2024 IEEE 10th International Power Electronics and Motion Control Conference (IPEMC2024-ECCE Asia) | 979-8-3503-5133-0/24/$31.

00 ©2024 IEEE | DOI: 10.1109/IPEMC-ECCEAsia60879.2024.10567746

BLDC Motor Zero-Speed Startup Position

Sensorless FOC Control
Nicholas Chiang Wei Jian1, Shuyu Cao1, K. J. Tseng1, Liang Boon Wee1, Shan Yin2
1
Singapore Institute of Technology, Singapore
2
Huawei Technologies Co., Ltd, China

Abstract-- This paper presents the BLDC motor BEMF it less attractive in practical application.
observer-based position sensorless control method. Online
parameter identification of the stator inductance and The BEMF voltage-based speed and position software
resistance is implemented to improve the observer accuracy. observer implemented in DQ synchronous reference
Before BLDC motor startup, the rotor is aligned to a known scheme [3] is a feasible and simpler solution for the BLDC
rotor field position using the stator phase current feedback
control. Afterwards, the BLDC speed ramps up to the position sensorless control in high-speed operation. There
minimum target speed in speed open loop control. When the are two major challenges to the BEMF observer based
BEMF voltage is high enough, speed close loop control is sensorless control method. The first challenge is to start the
enabled using the speed and position signals obtained from
BLDC motor at zero speed when the BEMF is too low for
the observer. To minimize the controller state transition, the
stator current vector FOC control is applied in both speed the observer to work. The second challenge is to improve
open loop control and speed close loop control. The observer the accuracy and transient performance of the speed and
is continuously initialized using the speed and position signals position observer. The method for starting up the BLDC
from the speed open loop control. The proposed control
algorithm is implemented in PLECS RT box and verified in motor from a known rotor position was explained in [2]
a BLDC motor 3-phase inverter hardware setup. The and [4]. The basic idea is to align the rotor to a known rotor
measurement results show that the speed and position of the field angle position first. Afterwards, ramps up the speed
observer converge quickly to the speed and position
in open loop control to the minimum speed for the BEMF
measured from the incremental encoder in the speed close
loop control. The proposed BLDC sensorless control scheme observer or zero-crossing BEMF detection circuit to work.
can achieve smooth speed acceleration with minimal stator
current and the stable steady state FOC control with small This paper proposed a reliable control solution to
position error in wide speed operation range. implement BLDC motor position sensorless control based
on the rotor field oriented DQ frame BEMF observer. The
Index Terms—BLDC, position sensorless control, field- on-line BLDC motor parameter identification is
oriented control (FOC), speed and position observer.
implemented to improve the accuracy of BEMF observer.
I. INTRODUCTION The stator phase current close loop control is proposed to
realize the rotor alignment to a known rotor field angle
There are two important position sensorless control
position. The same stator current vector FOC controller is
technologies successfully implemented in some BLDC
applied in both open loop speed ramping up control and
drive industry applications, the hardware-based BEMF
the speed close loop control to minimize the transit
zero-crossing detection technology [1], [2] and the
response of the phase current controller. The DQ frame
software-based rotor BEMF volage observer technology
BEMF voltage-based observer is continuously initialized
[3]. [4]. For the hardware-based BEMF zero-crossing
in the speed open loop control process to minimize the
method, the PWM switching on the selected stator phase
speed and position converge time and reduce the transient
is disabled in a narrow window period to measure the zero-
observer error in the start of the speed close loop control.
crossing instant of the BEMF volage signals to derive the
rotor speed and rotor field angle. This method normally II. BLDC MOTOR ZERO-SPEED STARTUP POSITION
requires a dedicated BEMF zero-crossing detection circuit SENSORLESS CONTROL METHODOLOGY
to predict the next BEMF zero crossing detection window
A. BLDC Motor Parameter Identification for Position
time. The maximum 300 diode freewheeling period could Sensorless Control
limit the feasible operation range of BEMF zero-crossing
method at high speed and heavy load condition [1]. The accuracy of the BLDC motor parameters is critical
to achieve stable and accurate rotor field angle position for
The αβ frame BEMF based position and speed observer sensorless control. For non-salient rotor BLDC motor, the
does not require observer state re-initialization when the stator voltage dynamic equations in the rotor flux field
speed close loop control is enabled [4]. However, the αβ oriented DQ synchronous rotating reference frame can be
frame BEMF observer has heavy computation load and is expressed in (1) and (2).
sensitive to the current harmonic distortion, which makes

979-8-3503-5133-0/24/$31.00 ©2024 IEEE

Authorized licensed use limited to: Amrita School of Engineering. Downloaded on June 04,2025 at 05:15:58 UTC from IEEE Xplore. Restrictions apply.
1282
 = + − (1) The BLDC rotor flux magnitude “ ” can be calculated
from the offline measurement of the rotor BEMF voltage
= + + + (2) when the BLDC machine operates in generator mode in a
test bench with its shaft rotating by a primer mover.
The BLDC motor phase resistance “R”, phase
inductance “L”, and rotor flux magnitude “ ” are three Fig. 1 shows the stator phase voltage control
implemented in a 3-phase inverter for BLDC motor stator
important parameters for the FOC based BLDC position
phase resistance and stator inductance parameter online
sensorless control.
identification.

Fig. 1: Control algorithm for stator phase inductance and stator resistance on-line identification.

inductance identification for the BLDC motor used in the

In the parameter identification process, the same
study. The voltage frequency is 5 kHz. The online
modulation index signals are applied for leg 2 and leg 3
identified phase inductance is 23.575 uH, which is close to
switches. The 3-phase PWM inverter is equivalent to a
30.2 uH measured offline from RCL meter.
single-phase H-bridge inverter. The current flows from the
Phase current (A)
testing phase to the other two phases. There is no BEMF Ia=I Ib=-I/2 Ic=-I/2
10
voltage induced in the stator phases because the rotor is
stationary. 0

-10
For stator inductance identification, high frequency H-bridge AC voltage (V)
U
inverter sine voltage reference “Uac” is applied. When the 10

phase voltage frequency is high enough, the phase shift 0

angle between phase voltage and phase current is close to

-10
900. The ratio of phase reactance and resistance is large 8.90 8.92
Time / s
8.94 8.96 × 1e-2

enough to safely ignore the phase resistance and to Fig. 2: High frequency inverter voltage applied for stator phase
calculate the phase inductance from the measured phase inductance on-line identification.

impedance from (3), where “fH” is the frequency of the Fig. 3 shows the measured stator phase current and the
volage signal. U and I are RMS voltage and RMS current. 1 Hz square voltage reference with change rate limit
applied in the stator resistance online identification
= ∗ (3) process for the BLDC used in the study. The online
identified phase resistance is 128.5 mΩ, which is slightly
The BLDC motor phase resistance is identified online
larger than 85.7 mΩ measured offline using RCL meter.
in the inverter control algorithm using a similar method by
phase current
applying low frequency square waveform with change-rate Ia Ib Ic

10
limit as the voltage reference signal “U*”. In this case, the
volage and current values in the constant high current 0

portion is used to calculate the phase resistance from (4) to

eliminate the voltage across the inductance due to non-zero Va
phase voltage

“LdI/dt”. 2

=
1

(4)
0
6.0 6.5 7.0 7.5 8.0 8.5

Fig. 2 shows the measured stator phase current and Time / s

Fig. 3: Low frequency inverter voltage applied for stator phase

inverter voltage reference signal applied in the online resistance on-line identification.

Authorized licensed use limited to: Amrita School of Engineering. Downloaded on June 04,2025 at 05:15:58 UTC from IEEE Xplore. Restrictions apply.
1283
B. BLDC Motor Zero Speed Startup Position Sensorless sensorless FOC control, the d-axis current reference is
Control Methodology always set to zero to keep the BLDC motor d-axis airgap
Fig. 4(a) shows the control function block diagram of flux magnitude to be same as the rotor flux magnitude.
the proposed BLDC motor zero speed startup FOC-based Fig. 4(b) shows the control algorithm implemented
position sensorless speed control. The motor startup PLECS RT-BOX real-time controller. BLDC motor
control functions are implemented in three startup stages. control state machine is implemented to control the BLDC
The control function blocks are selectively activated or motor startup process and activate different control
deactivated in the different startup stages. In the position algorithms in each BLDC motor startup operation stage.

(a) Sensorless control block diagram (b) Algorithm implementation in PLECS RT-BOX

Fig. 4: BLDC motor zero-speed startup position sensorless FOC control block diagram and implementation algorithm

In the first stage, the phase current feedback controller signal to estimate the speed and rotor position. The
is activated to align the rotor magnet pole to a known rotor estimated speed from the observer is applied for speed
position with a fixed DC stator current vector applied. In feedback control. The estimated position from the observer
the second stage, the BLDC motor will slowly ramp up in is applied for FOC current vector control. After observer
speed open loop control with the current vector FOC is stabilized, the BLDC motor runs in normal FOC speed
control enabled. In this stage, the ramping up speed profile control condition.
is generated in control algorithm, the position profile is C. BLDC Motor Rotor Position Alignment Before Zero
obtained by integrating the speed reference signal from the Speed Startup
known aligned rotor position. The q-axis current reference
Fig. 5 shows the phase current feedback proportional-
is set to a predefined constant to generate the acceleration
integral (PI) control algorithm implemented in the PLECS
torque. In the final stage, the speed and position observer
RT-Box when the phase “a” is selected to align the rotor
and speed feedback PI control are enabled simultaneously.
flux angle to the known rotor position with a predefined
The q-axis current reference of FOC current vector control
phase current magnitude applied. The PWM modulation
comes from the speed control output. The observer uses
index signals are generated according to the phase
the measured phase currents and phase voltage reference
selected.

>0
1 Iph_Polarity
PWM modulation index generation
a, b, c (1, 2, 3) Align with Phase "a" phase_select
phase selection 1
2
3
Iabc_inv Iph
Vph
18 Iph_ref 1
Switch1 2
phase current ref Align with Phase "b" 3

+− e PI(s) u × m_index_phase_alignment
÷
0 Rate Limiter
Divide2
Phase current Zero Phase current
PI control
~= 2 -1
mode 1/2
mode = 2 positon alighment Gain1
Vdc
Gain9 Saturation4
UpperLimit: inf
LowerLimit: 0.000001
Phase current feedabck control
Align with Phase "c"

Fig. 5: Phase current feedback control and PWM modulation index generation for rotor position alignment

from positive DC bus to phase “a”. The inverter current

Fig. 6 illustrates stator and rotor flux vector location
returns to the negative DC bus from phase “b” and “c”. The
after the rotor alignment by the positive phase “a” current
DC phase currents generate the combined stationary stator
feedback control. In this case, the inverter current flows
flux field “ ! ” in same direction as the phase “a” current

Authorized licensed use limited to: Amrita School of Engineering. Downloaded on June 04,2025 at 05:15:58 UTC from IEEE Xplore. Restrictions apply.
1284
vector. The magnet rotor will be rotated by the magnetic The rotor position is obtained by integrating the electrical
attraction force between the stator flux field with rotor angular speed from the known rotor flux angle position
magnet. After rotor alignment, the rotor flux vector “Ψr” obtained from the phase alignment.
opposes the stator flux vector “Ψs”. In this example, the
rotor flux angle is aligned to 1800.

Fig. 7: Speed and position signal generation in open loop control

Fig. 6: Stator and rotor flux vector location after rotor alignment by the E. Speed and Position Observer Algorithm for Speed
positive phase “a” current feedback control Close Loop Control
D. Speed and Position Signal Generation for Speed Open Fig. 8 shows the dq frame extended BEMF volage
Loop Variable Frequency Control
based speed and position observer modified based on the
Fig. 7 shows the algorithm for the rotor position and method presented in [3]. To achieve fast observer
speed signal generation in the speed open-loop control convergence, the observer speed and position are
stage. In the start of the speed open loop control, the speed continuously initialized to the speed and position signal
reference signal will slowly ramp up from zero to the used in the speed open loop. The speed and position
minimum speed required for observer to work with a pre- observer and the speed feedback control are enabled
defined change-rate limit. The mechanical speed in unit of simultaneously at the end of the open loop control when
rpm is converted to the electrical angular speed in “rad/s”. the minimum low-speed target is reached.

Fig. 8: BLDC speed and position observer for speed close loop FOC control

The rotor field angle is estimated from the calculated the stator inductance “L” and stator resistance “R” are
stator bemf voltage generated by the rotor flux from (5) critical BLDC parameters to achieve accurate speed and
and (6) in the estimated rotor flux field synchronous position observation. The impact of the rotor flux
rotating reference DQ reference frame. amplitude is less significant for the accuracy of the speed

#=
and position observer because it is not directly used in the
" − − +$ (5) observer algorithm. The rotor flux amplitude mainly
#=
affects the speed control loop bandwidth and the stator
" − − −$ (6)
BEMF voltage feedforward compensation in FOC current
If the observer position and speed are accurate, the vector control algorithm.
relationship between the actual rotor flux BEMF voltage The position error between the estimated DQ reference
vector and estimated BEMF voltage vector can be derived frame and actual DQ frame can be estimated from (9). For
as (7) and (8) from the DQ frame stator dynamic voltage positive speed reference, the estimated rotor field-oriented
equation (1) and (2). reference frame lags the actual DQ frame. For negative
#=
" =" − − + =0 (7)
speed reference, the polarity of the estimated reference
frame position error becomes positive because the
#=
" = " (8) estimated DQ frame is lagging the actual DQ frame in the
negative rotating direction.
In the rotor flux BEMF voltage-based observer scheme,

Authorized licensed use limited to: Amrita School of Engineering. Downloaded on June 04,2025 at 05:15:58 UTC from IEEE Xplore. Restrictions apply.
1285
 ∗) ,#
& ' sin(& ) ∗ signs( =− sign( ∗
) (9) towards target speed in speed close loop control with
-,# . /,
#. frequency and voltage magnitude both increasing together.

The observer frame error & is minimized by adjusting Fig. 11 shows the FOC control signals and observer
the estimated rotor speed through PI regulation of & . The signals in the speed acceleration process. The observer
estimated speed signal is integrated to generate the position and speed signals are compared with the position
estimated rotor field position. and speed signals from encoder measurement to validate
the accuracy of the rotor flux BEMF voltage-based
III. EXPERIMENTAL RESULTS position and speed observer.
Fig. 9 shows the hardware setup for the BLDC position
sensorless control. The proposed position sensorless
control algorithm is implemented in the PLECS RT box.
The BLDC motor (T-Motor U12 kV60) is driven by a 3-
phase GaN inverter (EPC9173). The inverter DC bus is
connected to a DC power supply. The DC bus voltage is
set to 40 V in the experimental test. An optical incremental
encoder with 1024PPR (RE30E-1024-213-1) is installed
on BLDC shaft to verify the accuracy of the position and
speed observer.

(a) FOC control signals (b) observer signals

Fig. 11: Sensorless control acceleration from 0 rpm to 1000 rpm

In the speed open loop control ramping up stage, the

observer is continuously initialized by the speed and
position signals from the speed and position profile
Fig. 9: BLDC motor position sensorless control hardware setup
generation block. The q-axis current reference is set to a
small constant value to provide sufficient acceleration
Fig. 10 (a) shows stator phase currents measured in
torque. In the speed open loop control, the position signal
oscilloscope from rotor position alignment to speed close
used in FOC control has obvious difference from the
loop control. Fig. 10 (b) shows the phase current signals
encoder measured position. Ud is slowly drifting away
and the voltage command signals from speed open loop
zero in no-load speed open loop control speed acceleration
control to speed close loop control.
process at low speed with a small q-axils current, which
means the rotor angle used in FOC control has larger error,
When the open loop speed reaches 150 rpm, the
observer and speed controller are enabled. The position
error between observer and encoder measurement is
greatly reduced in the speed close loop control. The
observer signals show a larger initial position error
dynamic transition at the start instant of close loop control.
(b) currents & voltages The observer converges quickly to reduce the position
(a) Stator phase currents
error. In the speed close loop control, Ud is close to zero
Fig. 10: Currents and voltages in sensorless control acceleration process in the no-load low-speed acceleration process with the
from 0 rpm towards 1000 rpm
accurate position and speed observation achieved in speed
In the speed open loop control stage, the constant V/F close loop FOC control. The BLDC speed continues to
ratio control is achieved by the FOC current vector ramp up to 1000 rpm, and Uq increases with BLDC speed.
controller with frequency and voltage magnitude both
slowly increase while the stator current magnitude is kept Fig. 12 shows the sensorless speed close loop control in
constant. In the speed close loop control stage, the current steady state operation at 300 rpm. Fig. 12 (a) shows stator
magnitude is greatly reduced after the initial dynamic phase currents measured from oscilloscope. Fig. 12 (b)
transition. The BLDC motor continues to accelerate shows the sensorless control internal signals. The RMS

Authorized licensed use limited to: Amrita School of Engineering. Downloaded on June 04,2025 at 05:15:58 UTC from IEEE Xplore. Restrictions apply.
1286
values of the phase currents from oscilloscope tally with IV. CONCLUSIONS
the values form the control algorithm. The speed and This paper presented a reliable control solution to start
position signals from the speed and position observer the BLDC motor with position sensorless FOC control
match well with the encoder measurement with noticeable from zero speed.
a small constant position error possibly due to rotor
position alignment error. The online parameter identification is implemented to
improve the accuracy of stator inductance and stator
resistance used in the BEMF observer algorithm.
The experimental results show that the proposed BLDC
motor sensorless control can start up the BLDC motor
from zero speed smoothly and reliably from a known rotor
position aligned by the phase current close loop control
before startup. The FOC-based speed open loop control is
stable, and the rotor speed smoothly ramps up to the open
(a) 3-phase stator currents at 300rpm loop target reference. The stator voltage increases with the
stator frequency and the stator current is kept constant in
the speed open loop speed acceleration process. When the
FOC-based speed close loop control and the position and
speed observer are enabled, the rotor speed and position
from the observer quickly converges to actual speed and
position signals measured from incremental encoder
sensor. In steady state operation, the position error and
speed error from the BEMF observer are both very small
in different speed operation conditions.
The future work will focus on design optimization of
the PI controller applied the observer algorithm for speed
signal generation to minimize the transient speed and
(b) Sensorless speed close loop control signals at 300 rpm position error in BLDC motor dynamic operation.
Fig. 12: Sensorless control steady state operation at 300 rpm
REFERENCES
Fig. 13 shows the sensorless control internal signals in
steady state operation at 1000 rpm. In high-speed [1] L. Yang, Z. Q. Zhu, H. Bin, Z. Zhang and L. Gong, "Safety
Operation Area of Zero-Crossing Detection-Based
operation, BLDC motor low frequency speed variation is Sensorless High-Speed BLDC Motor Drives," in IEEE
reduced. The small position error is constant in both high Transactions on Industry Applications, vol. 56, no. 6, pp.
speed and low speed operation. Compared to the RMS 6456-6466, Nov.-Dec. 2020.
[2] C. L. Baratieri and H. Pinheiro, "An I-F starting method for
values of 300 rpm, the stator currents are slightly increased
smooth and fast transition to sensorless control of BLDC
in 1000 rpm, and the stator voltages are obviously motors," 2013 Brazilian Power Electronics Conference,
increased in 1000 rpm. Gramado, Brazil, 2013, pp. 836-843.
[3] S. Morimoto, K. Kawamoto, M. Sanada and Y. Takeda,
"Sensorless control strategy for salient-pole PMSM based
on extended EMF in rotating reference frame," in IEEE
Transactions on Industry Applications, vol. 38, no. 4, pp.
1054-1061, July-Aug. 2002.
[4] A. Ungurean, V. Coroban-Schramel and I. Boldea,
"Sensorless control of a BLDC PM motor based on I-f
starting and back-EMF zero-crossing detection," 2010 12th
International Conference on Optimization of Electrical and
Electronic Equipment, Brasov, Romania, 2010, pp. 377-
382.
[5] A. Bosso, C. Conficoni, D. Raggini and A. Tilli, "A
Computational-Effective Field-Oriented Control Strategy
for Accurate and Efficient Electric Propulsion of Unmanned
Aerial Vehicles," in IEEE/ASME Transactions on
Fig. 13: Sensorless control steady state operation at 1000 rpm Mechatronics, vol. 26, no. 3, pp. 1501-1511, June 2021.

Authorized licensed use limited to: Amrita School of Engineering. Downloaded on June 04,2025 at 05:15:58 UTC from IEEE Xplore. Restrictions apply.
1287