3372

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 6, JUNE 2015

Control of Large Salient-Pole Synchronous

Machines Using Synchronous Optimal
Pulsewidth Modulation
Joachim Holtz, Life Fellow, IEEE, Gilberto da Cunha, Student Member, IEEE,
Norton Petry, and Paulo Jos Torri
AbstractHigh-power grinding mills are used in the
cement and mining industries to crush clinker or copper
ore and grind these materials to ne powder. The multimegawatt speed-controlled mill drives operate at a very low
angular speed. Synchronous motors with a high number of
pole pairs are used as the prime movers. They are traditionally fed by load-commutated thyristorized cycloconverters.
These are prone to failure modes that can lead to excessive
torque pulsations and high overcurrents. The huge stator,
which was built as a separate ring-shaped structure around
the tubular mill, may then get mechanically displaced, and
the operation of the plant is interrupted. A novel and
reliable direct drive uses a voltage source inverter that
operates at the unity power factor for increased efciency.
Synchronous optimal pulsewidth modulation ensures a low
harmonic current distortion and reduced switching losses
at a very low switching frequency. The optimization of the
pulse patterns takes the anisotropic magnetic properties of
a separately excited synchronous motor into account. The
implementation in a 23-MW semiautonomous grinding mill
installed in a Zambian copper mine is intended.
Index TermsAlternators, ball milling, Fourier transform,
gradient methods, pulse width modulation inverters, software algorithms, variable speed drives.

Fig. 1. Construction of a SAG mill with the stator and the rotor
separately mounted.

I. I NTRODUCTION

HE production of cement requires machinery for crushing the cement clinker produced by a rotary kiln and
to subsequently grind it to powder. The same procedure is
followed in copper mining [1]. Crushed copper ore is ground
to powder, from which the copper content is extracted in a
chemical process.
The production of fine powder is efficiently done in semiautonomous grinding (SAG) mills. Fig. 1 shows the construction
of a SAG mill [2]. It consists of a rotating hollow cylinder of
about 12 m in diameter. Two sets of bearings in ring-shaped
structures support the cylinder to let it rotate. State-of-the-art
gearless drives have about 76 salient poles attached around
the mill cylinder. They form the rotor of a separately excited
Manuscript received May 22, 2014; revised July 22, 2014, August 29,
2014, and October 19, 2014; accepted November 7, 2014. Date of
publication December 9, 2014; date of current version May 8, 2015.
J. Holtz is with Wuppertal University, 42119 Wuppertal, Germany
(e-mail: j.holtz@ieee.org).
G. da Cunha, N. Petry, and P. J. Torri are with WEG Automaao,
89256-900 Jaragu do Sul, Brazil.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIE.2014.2378732

Fig. 2.

Cross section of a SAG mill.

synchronous ring motor. The stator is separately mounted to the

basement. It is fed by a load-commutated cycloconverter with a
power rating of typically 20 MW at a maximum fundamental
frequency of 6 Hz [3], [4]. The converter is equipped with
thyristors, which are the most powerful semiconductor devices
available.
The interior of a SAG mill is shown as a cross section in
Fig. 2 [5]. The shaded area in the lower right portion indicates

0278-0046 2014 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/
redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

HOLTZ et al.: CONTROL OF SALIENT-POLE SYNCHRONOUS MACHINES USING SYNCHRONOUS OPTIMAL PWM

where the grinding material and a number of steel balls are

encountered. The cylinder rotates in the anticlockwise direction, which elevates the steel balls and the major-sized lumps
of ungrounded material up to a certain level. These objects
will eventually fall down into the lower portion of the cylinder,
smashing the ungrounded material there to smaller pieces and,
finally, to powder with a particle size of 180 m. This type
of mill is called semiautonomous as it is the inertial forces
of falling steel balls and ungrounded material that provide the
grinding effect.
Ring-motor-driven SAG mills in copper mines have shown
frequent malfunctions in the past [3], [6], [7]. The problem
is owed to the characteristics of the cycloconverter that feeds
the synchronous motor. Cycloconverters are composed of two
parallel-connected six-phase bridges per phase, with one bridge
conducting the positive current and the other conducting the
negative current. Zero-current intervals are inserted whenever
the direction of a phase current changes. This allows the
thyristor of the outgoing bridge to regain its voltage-blocking
capability. Zero-current intervals produce zero torque. Electromechanical oscillations are the consequences [3], accompanied by extreme mechanical forces that tend to break the
fixtures of the stator.
Even more critical are commutation failures that occur when
a sudden voltage sag appears in the feeding mains while the
converter operates in the regeneration mode. A commutation
failure leads to a short circuit between the terminals of the stator
winding.
In any of these failure modes may the stator get displaced on
its basement, and the minimum airgap may not be maintained.
The production of the mill must be interrupted for extensive
repair work. The standstill of production entails considerable
cost.
To overcome the aforementioned problems, a new drive
system was developed for Cleantech Company Ltd., Finland. It
was introduced as the worlds largest SAG mill to be installed at
First Quantums Kansanshi Copper Mine in Zambia in
2014 [8].

II. A LTERNATIVE D RIVE S YSTEM FOR SAG M ILLS

A. Medium-Voltage Three-Level Inverter
The failures that have occurred during the operation of SAG
mills were caused by using a mechanically sensitive ring motor
and a cycloconverter for its control. Improvement is sought by
installing a classical drivetrain, with the drive motor and the
SAG mill having their own bearings. A 23-MW salient-pole
synchronous motor, as shown in Fig. 3, was developed for that
purpose. The huge dimensions of this motor are demonstrated
by a person standing in the foreground. Fig. 4 shows the rotor
poles and the armature winding.
The unreliable cycloconverter is replaced by a pulsewidthmodulated (PWM) voltage source inverter for energy conversion and control [9]. Pairs of parallel-connected 6.5-kV
600-A insulated-gate bipolar transistors (IGBTs) serve as the
switching elements in a three-level topology [10], [11]. Using
optimal PWM permits operating at an extremely low switching

3373

Fig. 3.
Brazil.

Two synchronous motors coupled for testing in a factory in

Fig. 4.

Rotor poles and armature winding.

frequency without sacrificing on harmonic current distortion

[12]. Switching losses are thus reduced, and the nominal power
rating of the inverter is increased [13].
The minimization of the harmonic currents by synchronous
optimal PWM has become an established technology for
medium-voltage induction motor drives. For SAG mill drives
operating at a very low mechanical speed, synchronous machines are a better choice. The absence of slip reduces machine
losses and leads to a better tradeoff between the installation cost
and efficiency. A large airgap increases the reliability.
Different from induction motor drives, salient-pole synchronous machines exhibit anisotropic magnetic properties.
The reactions of this machine to transients caused by the
repeated voltage step changes of a PWM waveform is analyzed
next.
B. Dynamics of Synchronous Machines
The anisotropic magnetic properties of separately excited
synchronous machines require using a rotor-fixed dq reference
frame for dynamic analyses. The d-axis is aligned with the

3374

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 6, JUNE 2015

State equations (1) will be used to compute the harmonic

currents. These are excited by the harmonic content of the
inverter pulse sequences and represented by harmonic spectra
in the frequency domain. Accordingly, the solution of (1) and
(2) is obtained in the frequency domain, for which the algorithm
in [14] is followed. The approach relies on the definition of the
complex inductance operators d (s) and q (s), each of which
completely describes the magnetic properties in the respective
reference axis, i.e., d or q. The inductance operators are

Fig. 5.

Winding structure of a synchronous motor.

direction of the rotor pole of an equivalent two-pole machine.

The complex-plane topology of a synchronous machine is
shown in Fig. 5.
The following differential equations hold:
did
dif
diD
+Lq iq + Lf
+ LDd
+ LQq iQ
dt
dt
dt
diq
diQ
Ld id Lf if LDd id + LQq
u q = R a iq + L q
dt
dt
did
dif
diD
uD = RD iD + LdD
+ Lf D
+ LD
dt
dt
dt
diq
diQ
uQ = RQ iQ + LqQ
+ LQ
dt
dt
did
dif
diD
uf = Rf if + Ldf
+ Lf
+ LDf
(1)
dt
dt
dt
ud = Ra id +Ld

where ud and uq are the components of the armature (stator)

voltage, uf and if are the field windings excitation voltage
and respective current, id and iq are the components of the
stator current, and iD and iQ are the components of the damper
winding current in the rotor. The respective winding resistances
Ri are marked by subscripts. Ld and Lq are the inductances
of the armature winding, and Lf is the inductance of the field
winding. Mutual inductances are marked by pairs of subscripts,
e.g., LDq describes the magnetic coupling between damper
current iD and the q-axis armature winding.
A second set of equations describes the flux linkages as
functions of the currents. We have

(3)
q (s) = Lq

L2qQ s
R Q + LQ s

mmm

(4)

where s denotes the complex frequency.

Using the inductance operators eliminates currents of minor
interest from state equations (1). This is done here for damper
currents iD and iQ , and for field current if . Only two equations
then remain. They describe how currents id and iq depend on
armature voltages ud and uq as follows:
ud (s) = (Ra + sd (s)) id (s) + 1 q (s)iq (s)

(5a)

uq (s) = 1 d (s)id (s) + (Ra + sq (s)) iq (s). (5b)

Moreover, the flux linkage components are solely expressed
by id and iq as follows:
d (s) = d (s) id (s)

(6a)

q (s) = q (s) id (s).

(6b)

Equations (5) and (6) serve for computing the harmonic currents generated by the three-level inverter voltage waveforms.
III. S YNCHRONOUS O PTIMAL PWM

d = Ld id + Lf d if + LDd id

A. Modulation Index

q = Lq iq + LQq iQ
D = LD iD + LdD id + Lf D if
Q = LQ iQ + LqQ iq
f = Lf if + Ldf id + LDf iD .

d (s)


Lf LD L2f D Ld L2dD Lf L2f d LD +2Lf d LdD Lf D


s2
=
2
2
Lf Ld Lf D s +(Rf LD +RD Lf )s+Rf RD


(Rf LD +RD Lf )Ld L2f D RD L2dD Rf s+Ld Rf RD


+
Lf Ld L2f D s2 +(Rf LD +RD Lf )s+Rf RD

(2)

C. Calculation of Harmonic Currents

The objective of this paper is to optimize the PWM pulse sequence of the feeding inverter for minimum harmonic distortion
of the armature current. The structure of (1) indicates that this
also minimizes the harmonic currents in the damper winding
and in the field winding.

Modulation index m is the normalized fundamental voltage,

which is defined as
u1
m=
(7)
u1 six-step
where u1 is the fundamental voltage component of a given pulse
pattern, and
u1 six-step =

4ud

(8)

is the fundamental voltage at the six-step operation of the inverter. The modulation index ranges within 0 m 1, where
the unity value corresponds to the six-step mode.

HOLTZ et al.: CONTROL OF SALIENT-POLE SYNCHRONOUS MACHINES USING SYNCHRONOUS OPTIMAL PWM

3375

B. Current Harmonics
The machine currents are composed of a fundamental current
component i1 and the harmonic currents ih generated by the
inverter switching. Harmonic currents substantially contribute
to the machine losses. Minimizing the RMS harmonic current
is therefore an objective of synchronous optimal PWM [15].
The RMS harmonic current, i.e.,

1
[i(t) i1 (t)]2 dt
(9)
Ih rms = 

T
(T )

does not only depend on the respective pulse pattern but also
on the internal impedance of the machine. This influence is
eliminated when the distortion factor is used as a figure of
merit. The distortion factor is derived from the normalized RMS
harmonic content, i.e.,



 
2

Uk
l
1

Ih rms
1

2


=
Ik =
(10)
I1
I1
U1
k1 l
k=2

k=2

of a periodic current waveform, where I1 is the fundamental

RMS current, Ik and Uk are the respective Fourier components
of the RMS machine current and the inverter output potential
per phase, 1 = 2f1 is the fundamental angular frequency,
and I is the effective leakage impedance of the machine.
A new variable Ih rms six-step /I1 is obtained by inserting
in (10) the Fourier components Uk of the rectangular six-step
voltage waveform. This serves to define the distortion factor, i.e.,
d=

1h rms
Ih rms six-step

(11)

which is a quantity that represents the harmonic content of a

PWM waveform independently of the properties of the machine. We have d = 1 at the six-step operation by definition.
C. Optimum Pulse Patterns
The optimization of the PWM pulse patterns is done under
the restriction that steady-state conditions exist. The approach
implies that switching frequency fs and fundamental frequency
f1 = 1 /2 are synchronized. Their ratio, which is the pulse
number as follows:
N=

fs
f1

(12)

is then an integer number.

Constructing the synchronous optimal pulse pattern for a
three-level inverter starts with establishing the half- and quarterwave symmetries of the phase voltages using the following
conditions:
ua (1 t) = ua ( 1 t),

0 1 t /2

ua (1 t) = ua (1 t ),

0 1 t .

(13a)
(13b)

A pattern thus defined does not contain harmonics of even order

or subharmonics. It is characterized by N switching angles ai ,

Fig. 6. Definition of the optimal switching angles i in phase a. The

angles reappear in phases a and b at phase displacements of 2/3 and
4/3, respectively.

i 1, . . . , N , as illustrated in Fig. 6 for N = 5, where ud is

the dc-link voltage. The three waveforms show the standard
displacement angles 2/3 between the respective phases.
The optimum switching angles ai are determined such that
the distortion factor is minimum and a predefined low value
fs max of the switching frequency is not exceeded. Operating
at the synchronous optimal modulation instead of the conventional space vector modulation, both at an identical distortion
factor, can almost double the power rating of a given inverter
[16]. Skin effect losses are not considered in this comparison.
Satisfying the objective function, i.e.,
d min

(14)

for every steady-state operating point N, m defines the respective sets of N switching angles ai per fundamental period
T1 = 2/1 . The optimal sets are precalculated offline and
stored in a memory table of the controlling microprocessor. The
switching angles are retrieved during the operation in real time
and used for inverter control.
A pulse pattern is now considered the superposition of alternating positive and negative voltage steps at the respective
switching angles i . The waveform of phase a is examined in
the following. Waveforms ub and uc , as shown in Fig. 6 for
N = 5, are identical except for their respective phase displacements. The fundamental voltage component of the pattern is
ua1 (i ) =

N
4ud

(1)i+1 cos(ai ).
i=1

(15)

Using (8) to normalize (15) yields the modulation index of

pattern P (m, N ) as follows:
m=

(1)i+1 cos(ai )

(16)

i=1

which enters as one condition to determine the sets of optimum

switching angles that define pulse patterns P (m, N ). A second
condition is the objective function given by (11) and (14).

3376

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 6, JUNE 2015

The harmonic voltage components of the phase a waveform

in Fig. 6 are written as a Fourier series expansion as follows:
uah (S) (1 t) =

4ud
1

n=1 6n 1

The orthogonal dq components of space vectors (19) are

N


1
4ud

i+1

(1)
cos(6n 1)i
udk =
n=1 (6n 1)
i=1

(1)i+1 cos ((6n 1)ai ) sin ((6n 1)1 t) .

(17)

i=1

uqk

Operator addresses both the harmonic voltage components in

the clockwise rotation, which was represented by the negative
sign of the operator, and those in the anticlockwise rotation,
which was represented by the positive sign of the operator.
Superscript (S) denotes stationary coordinates.
The three harmonic phase voltages uah , ubh , and uch define
the stator voltage space vector in terms of its n harmonic vector
components, where n 1, . . . , . Each value of n stands for
a pair of spectral vector components that rotate in opposed
directions. Harmonic components having positive phase sequences rotate in the positive direction at angular velocities
(6n 1)1 . The harmonic components with negative phase
sequences rotate in the negative direction at angular velocities
(6n + 1)1 . The resulting harmonic orders are
1 6n,

n 1, . . . , .

(18)

Existing half-wave and quarter-wave symmetries (13) eliminate

the harmonics of even and triplen order.

(20a)
(1) 2 (31) cos ((6n 1)1 t)



1
4ud

=
(1)i+1 cos(6n 1)i
n=1 (6n 1)
i=1
1

(1) 2 (31) sin ((6n 1)1 t) .

(20b)

Pulse pattern optimization is done offline assuming steadystate conditions. The harmonic currents are represented by discrete spectra ik , k = 1 6n, n 1, . . . , . They are computed
by replacing Laplace operator s jk1 in (3) and (5). The
harmonic components of the machine current are then with
good approximation obtained by solving (5) for the respective
currents as follows:
idk =

uqk jk udk
1 d (k 2 + 1)

(21a)

iqk =

udk j k uqk
.
1 q (k 2 + 1)

(21b)

They define the distortion factor, i.e.,


d = idk 2 + iqk 2 .

(22)

D. Transformation to Synchronous Coordinates

As the synchronous machine is modeled in synchronous
coordinates, (17) and harmonic phase voltages ubh and uch
in their respective displacements are transformed to obtain the
voltage space vectors of the harmonics in the synchronous
coordinates [superscript (SY )]. The transformation yields
(SY )

uh

(1 t) =

4ud
1
n=1 6n 1

i=1

exp (1)

j6nw1

(23)

2) A maximum permitted value fs max of the switching

frequency is chosen. Every pattern P (m, N ) is then valid
in the following range:

(1)i+1 cos ((6n 1) i )

1
2 (31)

The optimization of pulse patterns is done in the following

steps:
1) For a constant stator flux, the fundamental voltage u1
f1 ; hence, m f1 , and from (12), we have
fs = m N.

E. Optimization Procedure


(19)

where expressions 6n 1 and exp[(1)1/2(31) create two

summing terms each, with the first term using 6n 1 to
which exp((1)1/2(31) ) associates and with the second term
using 6n + 1 to which exp[(1)1/2(3+1) ] associates. Term
(1)1/2(31) defines the direction of the rotation of the respective harmonic.
These components rotate at angular velocities 6n1 , either
in a positive or in a negative direction, going past the d-axis
at a high frequency. Hence, the load angle of the synchronous
machine does not influence the space vector of the harmonic
voltages.

fs max
fs max
m
.
N +1
N

(24)

3) Switching angles i characterize harmonic voltages (20)

on which harmonic currents (21) depend. Objective function d(i , m), i 1, . . . , N , defines an N -dimensional
hyperplane. To find the optimum pattern P (m, N ) for
a given N , the hyperplane is searched for minimum
values, varying the N different switching angles i and
modulation index m within its range (24).
4) A large number of local minimums is found from which
the global minimum is selected.
5) An optimum pattern P (m, N ) only extends over its
pertaining range of m, as in (24). The full range of m
is built from all computed patterns.

HOLTZ et al.: CONTROL OF SALIENT-POLE SYNCHRONOUS MACHINES USING SYNCHRONOUS OPTIMAL PWM

Fig. 7.

3377

Signal flow graph of the drive control system.

6) The resulting function i (m) will show various discontinuities. These will cause undesired transients during
the operation of the drive. They are eliminated by a
postoptimization procedure.

Fig. 8. Optimal switching angles i , i = 1, . . . , N , versus modulation

index m. (a) Optimal switching angles i for N = 6. (b) Complete set of
optimal switching angles. (c) Resulting distortion factor.

Pulse pattern optimization is described in detail in [16].

IV. S YNCHRONOUS M OTOR D RIVE C ONTROL
The synchronous optimal PWM is embedded in the drive
control system, as shown in Fig. 7. The superimposed current
controller receives the fundamental stator current vector from
an estimator [17]. In addition, the stator flux linkage vector
is estimated. Its argument is used for field-oriented control,
whereas its magnitude acts through a field-weakening controller and a flux controller on the excitation of the synchronous
motor. Superimposed is a controller for speed and the power
factor.
V. R ESULTS
As an example, the full set of optimal pulse patterns for N =
6 is shown in Fig. 8(a). The range of m selected for fs max =
400 Hz according to (24) is indicated in Fig. 8(b).
Fig. 8(b) shows that the discontinuities of the switching
angles exist at the changes in pulse number N . These abrupt
changes may generate current transients, which are avoided
by employing the stator flux trajectory control [12]. Fig. 8(c)
shows that the optimization results in a low distortion factor.
The method permits a smooth transition to the overmodulation
mode at m > 0.9.
The waveforms in Fig. 9 were recorded from a downscaled
30-kW laboratory setup. The operation at a 200-Hz space vector
modulation produces a highly distorted stator current, as shown
in Fig. 9(a). An increase in the switching frequency to 1 kHz
reduces the current distortion, as shown in Fig. 9(b). The same
distortion is achieved with the synchronous optimal modulation
at a switching frequency of only 200 Hz, as shown in Fig. 9(c).

Fig. 9. Measured waveforms of the phase a stator current and inverter

output potential at a fundamental frequency of 33.5 Hz. (a) Space
vector modulation at a switching frequency of 200 Hz. (b) Space vector
modulation at a switching frequency of 1 kHz. (c) 200-Hz synchronous
optimal modulation at a switching frequency of 200 Hz.

3378

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 6, JUNE 2015

Fig. 10. Commutation at t1 = 40 ms of the stator current from the

inverter to the ac mains.

energy conversion and control. This permits the unity power

factor operation that increases the efficiency of the drive system.
Extreme low-frequency switching reduces the dynamic
losses of the medium-voltage devices. The harmonic current
distortion is minimized using synchronous optimal PWM.
The optimization of the switching angles makes an allowance
for the anisotropic magnetic properties of the synchronous
motor. The machine is modeled by a fifth-order set of state
equations. The methodology of inductance operators reduces
the order of the state equations. The relationship between the
harmonic voltages and currents is then represented by only two
state equations. This simplifies the optimization algorithm.
Experimental results are obtained from a downscaled laboratory drive and from a 23-MV drive system in the field.
ACKNOWLEDGMENT
The authors would like to thank Dr. N. Oikonomou and
C. Erbach for their contributions to this paper.
R EFERENCES

Fig. 11. SAG mill installation with a diameter of 10 m at First Quantums

Kansanshi Copper Mine in Zambia.

After acceleration from standstill, the SAG mill operates at

a constant speed. It is then decoupled from the inverter and
commutated to the ac mains to increase the efficiency. The
commutation is shown in Fig. 10. A photograph of that SAG
mill installation at First Quantums Kansanshi Copper Mine in
Zambia is shown in Fig. 11. The diameter of the mill cylinder
is 10 m.
VI. S UMMARY
SAG mills are widely used in the cement and copper mining
industries for mineral grinding. Separately excited synchronous
machines with a high number of pole pairs are used as drive
motors. They operate at a very low rotational speed.
State-of-the-art SAG mills are driven by ring motors with the
stator poles mounted on the circumference of the mill cylinder.
The stator and the rotor thus form mechanically separated units.
The stator windings are fed by a thyristorized load-commutated
cycloconverter. Present drive systems have shown undesired
outages owed to frequent malfunctions of the cycloconverters
required for extensive repair work.
High reliability is achieved with the synchronous motor constructed as a separate unit, coupled to the mill cylinder through
a common shaft. An IGBT voltage source inverter is used for

[1] J. Rodrguez et al., Technical evolution and practical experience of grinding mill drives in mining applications, IEEE Trans. Ind. Appl., vol. 41,
no. 3, pp. 866674, May/Jun. 2005.
[2] J. Pontt, J. Rodrguez, and G. Sepulveda, Accionamientos Gearless
Para Molinos de Alta Potencia. Santiago, Chile: LOM Edicines Ltda.,
Nov. 2003.
[3] V. Guerrero and J. Pontt, Oscillatory torque caused by dead time in the
current control of high power gearless mills, in Proc. IEEE Ind. Electron.
Conf., Melbourne, Australia, 2011, pp. 19661970.
[4] J. Rojas, G. Urrea, L. Gutirrez, and J. Navarrete, A review of SAG mill
performance at Chuquicamata, in Proc. SAG 96, Vancouver, Canada,
Oct. 1996.
[5] J. Pontt et al., Current issues on high-power cycloconverter-fed gearless
motor drives for grinding mills, in Proc. IEEE Int. Symp. Ind. Electron.,
Budapest, Hungary, 2003, pp. 369374.
[6] J. O. Pontt, J. P. Rodrguez, J. C. Rebolledo, K. Tischler, and N. Becker,
Operation of high-power gearless drives under abnormal conditions,
IEEE Trans. Ind. Appl., vol. 43, no. 3, pp. 814820, May/Jun. 2007.
[7] J. O. Pontt, Current topics on reliability of high-power electrical machines employed in minery, in Conf. Rec. IEEE IAS Annu. Meeting,
Houston, TX, USA, Sep. 2009, pp. 15.
[8] [Online].
Available:
http://www.cleantechfinland.com/content/
outotec-presents-worlds-largest-sag-mill
[9] A. S. Dias, D. B. Cundido, G. Cunha, M. Sari, and P. J. Torri, A NPC
based medium voltage inverter applied to salient pole wound rotor synchronous machine, in Proc. 12th Brazilean COBEP, Gramado, Brazil,
Oct. 2013, pp. 852857.
[10] J. Holtz, Self-controlled inverter, (three-level and multi-level), German
Patent DE 23 39 034 C2 (1975), Patent in Japan (1983).
[11] A. Nabae, I. Takahashi, and H. Akagi, A new neutral point clamped
PWM inverter, IEEE Trans. Ind. Appl., vol. 17, no. 5, pp. 518523,
Sep./Oct. 1981.
[12] P. J. Torri, G. da Cunha, T. Boller, A. K. Rathore, and J. Holtz, Optimal
pulsewidth modulation for multilevel inverter systems, Eur. Patent EP 2
312 739 A1, Aug. 25, 2013, filed Sept. 29, 2009.
[13] J. Holtz and N. Oikonomou, Synchronous optimal pulsewidth modulation and stator flux trajectory control for medium voltage drives, IEEE
Trans. Ind. Appl., vol. 43, no. 2, pp. 600608, Mar./Apr. 2007.
[14] K. P. Kovcs and I. Rcz, Transient Processes in AC Machines.
Budapest, Hungary: Hungarian Academy of Sciences, 1959.
[15] G. S. Buja, Optimum output waveforms in PWM inverters, IEEE Trans.
Ind. Appl., vol. 16, no. 6, pp. 830836, Nov./Dec. 1980.
[16] A. Rathore, J. Holtz, and T. Boller, Synchronous optimal pulsewidth
modulation for low switching frequency control of medium-voltage multilevel inverters, IEEE Trans. Ind. Electron., vol. 47, no. 7, pp. 23742381,
Jul. 2010.
[17] J. Holtz and N. Oikonomou, Estimation of the fundamental current in low
switching frequency high-dynamic medium voltage drives, IEEE Trans.
Ind. Appl., vol. 44, no. 5, pp. 15971605, Sep./Oct. 2008.

HOLTZ et al.: CONTROL OF SALIENT-POLE SYNCHRONOUS MACHINES USING SYNCHRONOUS OPTIMAL PWM

Joachim Holtz (M87SM88F93) received

the Dipl.-Ing. and Ph.D. degrees in electrical engineering from the Technical University
Braunschweig, Braunschweig, Germany, in
1967 and 1969, respectively.
In 1969, he was an Associate Professor with
the Indian Institute of Technology, Madras, India,
where he became a Full Professor with and the
Head of the Control Engineering Laboratory in
1971. In 1972, he joined the Siemens Research
Laboratories, Erlangen, Germany. From 1976 to
1998, he was a Professor with and the Head of the Electrical Machines
and Drives Laboratory, Faculty of Electrical, Information and Media
Engineering, Wuppertal University, Wuppertal, Germany. He is currently
a Professor Emeritus and a Consultant with Wuppertal University. He is
the author of two invited papers in the P ROCEEDINGS OF THE IEEE and
12 invited papers in other journals, and he is a coauthor of four books.
He is the holder of 32 patents.
Dr. Holtz was the Editor-in-Chief of the IEEE T RANSACTIONS ON I N DUSTRIAL E LECTRONICS , and he is a Distinguished Lecturer of the IEEE
Industry Applications Society and of the IEEE Industrial Electronics
Society, where he is also a Senior Administrative Committee Member.
He was a recipient of 16 Prize Paper Awards, the IEEE Industrial
Electronics Societys Dr. Eugene Mittelmann Achievement Award, the
IEEE Industry Applications Societys Outstanding Achievement Award,
the IEEE Power Electronics Societys William E. Newell Award, the IEEE
Third Millenium Medal, the Anthony J. Hornfeck Service Award, and the
IEEE Lamme Gold Medal.

Gilberto da Cunha (S13) was born in Ibirama,

Brazil, in 1972. He received the B.E. and M.E.
degrees in control engineering from the Federal University of Santa Catarina, Florianopolis,
Brazil, in 1996 and 2003, respectively.
Since 1997, he has been with the Research and Development Department, WEG
Automaao, Jaragu do Sul, Brazil. His current
research interests include medium-voltage inverters.

3379

Norton Petry was born in Porto Alegre, Brazil.

He received the Diploma in electrical engineering from the Federal University of Rio Grande
do Sul, Porto Alegre, Brazil, in 1981 and the
Postgraduate Diploma in control and power
electronics from the Federal University of Santa
Catarina, Florianopolis, Brazil, in 2003.
Since 1981, he has been with WEG
Automaao, Jaragu do Sul, Brazil, leading
research and development activities and new
product development in the field of variablespeed drives.

Paulo Jos Torri received the Diploma in electrical engineering from the Federal University
of Santa Maria, Santa Maria, Brazil, in 1984
and the M.Sc. degree in electrical engineering
from the Federal University of Santa Catarina,
Florianopolis, Brazil, in 1986.
In 1987, he joined WEG Automaao, Jaragu
do Sul, Brazil, where he worked on the research and development (R&D) of products for
power electronics systems, drives, and electric
machines. He is currently the Manager of the
R&D Department, WEG Automaao, where he is responsible for the
development of products in the field of low-voltage and medium-voltage
power converters, high power, and renewables.