Journal of ELECTRICAL ENGINEERING, VOL 68 (2017), NO2, 99–108

Analysis of sensorless control of brushless DC motor

using unknown input observer with different gains
∗ ∗∗ ∗∗∗
Mitesh B. Astik, Praghnesh Bhatt, Bhavesh R. Bhalja

A sensorless control scheme based on an unknown input observer is presented in this paper in which back EMF of the
Brushless DC Motor (BLDC) is continuously estimated from available line voltages and currents. During negative rotation of
motor, actual and estimated speed fail to track the reference speed and if the corrective action is not taken by the observer,
the motor goes into saturation. To overcome this problem, the speed estimation algorithm has been implemented in this
paper to control the dynamic behavior of the motor during negative rotation. The Ackermans method was used to calculate
the gains of an unknown input observer which is based on the appropriate choice of the eigenvalues in advance. The criteria
to choose eigenvalue is to obtain a balance between faster convergence rate and the least noise level. Simulations have
been carried out for different disturbances such as step changes in motor reference speed and load torque. The comparative
simulation results clearly depict that the disturbance effects in actual and estimated responses minimizes as observer gain
setting increases.
K e y w o r d s: back EMF estimation, BLDC motor, observer design, observer gain, sensorless control, speed estimation

1 Introduction method gives an error at low speeds [6-8]. In the flux link-
age estimation method, rotor position can be obtained by
The use of Brushless DC Motors is extensively in- estimating the flux. As integration of voltage is done for
creasing for various applications due to their compact comparatively large time, this scheme also give estima-
and robust structures. Compared to Permanent Mag- tion error at low speeds [9, 10]. Ogasawara and Akagi
net Synchronous Motors; the BLDC motors have higher [11] proposed an estimation scheme in which rotor po-
power/weight and higher torque/current ratio [1]. A sition is estimated by detecting switching status of free-
BLDC motor requires an inverter and position sensors to wheeling diode. In this scheme, current flow through the
perform commutation process as the motor does not have freewheeling diodes are used to identify the zero crossing
commutator and brushes. However, the position sensors of back EMF. This method requires six comparators and
present many disadvantages like increase in machine size, six isolated power sources.
reduction in reliability and higher noise. Thus, many re- Above mentioned methods are not suitable where the
searchers have suggested application of sensorless drive high estimation accuracy of speed and position are re-
which can regulate speed and position without using quired as they cannot provide continuous rotor position
shaft-mounted position sensors [2]. Main types of sen- estimation. Terzic and Jadric [12] proposed a new scheme
sorless control methods are the back-EMF measurement, for estimation of speed and rotor position of a BLDC
third harmonic sensing of back EMF, back EMF inte- motor using Extended Kalman Filter (EKF). Only us-
gration, flux linkage estimation, freewheeling diode and ing stator voltages and currents measurements, EKF can
estimator-based approach [3-12]. estimate the motor state variables. In this method, fil-
The method for measurement of back EMF requires tering of voltage and current signals do not required but
additional power supply for a comparator to detect free- this method needs complex computing algorithm and it
wheeling current [3, 4]. Shen et al [5] proposed a scheme suffers from the initial-value problem.
to detect the third harmonic of back EMF. This scheme In this work, sensorless control of the brushless dc mo-
is only valid for high speed as third harmonics detection tor has been proposed by designing unknown input ob-
is difficult due to constant addition of noise signal at server. The observer design requires the input voltage and
minimum speed range. Additional hardware is required current signals of BLDC motor to estimate back EMFs
to detect the third harmonic of back EMF at minimum which eliminate the need of rotor speed measurement.
speed. In integration of back EMF method, commuta- The observer should be designed to give the optimal per-
tion instants can be detected by comparing the threshold formance over wide speed range and in case of reversal
value with the integration of the back-EMF waveforms of the motor rotation. Generally, the observer fails when
of the unexcited phase. Due to integration process, this the motor reverses its direction of rotation. The perfect

* Department of Electrical Engineering, A. D. Patel Institute of Technology, New Vallabh Vidyanagar, India, ** Department of
Electrical Engineering, C. S. Patel Institute of Technology, CHARUSAT, Changa, India, ***Department of Electrical Engineering, Indian
Institute of Technology, Roorkee, India, [email protected]

DOI: 10.1515/jee-2017–0014, Print (till 2015) ISSN 1335-3632, On-line ISSN 1339-309X c 2017FEI STU

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
100 M. B. Astik, P. Bhatt, B. R. Bhalja: ANALYSIS OF SENSORLESS CONTROL OF BRUSHLESS DC MOTOR USING UNKNOWN . . .

Three-phase PWM inverter BLDC motor

Q1 Q3 Q5 Rs L ea
ia

Rs L eb
vdc ib
N
Q2 Q4 Q6 Rs L ec
ic

Fig. 1. PWM Inverter connected to BLDC Motor

Table 1. Switching table for IGBTs

Rotor
position Reference stator
electrical currents IGBT switching sequence
degree
ia ib ic Q1 Q2 Q3 Q4 Q5 Q6
0-30 0 -1 1 0 0 0 1 1 0
30-90 1 -1 0 1 0 0 1 0 0
90-150 1 0 -1 1 0 0 0 0 1
150-210 0 1 -1 0 0 1 0 0 1
210-270 -1 1 0 0 1 1 0 0 0
270-330 -1 0 1 0 1 0 0 1 0
330-360 0 -1 1 0 0 0 1 1 0

ea
E 2 Modelling of BLCD motor
ia
0 The brushless dc motor equivalent circuit with three-
Q
phase PWM inverter is shown in Fig. 1, and the switching
-E sequence of IGBTs is shown in Table 1. Gate signals
eb
E for IGBTs are generated using commutation signals and
ib comparing the actual and reference current through the
0
Q hysteresis current controller. The speed and torque can
be adjusted by varying the gate pulse of IGBTs.
-E
In case of BLDC motor, the nature of the back EMF is
E trapezoidal. At any time only two windings are conduct-
ec
ic ing and third winding current is zero [13]. Fig. 2 shows the
0
Q back-EMFs and phase currents waveform for BLDC mo-
-E I II III IV V VI I tor. To produce constant torque, controlling is required to
get quasi square waveform for three-phase stator currents
0 p/6 p/2 5p/6 7p/6 9p/6 11p/6 2p
Fig. 2. EMF and phase current waveforms for BLDC motor to coordinate with the trapezoidal back EMF [14].
The three-phase voltage equations for the BLDC mo-
tor can be written as
tracking of reference speed is required by the actual and
estimated speed to avoid the motor to go into saturation     
Va Rs 0 0 ia
especially for negative direction of rotation. Hence, speed  Vb  =  0 Rs 0   ib  +
estimation algorithm has been implemented to overcome Vc 0 0 Rs ic
this limitation and different case studies have been pre-       (1)
sented to validate the proposed approach. L 0 0 i e
d  a  a
The paper presents modelling of BLDC motor includ- + 0 L 0  ib + eb ,
dt
ing the observer design for back EMF estimation. The 0 0 L ic ec
validation of observer design, proposed algorithm and dif-
ferent observer gains, carried out by performing detailed Where: Va , Vb and Vc are the per phase stator volt-
simulations and the results at both transient, and steady- ages; ia , ib and ic are per phase stator currents; Rs is
state conditions are provided. per phase stator winding resistance; L = Ls − M ; LS is

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
 Journal of ELECTRICAL ENGINEERING 68 (2017), NO2 101

the per phase stator winding inductance; M is the mu- From the variations of estimated quantities one finds
tual inductance; ea , eb and ec are the per phase back
EMFs. db
e  
= g2 i − bi . (8)
All stator winding resistances as well as self and mu- dt
tual inductances are assuming constant [14-15]. Any of
the stator phase voltages in (1) can be represented as Observer error equation, which shows the difference be-
tween actual and estimated value, then is
di     
v = Ri + L + e. (2) d ξ1 a − g1 a12 ξ1
dt = 11 ,
dt ξ2 −g2 0 ξ2
with    
3 Observer design for back emf ξ1 i − bi
= . (9)
ξ2 e − eb
In this paper, a sensorless control scheme has been
proposed using back EMF estimation. The voltage and The objective is to find out the optimal value of the
current at the stator terminal are measurable quantities, observer gains and so that the observer error can be re-
hence they are considered as known state variables. The duced to approximately zero. The methods for calculating
back EMF of brushless DC motor cannot be measured the observer gains are discussed in next section.
directly; hence estimation is required with the help of
unknown input observer. The time derivative of the stator
4 Observer gains calculation
current with the use of (2) is

di R 1 1 Ackermans method can be used to determine the state

= − i + v − e, (3) variable feedback gain matrix after placing the system
dt L L L
poles at the preferred positions. The desired eigenvalues
while the general form of state space representation is of the observer are set as λ1 and λ2 . This will result in
following characteristic polynomial.
ẋ = Ax + Bu, y = Cx + Du (4)
α(s) = (s − λ1 )(s − λ2 ) =
where (10)
s2 − (λ1 + λ2 ) s + λ1 λ2 .
    | {z } | {z }
a11a12 b −α1 α0
A= , B = 11 , C = [ 1 0 ] ,
0 0 0
      According to Cayley-Hamilton theorem [11], the charac-
i i v
D = 0, x= , y= , u= , teristic equation can be expressed in terms of matrix A
e 0 0
as
and, α(A) = A2 + α1 A + α0 I, (11)
R 1 1 giving, by the Ackermans formula, the observer gain ma-
a11 = − , a12 = − , b11 = .
L L L trix G, with Q being the observability matrix for the pair
Thus, the current and back EMF of the brushless DC (A, C)
motor can be represented in state variable form as      
       g1 −1 0 a11 + α1
d i G= = α (A) Q =
a a12 i b11 g2 1 α0 /a12
= 11 + v,    
dt e 0 0 e 0
  (5) C 1 0 (12)
Q= =
i CA a11 a12
y = Cx = [ 1 0 ] = i.
e
For stator resistance Rs = 0.2 Ω and inductance
To design the observer, the back emf is supposed to be L = 8.5 mH, a11 = −23.53 and a12 = −117.65 are ob-
a constant disturbance, hence its time derivative de = 0 . tained. By choosing eigen values λ1,2 = −1000 ± j 1200
dt
As the system expressed by (5) is totally observable; the for the above mentioned motor parameters resulting in
state observer can be represented in full-order by α0 = 2440000 and α1 = 2000 the observer gains are
g1 = 197647 and g2 = −20740 . In this paper, differ-
ḃ x + Bu + G (y − b
x = Ab y) , (6) ent eignvalues ie −1000 ± j 1200 , −2500 ± j 4000 , and
−1500 ± j 8000 , were chosen to validate the effect of ob-
where G is observer gain matrix server gains on the motor performance.
  The selection of the gains and the eigenvalues for the
g1 observer design are interrelated. The choice of the eigen-
G= . (7)
g2 value is very critical to get the optimal performance. The

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
102 M. B. Astik, P. Bhatt, B. R. Bhalja: ANALYSIS OF SENSORLESS CONTROL OF BRUSHLESS DC MOTOR USING UNKNOWN . . .

Stator current (A) where, kT is torque constant [21] given by

20
ia ib
10 P
kT = λp (14)
0 2
(a)
-10 where, P – number of poles, λp – flux linkage established
ic
-20 by magnets.
Speed (rpm)
Figure 3 shows the dynamic responses of currents,
800 Estimated speed
speeds, torques and line to line back EMFs for change
in reference speed for every 0.4 s without the use of speed
400 (b) estimation algorithm. As shown in Fig. 3, the motor works
0 Actual speed
satisfactory for positive rotation ie up to 0.6 s. When
the reference speed signal becomes zero, the actual speed
Torque (Nm)
20 becomes negative as undershoot occurs in speed due to
Actual torque use of PI controller. If corrective action is not taken by the
0
observer during this situation, the said undershoot results
(c) into saturation and it deteriorates dynamic behavior of
Reference torque BLDC motor.
-20
Line to line back EMF (V)
Actual back EMF
100 Estimated back EMF Estimated values of back EMF from the observer (eab, ebc,& eca)

0
(d) Introduce low-pass filter to add Take the absolute and
-100 phase delay in back EMF max.value of back EMF

0 0.4 0.8 Time (s) 1.2

Convert back EMF from abc frame to
dq frame using (16) and (17)
Torque
Fig. 3. Dynamic responses of stator currents, speeds, torques and ×
÷ constant
back EMFs of phase-ab for the change in reference speed from 300 k T using (15)
Determine rotor position using (18)
rpm to 0 rpm to 300 rpm, keeping load torque ( TL ) = 0 without
using speed estimation algorithm
Ù
No wm
Rotating direction ×
smaller value of the eigenvalue makes the response slug- is positive? ×

gish. Theoretically, the eigenvalues can be moved arbi-

Yes
trarily toward minus infinity, yielding faster convergence
Ù Ù
rate. However, observer act like a differentiator and it Dθe+ (k) > 0 Dθe+ (k) < 0
& &
becomes very sensitive to noise and other disturbances Ù Ù
Dθe- (k) > 0 Dθe- (k) < 0
for very large eigenvalues [19]. Hence, the eigenvalues are
chosen to have a balance between the faster convergence Ù Ù Ù Ù
rate and the least sensitivity to noise. θe (k) = θe+ (k) θe (k) = θe- (k)

Ù Ù
w (k) > 0 w (k) < 0
5 Proposed algorithm and simulation
block diagram for BLCD motor

Fig. 4. Speed estimation algorithm for BLDC motor

First the motor speed is estimated with the ratio of
estimated back EMF to torque constant [20]. Then pro-
posed speed estimation algorithm is implemented cor- This problem cannot be solved by using only (13).
rectly to control the motor speed when reference speed Therefore, to overcome this difficulty in practical applica-
becomes negative. The complete block diagram represen- tions, proposed speed estimation algorithm is suggested
tation of the proposed sensorless control of brushless dc to control the dynamic behavior of BLDC motor. Fig-
motor is discussed further. ure 4 shows the speed estimation algorithm for BLDC
motor which is implemented in this paper.
5.1 Rotor speed and position estimation The low-pass filter is used to add phase delay in back-
EMF. Back EMF observer and low-pass filter are used
The estimated speed of the motor is expressed as the
for speed estimation algorithm. The line to line values of
ratio of estimated back EMF to torque constant and given
estimated back EMF is converted into dq form given by
by
eb 1
bm =
ω (13) ebds = (b
eab − ebca ) (15)
kT 3

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
 Journal of ELECTRICAL ENGINEERING 68 (2017), NO2 103

A
Three-phase Braking Three-phase Bruschless
B
recifier chopper inverter DC motor
C
220V,60Hz Gate ia ib ic
three-phase signals
AC source for Calculate
inverter line-to-line
Limiter
* Hysteresis ia currents
w*m PI Is
controller current ib
Vdc controller ic
Ù iab ibc ica
wm
Sa Calculate vab
vbc Back EMF
Sb line-to-line vca
voltage observer
Sc Sa Sb Sc
Develop Ù
commutation eab Ù
ebc Ù
eca
signal
Ù Rotor speed estimation
Q algorithm using
torque constant k T
1/s
Ù
wm

Fig. 5. Block diagram of the proposed sensorless control of BLDC motor

1 the difference between actual and desired value of speed

ebqs = − √ ebbc (16)
3 can be find out. This output is fed to the PI controller
The rotor position is determined from the dq components which decides the current set value depending upon the
of back EMF and is given by gain values of controller. The torque reference divided by
the torque constant kT would give a reference stator cur-
 
ebqs rent Is ∗ . The gate signals for IGBTs are generated using
θbe = tan−1 . (17) commutation signals and comparing the actual and ref-
ebds
erence current through the hysteresis current controller.
Now, suppose θbe+ (k) is the estimated rotor position If the difference between reference and actual current go
beyond a set hysteresis band, the upper switch of three-
for the positive direction of rotation and θbe− (k) is the
phase inverter is turned on and the lower switch is turned
estimated rotor position for the negative direction of ro-
off. As these difference crosses the lower band limit, the
tation. Angular displacements are obtained by comparing
these values with their respective position values in the upper switch is turned off and the lower switch is turned
previous sampling instant as expressed by on. The logic for the generation of gate pulses for the in-
verter is shown in Appendix. These gate signals decide
∆θbe+ (k) = θbe+ (k) − θbe+ (k − 1) (18) exact voltage to be applied across the BLDC motor.

∆θbe− (k) = θbe− (k) − θbe− (k − 1) (19)

If the rotating direction is positive, the angular dis- 6 Simulation results and discussions
placement ∆θbe+ (k) > 0 and ∆θbe− (k) > 0 . In this case,
the actual rotor position θbe (k) is equal to θbe+ (k) and Various simulations have been carried out on brushless
speed direction is positive. If the rotating direction is dc motor to validate the effect of observer and speed
negative, the angular displacement ∆θbe+ (k) < 0 and estimation algorithm on its dynamic performance. The
∆θbe− (k) < 0 . In this case, the actual rotor position θbe (k) effects of different values of observer gain along with the
variations in rotor reference speed and load torque have
is equal to θbe− (k) and speed direction is negative.
been evaluated in the proposed work. The parameters of
5.2 Block diagram representation of sensorless control BLDC motor used in MATLAB/Simulink environment
of BLDC motor are given in Table 2 (Appendix).
Figue 5 shows the overall block diagram of the pro- In this work, the gains of the observer are calculated
posed sensorless control for BLDC motor. The line to line by limiting the real part of the eigenvalue to the certain
voltage is found from DC-link voltage and commutation extent in order to reduce the effect of disturbance and
signals. The speed and rotor position are calculated using noise. Ackermans method was used to calculate the ob-
the estimated back EMF and proposed algorithm (Fig. 4). server gains. The following cases are considered for the
Estimated speed is fed to the error detector from which validation of the proposed methods.

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
104 M. B. Astik, P. Bhatt, B. R. Bhalja: ANALYSIS OF SENSORLESS CONTROL OF BRUSHLESS DC MOTOR USING UNKNOWN . . .

Stator current (A) Stator current (A)

ia ib ia
10 10
ib

0 0
(a) (a)
ic
-10 -10
ic
Speed (rpm) Speed (rpm)
400
Estimated speed 300
Estimated speed
200 200
Actual speed (b) (b)
100 Actual speed
0
0
Torque (Nm) Torque (Nm)
20 20
Reference torque Actual torque
Actual torque
10
0
(c) (c)
Reference torque 0
-20 -5
Line to line back EMF (V) Line to line back EMF (V)

40 Estimated back EMF Estimated back EMF Actual back EMF

40

0 0
(d) (d)

-40 Actual back EMF -40

0 0.4 0.8 Time (s) 1.2 0 0.2 0.4 0.6 Time (s) 1.0

Fig. 6. Dynamic responses of stator currents, speeds, torques and Fig. 7. Dynamic responses of stator currents, speeds, torques and
back EMFs of phase-ab for the changes in reference speed from 300 back EMFs of phase-ab for the changes in load torque ( TL ) from
rpm to 0 rpm and again to 300 rpm, keeping load torque ( TL ) = 0 10 Nm to 0 Nm to 10 Nm keeping reference speed at 300 rpm

6.1 Step changes in reference speed from 300 rpm to 0 6.2 Step changes in load torque from 10 nm to 0 nm to
rpm and again 300 rpm for every 0.4 s, with load 10 nm, with rotor reference speed set at 300 rpm
torque equal to zero
As shown in Fig. 7(b), the BLDC motor starts at 0 s
and gradually attains 300 rpm within 0.2 s and remains
Figure 6 shows the variations in stator currents, rotor
300 rpm up to 0.4 s. In Fig. 7(c), with the change in
speeds, torques and back EMFs of phase-ab for the case
load torque from rated steady state torque to 10 Nm at
under consideration. The gains of the observer are set at 0.4 s, an increase in stator currents is observed as shown
g1 = 1976.47 and g2 = −20740 . From Fig. 6(a), it is in Fig. 7(a). At 0.6 s, the load torque is set to 0 Nm
observed that only two windings are conducting at any with the associated changes in stator currents. Similar
time and third winding current is zero, which exactly responses are found for further change in load torque at
matches with the response shown in Fig. 2 and polarity 0.8 s. The momentarily change in rotor speed can be seen
of currents given in Table 1. from Fig. 7(b) with the change in load torque and after
The variations in actual and estimated rotor speed that the rotor attains its reference speed. The variations
can also be observed in Fig. 6(b), which indicate that for in actual back EMF of phase-ab also keep tracking with
every step change in rotor reference speed, the estimated estimated back EMF as shown in Fig. 7(d). Here, the
values of observer gains g1 and g2 remain as mentioned
and actual speed always keep tracking with the rotor
in section 6.1.
reference speed. This proves the effectiveness of the speed
estimation algorithm with the proper gain settings of
the observer. The actual torque successfully tracks the 6.3 Effects of variation in observer gains on the dy-
reference torque at all instants even during step changes namic performance of the motor
in the rotor reference speed which can be observed from The relative difference between actual and estimated
Fig. 6 (c). Similarly as shown in Fig. 6(d), the actual back values of rotor speed, back EMF and rotor position are
EMF tries to track the estimated back EMF at all instants depicted in Figs 8-13 with variation in observer gains.
even during step changes in rotor reference speed. Following two cases are considered.

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
 Journal of ELECTRICAL ENGINEERING 68 (2017), NO2 105

Speed diference (rpm)

12 EMF diference (V)
g1 = 1976.47 and g2 = -20740 g1 =1976.47 and g2 = -20740
4
8

4 2

0 0
12 3
g1 = 2976.47 and g2 = -49810 g1 = 2976.47 and g2 = -49810
8 2

4 1

0 0
12
g1 = 4976.47 and g2 = -189125 g1 = 4976.47 and g2 = -189125
1.2
8
0.8
4
0.4
0 0
12
g1 = 9976.47 and g2 = -756500 g1 = 9976.47 and g2 = -756500
0.6
8
0.4
4 0.2

0 0.4 0.8 Time (s) 1.2 0 0.4 0.8 1.2

Time (s)

Fig. 8. Dynamic responses of difference between actual and esti- Fig. 9. Dynamic responses of difference between actual and esti-
mated speed due to changes in reference speed from 300 rpm to 0 mated back EMF of phase-ab due to changes in reference speed
rpm to 300 rpm, keeping load torque ( TL ) = 0 Nm with different from 300 rpm to 0 rpm to 300 rpm, keeping load torque ( TL ) = 0
observer gains Nm with different observer gains

Rotor position difference (rad) Speed difference ( rpm)

0.03
g1 =1976.47 and g2 = -20740 g1 =1976.47 and g2 = -20740
4
0.02

2
0.01

0.00 0
0.02 g1 = 2976.47 and g2 = -49810 g1 = 2976.47 and g2 = -49810
4

0.01
2

0 0
4
g1 = 4976.47 and g2 = -189125 g1 = 4976.47 and g2 = -189125
0.006
2
0.003

0 0
4
g1 = 9976.47 and g2 = -756500
0.003 g1 = 9976.47 and g2 = -756500
2

0.001

0 0.4 0.8 Time (s) 1.2 0 0.2 0.4 0.6 0.8 1.0
Time (s)

Fig. 10. Dynamic responses of difference between actual and es- Fig. 11. Dynamic responses of difference between actual and es-
timated rotor position due to changes in reference speed from 300 timated speed due to changes in load torque ( TL ) from 10 Nm to
rpm to 0 rpm to 300 rpm, keeping load torque ( TL ) = 0 Nm with 0 Nm to 10 Nm, keeping reference speed at 300 rpm with different
different observer gains observer gains

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
106 M. B. Astik, P. Bhatt, B. R. Bhalja: ANALYSIS OF SENSORLESS CONTROL OF BRUSHLESS DC MOTOR USING UNKNOWN . . .

6.3.1 Rotor reference speed is changed from 300 rpm two back EMF starts decreasing during this change and
to 0 rpm to 300 rpm for every 0.4 s, with varying becomes zero as long as the rotor reference speed is set
observer gains to 0 rpm. The higher value of observer gains significantly
As we move from first window to last window of Fig. 8, reduces this difference of back EMF and improves the
the transient response reveals that the difference between dynamic performance of the motor as depicted in Fig. 9.
actual and estimated rotor speed gradually reduces with It has been observed from Fig. 10 that the responses of
the selection of higher values of gains. Also, it is revealed difference between actual and estimated rotor positions
from Fig. 8 that the oscillations are completely damped also vary with respect to change in observer gains. The
out and the difference between actual and estimated rotor choice of the higher values of observer gains again helps
speed becomes zero in steady state with higher values of to minimize the difference between actual and estimated
gains as compared to the lower one. rotor positions.
When the rotor reference speed is changed from 300
rpm to 0 rpm, the actual and estimated speeds become 6.3.2 Step changes in load torque from 10 Nm to 0 Nm
negative due to action of PI controller at t = 0.6 s. This to 10 Nm keeping rotor reference speed at 300
causes the sudden change in the difference between actual rpm with varying observer gains
and estimated rotor speed. The higher values of gains can Step change is applied in load torque at different in-
reduce this sudden change as shown in Fig. 8. terval of time ie rated steady state torque to 10 Nm at
The response of the difference between actual and es- 0.4 s, 10 Nm to 0 Nm at 0.6 s and 0 Nm to 10 Nm at
timated back EMF of phase-ab is shown in Fig. 9 with 0.8 s. The responses of the difference between actual and
variation in observer gains. Initially the motor attains its estimated rotor speed for the case under consideration is
rated speed of 300 rpm within 0.2 s and remains con- shown in Fig. 11. The choice of higher values of observer
stant up to 0.4 s. During this period when the motor gains results in reduced difference and also minimizes the
speed approaches its rated speed, the difference between oscillations.
actual and estimated back EMF increases gradually and In this case, the changes in load torque causes a slight
remains maximum. The step changes in rotor reference variation in rotor speed; hence the back EMF is also the
speed from 300 rpm to 0 rpm is applied at 0.4 s. It can least affected. As a result, the difference in actual and esti-
be seen from the Fig. 9 that the difference of the said mated back EMF of phase-ab remains maximum through-

EMF difference ( V) Rotor position difference ( rad)

6
0.03
g1 =1976.47 and g2 = -20740
4 0.02

2 g1 =1976.47 and g2 = -20740 0.01

0 0.00
0.02 g1 = 2976.47 and g2 = -49810
3

2
g1 = 2976.47 and g2 = -49810 0.01
1

0 0.00
0.008
g1 = 4976.47 and g2 = -189125
1.2

0.004
0.6 g1 = 4976.47 and g2 = -189125

0.0 0.000
0.8

0.003
0.4 g1 = 9976.47 and g2 = -756500
g1 = 9976.47 and g2 = -756500
0.001

0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Time (s) Time (s)

Fig. 12. Dynamic responses of difference between actual and esti- Fig. 13. Dynamic responses of difference between actual and esti-
mated back EMF of phase - ab due to changes in load torque ( TL ) mated rotor position due to changes in load torque ( TL ) from 10
from 10 Nm to 0 Nm to 10 Nm, keeping reference speed at 300 rpm Nm to 0 Nm to 10 Nm, keeping reference speed at 300 rpm with
with different observer gains different observer gains

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
 Journal of ELECTRICAL ENGINEERING 68 (2017), NO2 107

out the simulation as shown in Fig. 12. The higher val- the least noise level. Hence, observer gains cannot be in-
ues of observer gains significantly reduces the difference creased to infinity. It has been observed that the higher
in back EMF and improves the dynamic behavior of the values of the observer gain results in more accuracy. The
motor as depicted in Fig. 12 for the case under consid- effectiveness of the higher values of observer gains is ver-
eration. The satisfactorily responses are obtained for the ified for different disturbances such as changes in rotor
rotor position difference with the higher values of observer reference speed and load torque. The dynamic responses
gains as depicted in Fig. 13. of differences in rotor speed, back EMF and rotor position
clearly depicts the lower overshoot and reduction in oscil-
lations with higher value of observer gain. It is observed
Appendix that the proposed algorithm can give accurate speed even
near zero speed range. Moreover, the proposed algorithm
accurately estimates the rotor position which exactly co-
Table 2. The logic for the generation of gate pulses for the inverter
incides with the actual rotor position.

Currents Switches
(i∗a − ia ) > HB 1 ON 2 OFF References
(i∗a − ia ) < HB 1 OFF 2 ON [1] P. Pillay and R. Krishnan, ”Application characteristics of per-
(i∗b − ib ) > HB 3 ON 4 OFF manent magnet synchronous and brushless dc motors for servo
drives”, IEEE Trans. Ind. Appl. , 27, no. 5, pp. 986-996, Sep/Oct
(i∗b − ib ) < HB 3 OFF 4 ON 1991.
(i∗c − ic ) > HB 5 ON 6 OFF [2] N. Matsui, ”Sensorless PM brushless dc motor drives”, IEEE
Trans. Ind. Electron. , 43, no. 2, pp. 300-308, 1996.
(i∗c − ic ) < HB 5 OFF 6 ON
[3] K. Iizuka, H. Uzuhasashi, M. Kano, T. Endo and K. Mohri,
”Microcomputer control for sensorless brushless motor”, IEEE
Trans. Ind. Appl. , IA-21, no. 3, pp. 595-601, May 1985.
where: i∗a , i∗b and i∗c and ia , ib and ic are the reference [4] C. -T. Lin, C. -W. Hung and C. -W. Liu, ”Sensorless control for
and the actual three phase stator currents respectively four-switch three-phase brushless DC motor drive”, Conf. Rec.
IEEE IAS Annual Meeting, 4, pp. 2048-2053, 2006.
and HB is the hysteresis band.
[5] J. X. Shen, Z. Q. Zhu and D. Howe, ”Sensorless flux-weakening
control of permanent-magnet brushless machines using third
harmonic back EMF”, IEEE Trans. Ind. Appl. , 40, no. 6,
Table 3. BLDC motor parameters pp. 1629-1636, 2004.
[6] T. M. Jahns, R. C. Becerra and M. Ehsani, ”Integrated cur-
Stator resistance ( Rs ) 0.2 Ω rent regulation for a brushless ECM drive”, IEEE Trans. Power
Electron. , 6, no. 1, pp. 118-126, January 1991.
Stator inductance ( Ls ) 8.5 mH
[7] R. C. Becerra, T. M. Jahns and M. Ehsani, ”Four-quadrant
Rotor inertia ( J ) 0.089 kg m 2 sensorless brushless ECM drive”, Proc. APEC’91, pp. 202-209,
1991.
Friction ( B ) 0.005 Nm s
[8] J. X. Shen and S. Iwasaki, ”Sensorless control of ultrahigh-speed
Number of pole pair ( p ) 4 PM brushless motor using PLL and third harmonic back EMF”,
Rated speed ( ωs ) 300 rpm IEEE Trans. Ind. Electron. , 53, no. 2, pp. 421-428, April 2006.
[9] N. Ertugrul and P. Acarnley, ”A new algorithm for senseless op-
Flux linkage established by magnets 0.175 Vs eration of permanent magnet motors”, IEEE Trans. Ind. Appl.
Proportional Controller gain ( kp ) 1.22 , 30, no. 1, pp. 126-133, Jan/February 1994.
[10] R. Wu and G. R. Slemon, ”A permanent magnet motor drive
Integral Controller gain ( ki ) 46
without a shaft sensor”, IEEE Trans. Ind. Appl. , 27, no. 5,
Current hysteresis bandwidth ( HB ) 0.01 A pp. 1005-1011, Sep/Oct 1991.
[11] S. Ogasawara and H. Akagi, ”An approach to position sensorless
drive for brushless dc motors”, IEEE Trans. Ind. Appl. , 27,
no. 5, pp. 928-933, Sep/Oct 1991.
[12] B. Terzic and M. Jadric, ”Design and implementation of the ex-
7 Conclusion tended Kalman filter for the speed and rotor position estimation
of brushless DC motor”, IEEE Trans. Ind. Electron. , 48, no. 6,
pp. 1065- 1073, December 2001.
This paper presents sensorless control method for the
[13] I. Janpan, R. Chaisricharoen and P. Boonyanant, ”Control of
brushless dc motor with the use of the unknown input the brushless dc motor combine mode”, Procedia Engineering,
observer. The rotor position and speed can be estimated pp. 279 - 285, 2012.
during transient as well as in steady state period from the [14] Chang-Liang and Xia, ”Permanent magnet brushless DC motor
line to line estimated back EMF without any additional drives and controls”, John Wiley & Sons Singapore Pte. Ltd. ,
hardware. The speed estimation algorithm has been suc- 2012.
cessfully implemented in this paper and it gives guaran- [15] C. Debjyoti, C. Madhurima and R. Priyanka, ”Modeling and
simulation of cost effective sensorless drive for brushless dc
tee to control the dynamic behavior of the motor during motor”, International conference on Computational Intelligence
negative rotation. The criteria to choose eigenvalue is to Modeling, Techniques and Applications (CIMTA- 2013), Proce-
obtain a balance between the faster convergence rate and dia Technology 10, pp. 279 - 286, 2013.

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM
 108 M. B. Astik, P. Bhatt, B. R. Bhalja: ANALYSIS OF SENSORLESS CONTROL OF BRUSHLESS DC MOTOR USING UNKNOWN . . .

[16] R. S. Burns, ”Advanced control engineering”, Butterworth ME degree in electrical power system from BVM Engineering
-Heinemann, 2001. College, Sardar Patel University, Vallabh Vidhyanagar, Gu-
[17] A. K. Mandal, ”Introduction to control engineering - modeling, jarat, India in 2002 and received PhD from S.V. National In-
analysis and design”, New Age International Pvt. Ltd. , 2006.
stitute of Technology (SVNIT), Surat, Gujarat, India in 2012.
[18] G. S. Phadke, ”Control system design”, Vipul Prakashan, 2008.
He has published around 20 papers in journals and conferences
[19] D. G. Luenberger, ”An introduction to observer”, IEEE Transac-
at international and national levels. He is currently working
tion on Automatic Control, Ac-16, no. 6, pp. 596-602, December
1971. as a Professor and Head of Department of electrical engineer-
[20] J-W. Park, S. -H. Hwang and J. -M. Kim, ”Sensorless control of ing, C.S. Patel Institute of Technology, CHARUSAT, Changa,
brushless DC motors with torque constant estimation for home Gujarat, India. His areas of interest are Power System Dynam-
appliances”, IEEE Trans. Ind. Appl. , 48, no. 2, pp. 677- 684, ics and Stability, Grid Integration of Wind Power Generation,
March/April 2012. Distributed Generation and Power System Protection. He is a
[21] R. Krishnan, ”Permanent magnet synchronous and brushless member ofIEEE, IEEE Power and Energy Society and Indian
DC motor drives”, CRC Press, 2010.
Society of Technical Education (ISTE).
Received 23 August 2016 Bhavesh R. Bhalja was born in 1979 in Gujarat, In-
dia. He received BE degree in electrical engineering and ME
Mitesh B. Astik, was born in 1980 in Gujarat, India. He degree in power system engineering from B.V.M. Engineering
received the BE degree in electrical engineering from Sardar College, Sardar Patel University, Vallabh Vidyanagar, India in
Patel University, Gujarat, India in 2003, the ME degree in
1999 and 2001, respectively. He obtained PhD degree in elec-
electrical engineering from Mumbai University, Maharashtra,
trical engineering from Indian Institute of Technology, Roor-
India in 2006. He is currently working toward the PhD degree
kee, India in 2007 and currently working as an Associate Pro-
at C.S. Patel Institute of Technology, CHARUSAT, changa,
Gujarat, India. Since 2008, he has been with the Department fessor in the department of electrical engineering, IIT Roorkee,
of Electrical Engineering, A.D. Patel Institute of Technology, India. He has published more than 50 papers in journals and
India. His research interests are in the areas of electrical ma- conferences at international and national levels. His research
chines and electrical control systems. He is a member ofIndian interests include power system protection, automation, plan-
Society of Technical Education (ISTE). ning and design, system modeling and simulation and artifi-
Praghnesh Bhatt was born in 1978 in Gujarat, India. He cial intelligence. He is also involved in many research project
received BE degree in electrical engineering from L.D. College funded by various research agencies. He got Young Engineers
of Engineering, Gujarat University, Ahmedabad, India in1999, Award in 2009 by Institution of Engineers, India.

Brought to you by | provisional account

Unauthenticated
Download Date | 6/23/17 3:18 PM