many
pieces to reduce the otherwise induced eddy current losses. For
the machines with single layer windings this remedy is
probably not even enough to avoid reduced machine
efficiencies.

F. Unbalanced radial forces and magnetic noise

Irregular MMF waveforms can also cause unbalanced
radial magnetic forces and magnetic noise. Consider the
layouts of the phase coils and the flux distributions caused by
the magnetomotive forces in Figure 1. The machine in (a) has
one quarter of the circumference between the coils in the same
phase, while the phase coil displacement is one half of the
circumference for the other three machines. This means that the
radial magnetic force is in balance for every quarter of the air
gap circumference for the machine in (a), while it is in balance
only for every half of the circumference for the other machines.
The traditional machine with double layers windings will
therefore have the lowest magnetic noise of the four machines.
The effect of unbalanced radial forces can be exemplified by
considering the stator or rotor to be a metal ring with different
radial vibration mode shapes as illustrated in Figure 6. A
unidirectional deflection of the stator and rotor for instance
represents a radial vibration shape of mode one [5]. Some other
modes are shown in the same figure. The radial forces usually
cause undesired magnetic noise. The sound level depends on
several factors like rigidity of the machine parts, machine
dimensions and location of the natural frequencies of the
machine structures. Small machines are in general not as
sensitive to higher vibration modes as big machines, but
especially mode number one can be influential even in this
case, because relatively low vibration frequencies can easily
coincide with the natural frequencies of the machine structures
and result in resonance.
The radial magnetic force densities on the rotor core for the
current loading 350 A/cm (RMS) at four different steps in time,
while the rotors are rotating, for the four machines are shown
in Figure 7.
Figure 7. Radial magnetic stress on the rotor core for the current loading 350
A/cm (RMS) at four different steps in time, while the rotor is rotating, for the
machines with (a) 8 poles and double layers, (b) 8 poles and single layer, (c)
10 poles and double layers and (d) 10 poles and single layer.
It is shown that the force density is symmetric for each
quadrant of the machine with 8 poles and double layer
windings. The force density for the machine with 10 poles and
double layers windings is balanced at the opposite side of the
air gap, but has a small mode 2 vibration characteristic. The
machines with single layer windings have more alternating
magnetic stress.
G. Validity of 2D finite element analysis
The major part of the discussion in this chapter is based on
2D finite element analysis. Although such analysis is normally
very reliable, there are some conditions where weaknesses may
be experienced. If the length-to-diameter ratio is low, the
contribution of the end effects plays a larger role, like leakage
inductance, end-winding resistance and axial fringing. The
single layer winding has furthermore longer end-windings than
double layers windings, which is considerable in short motors.
III. EXPERIMENTAL VERIFICATION
A. Motor prototype
The accuracy of the finite element analysis is demonstrated
by measurements on a motor prototype with 10 poles, 12 slots
and double layers windings. The theoretical investigation of
this motor combination showed that it has an extraordinary low

IAS 2005 1047 0-7803-9208-6/05/$20.00 2005 IEEE

ripple torque, which therefore is of greatest interest in the
experimental verification. A CAD-drawing, the electro-
magnetic layout and the test facility of the prototype motor are
shown in Figure 8, Figure 9 and Figure 10, respectively. The
active length of the motor is 72 mm, while the outer diameter is
95 mm. The waveforms of the finite element computed and the
measured open-circuit voltages are shown in Figure 11, which
match very well.
Figure 8. A CAD-drawing of the prototype motor.
Figure 9. The cross-sectional design and the winding layout of the prototype
motor.
Figure 10. The prototype motor connected to a torque transducer in the
laboratory test bench.
IAS 2005
B. Measurement results
The instantaneous electromagnetic torque for a three-phase
electric machine is given by [6]
eA (t ) i A (t ) + eB (t ) iB (t ) + eC (t ) iC (t )
Tem (t ) = , (4)
mech
where eA(t) and iA(t) is the instantaneous values of the back-
EMF and current in phase A, respectively, and mech is the
angular speed of the rotor. In an electromagnetically balanced
servomotor, the most influential ripple torque component is
normally the 6th harmonic, which peak value when fed with a
purely sinusoidal current is given by
T6 =
3
2 mech
( )
E 7 + E5 I1 , (5)
where E7 is the 7th time harmonic of the phase EMF, and I1
is the fundamental time harmonic of the phase current. The
torque ripple factor (in percent) is defined as the ratio of the
peak-to-peak ripple torque to the average torque [6]
T pp 2 T6
TRF = 100% 100% . (6)
T0 T0
The ripple torque is then easily verified by measurements
of the open-circuit voltages and the use of (5) and (6), under
the assumption that the motors are not saturated. The measured
torque ripple factor is 0.8%, while the computed factor, as can
be extracted from Figure 12, is 0.4%. The agreement is
considered to be acceptable, considering that the motor is a
prototype manufactured with laser-cut laminated steel and
manually glued surface-mounted permanent magnets, which
result in poorer tolerances (e.g. rotor eccentricity) than in a
typical series-production, where the laminations are punched.
The ripple torque performance of the prototype motor is
furthermore compared with three state-of-the-art commercial
servomotors, designed for robot applications. A robot is
especially sensitive for ripple torque, since the ripple directly
affects the speed accuracy and the movement of the robot arms.
All the motors in the comparison have a rated torque around 5
Nm. Motors A and B have concentrated double layers windings
with 0.5 slots per pole per phase. Motor C has distributed
windings with 1.5 slots per pole per phase. All motors are
measured in the same test bench by the use of the same
instruments as for the prototype motor. The motors are made
with ripple torque reduction techniques as stepwise skewing of
the rotor, advanced shapes of the permanent magnets and the
rotor core, or sinusoidal magnetization. The results of the ripple
torque comparison is summarized in TABLE I. It is clearly
shown that the prototype motor has the lowest torque ripple
factor of 0.8%, followed by the motors C, B and A with 1.7%,
2.5% and 4.5%, respectively. The results from the
measurement comparison are in line with the theoretical
analysis in the previous chapter, which verifies the
1048 0-7803-9208-6/05/$20.00 2005 IEEE
computations. Regarding the potential alternating magnetic TABLE I. RIPPLE TORQUE COMPARISON BETWEEN THREE COMMERCIAL
SERVOMOTORS AND THE PROTOTYPE MOTOR.
fields in the rotor of the prototype motor, as discussed in the
previous chapter, heat tests have shown that the rotor Servomotor A B C Prototype
temperature is normal. The rotor core was laminated and the No. of poles 6 8 8 10
permanent magnets were divided in 8 pieces per pole (4 axially No. of slots 9 12 36 12
and 2 tangentially) in order to reduce eddy-current losses. The EMF, 5:th harmonic, (%) 1,5 1,2 0,4 0,0
rotor was also stepwise skewed twice an angle of 3 degrees for EMF, 7:th harmonic, (%) 0,8 0,1 0,4 0,4
Torque ripple factor, (%) 4,5 2,5 1,7 0,8
additional ripple torque reduction purposes, but this measure
seems rather superfluous, because of the nature of this pole-slot
combination. The cogging torque, which has not been measured, looks to
be influenced by manufacturing tolerances and end-effects. The
prototype motor experiences more unbalanced radial forces
than traditional motors, according to the theory. The magnetic
noise radiation from this relatively small motor is in practice
negligible. Only a weak almost unnoticeable tone is
experienced at load conditions. Larger motors may experience
higher sound radiations, because of the higher energies that are
involved of which a very small part is transformed into sound
waves.
The discussed modular motor topology can advantageously
be used as small high performance motors, e.g. in servo
applications.

(a)
IV. CONCLUSIONS
A comparative analysis of the traditional brushless and
modular PM machines has been performed. The traditional
machine with double layers windings has lower torque
capability and higher torque ripple, but less influential parasitic
effects like rotor losses and noise radiation than modular
machines. Machines with single layer windings have been
found inferior in high performance applications due to their
lower efficiency and higher inductance compared to the double
layers windings machines. Measurements on a prototype motor
and three commercial servomotors have demonstrated that
modular motors are favorable regarding ripple torque
minimization.
(b)
Figure 11. (a) The calculated and (b) the measured open-circuit voltage
waveforms of the prototype motor with 10 poles and 12 slots. REFERENCES
[1] T. Koch, A. Binder, Permanent magnet machines with fractional slot
5,41
winding for electric traction, Proc. of International Conference of
Electrical Machines, Brugge, Belgium, August, 2002.
[2] K. Atallah, J. Wang, D. Howe, Torque ripple minimisation in modular
permanent magnet brushless machines, IEEE Trans. on Industry
5,40 Applications, Vol. 36, No. 6, November/December, 2003.
Torque [Nm]

[3] J. Wang, Z.P. Xia, D. Howe, Analysis of three-phase surface-mounted

magnet modular permanent magnet machines, Proc. of IEE
International Conference on Power Electronics, Machines and Drives,
Edinburgh, UK, March 31-April 2, 2004.
5,39
[4] F. Magnussen, On Design and Analysis of Synchronous Permanent
Magnet Machines for Field-weakening Operation in Hybrid Electric
Vehicles, Doctoral Thesis, Royal Institute of Technology, ISBN 91-
7283-887-6, Stockholm, Sweden, 2004.
5,38 [5] K.C. Maliti, Modelling and Analysis of Magnetic Noise in Squirrel-
0 6 12 18 24 30 36 42 48 54 60 66 72 Cage Induction Motors, Doctoral Thesis, Royal Institute of
Rotor position [DEG] Technology, ISBN 91-7170-553-8, Stockholm, Sweden, 2000.
[6] S. Huang, M. Aydin, T.A. Lipo, Torque Quality Assessment and Sizing
Figure 12. Finite element computed ripple torque at rated operation for the Optimization for Surface Mounted Permanent Magnet Machines, The
prototype motor. (The rotor is stepwise skewed in 2 steps.) 2001 IEEE Industry Applications Society Annual Meeting, Chicago,
USA, October, 2001.

IAS 2005 1049 0-7803-9208-6/05/$20.00 2005 IEEE