POSITION ESTIMATOR AND SIMPLIFIED CURRENT CONTROL STRATEGY

FOR BRUSHLESS-DC MOTORS, USING DSP TECHNOLOGY

Juan Dixon (SM IEEE), Matías Rodríguez and Rodrigo Huerta

Department of Electrical Engineering
Pontificia Universidad Católica de Chile
Casilla 306, Correo 22, Santiago, Chile
phone 56-2-686-4278
fax 56-2-552-2563
e-mail jdixon@ing.puc.cl

Contact author. Dr. Juan W. Dixon

Preferred topical area: DSP control, BLDC Motor control
Subject area and keywords: machine drives, electric traction.
POSITION ESTIMATOR AND SIMPLIFIED CURRENT CONTROL STRATEGY
FOR BRUSHLESS-DC MOTORS, USING DSP TECHNOLOGY

Department of Electrical Engineering

Pontificia Universidad Católica de Chile
Casilla 306, Correo 22, Santiago, Chile
phone 56-2-686-4278
fax 56-2-552-2563
e-mail jdixon@ing.puc.cl

ABSTRACT

This paper describes a different way to sense the phase currents and to estimate de rotor
position of a Brushless DC (BLDC) motor. The current is sensed taking the absolute value of two of
the three phases, transforming this information in a dc current IMAX, which is finally compared with a
reference value from the accelerator pedal. With this method of control, all the transistors of the
inverter are commutated with the same PWM signal. Based on the last property of the current
controller developed, the paper proposes a method to estimate the instantaneous position of the rotor.
The method is based on the determination of the current slopes during the conduction periods. The
main characteristic of this type of motor, fed with quasi-square-wave currents, is that it only needs a
six-position sensor, and one current controller for its full torque control. The solution proposed
makes use of information contained in the slopes of the currents, calculating the six commutation
points required. This method is only applicable while current slope variations can be sensed, so it
needs to be complemented with a starting method. The system was implemented using a fast digital
signal processor (TMS320F241) which is programmed with a closed loop PI control for the phase
currents. The processor also makes all the calculations required for position estimation. Additionally,
the PWM signals are transmitted through a fiber optic link to minimize noise production and
possibilities of error on commutations. Because of a lack of BLDC in laboratory, the motor tested
was a 12kW (16HP) permanent magnet synchronous motor (PMSM), with sinusoidal back emf. This
machine has been tested with quasi-square current waveforms using the method proposed, and fed
with an IGBT inverter working at 15 kHz commutation frequency. Experimental results of the
currents and the way the slopes of current locus are obtained are shown.
POSITION ESTIMATOR AND SIMPLIFIED CURRENT CONTROL STRATEGY
FOR BRUSHLESS-DC MOTORS, USING DSP TECHNOLOGY

Juan Dixon (SM IEEE), Matías Rodríguez and Rodrigo Huerta

Department of Electrical Engineering
Pontificia Universidad Católica de Chile
Casilla 306, Correo 22, Santiago, Chile
phone 56-2-686-4278
fax 56-2-552-2563
e-mail jdixon@ing.puc.cl

Abstract. This paper describes a different way to sense the not have brushes, and their rotors are robust because
phase currents and to estimate de rotor position of a commutator and/or rings do not exist. That means very low
Brushless DC (BLDC) motor. For current control, the maintenance. This also increases the power-to-weight ratio and
current is sensed taking the absolute value of two of the the efficiency. For induction motors, flux control has been
three phases, transforming this information in a dc current developed, which offers a high dynamic performance for
IMAX, which is finally compared with a reference value electric traction applications [1,2]. However, this control type
from the accelerator pedal. With this method of control, all is complex and sophisticated. The development of brushless
the transistors of the inverter are commutated with the permanent magnet machines [3-5] has permitted an important
same PWM signal. Based on the last property of the simplification in the hardware for electric traction control.
current controller developed, the paper proposes a method Today, two kinds of brushless permanent magnet machines for
to estimate the instantaneous position of the rotor. The traction applications are the most popular: i) the Permanent
method is based on the determination of the current slopes Magnet Synchronous Motor (PMSM), which is fed with
during the conduction periods. The main characteristic of sinusoidal currents, and ii) the Brushless DC Motor (BDCM),
this type of motor, fed with quasi-square-wave currents, is which is fed with quasi-square-wave currents. These two
that it only needs a six-position sensor, and one current designs eliminate the rotor copper losses, giving very high
controller for its full torque control. The solution proposed peak efficiency compared with a traditional induction motor
makes use of information contained in the slopes of the (around 95 % and more in Nd-Fe-B machines in the range 20
currents, calculating the six commutation points required. to 100 kW). Besides, the power-to-weight ratio of PMSM and
This method is only applicable while current slope BDCM is higher than equivalent squirrel cage induction
variations can be sensed, so it needs to be complemented machines. The aforementioned characteristics and a high
with a starting method. The system was implemented using reliability control make this type of machine a powerful
a fast digital signal processor (TMS320F241) which is traction system for electric vehicle applications [6]. However,
programmed with a closed loop PI control for the phase sensing the phase currents and the position of the rotor are two
currents. The processor also makes all the calculations of the drawbacks that this type of machine have. In this paper,
required for position estimation. Additionally, the PWM a control system for brushless DC motors, based on a DSP
signals are transmitted through a fiber optic link to from Texas Instruments is proposed. The DSP used is the
minimize noise production and possibilities of error on TMS320F241, which has been programmed to produce a
commutations. Because of a lack of BLDC in laboratory, simple way to control the machine currents, and to evaluate the
the motor tested was a 12kW (16HP) permanent magnet instantaneous position of the rotor. The method is based on 1)
synchronous motor (PMSM), with sinusoidal back emf. the measurement of the currents, based on a common dc
This machine has been tested with quasi-square current current IMAX, which is obtained taking the absolute values of
waveforms using the method proposed, and fed with an two of the three real phase currents, and 2) the calculation of
IGBT inverter working at 15 kHz commutation frequency. the commutation instants based on the slope variations of this
Experimental results of the currents and the way the slopes current IMAX.
of current locus are obtained are shown. The most popular way to control BDCM for traction
applications is through the use of voltage-source current-
I. INTRODUCTION controlled inverters. The inverter must supply a quasi-square
current waveform, whose magnitude, IMAX, is proportional to
In electric traction, like in other applications, a wide the machine shaft torque [7]. Then, by controlling the phase
range in speed and torque control for the electric motor is currents, torque and speed can be adjusted. There are two ways
desired. The DC machine fulfils these requirements, but this to control the phase currents of a BDCM: 1) through the
machine needs periodic maintenance. The AC machines, like measurement of the phase currents, which are compared and
induction motors, and brushless permanent magnet motors do
forced to follow a quasi-square template, and 2) through the was mentioned before, the DSP calculates the currents from
measurement of the dc link current, which is used to get the two of the three phases, takes their absolute values and makes
magnitude of the phase-currents, IMAX. In the first case, the a process similar to a rectification inside the chip. Later on,
control is complicated, because it is required to generate three, this rectified current, called IMAX, is compared with a reference
quasi-square current templates, shifted 120º for the three coming from the accelerator or brake pedal, and the error
phases. Besides, these current templates are not easy to follow signal is processed through a PI controller. The output of the
for the machine currents, because of phase-shifts and delays PI controller is compared with a saw-teeth carrier signal, to
introduced [8]. In the second case, it is difficult to measure the generate the PWM for the power transistors. At the same time,
dc current, because the connection between transistors and the the position estimator discriminates which couple of the six
dc capacitors in power inverters are made with flat plates to transistors of the inverter should receive this PWM signal. The
reduce leakage inductance. Then, it becomes difficult to information about direction (neutral, forward or reverse) is
connect a dc current sensor. To avoid those drawbacks, the also defined at this stage.
equivalent dc current is obtained through the sensing of two
the three armature currents. From these currents, the absolute TaU TbU TcU
i
IMAX
t
value is taken, and a dc component, which corresponds to the
+ BLDC
amplitude IMAX of the original phase currents, is obtained. This VDC ia

dc component is then used to drive the BDCM. This solution, TaL TbL
ib MOTOR
TcL
which was proposed in [9] using discrete (analog and digital) |ia|, |ib|
components, has been implemented now using DSP Position
technology. Besides, all the PWM signals for the six transistors OPTO DRIVERS
Estimator
are transmitted from the control to the inverter through fiber (The Brake Control
reverses the position
optics. estimator signals)
IMAX
For PMDC motors various methods for obtaining rotor
Brake
position and speed have been proposed in the literature. These Digital Selector
control
methods are called observers and are based on mathematical _ IMAX
models of the motor. They need an accurate knowledge of + Accelerator
error Reference
PWM PI
motor parameters, and some of them are: 1) The Luenberger COMP e(t) IREF Command Brake
Observer [10], which is one of the simplest, and has a linear Saw-teeth
Carrier
feedback. It is of a deterministic type and does not take into Modulator Control Circuit
account system and measurement noise. 2) The Sliding Mode
DEAD TIME
Observer [11] that has a similar structure to the previous GENERATOR
observer, but the error feedback signal is a non-linear IMAX
switching function, and 3) The extended Kalman filter [12], PROTECTION
SYSTEM RESET
which is a non-linear, recursive, stochastic filter, based on
noise properties of systems. Its feedback gain is a function of
the covariance matrices of the system and measurement noise.
Figure 1. Hardware representation of the DSP controller, to
This last solution needs knowledge of noise properties and is
manage the phase currents of the machine
computationally more intensive than the previous methods.
For BLDC motors, a precise determination of the rotor
The phase currents of figure 1 are sensed using LEMs
angle is not necessary. It is only required to know the position
(current sensors), which are passed through analog-to-digital
of commutation points, because the objective is to achieve
converters. The digital information of the currents is then
quasi-square current waveforms, with dead time periods of
processed into the DSP.
60º.
It is important to mention that all the operations displayed
This paper has been divided into two main research areas;
in figure 1, with exception of some conditioning hardware, are
the first consists on the design and implementation of a current
executed inside the DSP. Figure 2 shows the block diagram of
control system for the BLDC machine, and the second part
the duties realized inside and outside the DSP. These
describes the design of the logic able to calculate the
characteristics give a great flexibility to the overall control
instantaneous position of the rotor. All this calculations are
system of the machine.
realized with the DSP TMS320F241. However, the main
The IMAX current, which carries information about the dI/dt
contribution that this paper wants to show is the way the
in each commutation produced by the PWM, is also used for
instantaneous position of the rotor is evaluated.
estimation of the instantaneous position of the rotor. This
procedure is explained in the next paragraph. However, it is
II. CURRENT CONTROL SYSTEM very important to say that the position estimator system
The circuit of fig.1 shows, in an equivalent hardware developed here becomes feasible because the PWM signal is
configuration, the basic topology of the current controller, the same for all the six transistors of the inverter. This fact
which has been implemented for software inside the DSP. As will be clarified later on.
 As can be seen in figure 3, the slopes of both, phase-to-
Iref + PI PWM PWM Fiber Optic Inverter - Motor phase and phase-to neutral voltages change abruptly during
Error Controller Generator Signal Transducer System
_ commutation. Then, either of these voltages can be used for
position detection. But, since the neutral point of the machine
I_max DSP TMS320F241 is floating and not accessible, the only way to use this
Analog Ia information is through the phase-to-phase induced voltages.
Max Digital
ABS(Ia,Ib) Low-Pass ADC Low pass
Filter Ib Looking at the phase-to-phase voltages of figure 3, the method
Filter proposed for commutation instances detection, is based on the
fact that, for optimum operation, the commutation points of
phase a (switching-on) are produced exactly at the point where
Figure 2. Control scheme based on DSP the induced phase-to-phase voltage between phases b and c
changes its slope from zero to a non-zero values very rapidly.
III POSITION ESTIMATOR SYSTEM The same happens for phases b and c respect to (Egc-Ega) and
(Ega-Egb) respectively. But the problem is that the change in
The position estimator for a brushless dc machine needs to slope of these voltages and the voltages themselves are not
detect six positions, which determine the commutation points. measurable, and then a special strategy has to be used. In this
The diagram of figure 3 shows an ideal commutation sequence kind of machine, with only two of the three phases conducting
for the trapezoidal emf of the machine, with their flat 120º at any time, the circuit to be considered will include the two
maximums, centered with the phase-to-neutral currents for phases in conduction. The model for a BLDC motor being
optimal commutation. used is shown in the figure 4.
Ia IMAX

Ib
a)
Ic

Ega

Egb
b)
Egc Figure 4. Motor Circuit Model

When a couple of phases of the motor are conducting, i.e.

Ega-Egb the current IMAX flowing from phase a to phase b, the
differential equation to be solved is:

dI MAX
Egb-Egc V DC = 2 R ⋅ I MAX + 2 L ⋅ + ( Ega − Egb ) (1)
dt
c) where VDC is the dc supply voltage of the inverter, R and L are
the stator winding resistance and inductance respectively, IMAX
the phase currents being controlled by the PWM, and Ega-
Egc-Ega
Egb, the phase-to-phase back emf. Similar equations can be
written for the other two phase-to-phase induced voltages.
As the current is kept at the reference value by the
controller, the term 2R·IMAX does not change significantly, and
a -c b -a c -b a -c then:
dI MAX
Commutation points for a BLDC motor 2L ⋅ + ( Ega − Egb) ≈ K (2)
dt
Figure 3. Ideal currents and induced voltages, and the
commutation points for a BLDC motor. where K=constant
a) phase currents
b) phase-to-neutral voltages d 2 I MAX d
c) phase-to-phase-voltages 2L ⋅ 2
≈ − ( Ega − Egb) (3)
dt dt
 Despite IMAX is kept around a constant reference value, de PWM signal makes the current increase with a positive slope
differential terms of IMAX does not disappear, because of the IMAX+. When a sudden change in the slope is produced, the
slopes. These slopes balance the instantaneous magnitude of corresponding phase is commutated. In the case of sinusoidal
the induced voltage (Ega-Egb). Then, the information about emf, when the minimum of dI/dt is reached, then a delay of 30º
the commutation points for position estimation can be obtained is applied to switch-on the current in the corresponding phase.
from the slopes of IMAX , avoiding the calculation of (Ega- This also means 30º after d2I/dt2=0. The algorithm is applied
Egb). sequentially to commutate the phases a, b, c and so on. The
Now, because the current controller implemented here uses figure shows the case for both the extreme situations: ideal
the same PWM for all the transistors, when two phases are in trapezoidal emf, and sinusoidal emf, but the method can also be
operation during their corresponding 60º, both the transistors applied to any emf waveform. To have success with any
commutate at the same time. Then, the positive slope of the induced voltage waveform, the first step should be to obtain
resultant phase current always satisfies equations (2) and (3). experimentally this waveform by an open-circuit test of the
For the negative slopes, there is only a change in the sign of machine.
these equations. Then, (2) and (3) allow determining the Ic is switched-on Ia is switched-on
commutation points through the positive slopes of the current IMAX
IMAX, and with the corresponding changes in the sign of these
equations, the negative slopes can also be utilized for the
calculations. The figure 5 shows the only possible topologies d/dt(IMAX + )
during the conduction period of 60º mentioned before. Trapezoidal signals

INVERTER BLDC MOTOR IMAX

IMAX 2L
Ega +
+ VDC V DC d/dt(IMAX + )
L R
2R
L R Egb
Ega-
Egb 30º
Positive slope
d 2 /dt2(IMAX + )
Equivalent 0
circuit
Commutation Commutation
INVERTER BLDC MOTOR instants instants
2L
IMAX
Ega Sinusoidal signals
+ 60º 60º
VDC L R 2R
-V DC + IMAX + : positive slopes IMAX -: negative slopes
L R Egb
Ega-
Negative slope
Egb Figure 6. Calculation of d/dt, and d2/dt2, through the variation
Equivalent of the positive slopes of IMAX
circuit

Figure 5. The two possible topologies during positive At this point, it is quite important to mention that the
conduction of phase “a” strategy proposed in this work, does not depends on the values
of R and L, because is based on current slope variations
The commutation points are always on the opposite phase, regardless of it's magnitude. Even more, they can change in
that means, the sudden variation in slope of voltage (Ega-Egb) time because of many factors, but the slope variations used in
will give information to switch-on phase c, and so. However, this method will be produced always at the same places. The
by looking at equation (3) it is easy to realize that any change system does not use the value of the slope. It uses the change
in (Ega-Egb) will produce an opposite change in the slope of of the slopes with time to detect the commutation points.
dI/dt. Then it is not required to evaluate the emf voltages but Since the inverter tested was operated at a frequency of 15
only the changes in the slope of dI/dt. The sudden change in kHz, in order to obtain phase current measurements that can
the slope of the voltage will also produce a sudden change in give information on dI/dt, the sampling frequency must be
the slope of dI/dt, and this point will indicate the instant of faster than the 15 kHz. The DSP used has a minimum
commutation. In the case of sinusoidal emf (like the machine conversion time of 2µs which is equivalent to 500 kHz
tested in our laboratory) the commutation points should be sampling rate. In order to allow time for calculations, samples
located 30º after the slope reaches the minimum value, or 30º are taken only for a period of 30 µs, either on the positive
after d2I/dt2 equals zero. After commutation is produced, a new slope (conducting semi period of the PWM) or on the negative
couple of phases will be conducting, and they will give slope (nonconducting semiperiod). This situation will depend
information for the next commutation. The figure 6 shows how on the PWM duty cycle (bigger or smaller than 0.5
dI/dt variation is obtained from the positive slopes of the phase respectively). Special care has to be taken in order to avoid
current, in agreement with eq. (2). The current controller of the sampling currents on noise transient instants, like the ones
machine maintains the phase-currents with a maximum value produced just after a commutation has occurred. For this
IMAX, and the slope of IMAX is being computed only when the reason, a delay of 4 us is implemented before acquisitions are
done. The figure 7 shows the instants where the samples are β̂ = S −1 X ′Y (8)
taken for the positive slope case. The current displayed here, is
a real current from the machine.
Where, X is the independent variable, Y the dependent and
S = X’X.
According to the Gauss-Markoff Theorem, the best
(minimum-variance) linear (linear functions of the yi) unbiased
estimation of ß is given by the least squares. Solving for ß as a
function of the measured data,

β  1  ∑ xi2 ∑ y i − ∑ xi ∑ xi y i 
β =  1  =  
 − ∑ x y + n∑ y x 
 β 2  n ∑ ( xi − x )
2
 i i i i 

(9)

ß2 is the slope of the line and ß1 the value of E(y) when

x=0. An important fact of this result is that for fast calculations
and a constant sampling rate (not necessarily uniform), ß can
be calculated as:

Figure 7. Current sensing points to evaluate positive slopes. β = (S −1 X ′)Y (10)

One of the important features of this algorithm is that it Where the first term S-1X’ depends entirely on the
needs the information of the corresponding phase current on independent values of xi, which can be precalculated and be
the period of time before commutation is applied. This is not assumed as a constant vector. Then in the DSP, ß is a simple
difficult to obtain because there are relatively clean signals, as vector product. This property makes real time calculations
was shown in figure 7. feasible.
To get a reliable information about the dI/ dt (d/dt[IMAX+]), As shown in figure 7, each sample is obtained averaging
a linear statistical model was used to calculate the desired two points taken every two microseconds, and a total of six
slope [13]. The general linear hypothetical model of full rank pairs of samples are taken to calculate the slope with the
used is, previous method. All the acquisition process takes just 28
µsecs that means, around 43% of the total period of the PWM
Y = Xβ + e (4) signal. The problem is when the duty cycle of the inverter
becomes smaller than 43%. In this case, the negative slope
The form of the frequency function of the error (e) due to should be evaluated for position estimation. This possibility is
acquisition errors and noises unspecified, is assumed to have a not yet implemented, but there is no problem in doing it.
zero mean, One of the drawbacks of the system is the impossibility to
have information of rotor position when the rotor speed is too
E (e ) = 0 (5) slow or the machine is not running. This problem is under
study through the knowledge of inductance variation related
and a covariance matrix, with the position of the rotor.

E (ee′) = σ 2 I (6) IV. EXPERIMETAL RESULTS

Using the method of least squares, that is, finding the value As our laboratory does not have a permanent magnet
of ß where the sum of squares of e is minimum, the equation to brushless machine with trapezoidal back emf, the experiments
minimize it is: were done with a machine with sinusoidal emf.
The oscillogram of figure 8 shows the phase currents of the
n machine, for a speed of 1000 rpm. The figure displays the
′
∑e 2
i = e′e = (Y − Xβ ) (Y − Xβ ) (7) currents in phases a and b, and the magnitude of IMAX is 50
i =1 amps. The current variations during conduction are not very
clear, but figure 7 gives some help in showing the cleanless of
The least-squares estimation of ß is, the slopes of the phase currents.
 IM A X IMAX

d/dt(IMAX+)
IM A X = 5 0 [ A ] theorical
S p e e d = 1 0 0 0 [rp m ]
Ia is switched-on

d/dt(IMAX+)
real
Hysteresis band
Figure 8. Phase currents of the machine
Figure 10. Switching instant of a BLDC motor, based on
hysteresis band crossing point.
As the motor tested has sinusoidal emf, the second
derivative of IMAX+ can give precise information of the
V. CONCLUSIONS
switching instants. The figure 9 shows the oscillogram of the
d/dt(IMAX+). It can be observed that the signal, obtained with A different way to sense the phase currents and to control, in
the method explained in figure 6, is very noisy. However, with a sensorless way, a Brushless DC (BLDC) motor for electric
a digital filter programmed inside the DSP, this signal can be vehicle applications, has been presented. The current is sensed
cleaned to get a better waveform of d/dt(IMAX+). It should be taking the absolute value of two of the three phases,
remembered that the instants of commutation are delayed 30º transforming this information in a dc current IMAX, which is
with respect to the minimum (or 30º delayed when finally compared with a reference value from the accelerator
d2/dt2(IMAX+)=0). Then, the delay of 1.3 milliseconds in the pedal. The paper also presents a solution of the problem of the
filtering (see figure 9) means a maximum period of operation position sensor. The method is based on the determination of
of the machine of 15.6 milliseconds. This is traduced to a the current slopes during the conduction periods. The solution
maximum frequency of 64 Hertz for the inverter. Higher proposed makes use of information contained in the slopes of
frequencies will mean an unavoidable delay in the switching the currents, calculating the six commutation points required
instants. For this reason, is still under study de improvement in for this kind of machine. This method is only applicable while
getting the position information to reduce the delay. current slope variations can be sensed, so it needs to be
The problem described above is for brushless motors with complemented with a starting method. The system was
sinusoidal emf. However, for a real brushless dc motor this implemented using a fast digital signal processor
problem does not exist, because in this case the detection (TMS320F241) which is programmed with a closed loop PI
method is based on a hysteresis band and not on signal control for the phase currents. The processor also makes all the
filtering. The figure 10 shows how the switching instant is calculations required for position estimation. Additionally, the
obtained for a real brushless dc machine. PWM signals are transmitted through a fiber optic link to
7 minimize noise production and possibilities of error on
commutations. Because of a lack of BLDC in laboratory, the
6
motor tested was a 12kW (16HP) brushless permanent magnet
5
motor, with sinusoidal back emf, which is fed with an IGBT
inverter working at 15 kHz commutation frequency.
4
d I [A ]

ACKOWLEDGEMENTS
3

The authors want to thank Conicyt through Project

2
Fondecyt 1020982, for the support given to this work.
1
APPENDIX
0
0.02 0.04 0.06 0.08 0.1 0.12 0.14
Tim e [s ] The specific linear model used for equations 4) to 9) was:
+
Figure 9. Oscillograms of the d/dt(IMAX )
a) unfiltered signal y i = β 1 + β 2 x i + ei (a)
b) filtered signal
Where ßi are unknown scalar constants and xi are known. [6] C. C. Chan and K. T. Chau, “An Overview of Power
Refering to eq. (4) the matrices needed are: Electronics in Electric Vehicles”, IEEE Trans. on
Industrial Electronics, Vol. 44, Nº 1, February 1997, pp.
1 x1   y1  3-13.
   
1 x2  β   y2  [7] J. M. D. Murphy and F. G. Turnbull, “Power Electronic
X = β =  1  Y =  (b)
... ...   β2  ... Control of AC Motors”, Wheaton & Co. Ltd., Exeter,
    1988, pp. 424-428.
1 x n  y 
  n
[8] J. Dixon, S. Tepper and L. Morán, “Practical
Then, calculating S, Evaluation of Different Modulation Techniques for
Current-Controlled Voltage Source Inverters”, IEE
Proceedings on Electric Power Applications, Vol.143,
 n
S = X ′X = 
∑x i 
 (c)
Nº4, July 1996, pp. 301-306.
 ∑ xi ∑x 2 
i  [9] J. Dixon and I. Leal, “Current Control Strategy for
Brushless DC Motors, Based on a Common DC
and Signal”, IEEE Transactions on Industrial Electronics,
March 2002.
1  ∑ xi2 − ∑ xi 
S −1 =   (d) [10] L. A. Jonesand and J. H. Lang, "A State Observer for
n∑ ( xi − x )
2 − ∑ x n 
 i the Permanent Magnet Synchronous Motor", IEEE
Transactions on Industrial Electronics, Vol.36, No.3,
With the previous results, and replacing it on eq. (8), eq (9) pp. 374-382, August 1989.
is obtained.
[11] T. Furuhashi, S. Sangwongwanich and S. Okuma, "A
REFERENCES Position-and-Velocity Sensorless Control for Brushless
DC Motors Using an Adaptive Sliding Mode Observer",
[1] L. Ben-Brahim y A. Kawamura, “A Fully Digitised IEEE Transactions on Industrial Electronics, Vol.39,
Field-Oriented Controlled Induction Motor Drive No.2, pp.89-95, April 1992
Using Only Current Sensors”, IEEE Trans. on
Industrial Electronics, Vol. 39, Nº 3, June 1992, pp. [12] R. Dhaouadi, N. Mohan and L. Norum, "Design and
241-249. Implementation of an Extended Kalman Filter for the
State Estimation of a Permanent Magnet Synchronous
[2] Y. Xue, X. Xu, T. G. Habetler y D. M. Divan “A Stator Motor", IEEE Transactions on Power Electronics,
Flux-Oriented Voltage Source Variable-Speed Based Vol.6, No.3, pp.491-497, July 1991
on DC Link Measurement”, IEEE Trans. on Industry
Applications, Vol. 27, Nº5, September/October 1991, [13] Franklin A. Graybill, "An Introduction to Linear
pp. 962-069. Statistical Models. Volume I", McGraw-Hill, United
States of America, 1961, pp.106-148.
[3] A. R. Millner, “Multi-Hundred Horsepower Permanent
Magnet Brushless Disc Motors”, IEEE Applied Power
Electronics Conference, APEC’94, February 13-17
1994, pp. 351-355.

[4] P. Pillay, R. Krishnan, “Application Characteristics of

Permanent Magnet Synchronous and Brushless DC
Motors for Servo Drives”, IEEE Trans. on Industry
Applications, Vol. 27, N°5, Sept/Oct. 1991, pp. 986-
996.

[5] T. Low and M.A. Jabbar, “Permanent-Magnet Motors

for Brushless Operation”, IEEE Trans. on Industry
Applications, Vol. IA-26, N°1, Jan/Feb 1990, pp. 124-
129.