You are on page 1of 6

AbstractPrevious work on selective harmonic

elimination/control has made fundamental assumptions that


enforce output waveform quarter- or half- wave symmetry,
presumably in order to reduce the complexity of the resulting
equations. However, the quarter- or half-wave symmetric
assumption is not required and it restricts the solution space,
which can result in sub-optimal solutions with regard to the
uncontrolled harmonic distribution. More general formulations
can be proposed which have varying degrees of additional
complexity. In order to understand how these more general
formulations can be obtained, a qualitative description of the
waveform construction process for the two-level waveform case
will be discussed followed by presentation of the resulting system
of equations. This two-level case is then generalized to the m-
level, n-harmonic control problem. Finally, this generalization is
used to analyze three-level waveforms. All solutions presented in
this paper are unattainable utilizing previous techniques.

I. INTRODUCTION
Harmonic elimination has been a research topic since the
early 1960s, first examined in [1] and developed into a
mature form in [2-4] during the 1970s. Harmonic
elimination, a reduced switching technique for inverters,
provides direct control over output waveform harmonics.
This makes it a viable alternative to standard voltage sourced
inversion (quasi-square wave switching signals) or low
frequency PWM in high power conversion applications. In
the past, the problem has been formulated from a few different
perspectives, all assuming quarter-wave symmetry (except [5]
which enforces half-wave symmetry). The most familiar
formulation constructs an output waveform, two-level or
three-level, by notching a pre-existing square wave with each
notch representing a harmonic controlled [1-3]. Another
approach uses a double pulse waveform as a basis function to
create the desired two- or three-level output waveform [4].
Both formulations result in the same Fourier series
representation for a two-level or three-level waveform. The
resulting equations based on the Fourier series for the two-
level waveform are described by
( ) ( )
1
sin
2
( ) 1 2 -1 cos
4
i
j
j
k
n j j i j
i j
n
f m n n N
n


=




= = +

(1)
where n
j
is the j
th
element in a set of controlled harmonics N
having k elements, is a vector of length k, and q
i
is the i
th

switching angle. The magnitude of the harmonic content in
the n
j
th
harmonic is m
j
. The three-level waveform equations
are
( ) ( )
1
1
sin
2
( ) -1 cos
4
i
j
j
k
n j j i j
i j
n
f m n n N
n


+
=



= =

(2)
The set of all equations generated from (1) and (2) have
multiple solutions, which can be obtained using iterative
methods [6, 7], elimination theory [8], homotopy methods
[15] or optimization theory [9, 10]. Another formulation [11,
12], which maintains quarter-wave symmetry, uses Walsh
functions as a basis to create the desired harmonic elimination
waveform. This method results in a set of algebraic matrix
equations.
The quarter-wave symmetry assumption guarantees that the
even harmonics will be zero and that all harmonics will be
either in phase or anti-phase with the fundamental [4].
Although this is convenient, the quarter-wave symmetry
constraint limits the solution space. If the quarter-wave
symmetry constraint is relaxed to a half-wave symmetry
constraint as presented in [16], the even harmonics are still
zero but now the harmonic phasing is free to vary. The
method proposed in this paper advances the work done in [13]
and uses general periodic switching functions (GPSF) [13] as
a basis for the harmonic elimination waveforms creation.
Unlike the quasi-square wave basis, this basis function gives
freedom to place every switching edge in the waveform
independently. It is shown that this more general problem
formulation results in an infinite number of solutions due to
the under-constrained nature of the resulting system of
equations. The non-half-wave symmetric two-level
waveforms have unique uncontrolled harmonic profiles which
vary significantly from traditional results [6, 8, 13]. As such,
some solutions may have merit relative to others with respect
to system losses, ripple characteristics, or some other system
aspect. For example, these unique harmonic profiles can be
chosen such that they place significant energy in triplen
harmonics, a benefit for applications with balanced three-
phase wye-connected loads.
J.R. Wells, P.L. Chapman, P.T. Krein
Grainger Center for Electric Machinery and Electromechanics
Department of Electrical and Computer Engineering
University of Illinois at Urbana-Champaign
Urbana, IL U.S.A.
jwells@uiuc.edu, plchapma@uiuc.edu, krein@uiuc.edu
Generalization of Selective Harmonic
Control/Elimination

1358 0-7803-9033-4/05/$20.00 2005 IEEE.
This paper presents the two-level problem formulation and
analyzes several solutions for cases where triplen harmonics
are controlled and not controlled. The problem formulation is
then generalized to the m-level, n-harmonic case. Finally, this
generalization is used to analyze three-level waveforms.
Several solutions to the harmonic control problem are
presented which are unobtainable with previous techniques
including the generalization of [5].
II. TWO-LEVEL WAVEFORMS
A. Problem Formulation
As discussed in the introduction, the restriction of quarter-
wave or half-wave symmetry constrains the selective
harmonic control problem. Although this reduces the
complexity of the transcendental equations, it reduces the
solution space which may result in a suboptimal solution for a
given application. To relax the symmetry constraints, general
periodic switching functions [13] are summed to obtain the
desired switching waveform as depicted in Figure 1. This
basis function gives freedom to place every switching event in
the entire waveform independently. The resulting system of
equations to be solved is now posed by (3) and (4), where n
j
is
the j
th
element in a set of controlled harmonics N, D is a vector
of length K with each element related to the duty ratio of the
k
th
switching function as described by [14], is a vector of
length K with each element describing the phase shift of the
k
th
switching function, and m
j
real
and m
j
imag
are the desired real
and imaginary components of the n
j
th
harmonic.
( )
( )
( )
1
sin
, 2 cos
2
j
K
j i
real real
n j j i j
i j
n D
f m n n N
n


=


= =

D (3)
( )
( )
( )
1
sin
, 2 sin
2
j
K
j i
imag imag
n j j i j
i j
n D
f m n n N
n


=


= =

D (4)
Note that the harmonic content can also be described in polar
coordinates such that

( )
cos
real
j j j
m m = (5)

( )
sin
imag
j j j
m m = (6)
where m
j
is the magnitude and g
j
is the phase of the j
th

harmonic in the set N. Due to the lack of half-wave symmetry
in this formulation, even harmonics are no longer guaranteed
to be zero. Thus, in addition to the equations described by (3)
and (4), it is now necessary to control the dc component of the
waveform by enforcing the additional equation
( )
0
1
1
,
2
K
real
i
i
f D
=
= =

D (7).
In the generalization presented in [5], controlling all
harmonics out to a particular harmonic required at least

1
4 2
2
max
switch
n
n
+
= +


(8)
switching events where n
max
is the maximum harmonic to be
controlled and is always odd. This is the same number of
switching events required in the traditional case presented in
(1). Using the general switching function approach, for the
same harmonic control, only

1
4 1
2
max
switch
n
n
+
= +


(9)
switching events are needed. Since each general periodic
switching function provides two switching events, this leaves
an additional degree of freedom. The best utilization of this
additional degree of freedom remains an open question.
Possibilities include simply fixing a switching event at a
0 50 100 150 200 250 300 350
0
0.5
1
1.5
2
Anglular Time Scale (degrees)
N
o
r
m
a
l
i
z
e
d

O
u
t
p
u
t

0 50 100 150 200 250 300 350
0
0.5
1
1.5
2
Anglular Time Scale (degrees)
N
o
r
m
a
l
i
z
e
d

O
u
t
p
u
t

0 50 100 150 200 250 300 350
0
0.5
1
1.5
2
Anglular Time Scale (degrees)
N
o
r
m
a
l
i
z
e
d

O
u
t
p
u
t

a) General Periodic Switching Function (GPSF
1
) b) GPSF
2
c) GPSF
3
0 50 100 150 200 250 300 350
0
0.5
1
1.5
2
Anglular Time Scale (degrees)
N
o
r
m
a
l
i
z
e
d

O
u
t
p
u
t

0 50 100 150 200 250 300 350
0
0.5
1
1.5
2
Anglular Time Scale (degrees)
N
o
r
m
a
l
i
z
e
d

O
u
t
p
u
t

0 50 100 150 200 250 300 350
-1
-0.5
0
0.5
1
Anglular Time Scale (degrees)
N
o
r
m
a
l
i
z
e
d

O
u
t
p
u
t

d) GPSF
4
e) GPSF
5
f)
5
1
1
i
i
GPSF
=


Figure 1. Waveform construction, fundamental, 2
nd
, 3
rd
, and 4
th
controlled
1359
convenient place such as the zero crossing of the wave form,
imposing an additional optimization constraint equation, or
eliminating one GPSF per every two fundamental periods.
The problem posed by (3) and (4) can be solved using any
of the techniques that have been previously discussed in the
context of the more restrictive quarter-wave symmetric
problem including iterative approaches [6, 7], elimination
theory [8], minimization techniques [9, 10], homotopy
methods [15] or genetic algorithms [15]. Any solution to the
quarter-wave symmetric problem will also be a solution to the
more complete formulation posed in this paper.
B. Example Solutions
Figure 2 presents several plots that illustrate one set of
solutions to the two-level problem with the 1
st
through the 11
th

harmonics controlled. This requires at least 12 general
switching functions to solve 23 equations (11 for the real
components, 11 for the imaginary components, and 1 for the
dc component). Since each switching function introduces two
degrees of freedom (D
i
and f
i
), this results in 24 degrees of
freedom and the problem is thus under-constrained. An
iterative based solver was used to obtain solutions for D and
across a range of modulation depth and the results are shown
in 2a and 2b. Figure 2c shows an example waveform and
harmonic content of the solution with a modulation depth of
1.00. Note that the waveform no longer possesses quarter- or
half-wave symmetry and energy exists in the even harmonics
of the uncontrolled spectrum. Figure 3 presents similar
information for the case where triplen harmonics are not
controlled as is typical when using harmonic elimination in
motor drive applications.
III. GENERAL MULTI-LEVEL WAVEFORMS
A. Problem Formulation
The previous section detailed a more general problem
formulation and selected solutions for the two-level harmonic
control problem which is typically encountered. This problem
is actually a special case of the more general m-level, n-
harmonic control problem defined by
( )
( )
( ) ,1
1
sin
, cos
2
j
K
j i
real real
n j i j i j
i j
n D
f m c n n N
n


=
= =

D (10)
( )
( )
( ) ,1
1
sin
, sin
2
j
K
j i
imag imag
n j i j i j
i j
n D
f m c n n N
n


=
= =

D (11)
( )
0 ,1 ,2
1
, 0
K
real
i i i
i
f c D c
=
= = +

D (12).
where n
j
is the j
th
element in a set of controlled harmonics N,
D is a vector of length K with each element related to the duty
ratio of the k
th
quasi-square wave as described by [14], is a
vector of length K with each element describing the phase
shift of the k
th
switching function, m
j
real
and m
j
imag
are the
desired real and imaginary components of the n
j
th
harmonic,
and c
i,j
describes the magnitude (j = 1) and offset (j = 2) of the
i
th
switching function. The magnitude and offset of the
switching functions, c
i,j
, take discrete values from a set C
which is defined by the available voltage levels in the
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
Modulation Depth of the Fundamental Component
D
u
t
y

R
a
t
i
o
,
D

0 0.2 0.4 0.6 0.8 1
0
50
100
150
200
250
300
350
Modulation Depth of the Fundamental Component
S
w
i
t
c
h
i
n
g

F
u
n
c
t
i
o
n

P
h
a
s
i
n
g
,


(
D
e
g
r
e
e
s
)

0 5 10 15 20 25 30 35 40
0
0.5
1
H
a
r
m
o
n
i
c

M
a
g
n
i
t
u
d
e
Harmonic Number
0 50 100 150 200 250 300 350
-1
0
1
Angular Time Scale (degrees)
S
w
i
t
c
h
i
n
g

S
t
a
t
e

a) Duty ratio vs. modulation depth b) Phasing of the switching functions vs.
modulation depth
c) Example switching waveform and the
magnitude of the resulting harmonic content
at a modulation depth of 1.0 (see Appendix)
Figure 2. Example solutions for controlling the 1
st
through the 11
th
harmonics for a two-level waveform
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
0.2
0.25
Modulation Depth of the Fundamental Component
D
u
t
y

R
a
t
i
o
,
D

0 0.2 0.4 0.6 0.8 1
0
50
100
150
200
250
300
350
Modulation Depth of the Fundamental Component
S
w
i
t
c
h
i
n
g

F
u
n
c
t
i
o
n

P
h
a
s
i
n
g
,


(
D
e
g
r
e
e
s
)

0 5 10 15 20 25 30 35 40
0
0.5
1
H
a
r
m
o
n
i
c

M
a
g
n
i
t
u
d
e
Harmonic Number
0 50 100 150 200 250 300 350
-1
0
1
Angular Time Scale (degrees)
S
w
i
t
c
h
i
n
g

S
t
a
t
e

a) Duty ratio vs. modulation depth b) Phasing of the switching functions vs.
modulation depth
c) Example switching waveform and the
magnitude of the resulting harmonic content
at a modulation depth of 1.0 (see Appendix)
Figure 3. Example Solutions for Controlling the 1
st
through 13
th
Non-triplen Harmonics of a Two Level Waveform
1360
converter as

1 1
m m
i j
i j
i j
j i
l l
C
l l = =

(13)
where m is the number of levels in the converter and l
i
and l
j

are elements from the set L which contains all possible
converter level magnitudes.
B. Example Solutions for a Three-Level Waveform
The three-level waveform problem which is often discussed
in harmonic elimination papers can be obtained using the
generalization of the previous section by defining the set L as
{ } 1, 0, 1 L = (14).
Then by (13), the set C is defined as

1 1 2
, ,
0 1 1
C



=



(15).
In order to solve (10-12), it is necessary to choose an
appropriate number of each type of GPSW from C. Although
C includes three possible types of GPSFs for mathematical
completeness, topology considerations will often limit which
choices should be used. In this example, the choice of

2
1
c

=


(16)
would require switching all four devices in a standard H-
bridge inverter contributing unnecessarily to increased
switching losses.
Figures 4 and 5 present solutions to (10-12) for the three-
level harmonic elimination problem with triplens controlled
and uncontrolled respectively. For both examples, at least 10
GPSFs are required to solve 19 equations (9 for the real
components, 9 for the imaginary components, and 1 for the dc
component). In these examples, 5 GPSFs were chosen with

1
0
c

=


(17)
and 5 were chosen with

1
1
c

=


(18)
Again, note that the solutions do not possess the quarter- or
half-wave symmetry that is found in previously published
solutions.
IV. CONCLUSION
This paper has presented a more complete formulation for
the selective harmonic control problem generalizing the basic
theory set forth by [1-4] and extending the generalization of
[16]. It was shown that several solutions exist which have not
been previously identified despite claims of completeness in
several papers (although these papers did likely find all
solutions to the problem they posed). In particular, these
solutions differ by not imposing half- or quarter-wave
symmetry constraints that previous solutions must have.
The work presented in this paper may have several
implications regarding traditional harmonic control
applications. First, as suggested by [13], there exists an
optimal switching waveform for a given harmonic control
goal with respect to the uncontrolled harmonic content. This
problem formulation gives the designer much more flexibility
in selecting the uncontrolled harmonic content and may result
in improvements in a system cost function such as system
efficiency, electromagnetic compatibility, acoustic properties,
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
0.5
Modulation Depth of the Fundamental Component
D
u
t
y

R
a
t
i
o
,
D

0 0.2 0.4 0.6 0.8 1
0
50
100
150
200
250
300
350
Modulation Depth of the Fundamental Component
S
w
i
t
c
h
i
n
g

F
u
n
c
t
i
o
n

P
h
a
s
i
n
g
,


(
D
e
g
r
e
e
s
)

0 5 10 15 20 25 30 35
0
0.5
1
H
a
r
m
o
n
i
c

M
a
g
n
i
t
u
d
e
Harmonic Number
0 50 100 150 200 250 300 350
-1
0
1
Angular Time Scale (degrees)
S
w
i
t
c
h
i
n
g

S
t
a
t
e

a) Duty ratio vs. modulation depth b) Phasing of the switching functions vs.
modulation depth
c) Example switching waveform and the
magnitude of the resulting harmonic content
at a modulation depth of 1.0 (see Appendix)
Figure 4. Example Solutions for Controlling the 1
st
through 9
th
Harmonics of a Three Level Waveform
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
0.5
Modulation Depth of the Fundamental Component
D
u
t
y

R
a
t
i
o
,
D

0 0.2 0.4 0.6 0.8 1
0
50
100
150
200
250
300
350
Modulation Depth of the Fundamental Component
S
w
i
t
c
h
i
n
g

F
u
n
c
t
i
o
n

P
h
a
s
i
n
g
,


(
D
e
g
r
e
e
s
)

0 5 10 15 20 25 30 35 40
0
0.5
1
H
a
r
m
o
n
i
c

M
a
g
n
i
t
u
d
e
Harmonic Number
0 50 100 150 200 250 300 350
-1
0
1
Angular Time Scale (degrees)
S
w
i
t
c
h
i
n
g

S
t
a
t
e

a) Duty ratio vs. modulation depth b) Phasing of the switching functions vs.
modulation depth
c) Example switching waveform and the
magnitude of the resulting harmonic content
at a modulation depth of 1.0 (see Appendix)
Figure 5. Example Solutions for Controlling the 1
st
through 13
th
Non-triplen Harmonics of a Three Level Waveform
1361
etc. Investigation of such optimizations is a subject for future
research. In fact, the additional degree of freedom may allow
direct optimization of such a cost function in the solution
process.
As a final remark, it is important to realize when searching
for harmonic elimination waveforms that multiple solutions
can arise as result of two phenomena. First, the problem itself
is fundamentally under-constrained. As such, a free
parameter exists which can be varied independently giving
rise to a continuum of solutions. Second, even if this free
parameter is fixed, the system of transcendental equations can
be rewritten as a system of polynomial equations using
multiple angle identities [8, 16-18] which inherently has
multiple solutions.
APPENDIX
Experimental verification of example waveforms displayed in
paper (Note all waveform fundamental frequencies are 60Hz):
ACKNOWLEDGMENT
The authors would like to thank the Grainger Center for
Electric Machinery and Electromechanics and Motorola
Communications Center for providing funding and/or
facilities required for this work.
REFERENCES
[1] F. G. Turnbull, "Selected harmonic reduction in static dc-ac inverters,"
IEEE Transactions on Communication and Electronics, vol. 83, pp. 374-
378, 1964.
[2] H.S.Patel and R.G.Hoft, "Generalized techniques of harmonic
elimination and voltage control in thyristor inverters: Part II-Voltage
control Techniques," IEEE Transactions on Industry Applications, vol.
IA-10, pp. 666-673, 1974.
[3] H.S.Patel and R.G.Hoft, "Generalized techniques of harmonic
elimination and voltage control in thyristor inverters: Part I-Harmonic
elimination," IEEE Transactions on Industry Applications, vol. IA-9, pp.
310-317, 1973.

Figure 2c


Figure 3c



Figure 4c


Figure 5c
1362
[4] I. J. Pitel, S. N. Talukdar, and P. Wood, "Characterization of
programmed-waveform pulsewidth modulation," IEEE Transactions on
Industry Applications, vol. IA-16, pp. 707-715, 1980.
[5] J. R. Wells, B. M. Nee, P. L. Chapman, and P. T. Krein, "Selective
harmonic control: A general problem formulation and selected
solutions," IEEE Transactions on Power Electronics, in press.
[6] P. N. Enjeti, P. D. Ziogas, and J. F. Lindsay, "Programmed PWM
techniques to eliminate harmonics: A critical evaluation," IEEE
Transactions on Industry Applications, vol. 26, pp. 302-316, 1990.
[7] J. Sun and H. Grotstollen, "Solving nonlinear equations for selective
harmonic eliminated PWM using predicted initial values," in
International Conference on Industrial Electronics, Control,
Instrumentation, and Automation, 1992, pp. 259-264.
[8] J. N. Chiasson, L. M. Tolbert, K. J. McKenzie, and Z. Du, "A complete
solution to the harmonic elimination problem," IEEE Transactions on
Power Electronics, vol. 19, pp. 491-499, 2004.
[9] D. A. Deib and H. W. Hill, "The advantages of harmonic-distortion
minimization over traditional harmonic-elimination techniques," in
Power Electronics Specialists Conference, 1993, pp. 1048-1054.
[10] A. B. V.G. Agelidis, I. Balousktsis, "On Applying a Minisation
Technique to the Harmonic Elimination PWM Control: The Bipolar
Waveform," in IEEE Power Electronic Letters, in press.
[11] F. Swift and A. Kamberis, "A new Walsh domain technique of harmonic
elimination and voltage control in pulse-width modulated inverters,"
IEEE Transactions on Power Electronics, vol. 8, pp. 170-185, 1993.
[12] T. J. Liang and R. G. Hoft, "Walsh function method of harmonic
elimination," in Applied Power Electronics Conference, 1993, pp. 847-
853.
[13] J. R. Wells, B.M. Nee, P. L. Chapman, and P. T. Krein, "Optimal
harmonic elimination control," in Power Electronics Specialists
Conference, in press, pp.
[14] P. T. Krein, Elements of Power Electronics. New York: Oxford
University Press, 1998.
[15] A. I. Maswood, S. Wei, and M. A. Rahman, "A flexible way to generate
PWM-SHE switching patterns using genetic algorithm," in Applied
Power Electronics Conference, 2001, pp. 1130-1134.
[16] J. Chiasson, L. Tolbert, K. McKenzie, and D. Zhong, "Eliminating
harmonics in a multilevel converter using resultant theory," in Power
Electronics Specialists Conference, 2002, pp. 503-508.
[17] J. Chiasson, L. M. Tolbert, K. McKenzie, and Z. Du, "Elimination of
harmonics in a multilevel converter using the theory of symmetric
polynomials and resultants," in IEEE Conference on Decision and
Control, 2003, pp. 3507-3512.
[18] J. N. Chiasson, L. M. Tolbert, K. J. McKenzie, and Z. Du, "Control of a
multilevel converter using resultant theory," IEEE Transactions on
Control Systems Technology, vol. 11, pp. 345-354, 2003.


1363

You might also like