A Study of Sensorless Commutation Methods for Brushless DC Motors

by

Jiahao Kou

A thesis submitted in partial fulfillment of the requirements for the degree of

Master of Science

in

Energy Systems

Department of Electrical and Computer Engineering

University of Alberta

c Jiahao Kou, 2017

Abstract

Due to advantages including high efficiency, linear speed torque characteris-

tics and better superior speed performance, Brushless DC (BLDC) motors
have been applied in many industrial areas, including automotives, aerospace,
HVAC, refrigeration, medicine and industrial automation. BLDC motors are
controlled in either sensor or sensorless condition, and many control methods
have been proposed to improve their performance. This thesis investigates
two sensorless control methods based on line voltage differences and the dis-
turbance observer respectively. Each of these two sensorless control methods is
first verified through Simulink simulation results, and then implemented in an
experimental environment based on the DRV8312 Digital Motor Control Kit
from Texas Instruments. To ensure the correct commutation of the BLDC mo-
tor, different phase delay compensation algorithms are applied to compensate
the time delay from the low pass filters for these two sensorless control meth-
ods, respectively. Finally, a comparison is made between these two sensorless
drives, and the robustness of the sensorless drives during loading transient is
demonstrated by simulation results.

ii
Acknowledgements

I would like to thank my supervisor, Dr. Alan Lynch, for his guidance and
support throughout my master program. I would also like to thank the mem-
bers of my thesis committee, Dr. Qing Zhao and Dr. Yunwei Li, for reviewing
my thesis and giving me valuable comments.
I would like to thank my colleagues in the Applied Nonlinear Controls Lab
(ANCL): Ning Cao, Hui Xie and Lianfeng Hou for their help in my courses
and research works.
My sincere gratitude is dedicated to my parents, my brother, and my girl-
friend Hui for their love, support and encouragement during my life at Uni-
versity of Alberta.

iii
Table of Contents

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Conventional Control Method Using Hall-effect Sensors . . . . 2
1.3 Sensorless Control Methods for BLDC Motors . . . . . . . . . 4
1.3.1 Terminal Voltage Sensing method . . . . . . . . . . . . 4
1.3.2 Freewheeling Diode Conducting State Detection method 5
1.3.3 Third Harmonic Component Integration method . . . . 6
1.3.4 Back-EMF Integration Method . . . . . . . . . . . . . 7
1.3.5 Optimization . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.6 Observer Based Methods . . . . . . . . . . . . . . . . . 11
1.4 Summary of Contributions . . . . . . . . . . . . . . . . . . . . 13
1.5 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 Fundamentals of the BLDC Motor 15

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 BLDC Motor Structure . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Magnetic Circuit Model . . . . . . . . . . . . . . . . . . . . . 18
2.4 Modeling of a BLDC Motor . . . . . . . . . . . . . . . . . . . 21
2.5 Equivalent Circuit for a BLDC Motor . . . . . . . . . . . . . . 22
2.6 Back-EMF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7 Electromagnetic Torque . . . . . . . . . . . . . . . . . . . . . 23
2.8 Electromechanical Model . . . . . . . . . . . . . . . . . . . . . 24
2.9 Comparisons between BLDC Motors and PMSMs . . . . . . . 25
2.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3 Experimental Environment 28
3.1 Hardware Overview . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Piccolo F28035 ISO controlCARD . . . . . . . . . . . . . . . . 30
3.2.1 ADC Module . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.2 ePWM Module . . . . . . . . . . . . . . . . . . . . . . 30
3.3 DRV8312 Digital Motor Control Board . . . . . . . . . . . . . 31
3.3.1 DRV8312 . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3.2 Macro Blocks . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.3 Power Domains of the Kit . . . . . . . . . . . . . . . . 34
3.3.4 PWMDAC Module . . . . . . . . . . . . . . . . . . . . 34
3.4 BLDC Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4 Sensorless BLDC Motor Drive Based on Line Voltage Differ-

ences 36
4.1 Six-step Commutation . . . . . . . . . . . . . . . . . . . . . . 36
4.1.1 Analysis of Six-step Commutation . . . . . . . . . . . . 36
4.2 Sensorless Commutation Method Based on Line Voltage Differ-
ences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Voltage Measurement . . . . . . . . . . . . . . . . . . . . . . . 44
4.4 Phase Delay Compensation . . . . . . . . . . . . . . . . . . . 45
4.5 Starting Procedure . . . . . . . . . . . . . . . . . . . . . . . 49
4.6 Electrical Noises . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.7 Proposed Sensorless Drive . . . . . . . . . . . . . . . . . . . . 50
4.8 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . 51
4.8.1 Experiment Results at 600 rpm . . . . . . . . . . . . . 51
4.8.2 Experiment Results at 1000 rpm . . . . . . . . . . . . 53
4.8.3 Phase Delay Compensation . . . . . . . . . . . . . . . 53
4.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

5 Sensorless BLDC Motor Drive Based on Disturbance Observer 58

5.1 Disturbance Observer . . . . . . . . . . . . . . . . . . . . . . . 58

v
 5.2 Sensorless Commutation Method Based on Disturbance Observer 60
5.3 Design of the Disturbance Observer . . . . . . . . . . . . . . . 63
5.4 Phase Delay Compensation . . . . . . . . . . . . . . . . . . . 66
5.5 Proposed Sensorless Drive . . . . . . . . . . . . . . . . . . . . 70
5.6 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . 70
5.6.1 Experiment Results at 600 rpm . . . . . . . . . . . . . 71
5.6.2 Experiment Results at 1000 rpm . . . . . . . . . . . . 72
5.6.3 Phase Delay Compensation . . . . . . . . . . . . . . . 72
5.7 Comparison with the Sensorless Drive based on Line Voltage
Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6 Conclusions 79
6.1 Summary of Research . . . . . . . . . . . . . . . . . . . . . . . 79
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

Bibliography 80

vi
List of Figures

1.1 World market growth for electric motors in automotive appli-

cations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Hall-effect sensors outputs at 700 rpm . . . . . . . . . . . . . . 3
1.3 Third harmonic flux linkage . . . . . . . . . . . . . . . . . . . 7
1.4 Integrated areas of the back-EMF . . . . . . . . . . . . . . . . 8
1.5 Current flowing direction during PWM off time . . . . . . . . 10

2.1 Brushless DC motor waveforms . . . . . . . . . . . . . . . . . 16

2.2 Fundamental BLDC motor structure . . . . . . . . . . . . . . 17
2.3 A BLDC motor with the flux paths . . . . . . . . . . . . . . . 18
2.4 A magnetic circuit for BLDC motor . . . . . . . . . . . . . . . 19
2.5 Simplification of the magnetic circuit . . . . . . . . . . . . . . 20
2.6 Magnetic circuit considering the mutual coupling . . . . . . . 20
2.7 Equivalent circuit for BLDC motor . . . . . . . . . . . . . . . 22

3.1 Diagram of the DRV8312 Digital Motor Control Kit . . . . . . 29

3.2 DRV8312 Digital Motor Control (DMC) board . . . . . . . . . 29
3.3 Piccolo F28035 ISO controlCARD . . . . . . . . . . . . . . . . 30
3.4 Differential amplifier circuit on DRV8312 DMC board . . . . . 33
3.5 BLY172S-24V-4000 motor . . . . . . . . . . . . . . . . . . . . 35

4.1 Stator space sectors diagram . . . . . . . . . . . . . . . . . . . 37

4.2 BLDC motor six-step commutation . . . . . . . . . . . . . . . 39
4.3 Simulation results of the line voltage differences and back-EMFs
waveforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Simulation results of the the combined commutation function . 44
4.5 Voltage resistor divider circuit . . . . . . . . . . . . . . . . . . 45
4.6 Software flowchart of the sensorless commutation method based
on line voltage differences . . . . . . . . . . . . . . . . . . . . 48
4.7 Proposed sensorless BLDC motor drive based on line voltage
differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.8 Line voltage differences at 600 rpm . . . . . . . . . . . . . . . 52
4.9 Speed response at 600 rpm . . . . . . . . . . . . . . . . . . . . 52
4.10 Line voltage differences at 1000 rpm . . . . . . . . . . . . . . . 53
4.11 Speed response at 1000 rpm . . . . . . . . . . . . . . . . . . . 54
4.12 Phase A current and Vabca at 1800 rpm . . . . . . . . . . . . . 55
4.13 Speed response at 1800 rpm . . . . . . . . . . . . . . . . . . . 55
4.14 Commutation signals at 1800 rpm . . . . . . . . . . . . . . . . 56

5.1 Structure of DEF disturbance observer . . . . . . . . . . . . . 59

5.2 Simulation results of the back-EMF differences and phase currents 61
5.3 Applied disturbance observer structure . . . . . . . . . . . . . 62
5.4 Modified disturbance observer structure . . . . . . . . . . . . . 64
5.5 Estimated back-EMF differences at 1000 rpm . . . . . . . . . . 67
5.6 Estimated back-EMF differences at 2000 rpm . . . . . . . . . . 68
5.7 Software flowchart of the sensorless commutation method based
on the disturbance observer structure . . . . . . . . . . . . . . 69
5.8 Proposed sensorless BLDC motor drive based on the distur-
bance observer structure . . . . . . . . . . . . . . . . . . . . . 71
5.9 Estimated back-EMF differences at 600 rpm . . . . . . . . . . 71
5.10 Speed response at 600 rpm . . . . . . . . . . . . . . . . . . . . 72
5.11 Estimated back-EMF differences at 1000 rpm . . . . . . . . . . 73
5.12 Speed response at 1000 rpm . . . . . . . . . . . . . . . . . . . 73
5.13 Phase A current and êab at 1500 rpm . . . . . . . . . . . . . . 74
5.14 Speed response at 1500 rpm . . . . . . . . . . . . . . . . . . . 74
5.15 Commutation signals at 1500 rpm . . . . . . . . . . . . . . . . 75
5.16 Vabca at 300 rpm and êab at 250 rpm . . . . . . . . . . . . . . . 77
5.17 Speed response at 500 rpm during loading transient . . . . . . 78

viii
5.18 Speed response at 1500 rpm during loading transient . . . . . . 78

ix
List of Tables

1.1 Commutation function for the unknown input observer . . . . 12

2.1 Magnetic circuit model parameters . . . . . . . . . . . . . . . 19

3.1 BLDC motor specifications . . . . . . . . . . . . . . . . . . . . 35

4.1 Commutation states in six-step BLDC motor control . . . . . 38

4.2 Commutation sequence for the method based on line voltage
differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Terminal voltage sensing circuit specifications . . . . . . . . . 46

5.1 Commutation sequence for the method based on disturbance

observer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Chapter 1

Introduction

1.1 Background
In [30], it mentions that BLDC motors are better than induction motors and
brushed DC motors in many areas. Firstly, BLDC motors are smaller in size,
and have higher efficiency because of no losses in the rotor. Secondly, due to
the elimination of brushes, BLDC motors have better high speed capability,
and they do not need any brush maintenance. Thirdly, BLDC motors have a
faster response and a lower radio frequency interference (RFI). Also, BLDC
motors are designed with internal shaft position feedback, so that they have
higher starting torque compared with AC induction motors. Lastly, compared
to induction motors, BLDC motors have much better controllability because
of their linear speed torque characteristics.
The world market growth for electric motors in automotive applications
is shown in Figure 1.1, derived from the IMS research. As for global issues
like fuel shortage and global warming, the market for electric vehicles and
hybrid vehicles will continue to increase in the next few decades. With the
development of electrically-driven automotive systems, the demand for electric
motors is expected to grow, especially for BLDC motor shipments, which are
predicted to grow at twice the rate of other motor types.
BLDC motors have been applied in many other areas. For example, in the
Heating, ventilation and air conditioning (HVAC) and refrigeration industries,
BLDC motors have been applied to reduce the operating power requirements
instead of various types of AC motors. For medical applications, BLDC mo-

1
 10
Stepper
DC Brushed
9 DC Brushless

8
Motor shipments growth (%)
7

1
2012 2013 2014 2015 2016 2017
Year

Figure 1.1: World market growth for electric motors in automotive applications

tors, previously not an option, are now surging due to improvements in effi-
ciency and reliability. BLDC motors are also applied to industrial robots and
aerospace areas. BLDC motors are taking center stages and will be the future
trend, which will drive the growth in the BLDC motor market.
Since BLDC motors play a more and more important role in modern so-
ciety, it is of great value for us to research on BLDC motor drive systems to
make them better for applications in various areas.

1.2 Conventional Control Method Using Hall-

effect Sensors
BLDC motors need rotor positions for proper commutation of stator currents.
For a permanent magnet AC synchronous (PMAC) motor, a resolver or a shaft
encoder is normally applied to detect rotor positions because constant infor-
mation of rotor positions is required for PMAC motor control [8]. However, for
a BLDC motor, one electrical cycle contains only six commutation instants,

2
 2
1

Hall A
0
−1
0.6 0.61 0.62 0.63 0.64 0.65 0.66
time (s)
2
1
Hall B

0
−1
0.6 0.61 0.62 0.63 0.64 0.65 0.66
time (s)
2
1
Hall C

0
−1
0.6 0.61 0.62 0.63 0.64 0.65 0.66
time (s)

Figure 1.2: Hall-effect sensors outputs at 700 rpm

so Hall-effect sensors are often used, which can reduce the costs [19].
BLDC motors are controlled electronically instead of using mechanical
commutation. Hall-effect sensors can be mounted in one of three locations
inside the motor to measure the motor’s positions. When permanent mag-
netic north or south poles pass near the Hall-effect sensors, the Hall-effect
sensors will generate low or high signals accordingly. Figure 1.2 shows a simu-
lation result of the outputs of Hall-effect sensors at a speed of 700 rpm. Based
on the output signals sent from three Hall-effect sensors, the current state can
be determined and currents will be applied to corresponding motor coils.
However, the control method based on hall sensors has some drawbacks.
High temperature will affect the performance of hall sensors thereby limiting
their application areas. The cost and weight of motors will also increase and
extra mechanical arrangement need to be mounted [6]. To increase system
reliability and reduce system cost, sensorless control of BLDC motors has
attracted wide attentions.

3
1.3 Sensorless Control Methods for BLDC Mo-
tors
If a PM motor drive only needs electrical measurements without any require-
ments for a position sensor, it is called a sensorless drive [2]. There are many
kinds of sensorless drives for BLDC motors. Among them, back electromotive
force (back-EMF) based methods are the most commonly used category.
Conventional back-EMF based methods utilize the terminal voltage in the
floating phase [20][19][13], detection of the conducting state of freewheeling
diode in the floating phase [31][6], third harmonic components of back-EMF
waveforms [9] and integration of the floating phase back-EMF [1].
However, back-EMFs values are zero at standstill, and it is difficult to
detect the zero-crossing points from noises if the motor speed is low, so the
low speed performance is limited. For methods based on back-EMFs, a starting
procedure is normally required to speed up BLDC motors to a relatively higher
speed, and then motors can be controlled by sensorless drives [8].

1.3.1 Terminal Voltage Sensing method

The terminal voltage sensing method is one of the common methods among
back-EMFs based sensorless techniques.
For the trapezoidal control of BLDC motors, back-EMFs are required to
be in phase with the phase currents, so the motor is able to produce a constant
torque [8]. The most commonly way is to detect the zero-crossing points of the
back-EMF in the floating phase [22]. There is a 30◦ offset between zero-crossing
points of back-EMFs and current commutation instants, so a 30◦ phase shift
needs to be applied to get the correct commutation instances [20].
Since the back-EMFs are determined with respect to the motor neutral
point, the neutral point voltage must be compare with terminal voltages to
get the zero-crossing points of back-EMFs. However, neutral points are not
available in most BLDC motors. The most commonly used approach is to
connect the terminals of the BLDC motor to a Y-connected resistive network.
The virtual neutral point voltage can be used to compare with the terminal

4
voltage in the floating phase to get the zero-crossing points.
However, the virtual neutral point voltage contains lots of electrical noises
because the pulse width modulation (PWM) technique is normally applied to
drive the inverter. Another approach is to compare terminal voltages with half
of the DC link voltage to get back-EMFs zero-crossing points [19]. Work in
[28] mentioned that the measurements do not contain any noises if sampling
during PWM on time.
Low pass filters are normally applied to attenuate the high frequency har-
monics components in terminal voltages introduced by PWM. However, low
pass filters will cause a time delay which becomes more obvious when a BLDC
motor runs at high speeds range. With no phase compensation, the time delay
will restrict the suitable high speed range of BLDC motors. Because the back-
EMF magnitude is in proportion to the motor speed, it is unable to detect
back-EMFs at standstill and low speed, which is another drawback of termi-
nal voltage sensing method [8]. Therefore, conventional open-loop starting
method will be applied to speed up a BLDC motor from standstill.

1.3.2 Freewheeling Diode Conducting State Detection

method
Work in [31] proposed an method which can estimate the commutation points
on the basis of the conducting state of the freewheeling diode which is in paral-
lel with the power transistor in the floating phase. Six commutation instants
for one electrical cycle can be obtained based on whether the freewheeling
diode is conducting or not. This detected point leads the next commutation
instant by 30 degrees, so a phase shift is required. The conducting condition
of the freewheeling diode at the negative side of phase C is derived as:
V 1 + V2
ec < − (1.1)
2
where V1 is the transistors voltage drop, V2 is the diodes voltage drop and ec
is the back-EMF of phase C. Compared with the back-EMF value, V1 and V2
are negligible. When ec changes from positive to negative, the current flowing
through the freewheeling diode at the negative side of phase C will be detected

5
as.
Because the conducting of the freewheeling diodes relies on back-EMFs
value, detection of the rotor position at low speeds is hard to realize, so a
starting process is applied [8]. The main drawback of this method is that for
detection of current flowing in each freewheeling diode, a separate power supply
for the comparator circuit is required [6]. These additional requirements will
limit this method in real applications.

1.3.3 Third Harmonic Component Integration method

Works in [29] mentioned that the third harmonic component of back-EMF
waveforms are applied to estimate the commutation points. The trapezoidal
waveforms of the BLDC motor back-EMFs consist of fundamental and higher
order harmonics. After summing the three terminal voltages, the third har-
monic component of back-EMF waveforms is obtained [20]:

Vresult = Van + Vbn + Vcn

 
d
= R+L (ia + ib + ic ) + ea + eb + ec (1.2)
dt
= ea + eb + ec ≈ 3e3 sin (3θr )

where Vresult is the summation result of the three terminal voltages, Van , Vbn
and Vcn are the phase voltages with respect to the motor neutral voltage, R
is the phase stator resistance, L is the phase stator inductance, ea , eb and ec
are the phase back-EMFs, ia , ib and ic are the phase currents, e3 is the third
harmonic component of back-EMF waveforms and θr is the electrical angle.
Because during each state, the sum of three phase currents (ia + ib + ic )
is equal to zero. Vresult is dominated by the third harmonic component and
contains only third-order harmonics (3rd, 6th, 9th). After integrating Vresult ,
the third harmonic flux linkage can be obtained as:
Z
λ = Vresult dt (1.3)

where λ is the third harmonic flux linkage [8].

Based on the results in [29], the zero-crossing points of λ are identical to
the current commutation points as shown in Figure 1.3. The main advantage

6
of this technique is that Vresult does not contain the switching noises caused by
PWM signals. In comparison with other sensorless control methods which need
a low pass filter, this method requires less filtering [20]. Due to insensitivity to
filtering delay, this method can work over a wider speed range. Theoretically,
this sensorless commutation method is suitable for a wide speed range and it
can work at various load conditions. Its performance at lower speeds is also
better because third harmonics can be detected even when the speed is low.

θ (degree)

Commutation
Signal

0 60 120 180 240 300 360 420 480 θ (degree)

Figure 1.3: Comparison of the zero-crossing points of λ and the commutation

signal

1.3.4 Back-EMF Integration Method

Works in [1] proposed another sensorless control method in which the float-
ing phase back-EMF integration result is used to estimate the commutation
points. In close proximity to the back-EMFs zero-crossing points, back-EMFs
vary approximately linearly with time. The integration process will start if
the back-EMF in the floating phase reaches its zero-crossing points. The in-
tegration result Vint can be derived as:
Z t
e(t)
Vint = dt
0 k
(1.4)
mt2
=
2k
7
where k is the integration gain, t is the integration end time, e(t) is the back-
EMF value and m is the slope of the back-EMF waveform.
The integration process will stop when Vint reaches the predefined thresh-
old. As the back-EMF magnitude is in proportion to the motor velocity, when
the motor speeds up or down, the conduction intervals will decrease or increase
due to the fixed predefined threshold voltage as shown in Figure 1.4 [21].

e(t) (V)
low speed

mid speed

high speed

time (s)

Figure 1.4: Integrated areas of the back-EMF

When Vint reaches the predefined threshold, the integration value will be
set to zero by a triggered reset signal. To ensure the integration process won’t
restart before the small current in the floating phase decays to 0, the reset
signal should be sustained for a predefined time [8].
However, the back-EMF integration based method has poor performance
at low speeds because the integration process will cause errors in estimating
the commutation instants when the back-EMF value is small [20].

1.3.5 Optimization
PWM Techniques

Some optimization methods have been proposed. One category is based on

PWM techniques. Low pass filters are normally applied in the conventional
terminal voltage sensing method to attenuate the high frequency harmonics

8
components in terminal voltages introduced by PWM. The low pass filter
will cause time delay which limits high speed performance, if no phase delay
compensation method is applied. However, PWM strategies can eliminate the
requirements for low pass filters and work well over a wide speed range [24].
For conventional PWM methods, one upper transistor in one phase and
one lower transistor in another phase are conducting at the same time. The
remaining phase is the floating phase [31].This approach will increase power
losses in the motor side because of the high harmonic components [25].
Work in [36] proposed a direct back-EMF detection method based on the
PWM strategy. The direct back-EMF detection method eliminates the need
for virtual neutral voltage. In this method, PWM only applies to upper tran-
sistors, and lower transistors will be left on. Considering the condition that
phase A and B are conducting current, PWM is applied to the upper transistor
of phase A and the lower transistor of phase C is on. During PWM off time,
the current flowing direction is shown in Figure 1.5 [36]. The terminal voltage
of phase C is:
3
Vcn = ec + Vn = ec (1.5)
2
where Vcn is the terminal voltage of phase C, Vn is the neutral voltage and ec
is the back-EMF of phase C.
From the equation above, we can see that Vcn is proportional to ec through-
out the PWM off time. The terminal voltage is free of switching noises and is
referenced to ground, so the requirement for the motor neutral point voltage
is eliminated [8].
Work in [35] proposed an improved method which can detect back-EMFs
during PWM on time. For the method mentioned in [36], a minimum PWM
off time required to detect back-EMFs, so the maximum duty cycle of this
method cannot reach 100% which is desired for some applications in which
low battery voltages are applied. To optimize motor operation, back-EMFs
are detected during PWM off time when a motor operates in startup and at
low-speed conditions. Then back-EMFs are detected during PWM on time
when a motor operates in high speed conditions.

9
 VDC

Va R L ea

Vn
L
Vb R eb

ec

Vc L
R
GND

Figure 1.5: Current flowing direction during PWM off time

Lai, Yen-Shin, and Yong-Kai Lin proposed a similar method to detect

zero-crossing points of back-EMF waveforms. For PWM on and off states,
different methods are applied to detect zero-crossing points, and this sensorless
commutation method can work over a wide speed range [24]. The difference
between these two methods is that the second one uses a digital controller to
calculate zero-crossing points instead of using logical comparisons.

Methods Based on Terminal Line to Line Voltages

Another category is based on terminal line to line voltages rather than using
the terminal voltage in the floating phase. In this category, the requirement
for the motor neutral voltage can be eliminated.
Based on the work in [4], we can directly get the commutation instants
from the average terminal line to line voltages of BLDC motors. The average
line to line voltages are in phase with the Hall-effect sensors, so that can be
used directly without the need of a phase shift. Also, it does not need the
motor neutral voltage compared with the conventional method.

10
 Line voltage differences can also be applied to estimate back-EMF zero-
crossing points. Line voltage differences have the same zero-crossing points
as the back-EMFs [6]. As line voltages are applied, we can eliminate the
requirement for the motor neutral voltage.
The commutation points of BLDC motor currents are identical to the zero-
crossing points of line to line back-EMFs, so line to line back-EMFs are applied
directly to estimate BLDC motor commutation points without any phase shifts
[39]. Work in [26] proposed a novel commutation function based on calculated
line to line back-EMFs to estimate commutation instants. Commutation in-
stants occur when the commutation function changes from positive infinity to
negative infinity or from negative infinity to positive infinity. Work in [19]
proposed a method which uses a circuit constituting a differential amplifier
and a comparator to detect zero-crossing points of line to line back-EMFs.

1.3.6 Observer Based Methods

For PMAC motors, observer based methods are often used because they have
sinusoidal back-EMFs, and continuous rotor positions information is required
[8]. Different observers have also been applied to estimate the rotor positions
of BLDC motors.
Observer based methods often take measurements as inputs of the mathe-
matical model of BLDC motors, and take the output as the estimated results.
The difference between the measurements and the estimated results is sent
back to the mathematical model to make a correction.
The Extended Kalman Filter (EKF) can be employed to estimate rotor
positions and motor speed by using average line voltages and the measured
phase currents [37]. With the predictive current controller, average line volt-
ages are calculated at the beginning of the sampling interval. However, EKF
cannot obtain satisfactory estimation results when the motor speed is low.
Based on the work in [11], the computation burdens and difficulties to tuning
the control systems are significant drawbacks of EKF based methods, and the
sliding mode observer is an attractive method because it is robust despite the
measurement noises and parametric uncertainty of the system.

11
 Sliding mode observer is designed based on dynamic equations of BLDC
motors to estimate back-EMF values, rotor positions and the motor speed. For
conventional methods, a fixed sliding mode observer gain is used to estimate
back-EMF values which only suits a particular range of speed. The fixed
observer gain will produce multiple zero-crossings when the motor speed is
low. Also, it will cause a phase delay in many cases. An improvement has
been made to modify the observer gains according to the motor speed, so the
estimated back-EMF values are more accurate [7].
Back-EMF differences signals can also be estimated by observers. Work
in [22] proposed a new sensorless control method which utilizes an unknown
input observer. The unknown input observer is independent of the rotor speed
and can be used to estimate back-EMF differences. As shown in table 1.1, the
estimated back-EMF differences are used in the commutation functions to get
commutation instants.

Electrical Degree Commutation Function

0-60◦ êbc /êca
60◦ -120◦ êab /êbc
120◦ -180◦ êca /êab
180◦ -240◦ êbc /êca
240◦ -300◦ êab /êbc
300◦ -360◦ êca /êab

Table 1.1: Commutation function for the unknown input observer

For methods above, continuous rotor positions information is estimated.

However, for trapezoidal control of BLDC motors, only 6 commutation points
are needed for one electrical cycle, so the complex observer based methods
above are not necessary for the control of BLDC motors. Disturbance ob-
server structure can be used to estimate the back-EMF differences signals
[39]. This method is easier to implement, and it can precisely estimate the six
commutation instants.

12
1.4 Summary of Contributions
This thesis implements and verifies two sensorless control methods based on
line voltage differences and the disturbance observer respectively. Instead of
sampling terminal voltages during PWM on time to avoid noises [6], this the-
sis uses a simpler method in which low pass filters are applied to filter out
the high frequency harmonics components introduced by PWM. Compensa-
tion methods are included in the sensorless drives to compensate the phase
delay introduced by the low pass filters, which will guarantee the high speeds
operation of the BLDC motors. Also, the robustness of the sensorless control
methods during loading transient is demonstrated by simulation results.

1.5 Thesis Outline

Chapter 2 makes a detailed analysis of the fundamentals of BLDC motors.
Firstly, it provides an introduction to the BLDC motor structure with empha-
sis on the stator and rotor. Chapter 2 also presents the equivalent circuit and
the mathematical model of a BLDC motor, which are applied in the derivations
of Chapter 4 and Chapter 5. Back-EMF, which is the basis of the sensorless
control methods applied in this thesis, is analyzed. Finally, a comparison is
made between BLDC motors and PMSMs with respect to power density, power
losses, control complexity, and torque ripples.
Chapter 3 outlines the experimental environment of this thesis. The DRV8312
Digital Motor Control Kit, in essence a Piccolo F28035 controlCARD, a DRV8312
Digital Motor Control board and a BLDC motor, on which all the experiment
work is based, is introduced and discussed.
Chapter 4 presents a six-step sensorless control method based on the line
voltage differences. The Simulink simulation results verify that the line voltage
differences have the same zero-crossing points as the back-EMFs. In order to
get the six commutation instants for one electrical cycle of the BLDC motor,
a phase delay compensation method is applied in consideration of the time
delay caused by low pass filters. This sensorless drive is implemented in the

13
experimental environment under different speed conditions.
Chapter 5 introduces another six-step sensorless control method based on
the disturbance observer. The disturbance observer is applied to estimate
back-EMF differences. The Simulink simulation results show that the zero-
crossing points of the back-EMF differences are identical to the BLDC motor
commutation instants. A different phase delay compensation method is ap-
plied to compensate for the time delay caused by the disturbance observer
structure. The sensorless control method is further verified in the experimen-
tal environment. Finally, we make a comparison of these two sensorless drives.
Chapter 6 summarizes the thesis and makes discussions on possible future
research directions.

14
Chapter 2

Fundamentals of the BLDC

Motor

2.1 Introduction
In this chapter, we will make an introduction to the fundamentals of the
BLDC motor mainly based on the theories in these three references [10][23][40].
Based on the back-EMF waveform shapes, (i.e., sinusoidal and trapezoidal),
the permanent magnet synchronous machines can be classified into two differ-
ent types. The trapezoidal type is called the BLDC motor and the sinusoidal
type named the PMSM.
Figure 2.1 [23] shows waveforms of BLDC motors. For one electrical cycle
of a BLDC motor, it contains six states. During each state, two phases are
conducting and each phase will be excited for 120◦ . The induced back-EMFs
are in the trapezoidal shape, and the constant magnitude remains for 120◦ .
During the change states (from maximum to minimum, or from minimum to
maximum), the induced back-EMFs vary linearly.

2.2 BLDC Motor Structure

Figure 2.2 [12] shows the fundamental structure of a BLDC motor. A BLDC
motor contains a rotating part, called the rotor, and a stationary frame, called
the stator. In this example, the rotor has two pairs of north and south per-
manent magnet poles. The stator acts as the armature because it will be
energized, and the stator current will generate the rotating magnetic field. As

15
ea
30 150

210 330

eb
270
90

ec

ia

ib

ic

Te

θr (degree)

Figure 2.1: Brushless DC motor waveforms

16
m
ent
ion
edabo
ve,BLDC m
oto
rsa
resyn
chr
onou
sma
chin
es,wh
ich m
ean
sth
at
th
est
ato
randth
ero
torg
ene
rat
ema
gne
ticfi
eld
softh
esam
efr
equ
ency[
40]
.

S
tat
or

N S

R
oto
r S N

F
igu
re2
.2:F
und
amen
talBLDC m
oto
rst
ruc
tur
e

A
smen
tion
edin[2
7],th
ewind
ing
sdi
str
ibu
tioni
sdiff
eren
tfr
omth
ato
fan
indu
cti
on m
oto
r.T
oin
cre
aseeffi
cien
cyandt
orqu
e, BLDC m
oto
rsn
orm
ally
u
ses
tarc
onn
ect
eds
tat
orw
ind
ing
s.F
ore
achw
ind
ing
,itc
ons
ist
sofnum
erou
s
o
fco
il
s,wh
icha
rein
ter
conn
ect
edw
ithe
acho
the
r. Th
est
ato
rcanb
esl
ott
ed
o
rsl
otl
ess
. Th
esl
otl
esss
tat
orh
asal
owe
rindu
ctan
ce,s
oth
eBLDC m
oto
ris
ab
let
oruna
thi
ghe
rsp
eed
s.
P
erm
anen
tma
gne
tst
ypi
cal
lyf
ormth
ero
toro
faBLDC m
oto
r.D
iffe
ren
t
BLDC m
oto
rsm
ayh
avev
ari
ousnumb
ero
fpo
lep
air
s.In
cre
asin
gth
enumb
er
o
fpo
lep
air
swi
llin
cre
aseth
emo
tort
orqu
e,bu
tth
isw
il
llim
itth
emax
imum
po
ssib
lesp
eed[1
4].
Th
ema
ter
ialu
sedf
orth
econ
stru
cti
ono
fpe
rman
entm
agn
etsw
il
lal
so
aff
ectth
emax
imumt
orqu
eofBLDC m
oto
rsa
sde
scr
ibedin[
40]
.Tr
adi
tion
-
a
lly
,Fe
rri
tem
agn
etsa
reth
ema
ter
ialu
sedf
orp
erm
anen
tma
gne
tsb
ecau
seo
f
th
eirl
owp
ric
e.F
err
item
agn
etsh
avel
owfluxd
ens
ityf
orag
ivenv
olum
e,and
th
isd
isadv
ant
agew
il
llim
itth
emax
imumt
orqu
e.N
owad
ays
,ra
ree
artha
llo
y
m
agn
ets(
e.g
.,S
ama
riumC
oba
lt
,Ne
odym
iumandth
eal
loyo
fNe
odym
ium
,
F
err
iteandB
oron
)ar
ebe
com
ingm
orepopu
larb
ecau
seo
fth
eirh
igh
erm
ag-
n
eti
cden
sit
y,and m
oto
rsu
sin
gth
esea
llo
yma
gne
tsh
aveasm
all
ers
izewh
en

1
7
c
omp
arin
gwi
th m
oto
rso
fsam
ewe
igh
tus
ingf
err
item
agn
ets
.

S
tat
or

N S

R
oto
r S N

F
igu
re2
.3: ABLDC m
oto
rwi
thth
efluxp
ath
s

2
.3 Magne
tic C
ircu
it Mode
l
Th
iss
ect
ionm
ake
sanin
trodu
cti
ont
oth
ema
gne
ticc
ircu
itm
ode
lofth
eBLDC
m
oto
rba
sedonth
ere
fer
enc
e[1
0].F
igu
re2
.3[
10]sh
owstw
otyp
eso
ffluxp
ath
s.
On
ema
gne
tfluxs
tar
tsf
romth
eno
rthpo
le
,andth
eni
tcr
oss
esth
eai
rgap
.
Wh
enth
ema
gne
tfluxr
each
esth
est
ato
r,i
twi
llb
esep
ara
tedin
totw
opa
rts
.
Th
ene
acho
fth
esep
ara
ted m
agn
etfluxw
il
lcr
ossth
eai
rgapa
gainandt
rav
el
t
owa
rdsth
ead
jac
enttw
osou
thpo
les
.
Inadd
it
iont
oth
ema
gne
tfluxp
ath m
ent
ion
edabo
ve,s
omem
agn
etflux
w
il
lno
tcr
ossth
eai
rgapt
oth
est
ato
r,andi
twi
lld
ire
ctlyt
rav
elt
owa
rdsth
e
ad
jac
entm
agn
etpo
lein
ste
ada
ssh
owninF
igu
r .
e23[
10]
. Th
isk
indo
ffluxi
s
c
all
edth
ema
gne
tle
aka
geflux
.
W
ecan m
ode
lon
ead
jac
entpo
lep
aira
ssh
owninF
igu
re 2
.4[
10]
. Th
e
m
agn
etpo
lesi
srep
res
ent
edb
yafluxs
our
ce φr w
ithi
tsm
agn
etr
elu
ctan
ce
Rm. Th
emod
el
ingp
aram
ete
rsa
resh
owninT
abl
e2.
1.
Th
ele
aka
geflux φli
sve
rysm
all wh
ichc
anb
eign
ored
.Rr andRs a
re
sm
allwh
enc
omp
arin
gwi
thRg.Rr andRs c
anb
etr
eat
eda
sasm
allp
arto
f

1
8
 Name Description Name Description
φr flux source Rm flux source reluctance
Rg air gap reluctance Rl leakage reluctance
Rr rotor steel reluctance Rs stator steel reluctance
φg air gap flux φl leakage flux
φ magnet flux

Table 2.1: Magnetic circuit model parameters

φg

Rg Rs Rg

φl
φ Rl
N S

φr Rm Rm φr

S N

Rr

Figure 2.4: A magnetic circuit for BLDC motor

Rg . The equivalent reluctance (Rr + Rs + 2Rg ) can be simplified as:

Rr + Rs + 2Rg = 2Kr Rg

where Kr is the reluctance factor and it is slightly bigger than one.

By determining Norton equivalent circuit, the magnetic circuit can be sim-
plified into Figure 2.5 [10].
The mutual flux between one pole pair and a stator coil can be showed in
Figure 2.6 [10]. The total flux φ linking the coil comes from two sources: φ1 is
the flux from the coil current and φm is the flux from the permanent magnet.

19
 N

φr 2Rm 2Kr Rg

Figure 2.5: Simplification of the magnetic circuit

These fluxes can be expressed as following:

φ = φ1 + φm

Ni
φ1 =
R + Rm
Rm φr
φm =
R + Rm
where R is the reluctance seen from the stator coil, N is the winding turns, i
is the current flowing in the winding.
The flux linkage λ is:
λ = Li + N φm

where L is the self inductance.

Ni Rm φr

Figure 2.6: Magnetic circuit considering the mutual coupling

20
2.4 Modeling of a BLDC Motor
The BLDC motor modeling is based on the reference [23]. For the three stator
windings of a BLDC motor, the electrical relationships can be represented with
the equation below:
         
Van R 0 0 ia Laa Lab Lac ia ea
 Vbn  =  0 R 0   ib  + d  Lba Lbb Lbc   ib  +  eb  (2.1)
dt
Vcn 0 0 R ic Lca Lcb Lcc ic ec

where Van , Vbn and Vcn are the phase voltages referenced to the motor neutral
voltage, ia , ib and ic are the phase currents, ea , eb and ec are the phase back-
EMFs, R is the phase stator resistance which is same for three phases, Laa ,
Lbb and Lcc are the self-inductances, Lab , Lba , Lac , Lca , Lbc and Lcb are the
mutual inductances.
If three phases are symmetric and the rotor reluctance is identical with
the change of electrical angle, the three self-inductances are same and the six
mutual inductances are equal to each other as:

Laa = Lbb = Lcc = Ls (2.2)

Lab = Lba = Lac = Lca = Lbc = Lcb = M (2.3)

The equivalent inductance L can be calculated as:

L = Ls − M (2.4)

Assuming that the phase currents are symmetric, that is

ia + i b + ic = 0 (2.5)

Then the electrical relationships can be simplified and the BLDC motor can
be modeled as:
         
Van 1 0 0 ia L 0 0 ia ea
 Vbn  = R 0 1 0  ib  + d  0 L 0   ib  +  eb  (2.6)
dt
Vcn 0 0 1 ic 0 0 L ic ec

21
2.5 Equivalent Circuit for a BLDC Motor
For a three-phase BLDC motor, the electrical circuit model can be shown
in Figure 2.7. In this circuit, R represents the resistance component of the
stator winding and L represents the inductive component of the stator phase
winding.

+ a

ea
eb n
L
ec
R L
+ R c
b +

Figure 2.7: Equivalent circuit for BLDC motor

ea , eb and ec are the phase back-EMFs. Because the resistance and in-
ductance are very small for the BLDC motor, the phase currents need to be
limited during starting up. Normally, PWM technique is applied for BLDC
motor control, and a small duty cycle is set during starting up.

2.6 Back-EMF
This section introduces the back-EMF of a BLDC motor based on the reference
[23]. When starting the BLDC motor, stator coils will be energized and a
torque will be generated on the rotor. Then the rotor will begin to rotate
and the flux linkage in the stator coils will vary depending on permanent
magnets locations. According to Faraday’s law of induction, a voltage will
be induced across the stator windings due to the varying flux linkage. This
induced voltage is referenced to back-EMF. The polarity of voltage induced is

22
governed by Lenz’s law. That is, the back-EMF will always try to keep the
flux linkage from changing its present value.
For BLDC motors, the back-EMFs are in trapezoidal shape. The peak
value of the back-EMFs is Ep which can be calculated as:

Ep = K e ω (2.7)

where ω is the mechanical speed of the motor in rad/s, Ke is back-EMF con-

stant and its unit is V/(rad/s).
Ke can be calculated as:
Ke = N Blr (2.8)

where B is the flux density, N represents the conductors number per phase, l
is the conductor length, r is the air gap radius between the motor rotor and
stator.
From the equations above, the peak value of the back-EMFs depends
mainly on ω, B and N . Once the motor is designed, the back-EMF con-
stant Ke remains the same. The peak value of the back-EMFs is proportional
to the mechanical speed of the rotor.
The instantaneous back-EMFs can be written as:

ea = fa (θr ) Ke ω (2.9)

eb = fb (θr ) Ke ω (2.10)

ec = fc (θr ) Ke ω (2.11)

where θr is the electrical degree, fa , fb and fc are in the trapezoidal shape as

the back-EMFs, and their peak magnitude is 1.

2.7 Electromagnetic Torque

From the energy conversion point, a BLDC motor converts electrical power to
mechanical power to make the motor rotate. Each phase of a BLDC motor is
composed of a resistive component and an inductive component [23].
The stator current flowing through the winding will create ohmic losses
or heat in the winding resistor. The current will also create a magnetic field

23
that stores energy. When phase currents flow through back-EMF sources, the
instantaneous power Pinstant absorbed by sources can be derived as:

Pinstant = ea ia + eb ib + ec ic (2.12)

To satisfy the conservation of energy, this power will be converted to me-

chanical power Te ω. The electromagnetic torque is given by:

1
Te = [ea ia + eb ib + ec ic ] (2.13)
ω

Because back-EMFs are proportional to speed, the electromagnetic torque

can be derived as [23]:

Te = Kt [fa (θr ) ia + fb (θr ) ib + fc (θr ) ic ]

(2.14)
= Kt (2Ipd )

where Ipd is the peak value of the phase current, Kt is the torque constant and
its units are Nm/A. Once the motor is designed, Kt is a constant. As a result,
the electromagnetic torque is directly proportional to the stator currents of a
BLDC motor.
Ke and Kt are numerically equal. The units for back-EMF constant Ke
and the torque constant Kt are equivalent if they are all in SI units as shown
in the equation below:

Nm/A = V/(rad/s) = kg · m2 /A · s2 (2.15)

2.8 Electromechanical Model

For BLDC motors, the absorbed instantaneous power of back-EMF sources
will be converted to mechanical power Te ω.
The electromagnetic torque is used to drive the motor rotor to rotate. As-
suming that the system inertia is J, friction coefficient is B, and the load torque
is T1 if the BLDC motor drives a load, the equation for the electromechanical
model of the BLDC motor can be shown as [23]:

dω
J + Bω = (Te − T1 ) (2.16)
dt

24
and the relationship between the motor synchronous speed and the rotor po-
sition can be expressed as:
dθr p
= ω (2.17)
dt 2
where p is the rotor poles number, ω is the mechanical speed of the motor in
rad/s, θr is the rotor position in rad.
At the moment a BLDC motor starts, there is no back-EMF, and the
stator current will produce maximum electromagnetic torque on the motor
rotor. The motor will speed up, and the peak values of back-EMFs will also
increase proportionally. During this process, the stator currents will decrease,
and the motor speed will become stable finally.

2.9 Comparisons between BLDC Motors and

PMSMs
In this section, some comparisons will be made between BLDC motors and
PMSMs on their output power and torque ripples based on the reference [23].
The rms values of the stator currents are:

Ipm
Im = √ (2.18)
2
r
2
Id = Ipd (2.19)
3
where Ipm and Ipd are the peak values of the stator currents in PMSMs and
BLDC motors, respectively. Im is the rms value of the stator current in
PMSMs, and Id is the rms value of the stator current in BLDC motors.
From the equations above, we can see that BLDC motors have a smaller
rms value than that of the sinusoidal PMSMs if their stator currents have the
same peak value. Assuming that PMSMs and BLDC motors have the same
ohmic losses, the relationship can be derived as:

2
3Im R = 3Id2 R (2.20)

where R is the phase stator resistance which is the same for all three phases.

25
 Based on the analysis above, the relationship between the peak value of
PMSMs and BLDC motors currents can be derived as:
√
3
Ipd = Ipm (2.21)
2

From the current waveforms of a BLDC motor, we can see that only two of
the three phases are conducting at the same time. Consequently, the output
power of a BLDC motor is contributed by only two phases. For a PMSM, all
three phases have currents and contribute to output power. Assuming that
PMSMs and BLDC motors back-EMFs have the same peak value Ep , the ratio
of a BLDC motor and a PMSM output power can be calculated as:

BLDC motor power

Power output radio = (2.22)
PMSM motor power
2 × Ep × Ipd
= E Ipm
= 1.1547 (2.23)
3× √p × √
2 2

Based on the analysis above, If PMSMs and BLDC motors have same
ohmic losses, the output power of BLDC motors will be 15.4 % higher than
that of PMSMs. For BLDC motors, only two phases are conducting currents
at the same time, so the duty cycle of the phase currents is smaller than that
of PMSMs. The power losses are smaller in BLDC motors, and the inverter
thermal reliability is also better. Although, the currents of BLDC motors
and PMSMs both have variable frequencies, the rectangular phase currents of
BLDC motors have lower control complexity compared with PMSMs.
The torque ripples of BLDC motors are higher than PMSMs. During the
finite current transition time, torque ripples are created at each commutation
point. Torque ripples will be created whenever currents or back-EMF wave-
forms are not ideal. For low-performance applications, the torque ripples are
rarely significant. However, torque ripples are hard to eliminate and this will
limit the use of BLDC motors in applications where minimum torque ripples
are required. Knowing how to minimize torque ripples is very important for
high-performance applications.

26
2.10 Summary
In this chapter, we make an introduction to the BLDC motor structure, the
magnetic circuit model and the electrical model of a BLDC motor. Back-EMF
and electromagnetic toque concepts are also introduced. Finally a comparison
is made between PMSMs and BLDC motors. BLDC motors have a higher
output power but with larger torque ripples.

27
Chapter 3

Experimental Environment

In this thesis, all the experiment work is based on the DRV8312-C2 kit, which
is developed by Texas Instruments. This kit is a Medium Voltage Digital
Motor Control (DMC) kit. In this chapter, an introduction will be made to
the hardware details and the development environment.

3.1 Hardware Overview

The DRV8312 Digital Motor Control Kit includes a Piccolo F28035 control-
CARD, a DRV8312 Digital Motor Control (DMC) board, a BLDC motor and
a 24 V AC/DC supply with 2.5 A of current.
Figure 3.1 shows the diagram of the Kit. Figure 3.2 shows the picture of
DRV8312 DMC board, which has a socket for the Piccolo F28035 controlCARD
with a built-in isolated XDS100 emulator. The Piccolo F28035 controlCARD
acts as the control block and the DRV8312 DMC board is the motor driver.
Based on the Texas Instruments’ document [3], debuggers are useful tools
to debug software applications on hardware platforms, because the debuggers
are able to access the processors and the registers on hardware platforms.
On the F28035 controlCARD, the F28035 C2000 controller works with the
XDS100 emulation hardware. The emulator is applied to communicate with
the F28035 C2000 controller, which is the target processor, and the debugger is
the user interface to the communication information. A USB cable is applied to
connect the debugger and the laboratory computer. Joint Test Action Group
(JTAG) is the emulation logic protocol for the F28035 C2000 controller.

28
 DRV
831
2Di
git
alM
oto
rCon
tro
lBo
ard

DCBu
s
T1 T3 T5

PWMS
ign
als BLDC m
oto
r

H
allS
ign
als

T2 T4 T6 V
olt
a g
e
S
ensing
GND

Cur
ren
t
Sen
sin
g
I
DC I
a I
b I
c Va Vb Vc

ADC CAP

ePWM CPU

P
icc
oloF
280
35ISOc
ont
rolCARD

F
igu
re3
.1:D
iag
ramo
fth
eDRV
831
2Di
git
alM
oto
rCon
tro
lKi
t

F
igu
re3
.2: DRV
831
2Di
git
alM
oto
rCon
tro
l(DMC
)bo
ard

2
9
3.2 Piccolo F28035 ISO controlCARD
The Piccolo F28035 ISO controlCARD is the evaluation board for the F28035
C2000 controller. As shown in Figure 3.1, this controlCARD includes the
key peripherals like Analog-to-Digital converter (ADC), Enhanced pulse width
modulator (ePWM), CAP and GPIOs. The CAP peripheral is applied to read
the Hall-effect sensor signals. These peripherals signals communication is via
the controlCARD pins. Figure 3.3 shows the picture of Piccolo F28035 ISO
controlCARD.

Figure 3.3: Piccolo F28035 ISO controlCARD

3.2.1 ADC Module

Based on the Texas Instruments’ document [17], ADC is a single 12-bit con-
verter with a 16-bit ADC register, which is fed by two sample and hold circuits.
There are two different types of sampling modes:simultaneous and sequential
sampling modes. The analog input range is from 0 V to 3.3 V, and it can have
up to 16 inputs.

3.2.2 ePWM Module

The Pulse Width Modulation (PWM) strategy is used to chop the supply volt-
age. Because the PWM strategy has a fixed chopping frequency, the switching
noises are much easier to filter. In this thesis, the chopping frequency of PWM
is 20 kHz.
The ePWM module is a fundamental part in digital motor control. The
ePWM module is able to generate PWM signals with minimal CPU usage [18].

30
For the BLDC motor control, three ePWM modules are needed to generate
six PWM signals for the three-phase inverter.
In this thesis, one high side transistor is switched on and off according to
the control signals, and one low side transistor is left on. This mode is able to
reduce the switching losses and torque ripples.

3.3 DRV8312 Digital Motor Control Board

The DRV8312 DMC board is the motor driver board. The key component
is the DRV8312 which is a three-phase motor driver. On the DRV8312 DMC
board, some macros to implement different functions in the BLDC motor driver
system are included. In this section, an introduction will be made to the
DRV8312 DMC board based on the documents [15][16].

3.3.1 DRV8312
Because the BLDC motor is not self-commutated, the stator coils will be
energized in a predefined order to make the rotor rotate. The stator coils
needs to be powered by the external device. For a three-phase BLDC motor,
normally a three-phase inverter is used to power the stator coils. Figure 3.1
shows the diagram of the inverter. The Piccolo F28035 controlCARD will send
PWM control signals to the three-phase inverter.
DRV8312 is a three-phase BLDC motor driver in the Texas Instrument
DRV83x2 driver family. The DRV8312 is able to be applied in the energy-
efficient areas, because it has high energy efficiency which can reach maximum
97 %. In order to power DRV8312, two different power supplies are needed,
one at 12 V for gate-drive voltage supply and power supply for digital voltage
regulator, and another up to 50 V for power supply input for half-bridges [15].
The PWM switching operating frequency of the DRV8312 can be up to
500 kHz. This driver has an advanced protection system which can protect the
device from different fault conditions that could make a damage to the hard-
wares. The protection system includes the bootstrap capacitor undervoltage
protection which can protect the device from the possible damage caused low

31
switching frequency operation, overload protection, and short circuit protec-
tion. The overcurrent detectors allow users to set the current limits based on
the application requirements, and the current-limiting circuit will prevent the
BLDC motor from possible damage caused by overcurrent.

3.3.2 Macro Blocks

On the DRV8312 DMC board, there are several macro blocks to implement
different functions. The DC bus connection macro block provides DC bus
power for the control board. There are two options for the connection block:
one is using the power entry jack to connect a 24 V AC/DC supply, and another
is using PVDD/GND terminals to connect a different power supply.
From Figure 3.1, the DRV8312 DMC board can be connected with a BLDC
motor, and it can also receive the Hall-effect sensors or Quadrature Encoder
signals. The motor connection block is to connect the BLDC motor wires [16].
The Quadrature Encoder connection block can be applied to connect a shaft
encoder to receive motor positions information. Hall-effect sensor connection
block can be used to connect to the Hall-effect sensors. In this thesis, the
signals from the Hall-effect sensor connection block is applied to verify the
accuracy of the sensorless commutation methods.
From Figure 3.1, the DRV8312 DMC board has the terminal voltage sensing
block to measure the three phase terminal voltages Va , Vb and Vc . The terminal
voltages are measured by taking ground as the referenced point. The details
about the terminal voltage sensing block will be introduced in Chapter 4. The
current sensing block is to measure the BLDC motor currents. Ia , Ib and Ic
are the phase currents, and IDC are the DC bus current. In order to measure
the motor current, a reference voltage of 1.65 V is required. This reference
voltage can be provided by a voltage follower. On the DRV8312 DMC board,
the voltage follower circuit is applied to generate a 1.65 V reference voltage
from the 2.5 V voltage source.
A differential amplifier circuit is applied to convert the current input to
an output voltage from 0 to 3.3 V. Figure 3.4 shows the differential amplifier
circuit on the DRV8312 DMC board. The reference voltage 1.65 V is provided

32
 by the voltage follower circuit. The transfer function of the circuit can be
shown as:
R3
Vout = 1.65 + I · R1 · = 1.65 + 0.191I (3.1)
R2
where Vout is the output voltage from the differential amplifier circuit, I is the
motor current and R1 ,R2 and R3 are the resistances in the differential amplifier
circuit.
The transfer function is linear and the analog input range is from 0 V
to 3.3 V. Because the current is bidirectional, which has both positive and
negative currents, 0 V represents the minimum negative current and 3.3 V
represents the maximum positive current. When the output voltage is 0 V, the
current input is −8.65 A. When the output voltage is 3.3 V, the current input
is 8.65 A. The output voltage range (0 V to 3.3 V) represents the measured
phase current range (−8.65 A to 8.65 A).
When the current analog input is 3.3 V, the sampled ADC result value
would be 4096. The range of ADC result is from 0 to 4096. Based on the
ADC result, the current can be converted to the per unit format from −1.0 to
1.0. ±1.0 represents the measured motor current ±8.65 A.

1.65 V

R3 (19.1 kΩ)
Iphase
R2 (1 kΩ)
Vout
R1 (0.01 Ω)

R2 (1 kΩ)

R3 (19.1 kΩ)

Figure 3.4: Differential amplifier circuit on DRV8312 DMC board

33
3.3.3 Power Domains of the Kit
There are two power domains on this kit. One is the low voltage domain,
which is used to provide power for the F28035 C2000 controller and the logic
circuits. The low voltage domain has three levels voltages: 3.3 V, 5 V and
12 V. The low voltage domain can be sourced from two areas: one is the 12 V
DC control power entry, which needs external power supply, and another is
the on board buck regular, which will regulate the DC bus power to 12 V.
Another power domain is the medium voltage power domain, which will
provide medium voltages to the three-phase inverter. The medium voltages
are called the DC bus voltages. The three-phase inverter will provide power
to the BLDC motor and the medium voltages can be up to 52.5 V.
For each phase of the BLDC motor, an individual switch is used to deter-
mine the control mode of that phase. The switch can be manually set to three
positions. If the switch is in the middle position, the switch will be controlled
by a GPIO signal. If the switch is in the up position, the half-bridge output
will be enabled. If the switch is in the down position, the half-bridge output
will be disabled.

3.3.4 PWMDAC Module

The PWMDAC module can be applied to view the measurements waveforms
on an oscilloscope. To get the actual low frequency signals, a first order low-
pass filter is used to filter out the high frequency components. The PWMDAC
module can be used to observe three different signals at one time. The low-
pass filter time constant (τ ) and cut-off frequency (fc ) can be calculated using
following equations:

τ = R · C = 2.21 × 10−4 s (3.2)

1
fc = = 720.5 Hz (3.3)
2πτ
where R is the low-pass filter resistance and C is the low-pass filter capac-
itance.

34
3.4 BLDC Motor
Figure 3.5 shows the BLDC motor used in this thesis, which is made by Ana-
heim Automation and its model number is BLY172S-24V-4000. This BLDC
motor is a cost effective solution to the velocity control application. This mo-
tor has 4 pairs of rotor poles and three wires for the phases and five wires for
the hall sensors. Table 3.1 shows the motor specifications.

Parameter Values
P 8
R 0.4 Ω
L 0.6 mHz
Kt 0.041 Nm/A
Ke 3.35 V/kRPM
J 48 g − cm2

Table 3.1: BLDC motor specifications

Figure 3.5: BLY172S-24V-4000 motor

3.5 Summary
In this chapter, we present the experimental environment in this thesis. An
introduction is made to the Piccolo F28035 controlCARD, the DRV8312 DMC
board and also the basic macro blocks. Finally, we make a discussion on the
BLDC motor and its parameters.

35
Chapter 4

Sensorless BLDC Motor Drive

Based on Line Voltage
Differences

In this chapter, a sensorless BLDC motor control method based on line voltage
differences [6] is introduced and verified through experiment. As introduced
in Chapter 1, methods based on line voltage differences can eliminate the
requirement for the motor neutral voltage, which will increase the accuracy of
the commutation strategy.

4.1 Six-step Commutation

The six-step commutation is the most commonly used way to commutate a
BLDC motor. Figure 4.1 shows the stator space sectors diagram. A and
X represent the stator windings of phase A. B and Y represent the stator
windings of phase B. C and Z represent the stator windings of phase C. A−X,
B − Y and C − Z separate the space into six 60◦ sectors.

4.1.1 Analysis of Six-step Commutation

The six sectors represents six states of a BLDC motor. If Hall-effect sensors are
used to detect rotor positions, for each of the six states, three-digit numbers
based on the three Hall-effect sensors’ signals are used to represent the cur-
rent rotor position. When the rotor moves to the next sector, the three-digit
numbers will change to a different set of three-digit numbers representing the

36
 A

Y Z

3 4

2 5

1 0
C B

F
igu
re4
.1:S
tat
orsp
aces
ect
orsd
iag
ram

n
ews
tat
e.F
ors
ens
orl
essc
ont
rolm
eth
ods
,diff
eren
tte
chn
iqu
esw
il
lbeu
sedt
o
e
stim
ater
oto
rpo
sit
ion
s.
A
ssum
ingth
atth
ero
tori
sins
ect
or0
,th
ecu
rren
tdi
rec
tiono
fph
aseAi
s
f
romAt
oX.th
ecu
rren
tdi
rec
tiono
fph
aseBi
sfr
omY t
oB andph
aseCi
s
fl
oat
ing
.Nowth
est
ato
rfie
ldi
sinth
edi
rec
tiono
fC−Za
ssh
owninF
igu
re
4
.2(
a),andi
twi
llw
il
ldr
iveth
ero
tort
omo
vet
ose
cto
r1.
A
fte
rro
torm
ove
sin
tos
ect
or1
,th
ecu
rren
tdi
rec
tiono
fph
aseAi
sfr
omA
t
oX.th
ecu
rren
tdi
rec
tiono
fph
aseCi
sfr
omZt
oCandph
aseBi
sflo
atin
g.
Th
est
ato
rcu
rren
tssh
apeth
est
ato
rfie
ldwh
ichi
sinth
edi
rec
tiono
fB−Y
a
ssh
owninF
igu
re4
.2(b
).T
orqu
ewi
llb
egen
era
tedb
asedonth
eph
asesh
if
t
b
etw
eenth
ero
torfi
eldandth
est
ato
rfie
ld. Th
enth
eto
rqu
ewi
lld
riv
eth
e
r
oto
rtom
ovet
ose
cto
r2.
Ins
ect
or2
,th
ecu
rren
tdi
rec
tiono
fph
aseCi
sfr
omZt
oC. Th
ecu
rren
t
d
ire
cti
ono
fph
aseBi
sfr
omB t
oY andph
aseAi
sflo
atin
g. Th
est
ato
rfie
ld
i
sinth
edi
rec
tiono
fA−X a
ssh
owninF
igu
re4
.2(
c),andi
twi
lld
riv
eth
e
r
oto
rtom
ovet
ose
cto
r3.
Ins
ect
or3
,th
ecu
rren
tdi
rec
tiono
fph
aseAi
sfr
omX t
oA. Th
ecu
rren
t
d
ire
cti
ono
fph
aseBi
sfr
omB t
oY andph
aseCi
sflo
atin
g. Th
est
ato
rfie
ld

3
7
is in the direction of C − Z as shown in Figure 4.2(d), and it will will drive
the rotor to move to sector 4.
In sector 4, the current direction of phase A is from X to A. The current
direction of phase C is from C to Z and phase B is floating. The stator field
is in the direction of B − Y as shown in Figure 4.2(e), and it will drive the
rotor to move to sector 4.
In sector 5, the current direction of phase C is from C to Z. The current
direction of phase B is from Y to B and phase A is floating. The stator field
is in the direction of A − X as shown in Figure 4.2(f), and it will drive the
rotor to move to sector 0.
As analyzed above, the stator field moves in a discrete step of 60 electrical
degrees. The torque will continuously drive the rotor to rotate. Because the
phase shift between the rotor field and the stator field will change when the
rotor is rotating, this shift will cause torque ripples. The torque ripples can be
reduced if the field oriented control is applied. Then the phase shift between
the rotor field and the stator field is fixed to be 90 electrical degrees to generate
the maximum torque.
The commutation state pointer input (0 to 5) indicates one of the six states.
During each state, only two phases are conducting so two switches are active
and the other four will be turned off. If the commutation state pointer is
equal to 0, current flows from phase A to phase B. Table 4.1 shows the six
commutation states of the BLDC motor.

Commutation States ia ib ic
0 positive negative floating
1 positive floating negative
2 floating positive negative
3 negative positive floating
4 negative floating positive
5 floating negative positive

Table 4.1: Commutation states in six-step BLDC motor control

38
 ×A Fstator ×A
Y× Z Y ×Z
S S

N N
C ·B C. B
Fstator
·X .
X
(a
)Se
cto
r0 (b
)Se
cto
r1

Fstator
A
.A
Fstator
Y . ×Z Y.
N Z
N S

C. ×B
C
S
×B

X ×X
(
c)S
ecto
r2 (d
)Se
cto
r3

A
.A
Y× .Z
Y .Z
N
S N

C× .B
S
C× B

Fstator X
× Fstator
X
(
e)S
ecto
r4 (
f)S
ecto
r5

F
igu
re4
.2: BLDC m
oto
rsix
-st
epc
ommu
tat
ion

3
9
4.2 Sensorless Commutation Method Based on
Line Voltage Differences
In this section, a sensorless control method using the six-step commutation
sequence will be introduced. The proposed method are based on the work in
[6].
For the BLDC motor we used, stator windings of three phases are connected
in star. The terminal voltage of phase A referenced with the motor neutral
voltage is Van as:
dia
Van = Ria + L + ea (4.1)
dt
Similarly, terminal voltages of phase B and phase C are Vbn and Vcn as:

dib
Vbn = Rib + L + eb (4.2)
dt
dic
Vcn = Ric + L + ec (4.3)
dt
where Van , Vbn and Vcn are the phase voltages referenced to the motor neutral
voltage, L is the phase stator inductance, R is the phase stator resistance, ia ,
ib and ic are the phase currents and ea , eb and ec are the phase back-EMFs.
On DRV8312 DMC board, the terminal voltages are measured by taking
ground as the referenced point. The line voltage Vab of the terminal voltages
can be calculated as:

d(ia − ib )
Vab = R(ia − ib ) + L + ea − eb (4.4)
dt

Similarly, terminal line voltages Vbc and Vca are shown below:

d(ib − ic )
Vbc = R(ib − ic ) + L + eb − ec (4.5)
dt
d(ic − ia )
Vca = R(ic − ia ) + L + ec − ea (4.6)
dt
From equations above, the line voltages can be calculated based on the
terminal voltages measured with respect to ground, so the requirement for
motor neutral voltage is eliminated.

40
 We can calculate the line voltage differences based on the line voltages
equations above. The line voltage difference Vbcab can be calculated as:
Vbcab = Vbc − Vab
d(2ib − ia − ic ) (4.7)
= R(2ib − ia − ic ) + L + 2eb − ea − ec
dt
Similarly, Vabca and Vcabc can be calculated as:
Vabca = Vab − Vca
d(2ia − ib − ic ) (4.8)
= R(2ia − ib − ic ) + L + 2ea − eb − ec
dt
Vcabc = Vca − Vbc
d(2ic − ia − ib ) (4.9)
= R(2ic − ia − ib ) + L + 2ec − ea − eb
dt
If the commutation state pointer is equal to 0, the stator current flows from
phase A to phase B. Because phase C is open, the phase C current ic = 0.
Phase A current ia and phase B current ib have the same magnitude but
opposite. From Figure 2.1 [23], phase A back-EMF ea and phase B back-EMF
eb also have the same magnitude but opposite. The relationships can be shown
as:
i a + ib = 0 (4.10)

ea + eb = 0 (4.11)

When the commutation state pointer is equal to 0, the line voltage differ-
ence Vcabc can be calculated as:

Vcabc = Vca − Vbc = 2ec − ea − eb = 2ec (4.12)

From the equation above, the line voltage difference Vcabc is equal to an
amplified phase C back-EMF value and the gain is two. So we can use line
voltage differences to estimate zero-crossing points of back-EMFs. Theoreti-
cally, when the commutation state pointer is equal to 0, Vcabc starts from the
maximum positive value and goes to the minimum negative value. Because
the back-EMF waveform changes linearly, the zero-crossing point will occur in
the middle of the commutation state 0. As each state contains 60 electrical
degrees, a 30◦ offset exists between the back-EMFs’ zero-crossing points and
the current commutation instants.

41
 Similarly, when the commutation state pointer is equal to 3, the stator
current flows from phase B to phase A and phase C is open. Vcabc has the
same expression as in commutation state 0. The difference is that Vcabc starts
from the minimum negative value and goes to the maximum positive value.
In commutation state 1, the stator current flows from phase A to phase C,
and phase B is open. In commutation state 4, the stator current flows from
phase C to phase A and phase B is open. The phase B current ib = 0. Phase A
current ia and phase C current ic are equal and opposite. Phase A back-EMF
ea and phase C back-EMF ec are also equal and opposite. The line voltage
difference Vbcab can be calculated as:

Vbcab = Vbc − Vab = 2eb − ea − ec = 2eb (4.13)

In commutation state 2, the stator current flows from phase B to phase

C, and phase A is open. In commutation state 5, the stator current flows
from phase C to phase B, and phase A is open. The phase A current ib = 0.
Phase B current ib and phase C current ic are equal and opposite. Phase B
back-EMF eb and phase B back-EMF ec are also equal and opposite. The line
voltage difference Vabca can be calculated as:

Vabca = Vab − Vca = 2ea − eb − ec = 2ea (4.14)

To validate the proposed sensorless method, the relationship between the

line voltage differences and back-EMFs is verified in Matlab/Simulink envi-
ronment. For a real BLDC motor, the back-EMF waveform shape cannot be
ideal trapezoidal, which is the reason why the back-EMF voltage constant is
less than the torque constant for the BLDC motor used in the experimental
environment. The BLDC motor model for simulation has ideal trapezoidal
back-EMFs, and we choose the the torque constant to be same as the back-
EMF voltage constant instead of the one in Table 3.1. Except for the torque
constant, the BLDC motor in the Simulink model has the same parameters in
Table 3.1.
The line voltage differences are represented by the expressions of back-
EMFs to validate the zero-crossing points match. Figure 4.3 shows the simu-
lation results of the line voltage differences and back-EMFs waveforms. From

42
 20
2ea − eb − ec

Voltage (V)
ea
10
0
−10
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
20
Voltage (V)

2eb − ea − ec
eb
10
0
−10
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
20
Voltage (V)

2ec − ea − eb
ec
10
0
−10
0.6 0.62 0.64 0.66 0.68 0.7
time (s)

Figure 4.3: Simulations results of the line voltage differences and back-EMFs
waveforms at 700 rpm with 0.04 Nm load

the simulation results, we can see that at the zero-crossing points, the corre-
sponding line voltage difference is an amplified back-EMF, and the line voltage
differences have the same zero-crossing points as the back-EMFs.
In each state, one line voltage difference is applied to estimate the zero-
crossing points of the corresponding back-EMF waveform. A combined com-
mutation function is applied to estimate all the zero-crossing points of back-
EMFs as shown in Table 4.2. Table 4.2 also shows the polarity of the zero-
crossing points.
The combined commutation function can be shown in Figure 4.4. The
simulation results show that the combined commutation function has the same
zero-crossing points as the back-EMFs, which verifies the correctness of the
combined commutation function. The combined commutation function will be
applied in the experiment environment.

43
 Commutation States Commutation Function Polarity of Zero-crossing Points
0 Vcabc +Vmax to −Vmin
1 Vbcab −Vmin to +Vmax
2 Vabca +Vmax to −Vmin
3 Vcabc −Vmin to +Vmax
4 Vbcab +Vmax to −Vmin
5 Vabca −Vmin to +Vmax

Table 4.2: Commutation sequence for the method based on line voltage differ-
ences
Vcommtation (V)

5
0
−5
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
5
ea (V)

0
−5
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
5
eb (V)

0
−5
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
5
ec (V)

0
−5
0.6 0.62 0.64 0.66 0.68 0.7
time (s)

Figure 4.4: Simulations results of the combined commutation function and

back-EMFs waveforms at 700 rpm with 0.04 Nm load

4.3 Voltage Measurement

ADC modules convert the analog signals to digital signals. The analog input
range of the ADC module is from 0 V to 3.3 V but BLDC motor terminal
voltages are much higher than 3.3 V, so a voltage resistor divider circuit is

44
used to convert the terminal voltage to fit the analog input range of ADC
module.
Because of the inverter switching, BLDC motor terminal voltages contain
some high frequency harmonics. In order to eliminate the harmonics, a first-
order low pass filter is included in the voltage resistor divider circuit as shown
in Figure 4.5.

Vin

R1

Vout

R2 C

Figure 4.5: Voltage resistor divider circuit

The transfer function of the low pass filter is:

Vout R2
= (4.15)
Vin R1 + R2 + R1 R2 Cs

where C is the low pass filter capacitance, R1 and R2 are the voltage divider
resistances, Vout is the filtered waveform of Vin .
Table 4.3 shows the parameters of the motor terminal voltage sensing cir-
cuit.

4.4 Phase Delay Compensation

From the analysis above, line voltage differences can be used to estimate the
zero-crossing points of back-EMFs. Because a 30◦ offset exists between the

45
 Parameter Values
R1 95.3 kΩ
R2 4.99 kΩ
C 0.047 µF

Table 4.3: Terminal voltage sensing circuit specifications

back-EMFs’ zero-crossing points and the current commutation instants, a 30◦

phase shift should be made to ensure the correct operation of the motor.
However, because of the phase delay caused by the low pass filter, the actual
offset between the measured zero-crossing point of back-EMFs and current
commutation instants is not exactly 30◦ . The delayed angle caused by the
first-order low pass filter can be calculated as:
R1 · R2 · C · ωe
θ = − tan−1 ( ) (4.16)
R1 + R2
2π
ωe = (4.17)
T
where T is one electrical period and ωe is the synchronous speed of the BLDC
motor.
Due to the delayed angle caused by the low pass filter, the electrical degree
offset φ between the zero-crossing points of back-EMFs and current commu-
tation instants should be modified as

φ = 30◦ − θ (4.18)

Figure 4.6 shows the software flowchart of the proposed sensorless commu-
tation method with phase delay compensation. In order to calculate the φ in
software, the electrical period needs to be calculated. As the current electrical
period is unknown, the difference between two adjacent electrical periods is
small enough to ignore, so the last electrical period T is applied to calculate
Tdelay . At the beginning of state 5, T will be calculated based on the time
difference between the current time and the beginning time of state 5 in the
last electrical period. The phase delay time will be calculated and then used
in the next electrical period:
φ
Tdelay = ·T (4.19)
360
46
where T is the last electrical period and φ is the phase delay angle calculated
at the end of last electrical period.
For each commutation state, the time counter will start after detecting the
zero-crossing point of the corresponding line voltage difference. When the time
counter is equal to Tdelay , this point is treated as the commutation instant and
the BLDC motor will switch to the next state.
As mentioned in Chapter 2, the relationship between electrical frequency
and mechanical frequency is
p
fe = fm (4.20)
2
Nrpm
fm = (4.21)
60
where p is the number of poles, fm is the mechanical frequency, fe is the
electrical frequency and Nrpm is rotor speed in rpm.
The electrical period T from the commutation method will be applied to
estimate Nrpm as shown in equation below:

2
Nrpm = 60fm = 60 · fe
p
(4.22)
120
=
Tp
From the analysis above, the delayed angle caused by the low pass filter
has a positive relationship with the motor rotor speed. If the BLDC motor
speed is 1000 rpm, the delayed angle caused by the low pass filter is 5.33◦ using
the equation below:
R1 · R2 · C · ωe
θ = tan−1 ( )
R1 + R2
(4.23)
R1 · R2 · C Nrpm
= tan−1 ( · πp )
R1 + R2 60
When the speed is low or medium, the delayed angle won’t cause a com-
mutation mistake. But when motor speed is high, the delayed angle will cause
problems. Some compensation needs to be done to guarantee the correct op-
eration of BLDC motors. In this thesis, F28035 C2000 controller will be used
to compensate the phase delay φ, so the current commutation instants can be
in phase with the zero-crossing points of line voltage differences.

47
 If If If If If If
state=0 state=1 state=2 state=3 state=4 state=5

Read Tdelay Read Tdelay Read Tdelay Read Tdelay Read Tdelay Calculate Tdelay

Yes Yes Yes Yes

Yes Yes
If Vcabc > 0 If Vbcab < 0 If Vabca > 0 If Vcabc < 0 If Vbcab > 0 If Vabca < 0

No No No No No No

48
Time counter++
No

If time counter< Tdelay

Yes

Commutation instants

Figure 4.6: Software flowchart of the sensorless commutation method based on line voltage differences
4.5 Starting Procedure
For BLDC control methods with Hall-effect sensors or Quadrature Encoder
signals, it is not a problem for a BLDC motor to start up. For back-EMF
based methods like the method used in this chapter, back-EMFs’ values are
needed to achieve the position estimation. However, back-EMFs’ values are
zero during standstill, and they are also very small when the motor speed is
low.
Normally, a starting procedure is required for the methods based on back-
EMFs. In this chapter, the BLDC motor will be started blindly using the
conventional open-loop starting method [8].
During the first stage, any two phases of stator windings will be energized.
In this thesis, the stator current will firstly flow from phase A to phase B. The
stator currents will generate a flux whose amplitude and direction are both
constant. The stator flux will attract the rotor to point in the same direction,
so the rotor is forced to align at a predefined position [8]. Then, the stator
current flows from phase A to phase C so the BLDC motor will be in the
next state. The stator flux will point to a new direction, so the rotor will be
attracted to move to the next position. During this open-loop starting mode,
the stator windings will be energized in a predefined order:

AB → AC → BC → BA → CA → CB

When the BLDC motor speeds up, back-EMFs will also increase. When the
zero-crossing points of back-EMFs can be stably detected, the control system
will switch to the sensorless mode.

4.6 Electrical Noises

The noises included in the calculated line voltage difference waveforms may
cause false zero-crossing points detection. The main noises are from the
chopped pulses generated by PWM. Spikes caused by the free-wheeling diode
conduction current are another source [6].

49
 Assuming that the BLDC motor is in commutation state 1, now it will
switch to state 2. During commutation state 1, the upper transistor of phase
A (T1 ) and the lower transistor of phase C (T6 ) are on. At the commutation
instant, T1 will be turned off and the upper transistor of phase B (T3 ) will
be turned on. The stator current continues to flow through the free-wheeling
diode of phase A and decreases.
The free-wheeling current will cause spikes in the calculated line voltage
difference waveforms. The spikes may cross the zero thus cause false zero-
crossing points detection.
According to the reference [6], the free-wheeling current during time is
much smaller than the phase delay time Tdelay , so the spike will disappear
before the back-EMFs cross zero.
The calculated line voltage difference waveforms also contains noises from
the sensing circuit. For DRV8312-C2 kit, each terminal voltage is sensed with
a voltage resistor divider and a first order low pass filter. The sensing circuit
may bring noises to the measured terminal voltages so it will also affect the
line voltage differences.
Because the virtual neutral voltage is not used in this method, it will reduce
the affect of electrical noises. The low pass filter of the terminal voltage sensing
circuit can filter out the chopped pulses generated by PWM, but the calculated
line voltage difference waveforms still contain some noises which may cause
false zero-crossing points detection. In order to minimize the influence of the
noises, a counter can be used to calculate how many times the measured line
voltage difference value crosses zero [38]. The noises only remain for a very
short time. If the counter reaches a predefined value, this point is treated as
the real zero-crossing point of the line voltage difference. Otherwise, it will be
treated as noises and will not trigger the commutation to the next state.

4.7 Proposed Sensorless Drive

Assuming that ω̂ is the estimated motor speed and ωref is the reference motor
speed, a proportional integral (PI) controller is applied to determine the duty

50
cycle of the PWM signals based on the difference between ω̂ and ωref . Figure
4.7 shows the proposed sensorless drive for speed control of a BLDC motor.
The measured three phase terminal voltages are the inputs of the line
voltage differences calculation block. The calculated line voltage differences
and the motor speed are the inputs of the commutation trigger block. At the
beginning of each electrical period, the phase delay time will be calculated,
and it will be used in this electrical period.

Duty Cycle PWM Signals BLDC Motor

ωref + BLDC PWM Three-phase
PI inverter
− Drive
ω̂
Vcabc , Vbcab , Vabca
Commutation State Commutation
Method in
Figure 4.6

Speed Estimation
using (4.22)

Figure 4.7: Proposed sensorless BLDC motor drive based on line voltage dif-
ferences

4.8 Experimental Results

In this section, we presents the experiment results of the proposed sensorless
BLDC motor drive based on line voltage differences at different speeds, and
the effectiveness of the compensation method is also verified and analyzed.

4.8.1 Experiment Results at 600 rpm

Figure 4.8 shows the measured line voltage differences at 600 rpm. Figure 4.9
shows the speed response at 600 rpm, and the error is calculated using the
equation below:
|ω̂ − ωref |
error(%) = × 100% (4.24)
ωref

51
 5

Vabca (V)
0

−5
5 5.02 5.04 5.06 5.08 5.1
time (s)
5
Vbcab (V)

−5
5 5.02 5.04 5.06 5.08 5.1
time (s)
5
Vcabc (V)

−5
5 5.02 5.04 5.06 5.08 5.1
time (s)

Figure 4.8: Experiment results of line voltage differences at 600 rpm

700
ωref
ω̂
650
Speed (rpm)

600

550

500
5 5.5 6 6.5
time (s)

5
4
Error (%)

3
2
1
0
5 5.5 6 6.5
time (s)

Figure 4.9: Speed response at 600 rpm

52
4.8.2 Experiment Results at 1000 rpm
Figure 4.10 shows the measured line voltage differences at 1000 rpm. Figure
4.11 shows the speed response at 1000 rpm.
Based on the results in subsection 4.8.1 and subsection 4.8.2, we can see
that the line voltage differences waveforms are symmetric, and the speed of
the BLDC motor can remain stable with negligible error.
When the BLDC motor runs in open-loop commutation mode, there is a
big acoustic noise from the motor due to its inaccuracy commutation. During
the sensorless commutation mode, the acoustic noise is reduced.

10
Vabca (V)

−10
10.9 10.91 10.92 10.93 10.94 10.95 10.96
time (s)
10
Vbcab (V)

−10
10.9 10.91 10.92 10.93 10.94 10.95 10.96
time (s)
10
Vcabc (V)

−10
10.9 10.91 10.92 10.93 10.94 10.95 10.96
time (s)

Figure 4.10: Experiment results of line voltage differences at 1000 rpm

4.8.3 Phase Delay Compensation

For the trapezoidal control of BLDC motors, back-EMFs are required to be in
phase with the phase current. In this section, we will make comparison of the
proposed sensorless drive with and without phase delay compensation. From
the experiment results, the phase delay compensation method does not make
a big difference if the BLDC motor works in low or medium speeds. However,

53
 1100
ωref
ω̂
1050

Speed (rpm) 1000

950

900
10 10.5 11 11.5
time (s)

5
4
Error (%)

3
2
1
0
10 10.5 11 11.5
time (s)

Figure 4.11: Speed response at 1000 rpm

when the BLDC motor speed is high, the effectiveness of the compensation
method will be obvious.
In this section, we presents the experiment results at 1800 rpm. Because
three phases are symmetric, the experiment results of phase A is analyzed
to show the correctness of the compensation method. Figure 4.12 shows the
relationship between phase A current ia and Vabca . We can see that the com-
mutation points of ia lags 30◦ behind the zero-crossing points of Vabca . Figure
4.13 shows the speed response at 1800 rpm.
For the sensorless drive without phase delay compensation, the motor stalls
when its speed is over 1800 rpm, because of the significant phase delay error
(over 9.53◦ ) caused by the low pass filter. To verify the effectiveness of the
compensation method, Figure 4.14 shows the experiment results of the rela-
tionship between the ideal commutation signal (Hall-effect sensor A output)
and the estimated commutation signal (virtual Hall-effect sensor A output
obtained from the sensorless commutation method) at 1800 rpm.
From Figure 4.14, we can see that the estimated commutation signal lags

54
 20 1

15

10

5
Vabca (V)

i a (A)
0 0

−5

−10

−15

−20 −1
10.8 10.805 10.81 10.815 10.82
time (s)

Figure 4.12: Phase A current and Vabca at 1800 rpm

1900
ωref
ω̂
1850
Speed (rpm)

1800

1750

1700
10 10.5 11 11.5
time (s)

5
4
Error (%)

3
2
1
0
10 10.5 11 11.5
time (s)

Figure 4.13: Speed response at 1800 rpm

55
 Before compensation
1 4

Estimated signal (V)

Ideal signal (V)

2
0

0
0 0.005 0.01 0.015 0.02
time (s)
After compensation
1 4
Estimated signal (V)

Ideal signal (V)

2
0

0
0 0.005 0.01 0.015 0.02
time (s)

Figure 4.14: Ideal commutation signal and estimated commutation signal at

1800 rpm

behind the ideal commutation signal and the phase delay is significant. After
phase delay compensation, the estimated commutation signal matches well
with the ideal commutation signal.
Before phase compensation, the acoustic noise is noticeable due to inaccu-
racy commutation. After phase compensation, the acoustic noise is reduced.
Based on the analysis above, we can conclude that phase delay compen-
sation is necessary for high speed region. With phase delay compensation,
the high speed capability of BLDC motors can be improved, and the acoustic
noise from the motor will be reduced. The estimated commutation signal is
precise enough to replace the ideal commutation signal, so it can be applied
to realize sensorless commutation.

56
4.9 Summary
In this chapter, we make a detailed analysis on the six-step commutation
of the BLDC motors. Then we discuss the sensorless commutation method
based on line voltage differences, and it is validated by the simulation results.
The sensorless commutation method with phase delay compensation is further
implemented in the experimental environment.

57
Chapter 5

Sensorless BLDC Motor Drive

Based on Disturbance Observer

In Chapter 4, a sensorless control method based on terminal line voltage differ-

ences is introduced and verified through experiment. However, this sensorless
method requires a 30◦ phase shift to acquire the correct commutation instants,
which lowers the accuracy of the commutation method. The complexity of the
sensorless control method is also increased. In this chapter, a sensorless control
method based on the disturbance observer will be discussed, and this method
does not require a 30◦ phase shift [39].

5.1 Disturbance Observer

The disturbance observer has been widely applied in the motion control area
because of its ability to implement disturbance rejection and compensate the
plant uncertainties. The Disturbance Estimating Filter (DEF) disturbance
observer is introduced in [33].
The DEF disturbance observer is proposed based on the two-degree-of-
freedom controller. For the two-degree-of-freedom controller, one degree-of-
freedom is to implement disturbance rejection, and another is to design the ref-
erence response. The normal structure of DEF disturbance observer is shown
in Figure 5.1 [33]. The actual plant transfer function is P 0 (s) and the nominal
plant transfer function is P (s).
ui is the input signal and d is the disturbance signal. The actual input

58
 d
+
ui y
+ u P 0 (s)

− + n
−1
P (s) + −
uf

dˆ
Q(s)

Figure 5.1: Structure of DEF disturbance observer

signal u is the superposition result of ui and d. The output of the actual plant
model is y, and n is the measurement noise. The function of y is shown as:

y = u · P 0 (s) = (ui + d) · P 0 (s) (5.1)

The inverse of nominal plant transfer function is P −1 (s). The superposition

result of y and n will be passed to P −1 (s) to get the estimated input of the
actual plant model uf . The calculation function of uf is shown below:

uf = (y − n) · P −1 (s)
(5.2)
= (ui + d) · P 0 (s) · P −1 (s) − n · P −1 (s)

If there is no modeling error, P 0 (s) and P (s) are equal as:

P 0 (s) = P (s) (5.3)

uf can be simplified as:

uf = ui + d − n · P −1 (s) (5.4)

The difference between ui and uf is called the lumped disturbance [5].

From the equation above, we can see that the lumped disturbance contains
the measurement noise n, which may affect the performance of the disturbance

59
observer. The frequency of the measurement noise is usually high, and the
frequency of disturbance signal is low [5].
A low pass filter Q(s) is required in the disturbance observer to filter out
the measurement noise. For the Q(s), its magnitude should be designed with
the following characteristics [33]:

|Q(s)| ≈ 1, in the low frequency range (5.5)

|Q(s)| ≈ 0, in the high frequency range (5.6)

The filtered lumped disturbance dˆ is the estimated disturbance signal by the

disturbance observer.
dˆ = d · Q(s) (5.7)

The estimated disturbance signal by the disturbance observer dˆ can be fed

back to implement disturbance rejection.

5.2 Sensorless Commutation Method Based on

Disturbance Observer
Figure 5.2 shows the simulation results of the back-EMF differences and phase
currents. From the simulation results, we can see that the zero-crossing points
of the back-EMF differences (eab , ebc and eca ) are same as the commutation
instants of the phase currents, so back-EMF differences can be directly used
in phase currents’ commutation. In this section, a sensorless control method
based on the work in [38][39] will be discussed.
As mentioned in Chapter 2, the three terminal voltages of BLDC motor
can be shown below:
dia
Van = Ria + L + ea (5.8)
dt
dib
Vbn = Rib + L + eb (5.9)
dt
dic
Vcn = Ric + L + ec (5.10)
dt
where Van , Vbn and Vcn are the phase voltages referenced to the motor neutral
voltage, R is the phase stator resistance, L is the phase stator inductance, ia ,
ib and ic are the phase currents, and ea , eb and ec are the phase back-EMFs.

60
 2 4
1 2

eab (V)
ia (A) 0 0
−1 −2
−2 −4
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
2 4
1 2

ebc (V)
ib (A)

0 0
−1 −2
−2 −4
0.6 0.62 0.64 0.66 0.68 0.7
time (s)
2 4
1 2

eca (V)
ic (A)

0 0
−1 −2
−2 −4
0.6 0.62 0.64 0.66 0.68 0.7
time (s)

Figure 5.2: Simulations results of the back-EMF differences and phase currents
at 700 rpm with 0.04 Nm load

The three line voltages of BLDC motor can be calculated from the terminal
voltages as:
d(ia − ib )
Vab = R(ia − ib ) + L + ea − eb (5.11)
dt
d(ib − ic )
Vbc = R(ib − ic ) + L + eb − ec (5.12)
dt
d(ic − ia )
Vca = R(ic − ia ) + L + ec − ea (5.13)
dt
where Vab , Vbc and Vca are the three line voltages.
Laplace transform results of the line voltages are shown below:

Vab (s) = R(iab (s)) + Ls(iab (s)) + eab (s) (5.14)

Vbc (s) = R(ibc (s)) + Ls(ibc (s)) + ebc (s) (5.15)

Vca (s) = R(ica (s)) + Ls(ica (s)) + eca (s) (5.16)

61
 where Vab (s), Vbc (s) and Vca (s) are the Laplace transforms of the three line
voltages, eab (s), ebc (s) and eca (s) are the Laplace transforms of the back-EMF
differences and iab (s), ibc (s) and ica (s) are the Laplace transforms of the line
currents.
For each equation above, we take the line voltage as the inputs, the back-
EMF difference as the disturbances and the line current as the outputs, so we
can build different models for the three equations.
The disturbance observer structure can be applied to estimate distur-
bances. As shown in Figure 5.3 [38], The actual plant transfer function is
P 0 (s). −êab , −êbc and −êca are the estimated back-EMF differences. P (s) is
the nominal plant transfer function. The inverse of the nominal plant transfer
function is P −1 (s). n is the line currents measurement noise which can be
filtered out by the low pass filter. Q(s) is the low pass filter transfer function.

−eab /−ebc /−eca

+
Vab /Vbc /Vca Iab /Ibc /Ica
+ P 0 (s)

− + n
−1
P (s) + −

−êab /−êbc /−êca

Q(s)

Figure 5.3: Disturbance observer structure based back-EMF differences esti-

mation method

For the estimated disturbances, they are not used to implement distur-
bance rejection. Instead, estimated disturbances are used as the inputs of the
commutation algorithms, and their zero-crossing points will be commutation
instants of the BLDC motor.
In commutation states 0 and 3, the zero-crossing points of êbc are the

62
commutation instants of the BLDC motor. In commutation states 1 and 4,
the zero-crossing points of êab are the commutation instants of the BLDC
motor. In commutation states 2 and 5, the zero-crossing points of êca are the
commutation instants of the BLDC motor.

5.3 Design of the Disturbance Observer

For the design of a disturbance observer, the most important part is to design
the low pass filter Q(s). Q(s) used in this chapter is designed based on the
parameters of the DRV8312 DMC board.
As mentioned in Chapter 4, the terminal voltage of each phase is measured
by a voltage resistor divider circuit and a first-order low pass filter. The
transfer function of the terminal voltage measurement circuit is:
Vout R2
T (s) = = (5.17)
Vin R1 + R2 + R1 R2 Cs
where R1 and R2 are the divider resistances, C is the low pass filter capaci-
tance, Vin is the input signal and Vout is the filtered waveform of Vin .
The relationship between measured line voltages and real line voltages are
shown in equations below:

V̂ab (s) = T (s) · Vab (s) (5.18)

V̂bc (s) = T (s) · Vbc (s) (5.19)

V̂ca (s) = T (s) · Vca (s) (5.20)

where V̂ab (s), V̂bc (s) and V̂ca (s) are the measured line voltages, Vab (s), Vbc (s)
and Vca (s) are the Laplace transforms of the three line voltages.
The cutoff frequency fcut of the terminal voltage measurement circuit can
be shown below:
R1 + R2
fcut = (5.21)
2πR1 R2 C
With the parameters in Table 4.3, the transfer function and cutoff fre-
quency fcut can be calculated as:
Vout 223.3
T (s) = = (5.22)
Vin s + 4487.1

63
 fcut = 714.2 Hz (5.23)

The low pass filter Q(s) is designed to be the same as the transfer function
of the terminal voltage measurement circuit.

223.3
Q(s) = T (s) = (5.24)
s + 4487.1
For the measurement of three phase currents, three differential amplifier
circuits are used to measure three phase currents. The per unit range of -
1.0 to 1.0 is corresponding to the real current from −8.65 A to 8.65 A. The
relationship can be shown in equations below:

Iab (s) = K · Iˆab (5.25)

Ibc (s) = K · Iˆbc (5.26)

Ica (s) = K · Iˆca (5.27)

K = 8.65 A (5.28)

where iab (s), ibc (s) and ica (s) represent line currents. Iˆab , Iˆbc and Iˆca are per
unit line currents.

−eab /−ebc /−eca

V̂ab /V̂bc /V̂ca + Iˆab /Iˆbc /Iˆca

T −1 (s) P 0 (s) K −1
+ u

+
− +
TT−1
−1(s)
(s) KP −1 (s) n
−

−êab /−êbc /−êca

Q(s)

Figure 5.4: Modified disturbance observer structure considering voltage and

current sensing circuits

The modified disturbance observer structure is shown in Figure 5.4 [38].

The estimated back-EMF differences −êab , −êbc and −êca can be shown as

64
[38]:

−êab = −V̂ab · T −1 (s) · Q(s) + K · P −1 (s) · Q(s) · (Iˆab − n) (5.29)

−êbc = −V̂bc · T −1 (s) · Q(s) + K · P −1 (s) · Q(s) · (Iˆbc − n) (5.30)

−êca = −V̂ca · T −1 (s) · Q(s) + K · P −1 (s) · Q(s) · (Iˆca − n) (5.31)

The per unit line currents Iˆab , Iˆbc and Iˆca can be shown as:

Iˆab = [V̂ab · T −1 (s) − eab ] · P 0 (s) · K −1 (5.32)

Iˆbc = [V̂bc · T −1 (s) − ebc ] · P 0 (s) · K −1 (5.33)

Iˆca = [V̂ca · T −1 (s) − eca ] · P 0 (s) · K −1 (5.34)

Then −êab , −êbc and −êca can be modified as [38]:

−êab = −V̂ab · T −1 (s) · Q(s) + P −1 (s) · P 0 (s) · V̂ab · T −1 (s) · Q(s)

(5.35)
+ [−P −1 (s) · P 0 (s) · eab − nKP −1 (s)] · Q(s)

−êbc = −V̂bc · T −1 (s) · Q(s) + P −1 (s) · P 0 (s) · V̂bc · T −1 (s) · Q(s)

(5.36)
+ [−P −1 (s) · P 0 (s) · ebc − nKP −1 (s)] · Q(s)
−êca = −V̂ca · T −1 (s) · Q(s) + P −1 (s) · P 0 (s) · V̂ca · T −1 (s) · Q(s)
(5.37)
+ [−P −1 (s) · P 0 (s) · eca − nKP −1 (s)] · Q(s)
If there is no modeling error, P 0 (s) and P (s) are equal as:

1
P 0 (s) = P (s) = (5.38)
Ls + R

where R is the phase stator resistance and L is the phase stator inductance.
The low pass filter can filter out the measurement noise n. Then, −êab ,
−êbc and −êca can be simplified as [38]:

−êab = −eab · Q(s) (5.39)

−êbc = −ebc · Q(s) (5.40)

−êca = −eca · Q(s) (5.41)

65
 Thus, the modified disturbance observer structure can be applied to esti-
mate the back-EMF differences:

êab = V̂ab − K · P −1 (s) · Q(s) · Iˆab

êbc = V̂bc − K · P −1 (s) · Q(s) · Iˆbc (5.42)

êca = V̂ca − K · P −1 (s) · Q(s) · Iˆca

5.4 Phase Delay Compensation

Although the sensorless control method based on the disturbance observer
does not require a 30 electrical degree shift, phase delay will be introduced
because of low pass filters applied in the designed disturbance observer struc-
ture. Figure 5.5 shows the waveforms of the real back-EMF differences and
the estimated back-EMF differences at a speed of 1000 r/min. From the sim-
ulation results, we can see that the estimated back-EMF differences lag the
real back-EMF differences. Figure 5.6 shows the condition at a speed of 2000
r/min. Compared with Figure 5.5, we can see that the phase delay becomes
larger at a higher speed.
Ideally, the zero-crossing points of back-EMF differences can be treated
as the commutation instants of a BLDC motor. But due to the phase delay
introduced by low pass filters, phase delay compensation should be done to
ensure the correct commutation instants.
Based on the BLDC motor electrical period, the phase delay introduced
by low pass filters can be calculated as:
R1 · R2 · C · ωe
θ = tan−1 ( )
R1 + R2
(5.43)
R1 · R2 · C 2π
= tan−1 ( · )
R1 + R2 T
where θ is the phase delay, T is one electrical period, R1 and R2 are the divider
resistances and C is the low pass filter capacitance.
For each of the six commutation states, one back-EMF difference is chosen
to identify the commutation instant. The back-EMF difference will start from
a maximum positive value or a minimum negative value, then it will reach the

66
 1
êab

Voltage (V)
0.5 eab

0
−0.5
0.6 0.605 0.61 0.615 0.62
time (s)
1
Voltage (V)

êbc
0.5 ebc

0
−0.5
0.6 0.605 0.61 0.615 0.62
time (s)
1
Voltage (V)

êca
0.5 eca

0
−0.5
0.6 0.605 0.61 0.615 0.62
time (s)

Figure 5.5: Simulation results of the estimated back-EMF differences and the
real back-EMF differences at 1000 rpm with 0.04 Nm load

zero-crossing point. Theoretically, the back-EMF difference shape is trape-

zoidal, so we can assume that each chosen back-EMF difference varies linearly
with electrical angle during its commutation state.
To compensate the phase delay, threshold values ,which trigger commu-
tation instants, should be set based on the calculated phase delay instead of
zero [38]. With the similar triangles formula, the threshold values used in the
chapter to compensate the phase delay can be calculated as [38]:

θ
Vp = · Vpmax (5.44)
60◦
θ
Vn = · Vpmin (5.45)
60◦
where Vp is the threshold value when the back-EMF difference varies from the
maximum positive value to zero, Vn is the threshold value when the back-EMF
difference varies from the minimum negative value to zero, θ is the calculated

67
 1 êab

Voltage (V)
eab
0.5
0
−0.5
0.6 0.605 0.61 0.615 0.62
time (s)
Voltage (V)

1 êbc
ebc
0.5
0
−0.5
0.6 0.605 0.61 0.615 0.62
time (s)
Voltage (V)

1 êca
eca
0.5
0
−0.5
0.6 0.605 0.61 0.615 0.62
time (s)

Figure 5.6: Simulation results of the estimated back-EMF differences and the
real back-EMF differences at 2000 rpm with 0.04 Nm load

phase delay, Vpmax is the maximum positive value of the back-EMF difference
and Vpmin is the minimum negative value of the back-EMF difference.
Figure 5.7 shows the software flowchart of the sensorless commutation
method based on the disturbance observer. At the beginning of commuta-
tion states 0, 2 and 4, the estimated value of the chosen back-EMF difference
is regarded as the minimum negative value Vpmin . At the beginning of commu-
tation states 1, 3 and 5, the estimated value of the chosen back-EMF difference
is regarded as the maximum positive value Vpmax . Also, the phase delay θ is
calculated at this point based on the motor speed. Then the threshold values
(Vp and Vn ) can be calculated based on the equations above.
In order to minimize the influence of noises, a counter is applied to calculate
how many times the estimated back-EMF difference value crosses the threshold
value. When the counter reaches a predefined value Tnoises , this point is treated
as the commutation instant and the BLDC motor will switch to the next state.

68
 If If If If If If
state=0 state=1 state=2 state=3 state=4 state=5

Disturbance Disturbance Disturbance Disturbance Disturbance Disturbance

observer observer observer observer observer observer

êbc êab êca êbc êab êca

Calculate Vn Calculate Vp Calculate Vn Calculate Vp Calculate Vn Calculate Vp

Yes Yes Yes Yes

Yes Yes
If êbc < Vn If êab > Vp If êca < Vn If êbc > Vp If êab < Vn If êca > Vp

No No No No No No

69
Time counter++
No

If time counter< Tnoises

Yes

Commutation instants

Figure 5.7: Software flowchart of the sensorless commutation method based on the disturbance observer structure
5.5 Proposed Sensorless Drive
Figure 5.8 shows the proposed sensorless drive for speed control of a BLDC
motor. A proportional integral (PI) controller is applied to determine the duty
cycle of the PWM signals based on the difference between the measured motor
speed ω and the reference motor speed ω ∗ .
The measured three phase currents and the three phase terminal voltages
are the inputs of the disturbance observer. The estimated back-EMF differ-
ences êab , êbc and êca together with the measured motor speed are the inputs
of the commutation trigger block. The commutation sequence of the proposed
method is shown in Table 5.1.
Because the proposed method is based on back-EMF differences, a starting
procedure is required to speed up the BLDC motor to a higher speed. In this
chapter, the conventional open-loop starting method, as described in Chapter
4, is applied to blindly start the BLDC motor.

Commutation States Commutation Instants

0 êbc = Vn
1 êab = Vp
2 êca = Vn
3 êbc = Vp
4 êab = Vn
5 êca = Vp

Table 5.1: Commutation sequence for the method based on disturbance ob-
server

5.6 Experimental Results

In this section, we present the experiment results of the proposed sensorless
BLDC motor drive based on the disturbance observer structure at different
speeds. Compensation of the phase delay introduced by the disturbance ob-
server structure is also discussed.

70
 Duty Cycle PWM Signals BLDC Motor
ωref + Three-phase
PI BLDC PWM
− Drive Inverter
ω̂
Iˆab , Iˆbc , Iˆca V̂ab , V̂bc , V̂ca
Commutation êab Disturbance
Method in êbc Observer
Commutation State Figure 5.7 êca using (5.42)

Speed Estimation
using (4.22)

Figure 5.8: Proposed sensorless BLDC motor drive based on the disturbance
observer structure

5.6.1 Experiment Results at 600 rpm

Figure 5.9 shows the estimated back-EMF differences at 600 rpm. Figure 5.10
shows the speed response at 600 rpm.

5
êab (V)

−5
5.75 5.8 5.85
time (s)
5
êbc (V)

−5
5.75 5.8 5.85
time (s)
5
êca (V)

−5
5.75 5.8 5.85
time (s)

Figure 5.9: Experiment results of estimated back-EMF differences at 600 rpm

71
 700
ωref
ω̂
650

Speed (rpm) 600

550

500
5 5.5 6 6.5
time (s)

5
4
Error (%)

3
2
1
0
5 5.5 6 6.5
time (s)

Figure 5.10: Speed response at 600 rpm

5.6.2 Experiment Results at 1000 rpm

Figure 5.11 shows the estimated back-EMF differences at 1000 rpm. Figure
5.12 shows the speed response at 1000 rpm.
Based on the results in subsection 5.6.1 and subsection 5.6.2, we can see
that the estimated back-EMF differences waveforms are symmetric, and the
speed of the BLDC motor can remain stable with negligible error.

5.6.3 Phase Delay Compensation

In this section, we present the experiment results at 1500 rpm to verify the
effectiveness of the phase compensation method introduced in section 5.4.
Figure 5.13 shows the relationship between phase A current ia and êab .
We can see that the commutation points of ia are same with the zero-crossing
points of êab . Figure 5.14 shows the speed response at 1500 rpm.
Without phase compensation, the BLDC motor stalls sometimes when
it runs at 1500 rpm due to inaccurate commutation. Figure 5.15 shows the

72
 5

êab (V)
0

−5
10.9 10.91 10.92 10.93 10.94 10.95 10.96
time (s)
5
êbc (V)

−5
10.9 10.91 10.92 10.93 10.94 10.95 10.96
time (s)
5
êca (V)

−5
10.9 10.91 10.92 10.93 10.94 10.95 10.96
time (s)

Figure 5.11: Experiment results of estimated back-EMF differences at

1000 rpm

1100
ωref
ω̂
1050
Speed (rpm)

1000

950

900
10 10.5 11 11.5
time (s)

5
4
Error (%)

3
2
1
0
10 10.5 11 11.5
time (s)

Figure 5.12: Speed response at 1000 rpm

73
 10 1

5
êab (V)

i a (A)
0 0

−5

−10 −1
10.88 10.885 10.89 10.895 10.9
time (s)

Figure 5.13: Phase A current and êab at 1500 rpm

1600
ωref
ω̂
1550
Speed (rpm)

1500

1450

1400
10 10.5 11 11.5
time (s)

5
4
Error (%)

3
2
1
0
10 10.5 11 11.5
time (s)

Figure 5.14: Speed response at 1500 rpm

74
 Before compensation
1 4

Estimated signal (V)

Ideal signal (V)

0 2

−1 0
0.015 0.02 0.025 0.03 0.035 0.04
time (s)
Before compensation
1 4
Estimated signal (V)

Ideal signal (V)

0 2

−1 0
0.015 0.02 0.025 0.03 0.035 0.04
time (s)

Figure 5.15: Ideal commutation signal and estimated commutation signal at

1500 rpm

experiment results of the relationship between the ideal commutation signal

(Hall-effect sensor A output) and the estimated commutation signal (virtual
Hall-effect sensor A output obtained from the sensorless commutation method)
at 1500 rpm.
From Figure 5.15, we can see that the estimated commutation signal matches
well with the ideal commutation signal with phase compensation. For high
speed applications, it is necessary to compensate the phase delay caused by
the disturbance observer structure [38].
With phase compensation method, the sensorless commutation method can
get accurate commutation points, and the high speed range can be expanded.

75
5.7 Comparison with the Sensorless Drive based
on Line Voltage Differences
In this section, we make a comparison of the proposed sensorless BLDC motor
drive based on the disturbance observer structure and the sensorless drive
introduced in Chapter 4.
Based on the analysis above, both of these two sensorless drives can achieve
precise sensorless commutation. However, they all need compensation for the
phase delay caused by the low pass filter. The electrical frequency of the motor
must be smaller than the cutoff frequency fcut of the low pass filter, which is
714.2 Hz.
For the sensorless drive based on the line voltage differences, the maximum
allowable phase delay from the low pass filter is 30◦ (electrical frequency equals
412.5 Hz). In contrast, the maximum allowable phase delay is 60◦ (electrical
frequency equals 1237.4 Hz) for the sensorless BLDC motor drive based on the
disturbance observer structure. For the given motor drive board, we can see
that disturbance observer based sensorless drive allows the electrical frequency
of the motor to be up to the cutoff frequency of the low pass filter. Disturbance
observer based sensorless drive has a wider high speed range in comparison
with the sensorless drive based on the line voltage differences.
Both of these two sensorless drives are based on back-EMFs, so they all
have poor performance in the low speed range. Figure 5.16 shows Vabca and
êab waveforms at low speeds. We can see that the waveforms are distorted,
and we can not get accurate commutation points with either of the proposed
sensorless drives.
Figure 5.17 and Figure 5.18 show the simulation results of the speed re-
sponse at low speed (500 rpm) and high speed (1500 rpm) during loading tran-
sient, respectively. For 500 rpm, a step load (0.05 Nm) is applied at 0.3 s.
For 1500 rpm, a step load (0.05 Nm) is applied at 0.5 s. From the simulation
results, we can see that the speed response of the two sensorless drives are
similar. For both low and high speeds conditions, the motor speed decreases
after applying the load, and then it will keep on tracking the reference speed

76
 Vabca at 300 rpm
5

Vabca (V)
0

−5
5.8 6.3
time (s)
êab at 250 rpm
3
êab (V)

−3
5.8 6.3
time (s)

Figure 5.16: Vabca at 300 rpm and êab at 250 rpm

after the transient time.

5.8 Summary
In this chapter, we discuss a sensorless commutation method based on the dis-
turbance observer. The sensorless commutation method with is implemented
in the experimental environment, and it works well at different speeds. Com-
pared with sensorless commutation method based on line voltage differences,
disturbance observer based method has a wider high speed range, and it does
not require a 30◦ phase shift to get the correct commutation points.

77
 Speed response (line voltage differences)
1500

Speed (rpm)
1000

500

0
0 0.1 0.2 0.3 0.4 0.5
time(s)
Speed response (disturbance observer)
1500
Speed (rpm)

1000

500

0
0 0.1 0.2 0.3 0.4 0.5
time (s)

Figure 5.17: Speed response at 500 rpm during loading transient

Speed response (line voltage differences)

3000
Speed (rpm)

2000

1000

0
0 0.2 0.4 0.6 0.8 1
time(s)
Speed response (disturbance observer)
3000
Speed (rpm)

2000

1000

0
0 0.2 0.4 0.6 0.8 1
time (s)

Figure 5.18: Speed response at 1500 rpm during loading transient

78
Chapter 6

Conclusions

6.1 Summary of Research

This thesis focuses on the implementation of two sensorless control techniques
for a BLDC motor. Firstly, we discuss a sensorless control method, which uti-
lizes line voltage differences to estimate the zero-crossing points of back-EMFs.
The line voltage differences have the same zero-crossing points as back-EMFs,
which are verified by the simulation results. This method only requires three
motor terminal voltages without the need for the motor neutral voltage. This
sensorless control method is also implemented in the experimental environ-
ment, and a compensation method is applied to compensate the phase delay
introduced by the the low-pass filters in the terminal voltage sensing circuits.
Then we introduce a back-EMF differences based sensorless control method,
which does not require any phase shift to get the commutation instants. The
disturbance observer structure is applied to estimate the back-EMF differences.
Simulation results show that the estimated back-EMF differences matches well
with the real back-EMF differences except for a phase delay caused by the
low pass filter in the disturbance observer structure. This sensorless control
method with phase delay compensation is implemented in the experimental
environment.
From the experiment results, we conclude that both of these two sensorless
control techniques can be applied in commutating a BLDC motor. Phase delay
compensation methods are necessary for high speed operation. Disturbance
observer based method has a better high speed capability. Because both of

79
these two sensorless control techniques are based on back-EMFs, they have
limiting low speed capability. Based on the Simulink simulation results, both
of these two sensorless drives are able to work well even with load disturbance.

6.2 Future Work

In the past few decades, fuzzy logic controllers gained wide attentions and have
been applied in BLDC motor drives [32][34]. Fuzzy controller-based BLDC
motor drive has a better performance for the varying operating conditions of
the motor, such as motor parameters variation and load uncertainties [34].
For the future work, we will apply the fuzzy logic controller in the speed
control loop, together with the sensorless commutation techniques investigated
in this thesis, to build a combined BLDC sensorless control drive with better
performance.

80
Bibliography

[1] R. Becerra, T. Jahns, and M. Ehsani. Four-quadrant sensorless brushless

ecm drive. In Applied Power Electronics Conference and Exposition, 1991.
APEC’91. Conference Proceedings, 1991., Sixth Annual, pages 202–209.
IEEE, 1991.

[2] M. Bonfe and M. Bergo. A brushless motor drive with sensorless control
for commercial vehicle hydraulic pumps. In 2008 IEEE International
Symposium on Industrial Electronics, pages 612–617. IEEE, 2008.

[3] C. W. Brokish. Emulation fundamentals for tis dsp solutions. Texas

Instruments Application Report SPRA439C, 2005.

[4] C.-H. Chen and M.-Y. Cheng. A new sensorless commutation drive for
brushless dc motors and alternators. In 2006 IEEE International Sympo-
sium on Industrial Electronics, volume 3, pages 2116–2121. IEEE, 2006.

[5] W.-H. Chen, J. Yang, L. Guo, and S. Li. Disturbance-observer-based

control and related methodsan overview. IEEE Transactions on Industrial
Electronics, 63(2):1083–1095, 2016.

[6] P. Damodharan and K. Vasudevan. Sensorless brushless dc motor drive

based on the zero-crossing detection of back electromotive force (emf)
from the line voltage difference. IEEE Transactions on Energy Conver-
sion, 25(3):661–668, 2010.

[7] A. Deenadayalan, C. Dhananjai, and G. S. Ilango. Modified sliding mode

observer for wide speed range operation of brushless dc motor. Frontiers
of Electrical and Electronic Engineering, 7(4):467–476, 2012.

81
 [8] J. C. Gamazo-Real, E. Vázquez-Sánchez, and J. Gómez-Gil. Position
and speed control of brushless dc motors using sensorless techniques and
application trends. Sensors, 10(7):6901–6947, 2010.

[9] A. Girolkar and G. Bhuvaneswari. Control of pmbldc motor using third

harmonic back emf sensing with zigzag transformer. In Electrical Energy
Systems (ICEES), 2014 IEEE 2nd International Conference on, pages
110–115. IEEE, 2014.

[10] D. C. Hanselman. Brushless permanent magnet motor design. The Writ-

ers’ Collective, 2003.

[11] F. Hicham, D. Mohamed, R. Abdellatif, and B. Pierre. Sliding mode

observer for position and speed estimations in brushless dc motor (bldcm).
In Industrial Technology, 2004. IEEE ICIT’04. 2004 IEEE International
Conference on, volume 1, pages 121–126. IEEE, 2004.

[12] T. Hurd and I. Sever. Hardware-controlled brushless dc motors ease the

burden on cpus. EDN (Electrical Design News), 55(16):24, 2010.

[13] K. Iizuka, H. Uzuhashi, M. Kano, T. Endo, and K. Mohri. Microcomputer

control for sensorless brushless motor. IEEE Transactions on Industry
Applications, (3):595–601, 1985.

[14] M. Ikhlas. Rotor Position Identification for Brushless DC motor. PhD

thesis, The University of Western Ontario, 2015.

[15] T. Instruments. Drv83x2 three-phase pwm motor driver, december 2014.

URL http://www. ti. com/lit/ds/symlink/drv8332. pdf. Rev: SLES256E–
May, 2010.

[16] T. Instruments. Drv8312-c2-kit hardware reference guide. C2000 &

DRV8312 Systems and Applications Team, Apr, 2011.

[17] T. Instruments. Tms320x2802x, 2803x piccolo analog to digital converter

(adc) and comparator. Reference Guide, Dec, 2011.

82
[18] T. Instruments. Tms320x2802x, 2803x piccolo enhanced pulse width mod-
ulator (epwm) module. Reference Guide, Mar, 2011.

[19] T. Kim, C. Kim, and J. Lyou. A new sensorless drive scheme for a bldc
motor based on the terminal voltage difference. In IECON 2011-37th
Annual Conference on IEEE Industrial Electronics Society, pages 1710–
1715. IEEE, 2011.

[20] T. Kim, H. Lee, and M. Ehsani. Position sensorless brushless dc mo-

tor/generator drives: review and future trends. IET Electric Power Ap-
plications, 1(4):557–564, 2007.

[21] T.-H. Kim, H.-W. Lee, and M. Ehsani. State of the art and future trends
in position sensorless brushless dc motor/generator drives. In 31st Annual
Conference of IEEE Industrial Electronics Society, 2005. IECON 2005.,
pages 8–pp. IEEE, 2005.

[22] T.-S. Kim, B.-G. Park, D.-M. Lee, J.-S. Ryu, and D.-S. Hyun. A new ap-
proach to sensorless control method for brushless dc motors. International
Journal of Control, Automation, and Systems, 6(4):477–487, 2008.

[23] R. Krishnan. Permanent magnet synchronous and brushless DC motor

drives. CRC press, 2009.

[24] Y.-S. Lai and Y.-K. Lin. Novel back-emf detection technique of brushless
dc motor drives for wide range control without using current and position
sensors. IEEE Transactions on power electronics, 23(2):934–940, 2008.

[25] Y.-S. Lai, F.-S. Shyu, and Y.-H. Chang. Novel loss reduction pulsewidth
modulation technique for brushless dc motor drives fed by mosfet inverter.
IEEE transactions on power electronics, 19(6):1646–1652, 2004.

[26] Z. Li, S. Cheng, Y. Qin, and K. Cai. A novel line-to-line back emf calcu-
lation for sensorless brushless dc motor drives. In Electrical Machines and
Systems, 2008. ICEMS 2008. International Conference on, pages 1406–
1411. IEEE, 2008.

83
[27] P. Madaan. Brushless dc motors–part i: Construction and operating
principles. Cypress Semiconductor, 11, 2013.

[28] L. Mingyao, Z. Zhiyao, and L. Keman. A novel and easy-realizing initial

rotor position detection method and speedup algorithm for sensorless bldc
motor drives. ICEMS 2008, pages 2860–2865, 2008.

[29] J. C. Moreira. Indirect sensing for rotor flux position of permanent magnet
ac motors operating over a wide speed range. IEEE Transactions on
Industry Applications, 32(6):1394–1401, 1996.

[30] M. Mubeen. Brushless dc motor primer. Motion Tech Trends, 2008.

[31] S. Ogasawara and H. Akagi. An approach to position sensorless drive

for brushless dc motors. IEEE Transactions on Industry Applications,
27(5):928–933, 1991.

[32] A. Rubaai, D. Ricketts, and M. D. Kankam. Laboratory implementation

of a microprocessor-based fuzzy logic tracking controller for motion con-
trols and drives. IEEE Transactions on Industry Applications, 38(2):448–
456, 2002.

[33] E. Schrijver and J. Van Dijk. Disturbance observers for rigid mechanical
systems: equivalence, stability, and design. Journal of Dynamic Systems,
Measurement, and Control, 124(4):539–548, 2002.

[34] R. Shanmugasundram, K. M. Zakariah, and N. Yadaiah. Implementation

and performance analysis of digital controllers for brushless dc motor
drives. IEEE/ASME transactions on mechatronics, 19(1):213–224, 2014.

[35] J. Shao. An improved microcontroller-based sensorless brushless dc (bldc)

motor drive for automotive applications. IEEE Transactions on industry
applications, 42(5):1216–1221, 2006.

[36] J. Shao, D. Nolan, and T. Hopkins. A novel direct back emf detection for
sensorless brushless dc (bldc) motor drives. In Applied Power Electronics

84
 Conference and Exposition, 2002. APEC 2002. Seventeenth Annual IEEE,
volume 1, pages 33–37. IEEE, 2002.

[37] B. Terzic and M. Jadric. Design and implementation of the extended

kalman filter for the speed and rotor position estimation of brushless dc
motor. IEEE Transactions on Industrial Electronics, 48(6):1065–1073,
2001.

[38] S. Wang and A.-C. Lee. A 12-step sensorless drive for brushless dc motors
based on back-emf differences. IEEE Transactions on Energy Conversion,
30(2):646–654, 2015.

[39] S. Wang, C.-H. Lu, and A.-C. Lee. Disturbance observer structure ap-
plied to sensorless brushless dc motors drive. International Journal of
Computer Theory and Engineering, 7(2):92, 2015.

[40] P. Yedamale. Brushless dc (bldc) motor fundamentals. Microchip Tech-

nology Inc, 20:3–15, 2003.

85