This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication.

Back-EMF Sensorless Control Algorithm for High

Dynamics Performances PMSM
Fabio Genduso, Rosario Miceli member, IEEE, Cosimo Rando and Giuseppe Ricco Galluzzo

Abstract— In the paper a low time consuming and cost be submitted to Electro-Magnetic Interference (EMI) coming
sensorless control algorithm for high dynamics performances from external sources, producing an error in measurement that
Permanent Magnet Synchronous Motors (PMSM) both surface may be significant for feedback control. Sensorless drives are
or internal permanent magnet mounted for position and speed
estimation is introduced discussed and experimentally validated. a viable solution for all the mentioned situations in which the
This control algorithm is based on the estimation of rotor speed presence of a speed/position sensor is an element of weakness
and angular position starting from the back electromotive force in the drive.
space vector determination without voltage sensors by using the Various sensorless control techniques have been developed
reference voltages given by the current controllers instead of the for Brushless DC and Brushless AC machines and many
actual ones. This choice obviously introduces some errors that
must be vanished by means of a compensating function. are still in development. All the solutions proposed during
The novelties of the proposed estimation algorithm are the the last years have advantages and drawbacks and cannot be
position estimation equation and the process of compensation of considered resolutive for the widest range of applications and
the inverter phase lag that also suggests the final mathematical for the different types of motors. References [1]-[26] give an
form of the estimation. idea of the manifolds of the proposed solutions and of the
The mathematical structure of the estimation guarantees a
high degree of robustness against parameters variation as shown development during long time of the state of the art issued to
by the sensitivity analysis reported in the paper. sensorless control of Brushless drives.
Experimental verifications of the proposed sensorless control Fundamentally two big families of sensorless control for
system have been made with the aid of a flexible test bench for IPMSM are issued. The first develop the difference of saliency
Brushless Motor Electrical Drives. The test results presented in between direct and quadrature axis and the rotor position
the paper show the validity of the proposed low cost sensorless
control algorithm and, above all, underline the high dynamic information is derived from the signal conditioning of the
performances of the sensorless control system also with a reduced currents output due to the injection of high frequency voltage
equipment. on the stator winding (see e.g. [5],[6],[7]). In this way the
Index Terms— Brushless machines, Synchronous motor drives, machine is used itself as a resolver device. Signal injection on
Transducers, Control equipments the stator is effective and allows the control of the IPMSM
also at standstill, but it is quite complicated and require great
attention in the design of the control system and of the signal
I. I NTRODUCTION
conditioning system to extract rotor position. The latter is

I N FIELD ORIENTED control for Brushless Machines

the exact knowledge of rotor angular position is needed.
The rotor angular position and speed of an electrical ma-
based on back Electro Motive Force (EMF) calculated by
integration of the total flux linkage on the stator phase circuits
(see e.g. [1],[8],[11],[16]). This system is simpler but cannot
chine can be obtained in real time with specific sensors assure control at standstill or at very low speed and suffer
(encoders, resolvers), but their presence may introduce several also for flux integrator’s drift problem, especially in the analog
disadvantages. As a matter of fact, the choice of angular realizations. Furthermore it is very sensitive to the variations
position and speed sensors and the consequent choice of the of the stator resistance during operation. Actually due to fast
related signal conditioning circuit is not much influenced by enhancement in microcomputer technology the application of
electrical drive rated power. Then the influence of the above classical state observers of modern control theory such as
mentioned sensor’s cost grows up in percentage with respect Luemberger’s, optimal state observers, extended Kalman filter
to the electrical drive comprehensive cost, when the rated and so on, are gaining a renewed attention (e.g. [11],[25],[23]).
power of electrical machine is small or fractional. Finally the In this paper, a novel low time consuming and low cost sen-
signal transmission between sensor and control system can sorless control algorithm for permanent magnet synchronous
motor (PMSM) drives, both surface or internal permanent
Manuscript received 13-Jun-2009, Revised 20-Aug-2009. Copyright
°c 2009 IEEE. Personal use of this material is permitted. However, permission magnet mounted, based on the determination of the back elec-
to use this material for any other purposes must be obtained from the IEEE tro motive force (back-EMF) without the aid of voltage probes,
by sending a request to [email protected] is presented, discussed and experimentally verified. Estimation
The Authors are with the Department of Electrical Electronical
and Telecommunication Engineering, University of Palermo, Italy, Viale of the back EMF is made via the reference voltages given
delle Scienze zip code 90128, Palermo. Email: [email protected] by the current controller. Sensorless control with only current
[email protected] [email protected], [email protected]. This sensors is not new (see e.g. reference [19], [20], [21], and
work has been supported by MIUR (Italian Ministry of University and
Research) and by the Sustainable Development and Energy Saving Laboratory [22]), in particular in the reference [21] the reference voltages
(SDES Laboratory) part of the UNINETLAB of the University of Palermo. given by the current controllers are used for estimation with

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

identification of the machine parameter. This identification

concurs to make the estimation system quite still complex did
and computationally expensive. In a previous paper (see [22]) vd = Rid + Ld − Lq Pp ωiq
dt
an early version of the proposed control algorithm have been diq
discussed and experimentally tested. In the same paper the vq = Riq + Lq + Ld Pp ωid + Pp ωλP M
dt
main proposed innovation was essentially the way in which 3
the error introduced by reference voltage usage is corrected Tm = Pp [λP M iq + (Ld − Lq )id iq ] (1)
2
as well. This error is due to the time lag introduced by the dω 1
converter operation. The correction was there determined in = (Tm − TL − F ω)
dt J
real time, by implementing a compensating function. dθ
= ω
In this paper an enhanced version of the estimation is pro- dt
posed. The enhancements consists basically in the introduction in which:
of a control loop in the estimation that causes the vanishing vd vq are the stator voltages in the rotor reference frame;
of the error in position estimation due to currents transients id iq are the stator currents in the rotor reference frame;
effects. The presence of the control loop allows to increase the R is the stator phase resistance;
dynamic performances of the electrical drive and the precision Ld Lq are the machines inductances, respectively along
of the compensating functions introduced in [22]. direct quadrature axis;
It is clear that this correction method and estimation process Pp is the number of motor pole pairs;
is more suitable for high dynamic industrial electrical drives λP M is the flux generated by permanent magnets;
working near the rated speed and not for motion control Tm is the electromagnetic torque;
systems for the lack of the behavior at low speed, in particular TL is the load torque;
for the oscillation that the estimation exhibit at low speed. F is the viscous friction coefficient
As a matter of fact, the absence of voltage probes reduces J is the moment of inertia of all rotating masses;
the cost of the system and improves its reliability and electro- ω is the instantaneous angular speed;
magnetic suscectivity. The rotor position used for the field ori- θ is the instantaneous angular position.
ented control with null direct component of the stator current In industrial field oriented control for the sake of simplicity
(id ) is determined on the basis of the mathematical model of and for reduction of costs the condition id = 0 is used. In
the motor. The motor model allows to calculate the argument particular when the saliency ratio is not so high (<3) the
of the complex vector of the stator back EMF. From this one, contribution of the reluctance torque is hard to justify if
the real instantaneous rotor position is hereafter determined compared to the major complexity of the control algorithm
introducing some offset correction terms depending on the implemented to follow the condition of a maximum Torque -
actual stator currents and on rotor speed. current ratio.
The paper is summarized as follows: In section II after a The id = 0 condition is imposed also in the control
brief recall of the PMSM mathematical model, rotor position algorithm presented in this paper because of the industrial
and speed determination are discussed. Estimation method is target of the possible applications.
then the object of a further discussion in which the Authors In this case the torque expression in eq. (1) set becomes:
explain what are the electrical quantities and the machine 3
parameters on which position and speed offset terms depend Tm =
Pp λP M iq (2)
2
and in which way these must be taken into account.
Because of the constant permanent magnet flux, the torque
Experimental results, carried out with the aid of a PMSM
depends only on the quadrature component of the stator
drive flexible test bench, giving the validation of the proposed
current.
technique and showing the high performances reached, are
presented in Section III.
In section IV the degree of robustness of the estimation B. Description of the estimator.
method against stator resistance and quadrature inductance As already established in the introduction, a characteristic of
variations is analytically investigated. the proposed sensorless control technique is the determination
In section V conclusions are finally summarized. of the back e.m.f space vector without voltage measurements
by using the reference voltages instead of the actual ones, so
that the system is a low cost and reliable one. In the control
system, whose block scheme is shown in Fig. 1, motor supply
II. C ONTROL ALGORITHM DESCRIPTION . voltage values that constitute some inputs of the estimator,
are replaced with reference voltage signals (va∗ , vb∗ , and vc∗ )
A. PMSM mathematical model generated by the current regulator. Considering the inverter
reference voltage signals instead of the real supply voltages
The proposed control system is based on the motor model implies, obviously, some approximations in the back e.m.f
expressed in the dqo reference frame [5], [6], [7]. determination. In fact, by not considering the inverter op-
The equations of the motor model are: erations follows to neglect the voltage harmonics and the

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

Power ///
we refer as the “current offset term”) is the angle of the same
Iqref
+
Controller
+
Controller
vector computed in the dqo reference frame.
wref
- - dq/abc
PWM Eq. 4 presents two singularities when eSα or iq approach
Iq Converter
zero crossings. These singularities can be easily managed
+
Controller
because the function arctan(.) converges to ±π/2 when its
va,vb,vc
Id=0 - (actual) argument diverges to ∞. In general, in high level program-

est ming languages the singularity elimination is managed within
Id
the preprocessor directives and math libraries (namely the
abc/dq “math.h” library for C and C++). For the same eq. 4 the
usage of the four quadrant inverse tangent function (seldom
PMSM
ω est named atan2 in the field of technical computing) is a suitable
Speed and position
estimation v*a,v*b,v*c alternative to the simple arctan(.) function.
(reference) Eq. 4 gives only an early and rough estimation term to be used
in the control system.
ia,ib,ic In the following it is presented how to realize a satisfactory
Fig. 1. Electrical drive machine blocks diagram with sensorless control for estimation that allows to control the drive also during transient
IPMSM.
operations.
Equation (4) shows clearly as the rotor position depends on
delay time response due to power switches commutations and the quadrature current iq .
inverter control algorithm time consumption. It is worthwhile In eq. (4) evaluation of eSα and eSβ come out from the
to observe that better performances are obtained if rotor actual motor voltages, but in the real estimation scheme,
position and speed determination would be committed directly only the reference voltages given by the current controller
to the determination of the back EMF space vector e~S and not va∗ , vb∗ , and vc∗ are used so that eq. (4) is effectively used
to the determination of the flux space vector. This is essentially with e∗Sα = vα∗ − Riα and e∗Sβ = vSβ ∗
− Riβ . This
due to the absence of flux signals integrator. Furthermore, choice avoids the usage of voltage probes, but, in this way,
elimination of analog integrators improves the benefits in the term arctan[e∗Sβ /e∗Sα ] is clearly only approximated. The
analog circuitry for the absence of components suffering for actual motor voltages lag the reference voltage because of
thermal drift [16]. the presence of the delay introduced by the power converter
Assuming a balanced tree-phase system, the expression of [13],[14],[22].
the back EMF space vector e~S components is: In order to gain an affordable estimation process without
voltage probes, the effect due to actual voltage lag must be
· ¸ obviously compensated. Let T the lag time introduced by the
~ j inverter and:
e~S = v~S − RiS = eSα + jeSβ = va − √ (va + 2vb ) −
· ¸ 3 µ ¶
j vβ − Riβ
− R ia + √ (ia + 2ib ) = va − Ria + F(vα , vβ ) = arctan (5)
vα − Riα
3
1 the first term of eq. (4) i.e. the argument of e~S in the α − β
+ j √ [(va + 2vb ) − R(ia + 2ib )] (3)
3 reference frame. In eq. (5) the reference voltages vα∗ , vβ∗ are
substituted to the actual ones with the reference ones vα , vβ . A
in which:
relation between actual and reference voltage may be written
va , vb , ia , ib are, respectively, the voltages and the currents
in the form:
of phases “A” and “B”;
i~S is the space vector of the stator currents;
eSα and eSβ are, respectively, the components along the vα∗ = vα + δvα
stationary real and imaginary axis of the back electromotive vβ∗ = vβ + δvβ (6)
force space vector e~S .
The argument of the back EMF clearly is not the real rotor in which the variations δvα , δvβ are due to the phase differ-
position. The real rotor position is given by the difference ence.
between the argument of ~eS in the stator reference frame and Now, after substitution, considering a well known calculus
the argument of the same one in the rotating dqo frame. formula for the increment of functions, we can write:
A simple analysis on the machine model at steady state
∂F ∂F
with id = 0 gives the following expression for correct rotor F(vα∗ , vβ∗ ) = F(vα , vβ ) + δvα + δvβ (7)
position: ∂vα ∂vβ
· ¸ · ¸ and developing the partial derivatives of the incremental term
eSβ λP M ∂F ∂F
θ = arctan − arctan (4) ∂vα δvα + ∂vβ δvβ we get:
eSα Lq i q
(δv i − δv i ) R − δvα vβ + δvβ vα
where arctan [eSβ /eSα ] is the phase of the e~S vector in the ³ αβ ´ β α (8)
stationary reference frame and arctan[λP M /Lq iq ] (to which R2 i2α + i2β − 2R (iα vα + iβ vβ ) + vα2 + vβ2

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

Now consider that, neglecting all harmonics, vα and vβ are

respectively cosine and sine functions of ωt. The same can be
said for vα∗ and vβ∗ In particular it is:

vα∗ = V cos (ω(t + T ))

vβ∗ = V sin (ω(t + T ))
vα = V cos(ωt)
vβ = V sin(ωt) (9)

where V is the RMS of the stator voltage. It is now clear that,

being T very small compared to 2π/ω:

vα∗ = vα + δvα = vα − vβ ωT
vβ∗ = vβ + δvβ = vβ + vα ωT (10)

In this way our previous expansion of (8) reduces to: Fig. 2. Position offset at steady state as a function of speed and current.

− (iβ vβ + iα vα ) R + vβ2 + vα2

³ ´ ωT (11) Fig. 2 shows the sum of “speed” and “current” offsets
R2 i2α + i2β − 2R (iα vα + iβ vβ ) + vα2 + vβ2 V
+ arctan( λLPq IMq ) as a function of speed and current iq at
E ωT
steady state with reference to the IPMSM used for experiments
that in a more compact form becomes:
in this paper (see Table I above for nameplate ratings).
V 2 − V RI cos(ϕ) Speed and current offset may be easily calculated off line, or
ωT (12) better determined via measurements from which the offset cor-
V 2 + R2 I 2 − 2V RI cos(ϕ)
rections terms may be determined via polynomial interpolation
being cos(ϕ) the power factor in motoring operation and V , with data fitting procedures. Interpolated data can be stored in
I and E, respectively, the RMS values of the stator voltages discrete form in two-dimensional look up table (a single ROM
currents and back EMF. chip) from which they are recalled depending on the particular
Equation (12) may be written also in complex number form: motor operation state. This reduces the mathematical burden of
estimation and involve to bring an increased step of economy
~ · (V
V ~ − RI)
~∗ and simplicity in the electrical drive.
ωT (13)
~ 2
|E| In the next sections, the on line realization of the estimator
is considered.
where (*) denotes the complex conjugate:
eq (12) express the correction term to be used to estimate
C. Estimator realization.
the argument of the back EMF when the reference voltages
are used. Furthermore, as with id = 0 the motor drive operates Eq (15) may be rewritten in differential form by substituting
with near unity power factor, eq. (12) can be further simplified ω with the derivative of θest :
as follows:
vβ∗ − Riβ V dθest λP M
θest = arctan( )−T −arctan( ) (16)
V 2 − V RI cos(ϕ) VE V vα∗ − Riα E dt Lq iq
ωT ' 2 ωT = ωT (14)
2 2 2
V + R I − 2V RI cos(ϕ) E E It is important for the sake of a correct estimation to avoid
the occurrence of inverter overmodulation by oversizing the
However the complex products on space vector eq. (13) are
DC Link voltage.
not difficult to implement.
Equation 16 describes the estimation dynamic that converge
It is clear that E depends on the motor speed but also on the
to the value of the real rotor position with id = 0 field
current. So the term (V /E)ωT in (14) depends simultaneously
oriented control, but in practical realization it suffers for
on speed and on stator current (by means of E). This term is
steady state error due to variation of id during transients.
referred as the “speed offset” because ω appears explicitly,
In fact, even if the PI current regulators can theoretically
even if E depends on stator current.
vanish the cross coupling between the current channels via
Introducing these speed and current offset corrections in
feed-forewarding, the presence of the delay introduced by the
the (4), one finally get a suggestion for a first expression of
inverter make this decoupling not perfect so that id exhibit
“estimated” position:
some significant variations during speed change or load torque
insertion transients.
vβ∗ − Riβ V λP M The steady state error on θest may be vanished with a PI
θest = arctan( ) − ωT − arctan( ) (15)
vα∗ − Riα E Lq iq controller creating an estimation control loop.

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

TABLE I
To establish this control loop we consider that the estimated
IPMSM PARAMETERS AND RATED VALUE .
direct current iˆd = iα cos(θest ) + iβ sin(θest ) is forced to
zero at steady state only when θest = θ. The PI regulator maximum speed [rpm] 4000
so introduces a further correction term on θest vanishing the rated current [A] 3.6
phase rated voltage [V] 77
steady state position error. Taking the presence of the PI into rated torque [Nm] 1.8
account, the ultimate estimator equation form is: Number of poles 6
Stator phase resistance [Ω] 2.21
Direct axis inductance [mH] 9.77
vβ∗ − Riβ V dθest Quadrature axis inductance [mH] 17.94
θest = arctan( )−T +
vα∗ − Riα E dt permanent magnet flux [Wb] 0.084
Z Friction Torque [Nm] 0.04
λP M
− arctan( ) − kp iˆd − ki iˆd dt (17)
Lq i q TABLE II
where kp and ki are the proportional and integral PI parame- C HARACTERISTICS OF THE DPS 30 POWER ELECTRONIC INVERTER .
ters. Input phase rated voltage (at 50 Hz) [V] 230
Before the PI tuning it is mandatory to add a pole in the DC Link voltage [V] 310
origin of the complex plane with the aim to obtain a system Output max power [kW] 6.5
of the second type. This ensures that the position estimation Output max current (peak) [A] 30
Output max current (RMS) [A] 21
error θ − θest vanishes. In order to ensure stability to the PWM carrier frequency [kHz] 20
control loop the Routh Criterion is applied to the determinant Maximum braking power [W] 3200
of the approximated closed loop transfer function. The Routh
polynomial is a function of the kp and ki parameters. In this
way it is possible to establish a relational bound between these board with a sampling frequency of fs = 10kHz while the
parameters whose numerical determination is after made with down sample factor was chosen as n = 10.
the traditional methods of the control system theory. The output signals of the PWM generator that come out
The estimator block diagram is shown in Fig. 3. from the dSPACE board are then directly fed to the drivers of
the IGBT switches.
III. E XPERIMENTAL RESULTS In Table I and II the nameplate and per phase parameters
A. Description of the test bench. of the IPMSM and the electrical characteristics of the IGBT
power converter are respectively reported.
In order to validate the effectiveness of the proposed sen-
sorless control system a flexible test bench has been built and
set up. B. Results and discussion.
The electrical drive test bench is constituted by [17], [18]: Various tests have been made in order to validate the
• an internal permanent magnet synchronous motor; proposed estimation algorithm. In particular the results of the
• a controlled hysteresis brake; following tests are herein after reported:
• a dSPACE (digital Signal Processing And Control Engi- 1) step change in motor speed from 400 rpm up to 4000
neering) board; rpm (nominal speed) and back again to 400 rpm;
• a resolver (used only for comparison purpose). 2) sudden application of a 1.8 Nm load torque while the
In particular, the dSPACE, based on floating point micropro- motor run at 4000 rpm speed;
cessors allows the fast implementation, verification and real- In order to improve the strictness of the tests, the used PI
time simulation of algorithms; the control system here used is parameters in currents and speed control loops have been
implemented on the DS1103 board which is equipped with two locked to the values tuned with the sensored version of the
processors: a master Power PC 604E and a Texas Instruments electrical drive. It is clear that a PI re-tuning, taking the
slave DSP of the type TMS320F240, characterized by cycle presence of estimator into account, could reach a further
frequencies respectively of 400 kHz and 80 MHz. improvement of the performance in sensorless control.
The controlled hysteresis brake is a Magtrol - INC. model In the next figures dashed lines are used for the estimated
HD-705-8 (6 Nm) and the resolver is a MAGNETIC BLQ-41 rotor position and speed while solid line are used for the
used only to measure the real rotor position to be compared measured ones.
with the estimated one to verify the effectiveness of the Fig. 5 shows a comparison between the estimated speed
estimation. The test bench view is shown in the picture of and the measured one by the resolver of the test bench for
Fig. 4. verification purpose. Estimated and real motor speed are in
The basic drive module under test, composed of a tangential good accordance. Furthermore, the electrical drive shows a
flux IPMSM and an insulated-gate-bipolar-transistor (IGBT)- very good performance in settling and rise time. The maximum
based converter (the DPS 30) controlled by the dSPACE that estimation error did never exceeded 35 rpm during transient
manage the converter according to the id = 0 field-oriented and become lower (a small oscillation with about 2 rpm peak
control strategy. value) at steady state. Fig. 6 shows instead real and estimated
The sensorless control algorithm, including the speed and rotor position. It is evident that at steady state there is no
the current loops, has been fully implemented on the dSPACE estimation error and during the transient a small error appears

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

V
, + e + +
+
atan  1/s
e

- -
RI -

Current
R PM
IIR Filter atan 
I
, Lq Iq

est

 e e 2


2
 1
 v v 2


2

T

Iq I d =0
Id -
ki

, +
k p 1/s
dq s
est

Fig. 3. Estimator block diagram

5000
real
est
4000
ωreal, ωest

3000

2000

1000

0
0.5 1 1.5 2
time [s]
Fig. 4. The Test bench for IPMSM electrical drive machine. Fig. 5. Comparison between real (solid line) and estimated (dashed line)
speed during the execution of test n. 1.

4
est
as it can be seen by the little shifting between the two curves. 3 real
The estimation error vanishes within two cycles.
2
Test 2 is really significant because the motor works under
θreal θest [rad]

load condition and it can be observed during a transient 1

operation.
0
In Fig. 7 estimated and real angular position measured with
resolver are shown when an abrupt (step) load torque is applied −1
while the motor rotates at 4000 rpm. Real rotor position is
still measured with a resolver on the motor shaft. After the −2
load insertion the position recovery is really fast and only a
−3
high level zoom make the rotor position error visible. Real
and estimated speed are shown in Fig. 8. It can be observed −4
that suddenly, after torque application, the estimation error 0.5 1 1.5 2
time [s]
increases, but, as Fig. 9 shows, this error tends to vanish and
only residual small oscillation and noise due to the presence Fig. 6. Real (solid line) and estimated (dashed line) position during the
of the converter remain. execution of test n. 1.

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

3 real
est 5
2
0
1
θreal θest [rad]

∆ ω [rpm]
−5
0
−10
−1
−15
−2

−3 −20

0.98 0.99 1 1.01 1.02

time [d] 0.8 0.9 1 1.1 1.2 1.3 1.4
time [s]
Fig. 7. Comparison between real (sold line) and estimated (dashed line)
rotor position during the execution of test n. 2. Fig. 9. Estimation error for rotor speed during the execution of test n. 2.

4100 real
est i v − vα i β
SR = R ³ ´ α β =
4000 R2 i2a + i2β − 2R (ia va + iβ vβ ) + vβ2 + vα2
V I sin(ϕ)
ωreal ωest

3900 = R (19)
V 2 − 2RV I cos(ϕ) + R2 I 2
that can be further simplified in:
3800
R S sin(ϕ)
SR = (20)
3700 3 |E|2
where the symbols used are the same of eq. 14 and where
3600 S = 3V I is the machine apparent power.
Eq 20 shows that the sensitivity function SR is very small
1 1.1 1.2 1.3 at high speed when the back EMF reaches its highest values
time [t] and in each case results to be small at near unity power factor
(sin(ϕ) ' 0) that is assured by the id = 0 control.
Fig. 8. Comparison between real (solid line) and estimated (dashed line) A similar calculation lead to the sensitivity function on
rotor speed during the execution of test n. 2.
quadrature inductance SL :
dTL λP M Lq iq
This is a confirmation of the effectiveness and of the proper S L = Lq = −¡ 2 2 ¢ (21)
dLq Lq iq + λ2P M
action of the introduced offset terms.
from which it can be deduced that sensitivity to variation vs
IV. S ENSITIVITY TO PARAMETER VARIATIONS . quadrature inductance is smaller at high mechanical loads.
An estimation at the rated values for the motor under test
As eq. 16 shows the estimation process depends on the
shows that for a parameters variation of 100% the estimation
values of the stator resistance R and of quadrature inductance
error amount is about 1%. The proposed estimation process
Lq . In order to investigate the robustness of the estimation vs
appears to be very robust against parameters variation.
the parameter variations the relative sensitivity functions have
been calculated. It is clear that in the same equation the term
TR = arctan(eβ /eα ) depends only on the resistance while the V. C ONCLUSIONS
current offset term TL = arctan(λP M /Lq iq ) depends only on In this paper a low time consuming and low cost sensor-
the quadrature inductance. less control algorithm for PMSM without voltage probes for
The sensitivity function related to resistance variation is position and speed estimation was introduced, discussed and
defined as: experimentally verified.
The sensorless control system is based on the back electro-
dTR motive force space vector estimation. The use of the back EMF
SR = R (18)
dR space vector is advantageous respect to any other system using
Developing this formula the following form is obtained: flux estimation because of the integrator elimination avoiding

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

the problem of integration drift that requires opportune devices [2] T. D. Batzel, K. Y. Lee, "Electric Propulsion With the Sensorless
or sub systems for its compensation. This simplification also Permanent Magnet Synchronous Motor: Model and Approach", IEEE
transactions on energy conversion, Vol. 20, No. 4, Page(s):818 - 825,
include that the control system will be less susceptible against December 2005.
the EMI external sources. The proposed system in general [3] Jul-Ki Seok, Jong-Kun Lee, Dong-Choon Lee, "Sensorless Speed Con-
is more reliable and cheaper than a complicated one without trol of Nonsalient Permanent-Magnet Synchronous Motor Using Rotor-
Position-Tracking PI Controller", IEEE transactions on industrial elec-
loss of performance respect to other control systems proposed tronics, Vol. 53, No. 2, Page(s):399 - 405, April 2006.
in literature or in industrial applications and also it does not [4] Wallmark O., Harnefors L., "Sensorless Control of Salient PMSM Drives
require any voltage probe. in the Transition Region", IEEE transactions on industrial electronics,
Vol. 53, No. 4, Page(s):1179 - 1187, August 2006.
In fact in the proposed control systems the reference [5] Corley M.J., Lorenz R.D., "Rotor position and velocity estimation for
voltages, instead of the actual voltages, are used for the a salient-pole permanent magnet syncronous machine at standstil and
back EMF estimation so eliminating the presence of voltage hight speed", IEEE Transaction on Industry Applications, VOL. 34, No.
4, Page(s): 784-789, july/august 1998.
probes. Also the absence of voltage probes imply a significant [6] Ogasawara S., Akagi H., "Implementation and position control per-
smaller cost for the proposed system and increased reliability. formance af a position-sensorless IPM motor drive system based on
It is obvious that the choice of this structure, imply errors magnetic saliency" - IEEE Transaction on Industry Applications, VOL.
34, No. 4, Page(s): 806-812, july/august 1998.
in position estimation that must be compensated, but this [7] Ogasawara S., Akagi H., "An approch to real-time position estimation
compensation requires only the knowledge of the time delay at zero and low speed for PM motor based on saliency" - IEEE
introduced by the Voltage Source Inverter and the use of Field Transaction on Industry Applications, VOL. 34, No. 1, Page(s): 163-
168, january/february 1998.
Oriented Control for the brushless motor with id = 0. This [8] Pesse G., Paga T "A permanent magnet syncronous motor flux control
compensation may also be made by means of pre - collected scheme without position sensor", EPE, Trondheim, Norvegia, Page(s):
experimental data for the motor used in the drive. This could 553-568, 1997.
[9] Barrenscheen J., Flieller D., Kalinowski D., Louis J. P. "A new sensorless
allow to avoid on-line tasks if these reveal to be an obstacle speed and torque control for permanent magnet synchronous motors:
in some types of plant. realisation and modelling", EPE, Sevilla, Spain, Page(s): 839-844, 1995.
The comparison with traditional sensorless control meth- [10] J. Oyama, T. Abe, T. Higuchi, E. Yamada, K. Shibahara, "Position
sensor-less control of half-wave rectified brushless synchronous motor",
ods also based on back-EMF determination shows the high EPE, Sevilla, Spain, Page(s): 149-153, 1995.
dynamic performance of the proposed algorithm and the ef- [11] Peixoto Z. M. A., Sa F. M. F., Seixas P. F., Menezes B. R., Cortizo P.
fectiveness of the correction terms introduced in the estimation C., "Design of sliding observer for back electromotive force, position
and speed estimation of interior magnet motors", EPE, Sevilla, Spain,
process. Page(s): 833-838, 1995.
Furthermore a sensitivity analysis with respect to the stator [12] Ben-Brahim L., Kawamura A., "A fully digitized field-oriented controlled
resistance thermal variations have been carried out and the induction motor drive using only current sensor", IEEE Transaction on
Industry Applications, VOL. 39, No. 3, Page(s): 241-249, June 1992.
results have shown that the algorithm is robust against this [13] Holtz J. "Pulsewidh modulation a Survey", IEEE Transaction on Indus-
variations. For the typical range of paramenter variation the trial Electronics, Volume 39, Issue 5, Oct. 1992 Page(s): 410 - 420.
error on estimation is quite not appreciable. [14] Cataliotti, A.; Genduso, F.; Raciti, A.; Galluzzo, G.R.; “Generalized
PWM–VSI Control Algorithm Based on a Universal Duty-Cycle Ex-
Clearly the presented correction method is intended above pression: Theoretical Analysis, Simulation Results, and Experimental
all to make the electrical drive cheaper and suitable for indus- Validations”, IEEE Transactions on Industrial Electronics, Volume 54,
trial drives both surface or internal mounted permanent magnet Issue 3, June 2007 Page(s):1569 - 1580
[15] Matsui N., Shigyo M., "Brushless dc motor control without sensors"
working within the nominal speed range, as for example for IEEE Transaction on Industry Applications, VOL.28, No.1, Page(s):
spindle drives, while the field wakening is not taken into 120-127, January/February 1992.
account. [16] Wu R., Slemon G. R. "A permanent magnet motor drive without a shaft
sensor", IEEE Transaction on Industry Applications, VOL. 27, No. 5,
Drive starting is made with open loop operation. Speed Page(s): 1005-1011, September /October 1991.
loop is closed when speed has exceeded almost 10% of the [17] Alonge F. Di Tommaso, A.O.; Miceli, R. Rando C., "Sensorless Control
synchronous speed. of Permanent Magnet Syncronous Motors for wide speed range appli-
cations", Speedam 2006, Sect. S4 Page(s): 12-17 Taormina, Italy.
In conclusion the proposed algorithm may be considered [18] Di Tommaso, A.O.; Miceli, R., "A new high accuracy software based
a very good alternative in terms of economy and precision resolver-to-digital converter", 29th Annual Conference of the IEEE
without lack of performances and, furthermore exhibits an Industrial Electronics Society, IECON ’03, 2-6 Nov. 2003, Vol. 3,
Page(s): 2435 - 2440.
increase in reliability. [19] Ji-Liang Shi; Tian-Hua Liu; Yung-Chi Chang; "Position Control of an
Interior Permanent-Magnet Synchronous Motor Without Using a Shaft
Position Sensor", Industrial Electronics, IEEE Transactions on Volume
ACKNOWLEGEMENTS 54, Issue 4, June 2007 Page(s): 1989 - 2000;
This work has been supported by MIUR (Italian Ministry [20] Ji-Liang Shi; Tian-Hua Liu; Yung-Chi Chang; “Adaptive controller
design for a Sensorless IPMSM Drive System with a Maximum Torque
of University and Research) and by the “Sustainable Develop- Control” IEE proceeding on Electronics Power Applications, Vol. 6,
ment and Energy Saving Laboratory” (SDES Laboratory) part November 2006;
of the UNINETLAB of the University of Palermo. [21] Morimoto S.; Sanada S., Takeda M., Imai Y.; "Influence of Rotor Con-
figuration on Sensorless Control for Permanent-Magnet Synchronous
Motors", IEEE Transactions on Industry Applications, Volume 44, Issue
R EFERENCES 1, Jan.-feb. 2008 Page(s): 93 - 100
[22] Genduso F.; Miceli R.; Rando C.; Ricco Galluzzo G.; “A novel Cor-
[1] Rashed M, MacConnell P. F. A., Fraser A., Stronach, P. Acarnley rection Method for a Low Cost Sensorless Control System of IPMSM
S.; "Sensorless Indirect-Rotor-Field-Orientation Speed Control of a Electrical Drives”; International Symposium on Industrial Electronics
Permanent-Magnet Synchronous Motor With Stator-Resistance Estima- ISIE08, Cambridge, July, 2008, UK. Page(s): 509 - 514
tion", IEEE transactions on industrial electronics, Vol. 54, No. 3, [23] Sung-Yeol Kim; In-Joong Ha "A New Observer Design Method for HF
Page(s):1664 - 1675, June 2007. Signal Injection Sensorless Control of IPMSMs"; IEEE Transactions on

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

Industrial Electronics, Volume 55, Issue 6, June 2008 Page(s): 2525 - Cosimo Rando received the M.S. and Ph.D. de-
2529 grees in electrical engineering from the University
[24] Boussak M.; "Implementation and experimental investigation of sen- of Palermo, Palermo, Italy, in 2004 and 2008,
sorless speed control with initial rotor position estimation for interior respectively. Actually he is a post doctorate stu-
permanent magnet synchronous motor drive", IEEE Transactions on dent at the Department of Electrical Electronics
Power Electronics Volume 20, Issue 6, Nov. 2005 Page(s): 1413 - 1422 and Telecommunications Engineering of the the
[25] Andreescu G.-D.; Pitic C.I.; Blaabjerg F.; Boldea I.; "Combined Flux University of Palermo. His research interests in-
Observer With Signal Injection Enhancement for Wide Speed Range clude electrical drives, Brushless Machines, pho-
Sensorless Direct Torque Control of IPMSM Drives" IEEE Transaction tovoltaic systems, wind turbines and renewable
on Energy Conversion, Volume 23, Issue 2, June 2008 Page(s): 393 - energies.
402
[26] Imai N.; Morimoto S.; Sanada M.; Takeda Y.; "Influence of Rotor Con-
figuration on Sensorless Control for Permanent-Magnet Synchronous
Motors"; IEEE Transactions on Industry Applications, Volume 44, Issue
1, Jan.-feb. 2008 Page(s): 93 - 100
Giuseppe Ricco Galluzzo received the Electri-
cal Engineering degree from the University of
Palermo, Palermo, Italy. During 1992-2001, he
was an Associate Professor of Electrical Drives at
Fabio Genduso received the M.S. and Ph.D. de- the University of Palermo, where, since Septem-
grees in electrical engineering from the University ber 2001, he has been a Full Professor of Elec-
of Palermo, Palermo, Italy, in 1999 and 2004, trical Drives. His research involves the fields of
respectively. In 2005, he joined the Department mathematical models of electrical machines and
of Electrical Electronics and Telecommunications drive systems control and diagnostics.
Engineering, University of Palermo, as a Re-
searcher. His research interests include power
conversion, electrical drives, power electronics
control, mathematical modeling and simulation of
electrical machines and power conversion systems
and renewable energies.

Rosario Miceli (M’02) received the Electrical En-

gineering and Ph.D. degrees from the University
of Palermo, Palermo, Italy. He is currently an As-
sociate Professor of Electrical Machines with the
Faculty of Engineering, University of Palermo.
He is personnel in charge of the Sustainable
Development and Energy Savings Laboratory of
the Palermo Athenaeum. His main research in-
terests include mathematical models of electrical
machines, drive systems control, and diagnostics,
renewable energies and energy management. Dr.
Rosario Miceli is a IEEE member and a reviewer of IEEE Transaction
on Industrial Electronics and Industry Applications.

Copyright (c) 2009 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].
Authorized licensed use limited to: Universita degli Studi de Palermo. Downloaded on April 27,2010 at 10:56:56 UTC from IEEE Xplore. Restrictions apply.