A Comparison between Maximum Torque/Ampere and

Maximum Efficiency Control Strategies in IPM

Synchronous Machines

James Goss, Mircea Popescu, Dave Staton Rafal Wrobel, Jason Yon, Phil Mellor
Motor Design Limited Department of Electrical & Electronic Engineering
Ellesmere, UK University of Bristol, UK
[email protected]

Abstract— This paper presents a comparison between three different machines designed for the same small
maximum torque/ampere and maximum efficiency control electric vehicle propulsion application and a mixed
strategies for interior permanent magnet synchronous theoretical and experimental approach is applied to perform
machines (IPMSMs) in an electrical vehicle propulsion the analysis. The machines analysed are a 36 slot 10 pole
application. A mixed theoretical and experimental approach is
fractional slot per pole distributed winding, a 30 slot 10 pole
adopted to demonstrate how an improvement in performance
may be achieved when maximum efficiency control is utilised. integer slots per pole distributed winding design and a 12
The study is completed on three machines, a 36 slot 10 pole slot 10 pole concentrated winding design.
fractional slot per pole distributed winding, a 30 slot 10 pole
integer slots per pole distributed winding design and a 12 slot II. CONTROL STRATEGIES
10 pole concentrated winding design. The findings are Maximum torque per ampere control can be described by
experimentally validated on the 36 slot 10 pole IPM motor. It is
shown that the largest improved in efficiency is achieved with a
concentrated winding design however for all three designs a minimise , (1)
significant efficiency gain is achieved in the most common
operating regions of the standard international and US urban
drive cycles, WLTP and UDDS. subject to 0 (2)

I. INTRODUCTION
and 2 . (3)
IPMSMs are a preferred topology in the rapidly expanding
area of electric vehicle traction as the main design drivers
for these applications are power density, efficiency and cost. While maximum efficiency control is described by
A significant amount of effort is input into the design and
optimisation of these machines therefore it is particularly minimise , (4)
important that they are controlled in the way that maximises
performance. Maximum torque per ampere (MTPA) is the subject to 0 (5)
typical control strategy used to deduce the optimal operating
point for a permanent magnet synchronous machine [1],[2].
Maximum efficiency (ME) control has also been proposed and 2 . (6)
and applied by some authors [3], [4].
Where Tshaft is the torque produced at the shaft, Tdemand the
The aim of this paper is to apply the MTPA and ME control torque requirement, Id,q the direct and quadrature
strategies to a range of IPMSMs and highlight where components of the peak phase current, Vlim is the maximum
improvements in efficiency can be made by using a ME available peak phase voltage, the electrical frequency and
control strategy which considers iron and magnet losses in ψd,q are the direct and quadrature axis flux linkages. Wcu,Wfe
addition to joule losses in the winding when finding the and Wmagnet are the copper, iron and magnet losses
optimal operating point. The strategies are compared on respectively.

978-1-4799-5776-7/14/$31.00 ©2014 IEEE 2403

ME control differs from MTPA in that when finding the A model of the iron losses is generated, based on the
optimal operating point the iron and magnet loss modified Steinmetz equation,
components are also considered in addition to the joule
losses in the winding. When the current angle is advanced . (10)
from 0 toward 90⁰ in a PM synchronous motor the stator
reaction field opposes the permanent magnet excitation. The Here Wfe is the specifc core loss in W/kg under a sinusoidal
resultant reduction in flux linkage will limit the rate of excitation of frequency f, kh is hysteresis loss coefficient,
increase in iron loss with frequency [5]. The ME control keddy is the eddy current loss coefficient and Bpk is the peak
strategy returns a different operating point to MTPA when flux density in T.
an increase in current advance angle reduces the iron losses
by a greater amount than the corresponding increase in The variation of hysteresis and eddy current iron loss
current and joule losses required to compensate for the loss components can be modelled by mapping the losses against
in torque. This is most relevant to high frequency machines current magnitude, Is and current advance angle, γ, at an
where iron losses are the dominant loss component during arbitrary frequency, using a transient time stepping 2D FEA
high speed operation. solver. These calculated points are shown on the maps in
Figs.1-2. The hysteresis and eddy components of the iron
III. MODELLNG APPROACH loss, at each operating point, are then calculated using
A. Electromagnetic and loss modelling interpolation and scaled by f and f2 respectively,
To analyse the control strategies correlations for the
variation of torque, voltage and each loss component with . , , , (11)
current, advance angle and frequency are required. Using
models for the flux linkages and each component of loss the where n is the operating speed and Chys, Ceddy functions that
sets of equations detailed in (1-6) can be solved to find the perform the cubic spline interpolation shown by the surfaces
optimal operating point using a non-linear constrained displayed in Figs. 1-2.
optimisation algorithm. The flux linkage models, used to
calculate torque and voltage, are generated by evaluating ψq
and ψd using 2D FEA at various values of Id and Iq. Using a
least squares fitting method the FEA data is fitted to
Hystersis Iron Loss/Frequency (W/Hz)

polynomial functions [6], which describe the relationship 0.5

between the d-q axis flux components and the d-q axis 0.4

current components, 0.3

0.2
Ψ , 0.1
250
200

0 150

, (7) 80
60
100

40 50
20 0 Is (A pk)
0
Ψ , γ (°)

Figure 1. Hysteresis iron loss map including interpolation points

. (8)

In a brushless permanent magnet machine the main loss

components are joule losses in the winding including AC -4
effects, iron losses in the stator and rotor core pack, eddy x 10
Eddy Iron Loss/Frequency2 (W/Hz 2)

current losses in the magnets and frictional losses in the 6

bearings. To model the winding copper loss the DC

component is calculated from the measured or predicted DC 4

resistance value while AC loss effects are approximated as a

2
linear increase in the effective winding resistance with 250

speed,
200
0 150
. (9) 80
60
100
40 50
20
0 0 Is (Apk)
γ (°)
Here Rdc is the DC winding resistance per phase, m the
number of phases, n the rotational speed and n2ac the speed Figure 2. Eddy current loss map including interpolation points
at which the ac winding loss is equal to the dc winding loss.

2404
The magnet losses are again mapped against Is and γ at a IV. MOTOR DESIGNS
particular speed and interpolation used to calculate these at The analysis is performed on three machine topologies; a 36
each load point. The magnet losses are then scaled by nβ slot 10 pole (36s10p) fractional slot/pole distributed
where the co-efficient β is deduced from a number of winding, a 12 slot 10 pole (12s10p) concentrated winding
additional transient FEA calculations at various frequencies. and a 30 slot 10 pole (30s10p) integer slot/pole distributed
winding. Each machine is designed and optimised for the
The mechanical bearing losses are accounted for with same specification while the rotor design, active length,
outer diameter and flux density in the core have been kept
| , (12) consistent between all three variants. Only the number of
slots and winding layout is changed between the variants to
where n0 is the speed at which the bearing losses |
ensure a fair comparison
were calculated or measured at.
The 36 slot machine, shown in Fig. 6, has been
characterised while the other machine designs are paper
B. Efficiency maps and drive cycles studies only. A combined experimental and theoretical
Using the electromagnetic and loss models (1)-(6) can be approach is used to compare the MTPA and ME control
solved to produce efficiency maps for either MTPA or ME strategies on the 36 slot machine and the calibration factors
control. These are then compared to consider the resultant from the experimental measurements are carried over to the
improved in efficiency using ME control and to understand other machine designs. The motors are designed for a
the regions in the operational envelope in which the control continuous rating of 12kW and peak of 20kW with a
strategies differ. These models can also be used to calculate maximum speed of 10,000rpm. The specification for the
the improvement in efficiency, and the difference in energy machine designs is given in Table I, this is representative of
consumption, over typical driving cycles. Using a set of a small two person urban electric vehicle.
equivalent vehicle parameters the aerodynamic force at any
vehicle speed can be modelled by TABLE I. MACHINE SPECIFICATION

Maximum speed 10,000 rpm

, (13) Continuous Torque 30 Nm
Continuous Power 12 kW
while the rolling resistance of the vehicle is modelled using Peak Torque Requirement 60 Nm
Peak Power Requirement 20 kW
(14) DC Link Voltage 225 Vdc
Maximum Inverter Current 250 Apk
and the acceleration force is modelled using Cooling Type TENV

. (15) The vehicle model parameters, used to derive the motor

operating points from (13)-(17) are given in Table. II.
Here ρ is the air density, g acceleration due to gravity, v the
velocity of the vehicle in m/s, a the vehicle acceleration in TABLE II. VEHICLE MODEL PARAMETERS
m/s2, m the vehicle mass, Cd the drag co-efficient, kr the Mass 900kg
rolling resistance co-efficient, Af the vehicle frontal area, Drive ratio nd 10.13
and δ the mass correction factor. Using the forces predicted Rolling resistance co-efficient kr 0.008
in (13)-(15) the required machine torque, Tdemand is given Drag co-efficient Cd 0.34
from Mass correction factor δ 1.03
Effective frontal area (m2) Af 2.03
(16) Wheel radius (m) rω 0.254

and the machine speed from The lamination cross sections for the three machines are
shown in Figs. 3-5, some common machine design
. (17) parameters are shown in Table. III and a comparison of the
electromagnetic parameters of the three machines is shown
in Table IV. The three machines have relatively similar
Here is rω vehicle wheel radius and nd the drive ratio. The electromagnetic parameters with the 12s10p concentrated
machine torque and speed cycle can then be derived from winding design showing the lowest short circuit current,
(13)-(17) and standard vehicle speed Vs. time test cycles. saliency ratio and permanent magnet flux linkage.

2405
 TABLE III. COMMON DESIGN PARAMETERS

Stator Outer Diameter 175mm

Stack Length 90mm
Magnet Grade N38UH
Electrical Steel Grade M270-35A
Open Circuit Stator Tooth Flux Density 1.2T
Open Circuit Stator Yoke Flux Density 1.2T
Slot Fill Factor 0.45
Rated current density 4A/mm2

TABLE IV. ELECTROMAGNETIC PARAMETERS

36 Slot 30 Slot 12 Slot

PM flux linkage ,λm
42.5 38.1 32.4
(mWb)
Ld (mH) 0.33 0.26 0.29
Lq (mH) 0.52 0.39 0.42
Figure 3. Lamination cross section of the 36 slot 10 pole design Lq/Ld 1.6 1.5 1.46
Isc (Apk) 129 140 117
Irated/Isc (p.u) 0.73 0.84 1.0
Winding layout Distributed Distributed Concentrated

(a) (b)

Figure 6. Assembled rotor (a) and 36 slot stator (b).

V. EXPERIMENTAL SETUP
The 36 slot 10 pole machine has been characterised using
Figure 4. Lamination cross section of the 30 slot 10 pole design the test arrangement shown in Fig 7. The load machine is
coupled to the machine under test (MUT) and operated in
speed control mode. The MUT can be reconfigured to run at
open circuit and short circuit conditions or on load motoring
operation when connected to a drive and operated in torque
control mode. The machine is tested up-to a maximum
speed of 7000rpm and a maximum torque of 50Nm. The test
envelope was constrained by the inverter and torque
transducer available in the test laboratory. Efficiency and
loss measurements are computed from the difference of the
electrical input power at the MUT terminals to mechanical
output power at the MUT shaft. Electrical input power is
derived using current and voltage measurements at the
machine terminals fed into a power analyser. Mechanical
output power measurement is taken from the product of the
measured torque and rotational speed from the torque
transducer.

Figure 5. Lamination cross section of the 12 slot 10 pole design

2406
 Current 50
transducers Is = 140A rms
45
Is = 80Arms
40 Is = 20Arms
Measurements
35
FEA model
30

Torque(Nm)
25

MUT Load 20
Torque machine
transducer 15

10

Figure 7. Test rig set-up 0

0 10 20 30 40 50 60 70 80 90
Current advance angle ( ° electrical)
VI. MODEL CALIBRATION Figure 9. Measured and modelled torque vs. current angle
This analysis is performed using a mixed experimental and
theoretical approach. The model calibration and loss B. Loss model calibration
measurement and separation process is described here. All The DC resistance of the windings, Rdc, is derived from a
measurements here are completed transiently with little rise DC test where a current is applied to the 3 phases connected
in temperature from ambient and hence are compared to the in series and the voltage drop measured at a known winding
model prediction at ambient temperature, 25⁰C. temperature. This model is then calibrated to match this DC
resistance value. The bearing losses were measured by
A. Electromagnetic model calibration spinning the rotor in an empty housing without any stator.
The open circuit voltage measurements, measured at each The open circuit losses, measured at the torque transducer,
250rpm speed step up-to 7000rpm, are used to calibrate the minus the measured bearing losses are then used to calibrate
value of the magnet remenant flux in the 2D FEA model to the FEA iron loss model. A build factor of 2 was applied to
account for tolerances on the magnet and end leakage the eddy current component of the iron loss to account for
effects; in this instance no modification from the the degradation of the insulation coating on the electrical
manufacturer data was required. The measured and steel during the manufacturing process. No adjustment to
predicted back-emf waveform is shown in Fig. 8. Next the the hysteresis loss component prediction from the FEA
short circuit current measurements are used to calibrate the model was required. The measured total and modelled loss
value of the end winding inductance in the model. The split at open circuit shown in Fig. 10.
torque/amp vs. current angle is then checked against the
model. Small changes were made to the material B-H curves 700
Total measured OC loss
to ensure saturation effects were accurately accounted for. Measured bearing loss
600
A comparison between the FEA model and measurements is Total modelled OC loss

shown in Fig. 9 where the torque vs. current angle is plotted 500
at low, rated and high levels of current
400
Loss (W)

5
Measurements
4 FEA model 300
3

200
2
Phase voltage (V)

1
100
0

-1 0
0 1000 2000 3000 4000 5000 6000 7000
Speed (rpm)
-2

-3
Figure 10. Measured and modelled open circuit loss

-4
The losses measured from the short circuit test at various
-5
0 50 100 150 200 250 300 350
speed are used to derive the AC losses component and the
Rotor position ( ° electrical) value of n2ac. The total short circuit loss is measured from
Figure 8. Measured and modelled back-emf waveform at 200rpm the torque transducer across the speed range and the known
friction, DC copper, magnet and iron loss (from the

2407
calibrated model) are subtracted from the total. The iron and
50 96
magnet losses are calculated using the calibrated FEA
model at the measured short circuit current and each speed 45 9
95 39949 94
12
step. This remaining loss is attributed to AC copper loss

92

94
91
40

93
effects, which is shown in Fig 11. It can be seen that the AC 92

losses can be approximated to scale linearly with speed and 35

93 90
have a value of Rac/Rdc of 1.46 at 7000rpm which 99491

Efficiency (%)
95

Torque (Nm)
2
30

95
corresponds to n2ac equal to 15,175rpm. 88
25

92

94
95 86

91
93
20
200
94 94 84
AC loss model 15
180
W sc - (W fe +W mag+W cuDC+W friction) 93 93
160 10 92 92 82
92 91 91
91
140 5 80
1000 2000 3000 4000 5000 6000 7000
120 Speed (rpm)
Loss (W)

100
Figure 13. Modelled efficiency map

80 Very good correlation is achieved between the maps with a

60 small amount of distortion of the measured map at low
40
speed and torque. This distortion is due to a small amount of
measurement error on the input power and output torque
20
measurements at low values of power.
0
0 1000 2000 3000 4000
Speed (rpm)
5000 6000 7000
VII. RESULTS
Figure 11. Measured and modelled AC loss A. Efficiency maps
Efficiency maps are calculated for MTPA and ME control
C. Model validation
for each design, subtracting the MPTA efficiency map from
The calibrated electromagnetic and loss model is then the ME map shows the difference. The improvement in
finally validated by comparison of the measured and efficiency when ME control is used for the 36 slot 10 pole,
predicted efficiency maps, shown in Figs. 12 and 13 30 slot 10 pole and 12 slot 10 pole machines is shown in
respectively. The measured map is made up of 140 points Figs. 14, 15 and 16 respectively.
obtained using an automated test rig set-up, these points are
organised to limit the temperature rise in the machine and 50 5

all points are measured in under 2 minutes, resulting in a 45 4.5

maximum 20⁰C rise in the winding thermocouples over the 40 4
test procedure. This is compared to the modelled efficiency

Efficiency improvement (%)

map calculated at the ambient temperature. 35 3.5

30 3
Torque(Nm)

50 96
91 93 9294 25 2.5
91
45 93 94
20 2
992
954
40
92 15 1.5
95
92

35 91 10 1
90
91

93
Efficiency (%)
93
Torque (Nm)

992
94

30 4 5 0.5
88
95

25 0 0
0 2000 4000 6000 8000 10000
95 86 Speed (RPM)
95
20
Figure 14. Efficiency improvement for the 36s10p design
92

94
84
15 94 94
93

93 93
10 92 92 82
92 91 91
91
5 80
1000 2000 3000 4000 5000 6000 7000
Speed (rpm)

Figure 12. Measured efficiency map

2408
 50 5 50

45 4.5 40

40 4 30

Efficiency improvement (%)

35 3.5 20

30 3 10

Torque (Nm)
Torque(Nm)

25 2.5 0

20 2 -10

15 1.5 -20

10 -30
1
-40
5 0.5
-50
0 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
0 2000 4000 6000 8000 10000
Speed (rpm)
Speed (RPM)

Figure 15. Efficiency improvement for the 30s10p design Figure 18. UDDS driving cycle points

The computed average efficiency and total losses across

5 both cycles for MTPA and ME control is given in Tables.
45 4.5 V-VII for the 36s10p, 30s10p and 12s10p designs
40 4
respectively. 1
Efficiency improvement (%)

35 3.5 TABLE V. 36S10P EFFICIENCY OVER UDDS AND WLTP CYCLES

30 3
WLTP UDDS
Torque(Nm)

25 2.5 MTPA ME MTPA ME

20 2 Average Efficiency (%) 80.39 81.1 84.96 85.7
Total loss (Wh) 67.54 62.2 123.71 117.4
15 1.5
Average efficiency gain (%) 0.71 0.74
10 1

5 0.5 TABLE VI. 30S10P EFFICIENCY OVER UDDS AND WLTP CYCLES
0
0 2000 4000 6000 8000 10000
0 WLTP UDDS
Speed (RPM) MTPA ME MTPA ME
Figure 16. Efficiency improvement for the 12s10p design Average Efficiency (%) 80.13 81.59 85.02 85.94
Total loss (Wh) 69.27 71.42 121.39 112.33
B. Drive cycles Average efficiency gain (%) 1.46 0.92
The machine variants are also analysed across the standard
WLTP class 1 drive cycle and the UDDS drive cycles1. The TABLE VII. 12S10P EFFICIENCY OVER UDDS AND WLTP CYCLES
calculated torque, speed operating points for the WLTP and WLTP UDDS
UDDS cycle is given in Figs. 17 and 18 respectively. MTPA ME MTPA ME
50 Average Efficiency (%) 81.62 83.73 86.4 87.68
Total loss (Wh) 61.2 50.86 105.85 93.93
40
Average efficiency gain (%) 2.11 1.28
30

20 C. Experimental validation
10 To experimentally validate the findings the efficiency is
Torque (Nm)

measured at 2.5Nm and 5Nm and 4000rpm for the 36 slot

0
machine using MTPA and ME control. The results of this
-10
measurement along with model predictions are given in
-20 Table. VIII.
-30
1
-40 WLTP Class 1 is the world harmonized light vehicles test
procedure for low power vehicles with a rated power/weight
-50
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 ratio of <=22 kW/Tonne. UDDS is the US environmental
Speed (rpm)
protection agency (EPA) urban dynamometer driving
Figure 17. WLTP class 1 driving cycle points
schedule.

2409
 TABLE VIII. EXPERIMENTAL VALIDATION operating points 2.5Nm and 5Nm. These measurements were
Operating Point 1 Operating Point 2 repeated multiple times and the measurement gave consistent
MTPA ME MTPA ME values with little variation. An improvement in efficiency is
Torque (Nm) 2.5 5 shown at both points with a measured gain in efficiency that
Speed (rpm) 4000 4000 is slightly lower than the model predictions.
Current (Arms) 7.2 20.9 13.0 26.5
VIII. CONCLUSIONS
Current angle (⁰ elec) 3.9 71 5.9 64.5
Measured Efficiency (%) 78.6 80.3 87.5 88.5
This paper has shown that a ME control strategy can result
Measured efficiency gain in significant improvements in efficiency for high speed
1.74 0.98 IPM motors in traction applications. It has been found that
(%)
Modelled Efficiency (%) 78.3 80.3 87.04 88.4 the main difference between ME and MTPA control
Modelled efficiency gain
2.0 1.36
strategies occurs at low torque and just before the field
(%) weakening region which incidentally is a very common
Modelled copper loss (W) 2.3 25.8 9.3 41.3 operating region for traction machines over standard
Modelled iron loss (W) 206 149.1 220.2 152.2 automotive drive cycles. Three machine types have been
D. Discussion studied and it has been concluded that while all show some
difference between the strategies the effectiveness of
The modelled efficiency maps show that all three machines adopting ME over MTPA control is a function of the ability
offer a potentially significant gain when maximum of the machine to suppress iron losses from open circuit to
efficiency control is adopted. The largest improvements short circuit conditions. The benefit of maximum efficiency
occur at low torque and medium to high speeds just before control was greatest for the 12s10p concentrated winding
the field weakening region. The regions of highest design. The temperature dependency of the ME control
efficiency improvement tend to correspond with the most strategy as well as the practical implementation in a drive
frequent areas of operation on both the WLTP and UDDS has not been considered in this paper and should be included
cycles. The 12s10p concentrated winding machine as part of any future work.
consistently shows the largest difference between MTPA
and ME control. The 30s10p shows the next largest IX. REFERENCES
difference and the 36s10p the least. The differences are due
to 2 main factors, the proportion of iron loss to copper loss [1] A. Consoli, G. Scarcella, G. Scelba, and A. Testa, “Steady-State
in the machine and the amount the iron losses are and Transient Operation of IPMSMs Under Maximum-Torque-
suppressed between a 0 and 90⁰ advance angle. Table IX per-Ampere Control,” IEEE Transactions on Industry
Applications, vol. 46, no. 1, pp. 121–129, 2010.
shows the iron losses for all three machines at a speed of
7000rpm at both open circuit and short circuit conditions as
[2] P. . Niazi, H. A. Toliyat, and A. . Goodarzi, “Robust Maximum
well as the copper loss at rated current density. It can be Torque per Ampere (MTPA) Control of PM-Assisted SynRM for
seen that the iron losses are more effectively suppressed Traction Applications,” IEEE Transactions on Vehicular
from open circuit to short circuit conditions by the 30s10p Technology, vol. 56, no. 4, pp. 1538–1545, Jul. 2007.
and 12s10p designs. The DC copper loss at rated current is
also shown in Table. IX, differences here are mostly due to [3] C. Mademlis, I. Kioskeridis, and N. Margaris, “Optimal
differences in the end winding length, where the 30s10p has Efficiency Control Strategy for Interior Permanent-Magnet
Synchronous Motor Drives,” IEEE Transactions on Energy
the longest end windings and the 12s10p the shortest. Conversion, vol. 19, no. 4, pp. 715–723, Dec. 2004.
Therefore the effectiveness of the ME control strategy over
the MTPA for a particular machine type is dependent on [4] D. Pohlenz and J. Bocker, “Efficiency improvement of an
how much the iron loss is supressed from open circuit to IPMSM using Maximum Efficiency operating strategy,” in
short circuit operation as well as the ratio of copper loss to Proceedings of 14th International Power Electronics and Motion
iron loss in the machine. Control Conference EPE-PEMC 2010, 2010, pp. T5–15–T5–19.

TABLE IX. IRON AND COPPER LOSS COMPARISON [5] P. H. Mellor, R. Wrobel, and D. Holliday, “A computationally
efficient iron loss model for brushless AC machines that caters
36s10p 30s10p 12s10p for rated flux and field weakened operation,” in 2009 IEEE
Open circuit iron loss at 7000rpm (W) 449 447 408.3 International Electric Machines and Drives Conference, 2009,
Short circuit iron loss at 7000rpm (W) 219 172 145.5 pp. 490–494.
Ratio between OC and SC loss 0.49 0.38 0.36
DC Copper loss at rated current, [6] J. Goss, P. H. Mellor, R. Wrobel, D. A. Staton, and M. Popescu,
207.3 240 175 “The design of AC permanent magnet motors for electric
4A/mm2 (W)
vehicles: a computationally efficient model of the operational
envelope,” in 6th IET International Conference on Power
Electronics, Machines and Drives (PEMD 2012), 2012, pp. B21–
Table VIII. shows the experimentally measured efficiencies B21.
using MTPA and ME control at 4000rpm and two low torque

2410