You are on page 1of 10

SRMs Design Guide

This SRM design guide is a work in progress and not complete at the
present time. It is hoped that even in its present form it will be usefull to those
who are in the process of designing a SRM. See Predes.mcd for additional useful
information.
In designing electrical machines the machine!s area product defined as
the total slot area of times the area of the air gap is a useful si"ing metric. This
metric is denoted the area product of the machine and is found b# computing the
power converted to mechanical form
PF I BEMF
N nrep
P php p
ph
m

=
2
$%&
'ere P
m
is the mechanical power ()M*
p
is the peak machine back emf voltage
I
php
is the peak phase current P* is the machines power factor +
ph
is the number
of phases and nrep is the number of machine repetitions. The number of
machine repetitions is defined such that a three phase two,repetition machine is
the same as a si- phase machine. The machine!s power factor is assumed
known and is an e-perience factor defined b# ).. %. *or the nonsinusoidal SRM
the power factor is not e.ual to the cosine of the angle between the current and
voltage. /t the machines corner point speed $highest speed for constant tor.ue
operation& the peak ()M* is chosen to be e.ual to what will be defined as the
link voltage 0
link
. This is the D1 input voltage 0
dcin
minus the voltage drop across
the winding resistance and the voltage drop across the inverter. It is often the
case that dcin link V V though not alwa#s. In the case of the SRM and assuming
independent phases
mcorner
p
unaligned aligned
mcorner p link
d
d
dt
d
BEMF V


= = = $2&
'ere
p
is e.ual to the machines angular pole width measured at the air gap.
/ssuming that the unaligned flu- is a fi-ed fraction of the aligned flu- gives
( )
mcorner
p
BEMF aligned
mcorner
p
u aligned
mcorner
p
aligned u aligned
link V

1
$3&
'ere
u
is the fraction the unaligned flu- is of the aligned flu- and is a small
number $ 2 . 0 u &. Thus
()M*
is a number near one $ 8 . 0 BEMF &. ).uation 3
can be written in terms of field variables and the geometr# as
mcorner
p
BEMF p p
mcorner
p
BEMF pole p
link
stf lstk Rg B nser N A B nser N
V

max max
mcorner BEMF p link stf lstk Rg B nser N V max $4&
In ).. 4 (
ma-
is the ma-imum magnetic flu- densit# stf is the stacking factor +p
is the number of turns per pole and nser is the number of pole windings within a
phase that are in series. T#picall# the number of pole windings in series within a
phase is two. +ote that ).. 4 does not depend on the pole width and thus on the
number of machine phase. Thus if the number of phases is doubled with all of
the windings being put in series the number of turns per pole will have to be
halved. / specific case of the machine geometr# is shown in *ig. % where the
different geometric variables are defined. In particular the area of a single
Rg
Rr#
Rs#
p
5
Rs#
5-
5-
hw
pw
ph
/w
/ww
/wph62 /w
*ig. % Definition of geometr# terms
pole is computed as the pole width $pw& times the stack length $lstk&
p g
p
g R R pw

=
2
sin 2
with Rg e.ual to the radius to the air gap at the rotor and
p
is the pole!s angular
e-tent. If possible the machine is designed so that the iron saturates and (
ma-
is
e.ual to the iron!s saturation flu- densit# (
sat
. +ote that the machine!s ()M*
does not depend on the pole arc
p
. Plugging ).. 4 into ).. % for the peak ()M*
gives
PF I nser N stf lstk Rg B
N nrep
P php p mcorner BEMF
ph
m

= max
2
$7&
The peak phase current is related to the rms phase current b#
php rms phrms I I =
$8&
'ere
rms
is 9.:9: for sinusoidal machines. It is an e-perience factor defined b#
).. 8 for the nonsinusoidal SRM. The rms phase current can be written in terms
of the current densit# as
wph w rms
phrms
p A K J
npar
I
npar nser N =


$:&
'ere npar is the number of pole windings in a phase that are in parallel /
wph
is
the total area available to a phase winding and ;
w
is the fraction of this area
occupied b# copper. Thus /
wph
6 nser npar /
w
where /
w
$see *ig. %& is the area
available to one pole winding. +ote that /
w
and /
wph
onl# account for the area of
the wire on one side of a pole and thus one half of the slot area occupied b# the
winding. <sing ).. : ).. 7 becomes
PF
A K J
stf lstk Rg B
N nrep
P
rms
wph w rms
mcorner BEMF
ph
m

= max
2
$=&
).uation = can be solved for the geometr# in terms of the field .uantities and the
machines power rating.
mcorner w rms
m
BEMF ph
rms
wph
K J B
P
PF stf N nrep
A lstk Rg


=
max
2
$>&
+ow the area of the air gap is simpl# 2 Rg lstk. The total winding area or
e.uivalentl# the total slot area for all phases and repetitions of the machine is
wph ph w ph w ph wtotal A N nrep A N nrep A N nrep npar nser A = = = 2 4 2
$%9&
The 2 accounts for the fact that a winding occupies area on both sides of a pole.
+ote that total winding area 2 nrep +
ph
/
wph
does not actuall# depend on the
number of SRM repetitions or the number of phases. If the number of repetitions
or the number of phases changes /
wph
the area for one winding changes. The
greater the number of repetitions and phases the smaller the area of one
winding. The machine!s area product is defined as
( ) ( ) = = area slot total area gap air Ap
( ) ( ) = w ph A N lstk Rg 4 2
mcorner rms
m
BEMF Kw J B
P
PF stf
rms


max
8
$%%&
The area product is related to the si"e of the SRM being e.ual to the product of
the area the magnetic field passes through and the area the current flows
through. It has units of m
4
. /s e-pected the SRM!s si"e will be greater for higher
power machines. /lso increasing the ma-imum flu- densit# the current densit#
and the machine!s speed reduce its si"e. ).uation %% can be used to trade off
the iron area against the copper area or e.uivalentl# the radius of the machine
can be traded off against its stack length. +ote that the number of machine
repetitions and the number of phases do not directl# enter into the computation of
the SRM!s area product. The# do enter indirectl# as the# effect P*
()M*
and

rms
. The fact that the area product does not depend on the number of machine
repetitions and the number of phases does not mean the SRM!s si"e and weight
does not depend on the number of machine repetitions and the number of
phases. This is because increasing the number of machine repetitions and the
number of phases decreases the re.uired rotor and stator #oke thickness.
).uations % 8 and : can be combined to obtain
m m
rms p
wph w rms
p
ph
m PF
nser N
A K J
BEMF
N nrep
P

=
2
$%2&
which can be used to compute the tor.ue. 1ombining ). %2 with ).. %9 gives
PF
nser N
A K J
BEMF
rms p
wtotal w rms
p
m
m


=
2 2
1

$%3&
for the tor.ue. The peak machine back )M* is e.ual to the link voltage at the
corner fre.uenc# so
corner
m
link p V BEMF

=
$%4&
1ombining ).s. 4 %3 and %4 gives
PF A K J stf lstk Rg B wtotal w rms
rms
BEMF
m = max
4
1

$%7&
+ote that the tor.ue does not depend on the number of phases or the number of
repetitions. /lso the rotor?s tor.ue densit# is given b#
Rg
PF A K J B stf
lstk Rg

wtotal w rms
rms
BEMF m
densit!

= =
max
2
4
1


$%8&
If the window area is held constant the tor.ue densit# is largest for small
diameter machines. In general though the slot depth is proportional to Rg and
the slot width is proportional to Rg. Thus the window area is proportional to Rg
2
.
In this case the tor.ue densit# is proportional to Rg. @arger diameter machines
have a larger to.ue densit#. The tor.ue densit# can be rewritten using the area
product as shown in ).. %:. *or a fi-ed tor.ue T
m
and assuming the total window
area is proportional to Rg
2
the tor.ue densit# is largest when the rotor radius is
large and when the area product is small.
Ap Rg
A

m wtotal
densit!

=
4
$%:&
).uation %% for the machine?s area product onl# provides a constraint on
the product of its rotor radius and its stack length not on their individual values.
Thus to determine the SRM!s radius and stack length an additional constraint on
the machine!s geometr# be#ond ).. %% is re.uired. ).uation %: sa#s that to
ma-imi"e the rotor?s tor.ue densit# the rotor radius should be larger than the
stack length. *or high,speed machines the rotor is made as large as possible
subAect to the constraint imposed b# centrifugal stresses in the rotor. *or small
low speed SRMs where centrifugal stresses are not important and ma-imum
rotor tor.ue densit# is not the goal the losses in the copper can be used as a
constraint on the machine!s geometr#. The losses in the copper are
ph mean wph cu cu N nrep l A Kw
cu
Jrms
Vol
cu
Jrms
P = =

2 2
$%=&
'ere 0ol
cu
is the copper volume and l
mean
is the mean length of a turn. The mean
length of a turn is e.ual to twice the stack length plus the end,turn length and
given b#
( )

+ + = w p mean Rg lstk l

2
2
$%>&
where
p
and
w
are defined in *ig. %.
( )

+ + = w p ph wph
cu
rms
cu Rg lstk N nrep A Kw
J
P

2
2
2
$29&
Solving ).. %% for P
m
the mechanical power converted gives
wph ph mcorner rms
BEMF
m A N nrep lstk Rg Kw J B
rms
PF stf
P

max
2
$2%&
The ratio of the power dissipated in the copper to the power converted to
mechanical form is
( )
lstk Rg B PF stf
Rg lstk J
P
P
mcorner cu BEMF
w p rms rms
m
cu

+ +
= =

max
2
4
$22&
/s the ratio of the power dissipated in the copper to the power converted to
mechanical form goes to "ero the efficienc# goes to %99B. In fact the power ratio
in ).. 22 is related to the efficienc# b#
100
1
1

+
=

Eff 1
100
=
Eff

$23&
+ote from ).. 22 the SRM efficienc# $including onl# copper losses& does not
depend on the number of machine phases or repetitions to first order. The end
turn term which involves
p
and
w
does depend on the number of phases and
repetitions since these angles do. Since
p
and
w
decrease as the number of
phases and repetitions increase the machines end turn losses will be reduced in
such machines.
In the special case where the SRM saturates (
ma-
is e.ual to (
sat
so that
).. 22 becomes
( )
lstk Rg B PF stf
Rg lstk J
P
P
mcorner sat cu BEMF
w p rms rms
m
cu

+ +
= =

2
4
$24&
If the end turn losses can be neglected the stack length cancels and ).. 24 puts
strict limits on the SRM!s air gap radius and current densit# to achieve a specified
efficienc#. In this case Rg and lstk are chosen to minimi"e in ).. 24 subAect to
the constraint that the SRM!s area product given in ).. %% is fi-ed. This is the
same as choosing Rg and lstk to minimi"e the copper losses subAect to the area
product constraint. Solving the area product e.uation for the stack length and
then eliminating the stack length in the copper loss e.uation to obtain the copper
losses in terms of the single variable Rg can accomplish this minimi"ation. This
approach can be generali"ed to include minimi"ing a weighted sum of the copper
and iron losses.
The machine design procedure when the iron saturates is to make the
current densit# as large as possible without e-ceeding the efficienc# re.uirement.
/t this point the rotor radius and stack length are known from a solution of the
proposed minimi"ation problem. Cnce this calculation is completed the
machine?s geometr# is known. ).uation 4 can now be solved for +
p
the number
of turns per pole
mcorner BEMF sat
dcin
mcorner BEMF sat
link
p
stf lstk Rg B nser
V Eff
stf lstk Rg B nser
V
N


=
$27&
'ere the specified efficienc# times the D1 input voltage is used as an estimate of
the conservative electromechanical s#stem!s terminal voltage 0
link
. This is
reasonable for motoring though not for generating where the conservative
electromechanical s#stem!s terminal voltage is larger that the actual output
voltage. This design procedure is ade.uate if the various constants
()M*

rms

stf and P* are known and a method for determining the slot depth is known.
It is generall# considered a good practice to minimi"e the machines
airgap. *igure 2 shows the aligned and unaligned flu- linkage curves for two
identical machines accept that one has a large airgap and the other a small
airgap. The# have both been designed so that at the ma-imum current the# have
the same ()M* at the same speed. This is insured because the change in the
linked flu- from the aligned to the unaligned positions is the same.
u a
u sat
m BEMF

=
$28&
To appreciate that the airgap should be minimi"ed independent of the unaligned
inductance the aligned and unaligned flu- linkage curves in *ig. 2 have been
redraw in *ig. 3 for the case where the unaligned inductance is "ero. /gain the
two machines are identical accept for their airgaps and the# both are designed to
have the same ()M* at the ma-imum current and the same speed. The energ#
conversion per c#cles is e.ual to the area of the machines operating traAector# in
the flu- linkage current plane. *or a machine with no saturation the ma-imum
possible energ# conversion is
sat sat no I E max _
2
1
=
$2:&
*or a machine with no airgap $g 6 9& the ma-imum possible energ# conversion is

Unaligned
Aligned Small Air Gap (Saturated)
Aligned Large Air Gap (Unsaturated)
I
Operating
Trajectory 1
Operating
Trajectory 2
(I,)
sat u
sat
Ima!
*ig. 2 *lu- linkage curves for one machine with a large airgap and another with
a small airgap. (oth machines are identical accept for the airgap.

Aligned Small Air Gap (Saturated)
Aligned Large Air Gap (Unsaturated)
I
Operating
Trajectory 1
Operating
Trajectory 2
(I,)
sat u
sat
Ima! I
sat
*ig. 3 *lu- linkage curves for one machine with a large airgap and another with
a small airgap when the unaligned inductance is "ero. (oth machines are
identical accept for the airgap.
sat sat I E max = $2=&
Thus starting with a given non,saturating SRM design the total energ# converted
per c#cle and thus the ma-imum output power can be doubled Aust b# reducing
the SRM?s airgap to "ero. Inspection of *ig. 3 shows that reducing the SRM?s
airgap alwa#s increases its ma-imum energ# conversion potential even with "ro
unaligned inductance. Inspection of fig. 2 shows that a similar result holds for
finite unaligned inductance so that it is a good idea to minimi"e an SRM?s airgap.
To estimate when decreasing the airgap provides the greatest benefit
compute the ma-imum energ# conversion potential for the general case.

=
2
sat
max
I
I E sat con"
$2>&
(ut
sat
pole p o
sat I
npar
nser
g
A N
2

=
$39&
)liminating Isat from ). 2> using ).. 39 gives

= g
nser
npar
A N
I E
pole p o
sat
sat con"
2
max
2

$3%&
Reali"ing that
sat
does not depend on the airgap g the change in the potential
energ# conversion due to a change in the airgap can be found
dg
nser
npar
A N
dE
pole p o
sat
con"
2
2
2

=
$32&
)-pressing ).. 32 as in percent changes gives

=
g
dg
g
nser
npar
A N
I
nser
npar
A N
g
E
dE
pole p o
sat
sat
pole p o
sat
con"
con"
2
2
max
2
2
2
2

$33&
/ssuming that g will alwa#s be small so the second term in the denominator of
)..33 can be neglected compared to the first term allows ).. 33 to be simplified
to

=
g
dg
A K J
g
B
E
dE
wph w
o
sat
con"
con"
max 2

$34&
rewriting
sat
and I
ma-
in terms of field .uantities. +ow the area of the window
beside the pole can be simplified using the drawing in *ig. 4. *rom the figure the
area of the phase window /
wph
is e.ual to the window height times twice the
spacing between the rotor and stator poles in the unaligned position lr
%
6 lr
2
. 5ith
these definitions and assuming rms J J 2 max gives

=
g
dg
hs K J
B
lr
g
E
dE
w
o
sat
con"
con"
rms 2 2 4

$37&
).uation 37 defines a gain the gives the percentage increase in the potential
energ# converted resulting from a percentage decrease in the airgap.
hs K J
B
lr
g
#ain
w
o
sat
g
rms 2 2 4

=
$38&
+ote that this gain depends on the material properties the ratio of the airgap to
the spacing between the rotor and stator poles in the unaligned position and the
slot depth. +ote that the gain depends on the ratio of the airgap to the spacing
between the rotor and stator poles in the unaligned position even if the unaligned
inductance is made aero b# some means. The gain decreases as g decreases
with ever# thing else held constant so after awhile it does not pa# to decrease g
since the increase in the energ# converted per c#cle does not go up enough to

"otor Iron
Stator Iron
#inding
lr
Lr
1
Lr
2

p$
%r
& '
%s
2 Lr
2

*ig. 4 Diagram of a single stator pole defining dimensions.
Austif# the manufacturing effort incurred. The slot depth hs is a measure of the
si"e of the machine bigger machines will have bigger slot depths. Thus the gain
will be larger for smaller machines and thus it pa#s to make gDlr
2
smaller in
smaller machines. Thus the airgap will tend to be larger than desired in small
machines.

You might also like