IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

Brushless DC Motor Characterisation and Selection

for a Fixed Wing UAV
Darren Lance Gabriel Johan Meyer Francois du Plessis
Department of Electrical and Department of Electrical and Department of Electrical and
Electronic Engineering Electronic Engineering Electronic Engineering
University of Johannesburg University of Johannesburg University of Johannesburg
Johannesburg, South Africa Johannesburg, South Africa Johannesburg, South Africa

Abstract—The aim of the paper is to present a method for eval- Realising long endurance flight requires, in addition to a
uating brushless DC motors (BLDC) for fixed wing Unmanned high energy density power source, an efficient propulsion sys-
Aerial Vehicles (UAVs). Selection of a BLDC motor, which tem. An electric propulsion system converts electrical energy
operates in its efficient region for a UAV platform’s specifications,
will affect the endurance capabilities of the UAV platform. A to mechanical thrust, and consists of four main components:
four-constant model is presented as a means of modelling and an electrical power source, Electronic Speed Controller (ESC),
predicting the characteristics of BLDC motors. Further, physical electric motor and a propeller (Fig. 1). Each of these compo-
testing of each motor is performed on a constructed motor test nents has their own performance characteristics and regions
bench. Validation of the model’s results is compared to physical of maximum efficiency. The selection of each component is
testing. The methods presented provide a means of accurately
characterising a given BLDC motor. This data can be used to critical in achieving an overall propulsion system which is both
design the additional components used in the UAV’s propulsion efficient and suited to the application platform’s requirements.
system.

I. I NTRODUCTION
The Unmanned Aerial Vehicle market has grown into a
multi-billion dollar industry, with the UAV market worth over
$4.9 billion in 2010 — largely due to the growing application
domain for UAVs [1]. A UAV is an aircraft without an
on-board human pilot. Instead the aircraft may be piloted
remotely or using an autopilot system. Emerging applications
in South Africa require UAVs to operate for long uninterrupted
durations — typically between than 4 and 24 hours. Some of
these applications include:
• Wildlife tracking;
• Traffic monitoring;
• Border patrols;
• Environmental research;
• Geographical mapping; and
• Numerous military applications Fig. 1. Components of a typical electrical propulsion system. a) Battery, b)
Electronic Speed Controller (ESC), c) Motor, d) Propeller.
The above mentioned applications need to be performed
with a UAV which, is “invisible” to the application for which
it is used. Military reconnaissance for instance would benefit
from a platform with low heat and noise signatures in the The University of Johannesburg is currently developing
quest for “invisibility” to an enemy target. Data obtained a modular “Green Energy” UAV with power derived from
from environmental research of atmospheric gases may be batteries, solar power and a hydrogen fuel cell. In this paper,
compromised, should the UAV platform produce gases from the investigation of a means of modelling, characterising and
an internal combustion type propulsion system. Taking these testing of brushless DC motors is presented. With a model that
aspects and the application domain into account, electrical adequately represents BLDC motors, one can easily predict
propulsion becomes a better means of propulsion — over the characteristics of a given motor without the cost and time
internal combustion engines [2]. Electric propulsion systems consuming loaded bench testing. The data gathered during
produce zero gaseous emissions, low heat, and generate less this process aids in the selection of the remaining propulsion
audible noise than a internal combustion system. components needed for the UAV platform.

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

II. T HEORY B UILDING AND M ETHODOLOGY and rotating permanent magnets [3]. Fig. 3 illustrates the two
A. Propulsion Power Transfer physical BDLC motor configurations — inner-rotor and outer-
rotor. The inner-rotor places the magnets in the centre, sur-
The electric propulsion system converts electrical power to
rounded by the coil windings. The outer-rotor places the coil
mechanical power in the form of thrust. Fig. 2 illustrates this
windings enclosed by an outer casing composed of permanent
power transfer through the four components of the propulsion
magnets [4]. This configuration is favoured over the inner-
system.
rotor configuration in model aviation due to the need for lower
energy magnets, reduced copper losses, reduced production
costs, and greater rotor inertia [4].

Fig. 2. Transfer of power through propulsion components.

The power derived from the battery source is given by (1):

Pbatt = Vbatt × Ibatt (1)

Where:
Vbatt Battery voltage [V]
Ibatt Battery current [A]
The ESC receives Pulse Width Modulation (PWM) throttle
control signals either from an on-board Radio Frequency
Fig. 3. The two BLDC motor configurations: a) Outer-rotor BLDC motor,
(RF) receiver or an auto-pilot system. The percentage duty b) Inner-rotor BLDC motor [4].
cycle depends on the required throttle position. The ESC
controls the motor speed according to the percentage duty BLDC Stator Windings
cycle. Power is transferred from the ESC to the motor in the The two most common types of winding configurations for
form of a quasi-three-phase AC power architecture given by 3-phase BLDC motors are Wye (Y ) and Delta (∆) — with
(2): their own advantages and disadvantages. Most commercial
model aviation BLDC motors are Delta (∆) wound and excel
at high speed applications. Delta (∆) wound motors have
PESC (RM S) = VESC (RM S) × IESC (RM S) (2) some disadvantages, these being: additional ohmic losses
(over the Y configuration) and torque ripple from current
Where: circulation in the Delta loop [5].
VESC (RM S) ESC voltage [V]
IESC (RM S) ESC current [A] The Y winding configuration is favoured for lower speed
The brushless DC motor converts the electrical power to a operation where greater torque production is√needed. Y wound
√
torque (M) on its output shaft at a specific angular velocity motors produce 3 more torque and 1/ 3 less rotational
(ω) given by (3): speed than Delta wound BLDC motors [5]. The relationship
between Y and ∆ windings can be represented by (4) and (5).
PM = M × ω (3)
√
Where: MY = 3M∆ (4)
M Torque [N.m]
Where:
ω Angular velocity [rad/s] M Torque [N.m]
The propeller converts the torque (M ) and rotational speed
to an aerodynamic thrust (T ). The required thrust is deter- 1
ωY = √ ω∆ (5)
mined from the UAV platform specifications. Selection of 3
the correct electrical motor is essential for meeting the UAV Where:
platform application requirements. ω Angular velocity [rad/s]
B. Brushless DC Motor Theory C. BLDC Model
BLDC Configurations Most commercial “off-the shelf” BLDC motors used in
Brushless DC motors eliminates the need for brushes used UAV propulsion systems do not have available data with
in shunt DC motors, thereby reducing maintenance costs. respect to the internal parameters of the motor - number
Instead, BLDC motors use electronic commutation through of windings, core material, permanent magnet data etc.
switching electronics to change the current direction. BLDC Modelling of these motors can be difficult as a result. A
motors operate by means of stationary current-carrying coils “black-box” parameter based model is needed. The most

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

common steady-state model used for BLDC motors is the acquired from ESC and motor data sheets. Kirchhoff’s voltage
three-constant model. The constants being: Rm , I0 , and Kv . law is applied to the left-most loop of Fig. 4, which gives:
Rm represents the motor winding resistance; I0 the idle
no-load current due to friction losses and magnetic hysteresis; Ei = E − IRm (6)
and Kv the RPM per back-emf voltage [6].
Where:
Ei Back-EMF [V]
The three-constant model falls short of adequately
modelling all the losses of a BLDC motor [6]. The idle E Battery voltage [V]
current (I0 ) may increase according to the applied voltage, I Battery current [A]
rather than remaining constant as predicted by the three- Rm Combined winding and ESC resis-
constant model. Eddy currents flow within the stator core due tance [Ω]
to the varying magnetic field from electronic commutation. The rotational speed of the motor is proportional to the back
Eddy currents losses are not accounted for in the three- EMF (Ei ) with, proportionality constant (Kv ) given by:
constant model, these losses increase proportionally to the
rate of change of magnetic flux according to Maxwell’s N
Kv = (7)
equations i.e. rotational speed. Viscous damping (windage Ei
losses) are also neglected in this model — yet they can Where:
be substantial since windage power losses are quadratic in Kv RPM/Volt constant
relation to the rotational speed of the motor. N Rotational speed of motor [RPM]
Ei Back-EMF [V]
The proposed model for BLDC motors consists of four con- The value of I0 can be obtained from the current axis
stants — Rm , I0 , Kv and R1 . Rm is the combined ESC and interception of the battery current vs. battery voltage graph
winding resistance. I0 differs from the three-constant model, (I vs. E graph). The fourth constant (R1 ) is obtained from the
now representing the no-load current at zero battery voltage equation below:
[6]. Kv is identical to the three-constant model. The fourth
constant, R1 , is used to model the Eddy current and viscous E − I(Resc + Rm )
R1 = (8)
damping losses [6]. Fig. 4 is an electrical representation of the I − I0
four-constant model for BLDC motors. Where:
R1 Windage and Eddy current losses
[Ω]
The power delivered to the load is calculated with (9):
Ei
Pload = (9)
Ii
Where:
Pload Power delivered to load [Ω]
Ii Load current [A]
Fig. 4. Four-constant model for BLDC motors.
The characteristic model prediction (efficiency vs. RPM)
can be found using (1) and (9):
The electronic commutation is achieved by switch mode
Pload
DC to AC converters in the form of what is commonly η= × 100 (10)
referred to as an Electronic Speed Controller. Switch mode Pbatt
converter losses are generally assigned to on-state, switching Where:
and energy storage medium losses. On state and energy η Efficiency [%]
storage losses are assumed linearly dependent on the current
flowing through the converter. Switching losses are dependent D. Loaded Motor Characterisation
on switching frequency and voltage by current during the A loaded characterisation of BLDC is necessary in order
cross over periods. Switching losses are greatly dependent on to validate the four-constant model, and more importantly
the rising and falling times of the switching architecture. The determine if a particular motor will satisfy the needs the UAV
combined losses incurred by the ESC and the winding losses platform. A motor test bench is needed that has a stable means
are represented by Rm in Fig. 4. of mounting the motor under test; measurement of the output
torque and RPM; and a means of increasing/decreasing the
All four constants can be acquired by recording the no- amount of mechanical loading on the motor. One such method
load current and rotational speed in revolutions per minute is by means of a prony brake whereby, a pulley wheel (attached
(RPM) while varying the battery voltage (E). The Rm value is to the motor shaft) with a belt of fixed length connected at

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

The motor under test spins a 2 mm thick aluminium disk

(300 mm diameter). The left-hand upright (Fig. 6) houses a
freely rotating aluminium arm with two neodymium magnets
at 120 mm from the pivot point of the arm. The rotating
movement of the aluminium disk in close proximity to the
rotating arm (housing magnets) induces Eddy currents to
flow inside the aluminium disk. From Lenz’s law, these Eddy
currents will flow in such a manner as to oppose the change
in flux, causing the arm to rotate opposite to that of the disk.

The force at a displacement of 500 mm from the pivot of

the arm is measured with a load cell. Rotational speed of the
Fig. 5. Prony brake motor test setup [7]. motor is measured via a hall or optical sensor. The torque and
power output of the motor can be calculated from (12) and
(13) respectively.
each end to a load cell (Fig. 5). The motor load is increased
by tightening the belt tension adjustment screw V in Fig. 5) Marm = Floadcell × r (12)
The difference of the two load cell forces (tangential to Where:
the pulley), together with the known pulley radius is used to Marm Torque [N.m]
calculate the output torque (11) [7]: Floadcell Force[N]
r Displacement [m]
M = (F1 − F2 )r (11)

Where: Parm = Marm ω (13)

F1 Force reading on load cell 1 [N.m]
F2 Force reading on load cell 2 [N.m] Where:
Parm Power [W]
r Pulley radius [m]
Marm Torque [N.m]
The pulley method is conceptually very basic, but hard to ω Angular velocity [rad/s]
Bringing the motor upright towards/away from the rotating
implement since there is slip between the belt and pulley,
arm will increase/decrease the load to the motor under test
giving rise to noisy measurements on the load cells. The need
respectively. There is a power loss incurred due to the viscous
for two load cells increases the cost of the experimental set-up.
damping from the aluminium disc spinning in air which, needs
to be accounted for using a quadratic fitting to the no-load
A better experimental set-up is needed whereby experimen-
current vs. voltage plot with the aluminium disk attached to
tal errors are minimised. The experimental set-up for loaded
the motor shaft (14).
characterisation consists of two parallel uprights made of
aluminium (Fig. 6). The BLDC motor under testing is mounted
Pdisk ∝ ω (14)
to the right-hand upright, which has been designed to slide
towards/away from the left-hand upright in Fig. 6. From (13) and (14), the output power from the BLDC motor
under test is:

PM = Parm + Pdisk (15)

The motor efficiency is:
PM
η= × 100 (16)
Pbatt
III. R ESULTS AND D ISCUSSIONS
The motor characterisation was performed to ascertain
which motor would meet the requirements of the given UAV
platform. The requirements for the motor where: operate at
maximum efficiency above 80 % and provide a minimum of
0.5 N.m of torque at 6 A battery current — whilst supplied by
a 4-cell 14.8 V Lithium Polymer (LiPo) battery pack. Three
BLDC motors were tested and characterised using the four
constant model predictions and loaded testing. The following
Fig. 6. Motor test bench. motors were tested:

978-1-61284-993-5/11/$26.00 ©2011 IEEE

 IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

1) Scorpion S5525(∆); efficiency of 82 % — meeting the requirement of above 80

2) Scorpion S5525(Y ); % efficiency. The fact that the Scorpion S5525(Y ) reaches an
3) Axi 5320/34(∆); even higher efficiency of 84.4 % @ 7.5 A, indicates the motor
The no-load I vs. V plots for the Scorpion S5525(∆), is very efficient — should the motor load increase due to flight
Scorpion S5525(Y ), and Axi 5320/34 (∆) allows for conditions or additional UAV payload.
extraction of the I0 and R1 constants (Fig. 7). Analysing IV. C ONCLUSION
the plots one can see the idle current at zero battery voltage
The UAV industry has grown into a multi-billion dollar,
(I0 ) for all three motors tested i.e. the I-axis intercepts.
highly competitive market. It is of paramount importance
The Scorpion S5525 (Y ) therefore has the lowest hysteresis
to each UAV manufacturer to provide a UAV capable of
and friction losses (see Section II-C), followed by the Axi
sustaining level flight for very long durations. Apart from the
5320/34(∆) and the Scorpion S5525(∆). The Kv and R1
UAV platform’s aerodynamic design, the propulsion system
constants are found using (7) and (8) respectively. The slope
becomes the largest system that restricts long endurance
of each plot from Fig. 7 gives us insight into the R1 eddy
flight. In this paper a means of modelling, bench testing
current and windage losses (see Section II-C), i.e. the steeper
and characterising brushless DC motors (BLDC) has been
the slope gradient — the lower the R1 value.
presented. Knowing the characteristics of a BLDC motor aids
in the design of the additional components — in the quest
The constants from no-load testing are presented below
for an efficiently integrated propulsion system.
in Table I. Note the Scorpion S5525 motor was tested in
both Y and ∆ winding √configurations, where the Y Kv
Modelling of a BLDC motor becomes difficult without
constant is approximately 3 times larger than the equivalent
knowing the internal parameters of the motor. The four-
∆ connected motor (as predicted from Section II-B).
constant model provides a relatively easy, yet effective means
TABLE I of predicting a specific motor’s performance capabilities —
N O - LOAD MOTOR CONSTANTS . reducing the need for time consuming loaded bench testing.
This is validated against physical loaded bench testing of
Motor Kv Rm (Ω) I0 (A) R1 (Ω) each motor.
Scorpion S5525(∆) 180 0.125 0.982 19.443
Three typical production grade motors were tested, two of
Scorpion S5525(Y ) 104 0.16 0.474 52.140 which were identical apart from winding connection (Scorpion
S5525(∆) and S5525(Y )). The findings were that Scorpion
Axi 5320/34(∆) 220 0.14 0.511 85.608
S5525(∆) (low efficiency) and the Axi 5320/34(∆) (low
torque) were not suitable for the UAV platform requirements
The constants in Table I were used to predict each motor’s (see Section III). The Scorpion S5525(Y ) motor was chosen
efficiency vs. current curve. This is illustrated for the Scorpion since it reached higher efficiencies over and above the
S5525 ∆, Scorpion S5525 Y , and Axi 5320/34 Y in Fig. 8. required operating current, while providing sufficient torque.

Validation of these prediction results was done by means

R EFERENCES
the motor test bench as shown in Fig. 6. Input voltage, current
and power measurements were taken using a Yokogawa 2533 [1] tealgroup.com, “Teal group predicts worldwide uav market will
total over $80 billion in its just released 2010 uav market
Digital power meter. Force and RPM measurement were profile and forecast,” Internet Page, Accessed March 2011.
obtained by means of a load cell and hall sensor respectively. [Online]. Available: http://www.tealgroup.com/index.php?option=com
Output power measurement was obtained through equations content&view=article&id=62:uav-study-release&catid=3&Itemid=16
[2] C. Herwerth, U. Ofoma, C. Wu, S. Matsuyama, and S. Clark, “Devel-
11, 12, 13 & 14, with efficiency from (15). Each motor was opment of a fuel cell powered uav for environmental research,” in Proc.
tested over its intended operating load region with a constant 44th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada, Jan.
14.8V DC power supply source. The predicted plots follow 2006, pp. 2006–237.
[3] A. Emadi, Handbook of automotive power electronics and motor drives,
the results from loaded testing accurately, proving that the A. Emadi, Ed. Florida, United States of America: CRC Press, 2005.
four-constant model for BLDC motors can be used with [4] W. H. Yeadon, Handbook of small electric motors, W. H. Yeadon, Ed.
confidence when selecting a motor for a UAV platform. United States of America: McGraw-Hill, 2001.
[5] D. C. Hanselman, Brushless permanent-magnet motor design, H. B.
Crawford, Ed. United States of America: McGraw-Hill, 1994.
The Scorpion S5525(∆) motor produced 0.64 N.m of torque [6] J. Carri. (2007) A four-constant model for electric motors (draft no.
2). Internet draft. [Online]. Available: http://flbeagle.rchomepage.com/
and reached a maximum efficiency of 79 % @ 14 A — research/4 consts paper ver 02.pdf
making it unsuitable for the UAV platform requirements due [7] T. Wildi, Electrical Machines, Drives, and Power Systems, 6th ed.,
to the high operating current and low efficiency. The Axi A. Wolf, Ed. United States of America: Prentice Hall, 2006.
5320/34(∆) produced 0.46 N.m of torque at an efficiency
of 82.2 % @ 6 A. At the required operating current (6 A),
the Scorpion S5525(Y ) produced 0.52 N.m of torque with an

978-1-61284-993-5/11/$26.00 ©2011 IEEE

IEEE Africon 2011 - The Falls Resort and Conference Centre, Livingstone, Zambia, 13 - 15 September 2011

Fig. 7. No-load battery current vs. battery voltage for Scorpion S5525(∆), Scorpion S5525(Y ) and Axi 5320/34(∆).

Fig. 8. Loaded efficiency vs. battery current for Scorpion S5525(∆), Scorpion S5525(Y ) and Axi 5320/34(∆).

978-1-61284-993-5/11/$26.00 ©2011 IEEE