You are on page 1of 55

A New Topology for Unipolar Brushless DC Motor Drive

with High Power Factor

A new converter topology is proposed for driving a permanent magnet brushless dc
(BLDC) motor with unipolar currents. It is based on a front-end single-ended primary inductance
converter (SEPIC) and a switch in series with each phase. All the switches are ground-
referenced, which simplifies their gate drives. The available input voltage can be boosted for
better current regulation, which is an advantage for low voltage applications. For operation with
an ac supply, the SEPIC converter is designed to operate in the discontinuous conduction mode.
In this operation mode, it approximates a voltage follower and the line current follows the line
voltage waveform to a certain extent. The reduction in low-order harmonics and improved power
factor is achieved without the use of any voltage or current sensors. The simplicity and reduced
parts count of the proposed topology make it an attractive low-cost choice for many variable
speed drive applications

Cost minimization is the key to the large volume manufacture and application of brushless
dc (BLDC) motors in variable speed drives. BLDC motors are conventionally excited with
bipolar currents which requires a six-switch inverter as shown in Fig. 1. The unipolar motor
needs fewer electronic parts and uses a simpler circuit than the bipolar motor. For these reasons,
unipolar-driven motors are widely used in low-cost instruments [1]. The savings in converter
cost opens up a lot of applications for variable speed drives (VSD) such as HVAC, fans, pumps,
and appliances which have been dominated by constant speed drives.

The simplest unipolar drive consists of a single switch in series with each winding and
a zener diode or dump resistor in the freewheeling path as shown in Fig. 2 [2]. This drive is
inefficient because the stored energy in the phases is dissipated. Better performance can be
obtained by using topologies that have previously been used for driving switched reluctance
motors (SRM).An example is the C-dump converter shown in Fig. 3 [3], which offers full
regenerative control. However, it has the disadvantage of requiring a complicated control for the
dump capacitor voltage, the failure of which could be catastrophic. A buck converter- based drive
for the unipolar BLDC motor was proposed in [4]. Both these topologies require a higher voltage
on the dump capacitors than what is applied to the motor phases during turn-on. While this is a
requirement for the SRM motor in order to achieve a fast turn-off of the phase current to avoid
negative torque spikes, it is not so for the BLDC motor. In fact, by allowing the phase currents to
overlap during the commutation intervals, the commutation torque pulsations can be reduced.
The topology proposed in this paper takes advantage of this fact to use a smaller voltage on the
dump capacitor. A three-switch converter for the unipolar BLDC motor for ac supply operation
was investigated in [5], but it requires a modification in the machine windings and a split-
capacitor voltage balancing control scheme.
Fig. 1. Conventional six-switch inverter used for BLDC motor drive

Fig. 2. Simple unipolar converter for three-phase BLDC motor

Fig. 3. C-dump converter for unipolar BLDC motor drive.

Fig. 4. AC motor drive with active power factor correction stage

Fig. 5. AC motor drive with integrated PFC stage and inverter

For applications requiring operation from the utility supply, it is important to
design the equipment to satisfy harmonics standards such as the IEC 1000-3-2, which limit the
magnitude of current harmonics that can be injected into the utility. These standards are typically
not satisfied by the conventional method of ac/dc conversion using a bridge rectifier followed by
a large dc bus capacitor. Passive power factor correction (PFC) circuits based on the use of
reactive elements are impractical in 5060 Hz single-phase lines because of size, weight and cost
[6]. Active PFC methods are becoming increasingly popular because of the availability of low-
cost switches. They consist of a dcdc converter between the diode bridge and the bulk capacitor
which is controlled such that the input current is shaped to follow the input voltage. The
frequency spectrum of the input current would then consist of the fundamental plus easily
filtered higher order harmonics. The typical configuration of an ac motor drive with a power
factor correction stage is shown in Fig. 4. For low power levels, the extra cost and complexity of
the additional PFC stage is not justified by the improvement in power factor. These applications
need to take advantage of inherent features of the topology to improve the power factor. A
method of input current shaping utilizing inverter current control has been discussed in [7].
This paper makes use of the desirable properties of the singleended primary
inductance converter (SEPIC) [8] operating in the discontinuous conduction mode (DCM). At
constant duty cycle, the average input current automatically tracks to some extent the sinusoidal
shape of the input voltage [9]. This is realizedwithout the need of sensing and controlling the
input current, thus simplifying the control circuit. Such a feature can be used to integrate the PFC
stage with the output voltage regulation or inverter stage, which can lead to considerable cost
reduction. The schematic of an ac motor drive with the PFC stage integrated with the inverter is
shown in Fig. 5.
This paper is organized as follows. The unipolar excitation scheme of BLDC motors is
discussed in Section II. The proposed topology is introduced and its operation with a dc supply is
discussed in Section III. The operation with an ac supply to improve the utility interface is
discussed in Section IV. A design example is discussed in Section V, simulation and experimental
results are presented in Section VI, and Section VII gives concluding remarks
Fig. 6. Back-emf, phase current and output torque waveforms with: (a) bipolar
excitation and (b) unipolar excitation


Unipolar current conduction limits the phases to only one direction of current, and
the commutation frequency is half that of a bipolar or full-wave drive. The width of the current
pulse is 180' for 2-phase, 120" and 180'' for 3-phase, 90' and 180" for 4-phase, and 72" and 180"
for 5-phase motors. Fig. 1 shows the different current waveforms that are used to excite the
motors. The peak currents for unipolar conduction are higher than for the corresponding bipolar
schemes. The schematics of the full-wave and half-wave converters used with 3-phase BLDC
motors are shown in Fig. 2. In the simple unipolar converter, a single switch is placed in series
with each motor winding, while a reverse-parallel diode provides a freewheeling path at turn-off.
The motor windings can only be excited for a maximum of half of the total excitation cycle and
therefore the windings are poorly utilized. In addition, inductive energy stored in the windings at
turn-off must be dissipated within the motor rather than being returned to the supply. This drive
has no regenerative control. but with other unipolar converter topologies, it is possible to obtain
fonrquadrant operation of the drive
Fig.2. (a) 3-Phase Bipolar Converter, (b) 3-Phase Unipolar Converter

Another factor that has to be considered before choosing unipolar excitation is that the motor
neutral has to be available.
Bipolar circuits allow each winding to be excited in the positive and negative sense
during a cycle, giving aconsiderable improvement in winding utilization. All bipolar circuits are
capable of returning stored inductive energy at turn-off and of regenerative braking. However,
they have the following disadvantages [SI:

1.Winding current is conducted through two series devices, causing an increase in total
conduction losses.
2. There is a risk of shoot-through faults.
3. Switching of devices connected to the supply rails generally requires some isolation circuitry.


The motor models are created using the MAXWELL 2-D Field Simulator package [71. The
back-emf data obtained from Finite Element Method (EM) is then used in dynamic simulations
in MATLAB [81. To simplify the interpretation of results, it is assumed that the reference
currents in Fig. 1 flow through the wye-connected motor windings. The electromagnetic torque
output of the motors is given by 191 ~ , = ( e , i , + e , i , +.. ..+ e , i , ) /w, . (1)


We first consider a 2-phase motor with 4 slots as shown in Fig. 3. This is an integral slots/pole
motor and the coil span is 2 slots. The currents of Fig. l(a) are used to excite this motor. This
drive requires only one position sensor and a single current sensor. The number of turns/phase is
higher than the reference for the same amount of copper. This results in higher peak torque, but
as Table I shows, the torque ripple of this motor with unipolar currents is very high. This is

2-phase 4-slot BLDC motor

because commutation between the phases takes place during the zero crossing of the back-emf.
This can be avoided by increasing the number of phases.


Let us consider a 3-phase, 12-slot motor as shown in Fig. 4(a) and use the currents of
Fig. l(c). In this case, the commutation takes place before the back-emf of each phase reaches
zero, and the ripple is reduced as shown in Fig. 6(b). When we use the currents in Fig. l(d), the
torque ripple is worse as shown in Fig. 6(c). This is because different numbers of phases
contribute to the torque at different instants of time. From the previous two cases, we realize that
we need a combination of 180' unipolar currents and small back-emf width to reduce the torque
To investigate this case, we consider the 6-slot motor shown in Fig. 4(h). The
back-emf plots as a function of rotor position for both the 3-phase motors are shown in Fig. 5.
We find that the 6-slot motor has a higher peak and smaller backemf width than the 12-slot
motor. This can be explained as follows.
The end-turns are shorter for the 6-slot motor, because of which the number of
turns per coil is more for the same amount of copper. This increases the peak value of the
backemf. The maximum coil span or winding pitch is determined by dividing the number of slots
by the number of poles and rounding off to the next lowest integer [6]. For the 12-slot motor, the
slots/pole is 3, and th-, coils are full-pitched, which maximizes the width of the back-emf. In the
6-slot motor, the slots/pole is 1.5, and the coil span used is 1 because of which the width of the
back-emf waveform is smaller.
This effect can also be achieved by short-pitching the coils in an integral slots/pole
design. The possible short-pitch coil spans for the 12-slot motor are 1 and 2. Using a coil span of
1 would make the back-emf width too narrow and increase the torque ripple. A coil span of 2
would be ideal, but would leave half the slots unutilized. Using a fractional slots/pole motor has
the additional advantage of reducing the cogging torque [6]. If the number of slots is increased to
24 or 32, many more combinations are possible for obtaining smaller back-emf width, and the
designer can then make a choice based on other considerations. However, in general, the smallest
number of slots gives the lowest labor cost in winding, and a coil span of 1 or 2 slots minimizes
the end turns [6]. Note that similar results could also be obtained by using full-pitch stator coils
and a magnet pole arc of 120' electrical as discussed in [6].
From Table 11, we see that the 3-ph 6-slot motor excited with 180' unipolar currents
gives better performance in terms of torque ripple, with some loss in peak and average torque.
This is explained as follows. In Fig. l(d), we have one phase conducting during intervals 2,4 and
6 and two phases conducting during intervals 1,3 and 5. In particular, when the back-emf of
phase A reaches its peak during interval 2, only phase A is conducting. When the back-emf of
phase A starts decreasing in interval 3, phase B also comes into conduction. The decreasing
torque contribution of phase A is compensated by the increasing contribution from phase B. The
result is an almost constant torque over the entire cycle. The remaining case of the 6-slot motor
excited with 126 unipolar currents results in high torque ripple because the back-emf during the
commutation instants is low.
Both the unipolar and the bipolar drives require three hall-effect sensors, with the
second and third displaced by 120" and 240" electrical respectively from the first. The bipolar
drive requires six switches while the unipolar drive requires only three, albeit with higher current
ratings. The advantage of using 120" currents is that we require only one current sensor in the dc
link. However, the torque ripple is not low enough. It can be reduced further by increasing the
number of phases to four.

Two 4-phase motors are considered: One with I6 slots, and the other with 8 slots as shown in
Fig. 7. Both motors are integral slots/pole designs. The 16-slot motor is short-pitched by a factor
of 2, while the 8 slot motor is full-pitched, which
Fig.6. Torque outputs of 3-Phase Moton (a) 12-slot motor with 120" bipolar current, (b)
12-slot motor with 120" unipolar current, (c) 12-slot motor with 180' unipolar current. (d) 6-
slot motor with 120' unipolx current, (e) 6-slot motor with 180' unipolar current.

explains the difference in the width of their back-emf waveforms shown in Fig. 8. The number of
turns/phase is more in the 16-slot motor because of the shorter end-turns, which explains the
higher peak of its back-emf. These motors are excited with the current waveforms of Fig. l(e)
and (0. The torque outputs are shown in Fig. 9 and Table 111 gives the numerical values. For
both motors, using 90' conduction gives better results because the commutation between phases
takes place when the back-emfs are high. In addition, it requires the use of only a single current
sensor in the dc link. In the 180" conduction scheme, two phases conduct at all times, and the
back-emfs of the incoming and outgoing phases are low, resulting in large torque ripple. It also
requires the use of a current sensor in each phase.

The Single Ended Primary Inductance Converter (SEPIC)

The most basic converter that we looked at last month is the buck converter. It is so
named because it always steps down, or bucks, the input voltage. The output of the converter is
given by:
Interchange the input and the output of the buck converter, and you get the second basic
converter the boost. The boost always steps up, hence its name. The output voltage is always
higher than the input voltage, and is given by:

What if you have an application where you need to both step up and step down, depending on the
input and output voltage? You could use two cascaded converters a buck and a boost.
Unfortunately, this requires two separate controllers and switches. It is, however, a good solution
in many cases.

The buck-boost converter has the desired step up and step down functions:

The output is inverted. A flyback converter (isolated buck-boost) requires a transformer instead
of just an inductor, adding to the complexity of the development.

One converter that provides the needed input-to-output gain is the Sepic (single- ended
primary inductor converter) converter. A Sepic converter is shown in Fig. 1. It has become
popular in recent years in battery-powered systems that must step up or down depending upon
the charge level of the battery.
Fig. 2 shows the circuit when the power switch is turned on. The first inductor,
L1, is charged from the input voltage source during this time. The second inductor takes energy
from the first capacitor, and the output capacitor is left to provide the load current. The fact that
both L1 and L2 are disconnected from the load when the switch is on leads to complex control
characteristics, as we will see later.
When the power switch is turned off, the first inductor charges the capacitor C1
and also provides current to the load, as shown in Fig. 3. The second inductor is also connected
to the load during this time.
The output capacitor sees a pulse of current during the off time, making it inherently
noisier than a buck converter
Figure 1. The Sepic converter can both step up and step down the input voltage, while
maintaining the same polarity and the same ground reference for the input and output.

Figure 2. When the switch is turned on, the input inductor is charged from the source, and the
second inductor is charged from the first capacitor. No energy is supplied to the load capacitor
during this time. Inductor current and capacitor voltage polarities are marked in this figure.

Figure 3. With the switch off, both inductors provide current to the load capacitor

The input current is non-pulsating, a distinct advantage in running from a battery supply.
Figure 4. In order to take advantage of Vorprians PWM switch model, the circuit elements must
first be rearranged. The function of the original topology is retained when the capacitor is
moved, and the second inductor is redrawn.

Figure 5. For DC analysis, the small signal sources are set to zero, inductors become short
circuits, and capacitors become open circuits. After the circuit is redrawn, it is a trivial matter to
write KVL around the outer loop of the circuit to solve for the conversion gain of the converter.

The PWM Switch Model in the Sepic Converter

The best way to analyze both the AC and DC characteristics of the Sepic converter is by using
the PWM switch model, developed by Dr. Vatch Vorprian in 1986. Some minor circuit
manipulations are first needed to reveal the location of the switch model, and this is First,
capacitor C1 is moved to the bottom branch of the converter. Then, inductor L2 is pulled over to
the left, keeping its ends connected to the same nodes of the circuit. This reveals the PWM
switch model of the converter, with its active, passive, and common ports, allowing us to use
well-established analysis results for this converter.
For more background on the PWM switch model, the text book Fast Analytical
Techniques for Electrical and Electronic Circuits [1] is highly recommended.

Figure 2.a shows the basic structure of a Single Ended Primary Inductance Converter
(SEPIC) [ 11. The same circuit is represented in figure 2.b, but in this figure the transformer
magnetizing inductance is explicitly represented by the inductance L2.

PWM SEPIC (a) and this same converter with the transformer magnetizing inductance

First of all, let us calculate the average value of the current iz, i2a, as a function of the average
output current i, ia, ( ia= i because it is constant).
Thus, we can write:

and, therefore

Taking average values, we obtain

But ic1, and ic2, are flowing in two capacitors working in steady-state, and, therefore:
On the other hand, we can calculate similarly the average value of the voltage vcl , vcIa as a
function of the input voltage vg :

and taking average values, we obtain

But v1 and vz are voltages in two inductors, and therefore

Assuming very low ripple in C1 , equation [lo] becomes

Equations (6) and (1 1) are also valid in ZCS and ZVS QRC's because no considerations have
been made about the type of switch used. The dc voltage-conversion ratio in continuous
conduction mode,m , can be easily obtained from figure 3. Volts-second balance applied to the
inductor can be expressed by:

and, therefore:

where d is the duty cycle. From equations [14], 1151 y [16], we obtain:

In discontinuous conduction mode, the dc voltage-conversion ratio M4 can be obtained

averaging the current id, injected into the output cell r-cz (see reference [5]). This current can be
seen in figure 4
Volts-second balance appliedto L2 yields to the following equation

The maximun value of i, is (from Faraday's law):

Fig. 3 Equivalent circuits in a PWM SEPIC. A and b in continuous conduction (8) mode. a,
b and c in discontinuous
Fig.4 Current i, waveforms

And the output voltage can be calculated by multiplying the average value of i, by the output
load r :

From equations [14], 1151 y [16], we obtain:

Where K is a dimensionless parameter defined by

The boundary value of K between continuous and discontinuous conduction modes, K c ,, can be
easily calculated (mr = m):

(note that the values of m, md and K c , are the same in SEPIC, Cuk and flyback converters)

As usual, a SEPIC is operating in continuous conduction mode when:

and in discontinuous conduction mode when:

The waveforms in the circuit shown in figure 2.b, when it is operating in continuous conduction
mode, can be seen in figure 5, in which i1a, and i2a represent the average value of the currents
i1, and , i2, and io, the peak value of is and iD (neglecting ripple), and Vsn and Vdm the peak
value of the voltages Vs and v,, . The values of these magnitudes and of the dc voltage
conversion ratio m are summarized in Table 1, in which these values have been expressed as
functions of the input magnitudes vs, , i1a, and the output magnitudes V , i .
All this values can be easily obtained from figure 3 and from equation [13]. For
example, current i , flowing in the capacitor C, is i2 duringdr (fig. 3.a) and i, during
(fig. 3.b) This current can be seen in figure5 Likewise, thepeak current value (neglecting ripple)
inthe power transistor Si is the addition of i1a, and i2a. :

being i, and i related by

(power input-output balance):

From equations [13], [22] and [23] we obtain:

this value can be seen in Table 1.

AC Analysis of the Sepic Converter

You wont find a complete analysis of the Sepic converter anywhere in printed
literature. What you will find are application notes with comments like, the Sepic is not well-
understood. Despite the lack of documentation for the converter, engineers continue to use it
when applicable
Proper small-signal analysis of the Sepic converter is a difficult analytical task, only
made practical by advanced circuit analysis techniques originally developed by Dr. David
Middlebrook and continued by Vorprian. [1]
If youre going to build a Sepic, as a minimum, you need to understand the control
characteristics. Fortunately, Vorprians work is now available for this converter, and you can
download the complete analysis notes .[2]
The simplified analysis of the Sepic converter, derived in detail in [2], ignores
parasitic resistances of the inductors and capacitors, and yields the following result for the
control-to-output transfer function:

Figure 6. The small-signal AC sources are included in the switch model, and we can either solve
the analysis by
hand, or . The hand analysis is crucial for symbolic expressions and design equations.

Figure 7 This figure shows a specific design example for a 15 W converter.

As you can see from these expressions,+ the simplified analysis is anything but simple.
Including the parasitic resistances greatly complicates the analysis, but may be necessary for
worst-case analysis of the Sepic converter. The analysis of this converter involves the use of the
powerful extra element theorem, and Vorprians book
on circuit analysis techniques. [1]
In addition to the inevitable fourth order denominator of the Sepic, the most
important features to note in the control transfer function are the terms in the numerator. The first
term is a single right half- plane (RHP) zero. Right-half-plane zeros are a result of converters
where the response to an increased duty cycle is to initially decrease the output voltage.
When the power switch is turned on, the first inductor is disconnected from the load
and this directly gives rise to the first order RHP zero. Notice that the expression only depends
on the input inductor, L1, the load resistor, R, and the duty cycle.
The complex RHP zeros arise from the fact that turning on the switch disconnects the
second inductor from the load. These zeros will actually move with the values of parasitic
resistors in the circuit, so careful analysis of your converter is needed to ensure stability under all

Brushless DC (BLDC)

Brushless Direct Current (BLDC) motors are one of the motor types rapidly gaining
popularity. BLDC motors are used in industries such as Appliances, Automotive, Aerospace,
Consumer, Medical, Industrial Automation Equipment and Instrumentation. As the name implies,
BLDC motors do not use brushes for commutation; instead, they are electronically commutated.
BLDC motors have many advantages over brushed DC motors and induction motors. A few of
these are:

Better speed versus torque characteristics

High dynamic response
High efficiency
Long operating life
Noiseless operation
Higher speed ranges

In addition, the ratio of torque delivered to the size of the motor is higher, making it
useful in applications where space and weight are critical factors. In this application note, we
will discuss in detail the construction, working principle, characteristics and typical applications
of BLDC motors. Refer to Appendix B: Glossary for a glossary of terms commonly used
when describing BLDC motors.


BLDC motors are a type of synchronous motor. This means the magnetic field
generated by the stator and the magnetic field generated by the rotor rotate at the same frequency.
BLDC motors do not experience the slip that is normally seen in induction motors.
BLDC motors come in single-phase, 2-phase and 3-phase configurations. Corresponding
to its type, the stator has the same number of windings. Out of these, 3-phase motors are the most
popular and widely used. This application note focuses on 3-phase motors.


The stator of a BLDC motor consists of stacked steel laminations with windings
placed in the slots that are axially cut along the inner periphery (as shown in Figure 3).
Traditionally, the stator resembles that of an induction motor; however, the windings are
distributed in a different manner. Most BLDC motors have three stator windings connected in
star fashion. Each of these windings are constructed with numerous coils interconnected to form
a winding. One or more coils are placed in the slots and they are interconnected to make a
winding. Each of these windings are distributed over the stator periphery to form an even
numbers of poles. There are two types of stator windings variants: trapezoidal and sinusoidal
motors. This differentiation is made on the basis of the interconnection of coils in the stator
windings to give the different types of back Electromotive Force (EMF). Refer to the What is
Back EMF? section for more information.
As their names indicate, the trapezoidal motor gives a back EMF in trapezoidal
fashion and the sinusoidal motors back EMF is sinusoidal, as shown in Figure 1 and Figure 2. In
addition to the back EMF, the phase
current also has trapezoidal and sinusoidal variations in the respective types of motor. This
makes the torque output by a sinusoidal motor smoother than that of a trapezoidal motor.
However, this comes with an extra cost, as the sinusoidal motors take extra winding
interconnections because of the coils distribution on the stator periphery, thereby increasing the
copper intake by the stator windings.
Depending upon the control power supply capability, the motor with the correct
voltage rating of the stator can be chosen. Forty-eight volts, or less voltage rated motors are used
in automotive, robotics, small arm movements and soon. Motors with 100 volts, or higher
ratings, are used in appliances, automation and in industrial applications.

The rotor is made of permanent magnet and can vary from two to eight pole pairs with
alternate North (N) and South (S) poles.
Based on the required magnetic field density in the rotor, the proper magnetic material is
chosen to make the rotor. Ferrite magnets are traditionally used to make permanent magnets. As
the technology advances, rare earth alloy magnets are gaining popularity. The ferrite magnets are
less expensive but they have the disadvantage of low flux density for a given volume. In contrast,
the alloy material has high magnetic density per volume and enables the rotor to compress
further for the same torque. Also, these alloy magnets improve the size-to-weight ratio and give
higher torque for the same size motor using ferrite magnets.
Neodymium (Nd), Samarium Cobalt (SmCo) and the alloy of Neodymium, Ferrite and
Boron (NdFeB) are some examples of rare earth alloy magnets. Continuous research is going on
to improve the flux density to compress the rotor further.
Figure 4 shows cross sections of different arrangements of magnets in a rotor.

Hall Sensors
Unlike a brushed DC motor, the commutation of a BLDC motor is controlled
electronically. To rotate the BLDC motor, the stator windings should be energized in a sequence.
It is important to know the rotor position in order to understand which winding will be energized
following the energizing sequence. Rotor position is sensed using Hall effect sensors embedded
into the stator.
Most BLDC motors have three Hall sensors embedded into the stator on the non-driving
end of the motor. Whenever the rotor magnetic poles pass near the Hall sensors, they give a high
or low signal, indicating the N or S pole is passing near the sensors. Based on the combination of
these three Hall sensor signals, the exact sequence of commutation can be determined.

Figure 5 shows a transverse section of a BLDC motor with a rotor that has alternate N
and S permanent magnets. Hall sensors are embedded into the stationary part of the motor.
Embedding the Hall sensors into the stator is a complex process because any misalignment in
these Hall sensors, with respect to the rotor magnets, will generate an error in determination of
the rotor position. To simplify the process of mounting the Hall sensors onto the stator, some
motors may have the Hall sensor magnets on the rotor, in addition to the main rotor magnets.
These are a scaled down replica version of the rotor. Therefore, whenever the rotor rotates, the
Hall sensor magnets give the same effect as the main magnets. The Hall sensors are normally
mounted on a PC board and fixed to the enclosure cap on the non-driving end. This enables users
to adjust the complete assembly of Hall sensors, to align with the rotor magnets, in order to
achieve the best performance.
Based on the physical position of the Hall sensors, there are two versions of output.
The Hall sensors may be at 60 or 120 phase shift to each other. Based on this, the motor
manufacturer defines the commutation sequence, which should be followed when controlling the

Theory of Operation

Each commutation sequence has one of the windings energized to positive power
(current enters into the winding), the second winding is negative (current exits the winding) and
the third is in a non-energized condition. Torque is produced because of the interaction between
the magnetic field generated by the stator coils and the permanent magnets. Ideally, the peak
torque occurs when these two fields are at 90 to each other and falls off as the fields move
together. In order to keep the motor running, the magnetic field produced by the windings should
shift position, as the rotor moves to catch up with the stator field. What is known as Six-Step
Commutation defines the sequence of energizing the windings. See the Commutation
Sequence section for detailed information and an example on six-step commutation

Sensor less Control of BLDC Motors

Until now we have seen commutation based on the rotor position given by the Hall
sensor. BLDC motors can be commutated by monitoring the back EMF signals instead of the
Hall sensors. The relationship between the Hall sensors and back EMF, with respect to the phase
voltage, is shown in Figure 7. As we have seen in earlier sections, every commutation sequence
has one of the windings energized positive, the second negative and the third left open. As shown
in Figure 7, the Hall sensor signal changes the state when the voltage polarity of back EMF
crosses from a positive to negative or from negative to positive. In ideal cases, this happens on
zero-crossing of back EMF, but practically, there will be a delay due to the winding
characteristics. This delay should be compensated by the microcontroller. Figure 10 shows a
block diagram for sensorless control of a BLDC motor.
Another aspect to be considered is very low speeds. Because back EMF is
proportional to the speed of rotation, at a very low speed, the back EMF would be at a very low
amplitude to detect zero-crossing. The motor has to be started in open loop, from standstill and
when sufficient back EMF is built to detect the zero-cross point, the control should be shifted to
the back EMF sensing. The minimum speed at which back EMF can be sensed is calculated from
the back EMF constant of the motor.
With this method of commutation, the Hall sensors can be eliminated and in some
motors, the magnets for Hall sensors also can be eliminated. This simplifies the motor
construction and reduces the cost as well. This is advantageous if the motor is operating in dusty
or oily environments, where occasional cleaning is required in order for the Hall sensors to sense
properly. The same thing applies if the motor is mounted in a less accessible location.


BLDC motors find applications in every segment of the market. Automotive, appliance,
industrial controls, automation, aviation and so on, have applications for BLDC motors. Out of
these, we can categorize the type of BLDC motor control into three major types:

Constant load
Varying loads
Positioning applications

Applications With Constant Loads

These are the types of applications where a variable speed is more important than
keeping the accuracy of the speed at a set speed. In addition, the acceleration and deceleration
rates are not dynamically changing. In these types of applications, the load is directly coupled to
the motor shaft. For example, fans, pumps and blowers come under these types of applications.
These applications demand low-cost controllers, mostly operating in open-loop.
Applications With Varying Loads

These are the types of applications where the load on the motor varies over a speed
range. These applications may demand a high-speed control accuracy and good dynamic
responses. In home appliances, washers, dryers and compressors are good examples. In
automotive, fuel pump control, electronic steering control, engine control and electric vehicle
control are good examples of these. In aerospace, there are a number of applications, like
centrifuges, pumps, robotic arm controls, gyroscope controls and so on. These applications may
use speed feedback devices and may run in semi-closed loop or in total closed loop. These
applications use advanced control algorithms, thus complicating the controller. Also, this
increases the price of the complete system.

Positioning Applications

Most of the industrial and automation types of application come under this category.
The applications in this category have some kind of power transmission, which could be
mechanical gears or timer belts, or a simple belt driven system. In these applications, the
dynamic response of speed and torque are important. Also, these applications may have frequent
reversal of rotation direction. A typical cycle will have an accelerating phase, a constant speed
phase and a deceleration and positioning phase, as shown in Figure 11. The load on the motor
may vary during all of these phases, causing the controller to be complex. These systems mostly
operate in closed loop. There could be three control loops functioning simultaneously: Torque
Control Loop, Speed Control Loop and Position Control Loop. Optical encoder or synchronous
resolvers are used for measuring the actual speed of the motor. In some cases, the same sensors
are used to get relative position information. Otherwise, separate position sensors may be used to
get absolute positions. Computer Numeric Controlled (CNC) machines are a good example of
this. Process controls, machinery controls and conveyer controls have plenty of applications in
this category.

Matlab is a high-performance language for technical computing. It integrates
computation, visualization, and programming in an easy-to-use environment where problems and
solutions are expressed in familiar mathematical notation. Typical uses include Math and
computation Algorithm development Data acquisition Modeling, simulation, and prototyping
Data analysis, exploration, and visualization Scientific and engineering graphics Application
development, including graphical user interface building.

Matlab is an interactive system whose basic data element is an array that does not require
dimensioning. This allows you to solve many technical computing problems, especially those
with matrix and vector formulations, in a fraction of the time it would take to write a program in
a scalar no interactive language such as C or Fortran.

The name matlab stands for matrix laboratory. Matlab was originally written to provide
easy access to matrix software developed by the linpack and eispack projects. Today, matlab
engines incorporate the lapack and blas libraries, embedding the state of the art in software for
matrix computation.

Matlab has evolved over a period of years with input from many users. In university
environments, it is the standard instructional tool for introductory and advanced courses in
mathematics, engineering, and science. In industry, matlab is the tool of choice for high-
productivity research, development, and analysis.

Matlab features a family of add-on application-specific solutions called toolboxes. Very

important to most users of matlab, toolboxes allow you to learn and apply specialized
technology. Toolboxes are comprehensive collections of matlab functions (M-files) that extend
the matlab environment to solve particular classes of problems. Areas in which toolboxes are
available include signal processing, control systems, neural networks, fuzzy logic, wavelets,
simulation, and many others.

The matlab system consists of five main parts:

Development Environment. This is the set of tools and facilities that help you use matlab
functions and files. Many of these tools are graphical user interfaces. It includes the matlab
desktop and Command Window, a command history, an editor and debugger, and browsers for
viewing help, the workspace, files, and the search path.

The matlab Mathematical Function Library. This is a vast collection of computational

algorithms ranging from elementary functions, like sum, sine, cosine, and complex arithmetic, to
more sophisticated functions like matrix inverse, matrix eigenvalues, Bessel functions, and fast
Fourier transforms.

The matlab Language. This is a high-level matrix/array language with control flow
statements, functions, data structures, input/output, and object-oriented programming features. It
allows both "programming in the small" to rapidly create quick and dirty throw-away programs,
and "programming in the large" to create large and complex application programs.

Matlab has extensive facilities for displaying vectors and matrices as graphs, as well as
annotating and printing these graphs. It includes high-level functions for two-dimensional and
three-dimensional data visualization, image processing, animation, and presentation graphics. It
also includes low-level functions that allow you to fully customize the appearance of graphics as
well as to build complete graphical user interfaces on your matlab applications.

The matlab Application Program Interface (API). This is a library that allows you to
write C and Fortran programs that interact with matlab. It includes facilities for calling routines
from matlab (dynamic linking), calling matlab as a computational engine, and for reading and
writing MAT-files.

4.2.1 Introduction:

Simulink is a software add-on to matlab which is a mathematical tool developed by The

Math works,( a company based in Natick. Matlab is powered by
extensive numerical analysis capability. Simulink is a tool used to visually program a dynamic
system (those governed by Differential equations) and look at results. Any logic circuit, or
control system for a dynamic system can be built by using standard building blocks available in
Simulink Libraries. Various toolboxes for different techniques, such as Fuzzy Logic, Neural
Networks, dsp, Statistics etc. are available with Simulink, which enhance the processing power
of the tool. The main advantage is the availability of templates / building blocks, which avoid the
necessity of typing code for small mathematical processes.

Concept of signal and logic flow:

In Simulink, data/information from various blocks are sent to another block by lines
connecting the relevant blocks. Signals can be generated and fed into blocks dynamic /
static).Data can be fed into functions. Data can then be dumped into sinks, which could be
scopes, displays or could be saved to a file. Data can be connected from one block to another,
can be branched, multiplexed etc. In simulation, data is processed and transferred only at
Discrete times, since all computers are discrete systems. Thus, a simulation time step (otherwise
called an integration time step) is essential, and the selection of that step is determined by the
fastest dynamics in the simulated system.
Fig 4.1 Simulink library browser
Connecting blocks:

fig 4.2 Connectung blocks

To connect blocks, left-click and drag the mouse from the output of one block to the input

of another block.
Sources and sinks:

The sources library contains the sources of data/signals that one would use in a dynamic
system simulation. One may want to use a constant input, a sinusoidal wave, a step, a repeating
sequence such as a pulse train, a ramp etc. One may want to test disturbance effects, and can use
the random signal generator to simulate noise. The clock may be used to create a time index for
plotting purposes. The ground could be used to connect to any unused port, to avoid warning
messages indicating unconnected ports.

The sinks are blocks where signals are terminated or ultimately used. In most cases, we
would want to store the resulting data in a file, or a matrix of variables. The data could be
displayed or even stored to a file. the stop block could be used to stop the simulation if the input
to that block (the signal being sunk) is non-zero. Figure 3 shows the available blocks in the
sources and sinks libraries. Unused signals must be terminated, to prevent warnings about
unconnected signals.

fig 4.3 Sources and sinks

Continuous and discrete systems:

All dynamic systems can be analyzed as continuous or discrete time systems. Simulink
allows you to represent these systems using transfer functions, integration blocks, delay blocks

fig 4.4 continous and descrete systems

Non-linear operators:

A main advantage of using tools such as Simulink is the ability to simulate non-linear
systems and arrive at results without having to solve analytically. It is very difficult to arrive at
an analytical solution for a system having non-linearities such as saturation, signup function,
limited slew rates etc. In Simulation, since systems are analyzed using iterations, non-linearities
are not a hindrance. One such could be a saturation block, to indicate a physical limitation on a
parameter, such as a voltage signal to a motor etc. Manual switches are useful when trying
simulations with different cases. Switches are the logical equivalent of if-then statements in

fig 4.5 simulink blocks

Mathematical operations:

Mathematical operators such as products, sum, logical operations such as and, or, etc.
.can be programmed along with the signal flow. Matrix multiplication becomes easy with the
matrix gain block. Trigonometric functions such as sin or tan inverse (at an) are also available.
Relational operators such as equal to, greater than etc. can also be used in logic circuits
fig 4.6 Simulink math blocks


In complicated block diagrams, there may arise the need to transfer data from one portion
to another portion of the block. They may be in different subsystems. That signal could be
dumped into a goto block, which is used to send signals from one subsystem to another.

Multiplexing helps us remove clutter due to excessive connectors, and makes

matrix(column/row) visualization easier.
fig 4.7 signals and systems

4.6.3 Making subsystems

Drag a subsystem from the Simulink Library Browser and place it in the parent block
where you would like to hide the code. The type of subsystem depends on the purpose of the
block. In general one will use the standard subsystem but other subsystems can be chosen. For
instance, the subsystem can be a triggered block, which is enabled only when a trigger signal is

Open (double click) the subsystem and create input / output PORTS, which transfer
signals into and out of the subsystem. The input and output ports are created by dragging them
from the Sources and Sinks directories respectively. When ports are created in the subsystem,
they automatically create ports on the external (parent) block. This allows for connecting the
appropriate signals from the parent block to the subsystem.

4.6.4 Setting simulation parameters:

Running a simulation in the computer always requires a numerical technique to solve a

differential equation. The system can be simulated as a continuous system or a discrete system
based on the blocks inside. The simulation start and stop time can be specified. In case of
variable step size, the smallest and largest step size can be specified. A Fixed step size is
recommended and it allows for indexing time to a precise number of points, thus controlling the
size of the data vector. Simulation step size must be decided based on the dynamics of the
system. A thermal process may warrant a step size of a few seconds, but a DC motor in the
system may be quite fast and may require a step size of a few milliseconds.


The motor under investigation is of the surface-mount permanent magnet variety
with concentrated stator windings such that the induced back-emfs with respect to rotor position
are trapezoidal with a flat-top width that is as wide as possible. Smooth torque production
requires forcing a constant current through each phase winding when its back-emf is at its peak
value and turning off the current when the back-emf is changing. For bipolar excitation, positive
current is injected when the back-emf is positive, and negative current when the back-emf is
negative, with each conduction period lasting 120 . This results in two phases conducting current
and producing torque at any instant of time as shown in the waveforms of Fig. 6(a). Unipolar
current conduction limits the phases to only one direction of current as shown in Fig. 6(b).
Constant torque production is still possible because one phase is conducting current at any
instant. It is of course possible to have an overlap in the phase conduction to have a smoother
torque production [10]. In any case, the motor windings are poorly utilized compared to the
bipolar case. This is reflected in the lower output torque of the unipolar motor for the same peak
phase currents. The primary motivation for choosing unipolar excitation is that in practice, the
inverter typically costs more than the motor and there is a great potential for reducing its cost and
hence the overall cost of the drive. In addition to cost reduction, unipolar excitation offers the
following advantages:

There is only one device in series with each phase, minimizing conduction losses.
The risk of shoot-through faults is eliminated.
Switching of devices connected to the supply rails, which generally requires some isolation
circuitry, can
be avoided.

Another factor that has to be considered before choosing unipolar excitation is that the
motor neutral has to be available because the phase currents are no longer balanced.
Fig. 7. Schematic of SEPIC converter based BLDC motor drive.


The proposed converter with four controlled switches and diodes is shown in Fig. 7. The front-
end consists of a SEPIC dc/dc converter comprised of inductors L1and L2 , switch S1 ,
intermediate capacitor C1, diode D1 and output capacitor C2. The modification from the usual
SEPIC configuration is that the diode D1 is placed in the return path instead of in the ositive ail.
This is to block the flow of current through the phases during the periods of negative back-emf.
A, B, and C are the three machine windings, and the currents through them are controlled by
turn-on and turn-off of the switches Sa Sb, and Sc, respectively. Since there is only one switch
per phase, the currents through them are unidirectional. The diodesDa, Db and Dc serve to
freewheel the winding currents when the switches are turned off during current regulation and
phase commutation. The output of the converter is used to energize the three phases of the motor,
and the voltage of capacitor C1 is used to demagnetize the phases during turn-off and for current
Each phase is energized by turning on the corresponding switch in series with it. The
equivalent circuit of phase A when switch Sa is turned on is shown in Fig. 8(a). To regulate the
current,Sa is turned off, which forces the turn-on of diode Da , and the flow of current through
C1as shown in the equivalent circuit of Fig. 8(b). This applies a voltage of across the
machine winding, enabling a fast decay of the phase current. For proper demagnetization of the
phase after each conduction interval and to prevent conduction during periods of negative back-
emf, the instantaneous value of should be greater than the peak value of the back-emf E, or

By applying Kirchoffs voltage law to the SEPIC front-end, we obtain

Since the average voltages in the two inductors are zero, we get

From (1) and (2), we obtain the peak back-emf at the maximum speed of the motor, which is
given by , assuming that the ripple in the intermediate capacitor voltage

Fig. 8. Equivalent circuits of each machine phase when (a) the switch is on and (b) when the
diode is conducting.

is negligible. The maximum operating speed is then given by

where Kc is the phase back-emf constant of the motor.
If the motor is operated beyond this speed, it would result in negative torque spikes
because of conduction during periods of negative back-emf.

The minimum voltageVdc required is where Rs and Ls are the

phase resistance and inductance, and I is the phase current. At low speeds, when the back-emf is
low, the switching frequency of the phase switches increases in order to regulate the phase
current. The switching frequency and hence the losses at low speeds can be minimized by
bucking the input voltage to lower levels at the output Vdc . At higher speeds, the current
regulator loses its ability to force current into the phases especially during turn-on because of the
high back-emf voltage. The ability of the SEPIC front-end to boost the available input voltage
makes it possible to maintain current-regulated operation of the drive at higher speeds. This
feature makes the proposed topology particularly suitable for low voltage dc applications such as
automotive circuits. The key operating waveforms at low speeds and at high speeds
are illustrated in Fig. 9(a) and (b), respectively.
The front-end SEPIC converter can be designed for operation either in the continuous
conduction mode (CCM) or in the discontinuous conduction mode (DCM). In CCM, its voltage
conversion ratio is given by

where D is the duty cycle of the switch S1 .

In DCM, its voltage conversion ratio is given by

where , R being the equivalent load resistance and T the time period of
switch S1 .The boundary value of K between continuous and discontinuous conduction modes,
can be calculated

The converter operates in CCM when and in DCM when . In both modes
of operation, can be regulated at a value higher (Boost operation) or lower (Buck operation)
than the input voltage . From the controls viewpoint, it is advantageous to have the converter
operating in the same mode under all load conditions. In addition, the size of the inductors and
hence the overall converter can be reduced if it is operated in DCM [12]. Hence it is proposed
that the converter be designed for operation in the critical conduction mode at maximum load, so
that it operates in DCM at rated load and all values less than rated load.


For applications requiring operation from an ac supply, it is desired to obtain
improved power factor by using the proposed topology as shown in Fig. 10. By operating the
SEPIC front-end in DCM, the following desirable characteristics are obtained

Fig. 10. Schematic of the proposed converter operating from an ac supply.

[9]: The converter works as a voltage follower, meaning that the input current naturally follows
the input voltage profile (No current loop is needed), and the theoretical power factor is unity.
For ideal voltage follower operation, the intermediate capacitor voltage should follow the half-
sinusoidal input voltage, and goes to zero in each half-cycle. This is illustrated for the case of a
resistive load in Fig. 11. However, with a unipolar BLDC motor load, the intermediate capacitor
voltage has to be greater than the phase back-emf for proper demagnetization of the phases. This
causes a distortion of the input current waveform around his zero-crossings of the input voltage.
This is acceptable because the input current shaping is achieved at no cost to the drive, and as
will be seen, the resulting power factor is better than with the conventional circuit configuration.
There is a practical limit to the power level up to which dc/dc converters can be
operated in DCM. This limit is reached around 300W. The use of unipolar excitation for BLDC
motors beyond this power rating is also not recommended, as bipolar excitation would better
utilize the machine windings. So the proposed topology is well-suited to low-power, low-
performance applications where cost is a major consideration
Fig. 11 Operation with resistive load: (a) input voltage and current

A commercially available fractional horsepower BLDC motor with a phase back-

emf constant of 12 V/Krpm is used in the design example. Because of the low back-emf
constant, the input voltage is chosen to be 50 V peak. A drive with a power rating of 100 W is
designed. The following equations are used for the design [9]
The actual value of C1 should be higher to minimize the voltage ripple caused by the
freewheeling phase currents and is determined by simulation to be 10 .
Fig. 14. Speed response to step change in load torque at 0.3 s.
Fig. 13. Speed reference and speed (rpm).


The operation of the proposed topology has been verified both by simulations and
experiments. A block diagram of the drive system implementation is shown in Fig. 12. The rotor
position is sensed by means of three hall sensors, and the position information is used to
determine the phase winding to be excited. The motor speed is derived from the position inputs
and is compared with the speed reference to generate the current references. Hysteresis control
is used to regulate the phase currents to the reference current waveforms of Fig. 6(b). The dc bus
voltage is regulated by PWM control of the switch S1.
The motor shaft is coupled to a hysteresis brake acting as a load. The controller is
implemented using a TMS320C240 DS evaluation module board from Texas Instruments. It has
two built-in analog to digital converters (ADCs) with eight multiplexed channels each that are
used for converting the analog signals from the current and voltage sensors into the digital values
required by the DSP. Hall-effect based current transducers are used for the phase current sensing,
and an isolation amplifier is used to sense the output voltage. The inbuilt PWM outputs of the
DSP are used to derive the gate signals for the MOSFETs used in the power converter. Opto-
isolators are used to interface the PWM outputs with the MOSFET gates. The commutation
sequence, current, voltage and speed control loops are implemented in DSP software.
A PI controller is used to compare the reference and actual
speed and generate the current reference. The resulting speed response is shown in Fig. 13, and
the speed response to a step change in load torque in Fig. 14. The input current plotted in Fig.
15(a) is seen to follow the input voltage waveform. Fig. 15(b) shows the intermediate capacitor
voltage waveform. In an ideal PFP, this would go to zero in each half-cycle of the input voltage,
but in this case, its minimum value is limited to the peak phase back-emf. This results in some
distortion of the input current around the zero-crossing of the input voltage. The phase currents at
500 rpm are shown in Fig. 15(c).
The experimentally obtained waveforms are shown in Fig. 16. The unipolar operation
of the motor is shown by the current waveform of Fig. 16(a), which also shows the output
voltage. The intermediate capacitor voltage waveform is shown in Fig. 16(b). As seen from Fig.
16(c), the input current follows the input voltage waveform except around the zero-crossings.
The experimentally measured harmonics of the input current are given in Table I. They are
expressed as a percentage of the fundamental current so that the data is independent of the line
voltage magnitude. The corresponding total harmonic distortion (THD) and power factor data are
tabulated in Table II. The performance improvement achieved by using the proposed topology is
evident. A high power factor is also achieved over the entire speed range.

In1 Out1


In1 Out1


In1 Out1

9TH 4.833

In1 Out1 27.17


In1 Out1


Discrete ,
Ts = 2e-006 s In1 Out1
3RD signal THD Scope 5 Scope6

Total Harmonic In1 Out1
Distorsion Display

Scope 7 Diode 1
Diode 2 gates

Diode 3 Goto
Gates Decoder
Scope 3
l1 c1
Gates emf_abc emf_abc Hall
+ i
- A a
Current Measurement
Series RLC Branch 5

B is_ a e_ a

A +
+ Mosfet C <Stator current is _a (A)>
+v - v c2 c Step Tm
- l2

25V/50HZ Series RLC Branch 6 Voltage Measurement1 Three -Phase A <Stator back EMF e _ a (V)> N (rpm)

B -

Voltage Measurement2 V-I Measurement m

Mosfet1 Mosfet3 B
Mosfet2 K-
Universal Bridge <Rotor speed wm (rad/s)>
C rad2rpm



Permanent Magnet <Electromagnetic torque Te (N*m)>

Synchronous Machine
Te (N.m)


+ From
- v Goto1
Voltage Measurement

Scope 1
Scope Goto2

Add1 speed error
f(u) PI
Fcn Add 24
Relational Discrete
Operator PI Controller Constant

Scope 4

A new converter topology based on a SEPIC converter operating in DCM has been proposed for
unipolar excitation of brushless dc motors. The proposed scheme has the following advantages.
1) The proposed converter uses only four controlled switches, all of which are referenced to
ground. This considerably simplifies their gate drive circuitry and results in low cost and
compact packaging.
2) It is capable of bucking or boosting the available input dc voltage to maximize the current-
regulated operation of
the drive.
3) The input current naturally follows the input voltage to a certain extent, reducing the amount
of low-order harmonics and resulting in a high power factor.
4) Eliminates the possibility of shoot-through faults which could occur in bipolar converters.
5) Lower conduction and switching losses because of the presence of only one switch and diode
per phase as opposed to two in the bipolar case





[1] T. Kenjo and S. Nagamori, Permanent-Magnet and Brushless DC Motors. Oxford, U.K.:
Clarendon Press, 1985.
[2] J. R. Hendershot Jr. and T. J. E. Miller, Design of Brushless Permanent Magnet Motors.
Hillsboro, OR: Magna Physics Publishing, 1994.
[3] R. Krishnan and S. Lee, PM Brushless dc motor drive with a new power converter
topology, in Proc. IEEE IAS Annu. Meeting, Oct. 1995, pp. 380387.
[4] R. Krishnan and P. Vijayraghavan, A new power converter topology for PM Brushless dc
motor drives, in Proc. IEEE IECON98 Conf., vol. 2, 1998, pp. 709714.
[5] R. Krishnan, A novel single switch per phase converter topology for four-quadrant PM
Brushless dc motor drive, in Proc. IEEE IAS Annu.
Meeting, vol. 1, Oct. 1996, pp. 311318.
[6] J. Sebastin, M. Jaureguizar, and J. Uceda, An overviewof power factor correction in single-
phase off-line power supply systems, in Proc. IEEE IECON94 Conf., vol. 3, 1994, pp. 1688
[7] J. Skinner and T. A. Lipo, Input current shaping in Brushless dc motor drives utilizing
inverter current control, in Proc. 5th Intl. Conf. Elect.
Mach. Drives, 1991, pp. 121125.
[8] R. P. Massey and E. C. Snyder, High voltage single-ended dc-dc converter, in Proc. IEEE
PESC77 Conf., 1977, pp. 156159.
[9] D. S. L. Simonetti, J. Sebastin, and J. Uceda, The discontinuous conduction mode SEPIC
and Cuk pow er factor preregulators: analysis and
design, IEEE Trans. Ind. Electron., vol. 44, pp. 630637, Oct. 1997.
[10] T. Gopalarathnam, S. Waikar, H. A. Toliyat, M. S. Arefeen, and J. C. Moreira,
Development of low-cost multi-phase Brushless dc (BLDC)
motors with unipolar current excitations, in Proc. IEEE IAS Annu. Meeting, Oct. 1999, pp. 173