Você está na página 1de 30

[I] [2]

H.B. Dwight, Table of Integrals and Other Mathematical Data 3rd ed MacMillan, New York, 1957. ' ., T.A. Lipo, Introduction to AC Machine Design, Vol. 1, University of Wisconsin Power Electronics Research Center, University of Wisconsin, Madison, 1996.

Modulation of One Inverter Phase Leg


One of the most widely utilized strategies for controlling the AC output of power electronic converters is the technique known as pulse width modulation (PWM), which varies the duty cycle (or mark-space ratio) of the converter switch( es) at a high switching frequency to achieve a target average low-frequency output voltage or current. Modulation theory has been a major research area in power electronics for over three decades and continues to attract considerable attention and interest. This is not surprising, since modulation is at the heart of nearly every modern power electronic converter, but the enormous amount of material published makes it challenging for a user to identif).r basic modulation principles and apply them to particular implementations. Much reported research has presented "new" or "improved" PWM techniques, which are often only a straightforward variation of a previous approach. But it can at times be quite difficult to see how they are related. On the other hand, there have been a number of clear trends in the development of PWM concepts and strategies since the 1970s, addressing th.e main objectives of reduced harmonic distortion and increased output magnitudes for a given switching frequency and the development of modulation strategies to suit different converter topologies. Since the work presented in this book primarily relates to medium and higher power level hard-switched inverters (i.e., above the 1 kW power level), this chapter considers only modulation strategies which are appropriate for the two major converter topologies in this power range, being the voltage source inverter (VSI) and the current source inverter (CSI). While this chapter concentrates on VSI modulation, the principles are equally applicable to a CSI as will be discussed in Chapter 9. While there has been a wealth of research investigating the modulation and control of lower power DC/DC converters, the actual PWM process for these converters is usually a simple comparison between a reference wavefonn and a

Evaluation of PWM Schemes

sawtooth or a triangular carrier waveform. The operation and performance of this strategy is straightforward and needs little elaboration in this text. A more recent area of research in power electronic converters has been the development of soft-switched resonant converters, and the selection of appropriate modulation strategies for this class of converters is a further area of interest. Generally, closed-loop feedback modulation strategies are used for this class of inverter, and their performance is more dependent on the performance of the closed-loop controller than any intrinsic principles of PWM. Since the focus of this book is more directed toward fixed-frequency hardswitched PWM schemes, the modulation and control of resonant converters and similar topologies will not be considered further in this book. ) { IYlt<' c! <"It' e.; Sf(n t ..e (nO: f (~11< f,Jrr;6 IS', (l'Uf 1~I(d( c
I

Many variations of these three alternatives have been published, and it sometimes can be quite difficult to see their underlying commonality. For example, even the well-known space vector modulation strategy, which is often claimed to be a completely different approach to modulation, is really only a variation of regular sampled PWM which specifies the same switched pulse widths but places them a little differently in each carrier interval. In this chapter, the principles of PWM are developed in a generalized way which can be used with any converter topology and modulation implementation. This development offers a more integrated perspective than that presented in most technical papers and textbooks.

I I
~r()

The fundamental basis of this development is that all fixed-frequency open-loop PWM strategies can be explained in terms of Switched pulse width determination. Switched pulse position within a carrier interval. Switched pulse sequence within and across carrier intervals. The harmonic performance of a particular PWM implementation then results from the effect of these three factors on The harmonics generated by the phase leg switched waveform. Harmonic cancellation that may occur between individual phase leg switched outputs. Note that switching frequency in this context is not regarded as a significant parameter which changes harmonic performance - changes in switching frequency affect the influence of the sideband harmonics because of attenuation through the low-frequency filtering action of loads but have no direct impact on the magnitude of these sideband harmonics. In this chapter, the means by which naturally sampled and regularly sampled PWM determine the switched pulse width of one phase leg of an inverter is developed. The results of this development are then used in later chapters to investigate the performance of multiple phase legs arranged as single- and three-phase inverter systems.

()..wrs/w,lb,.i::lWlae

f1ef

i:f.'

cC'VI~(~v(

,rt

ft'

111 {tll(

lut;:Ok...

3.1

Fundamental Concepts of PWM

In principle, all J;I1odulationschemes aim to create trains of switched pulses which have the same fundamental volt-second average (i.e., the integral of the voltage waveform over time) as a target reference waveform at any instant. The major difficulty with these trains of switched pulses is that they also contain unwanted harmonic components which should be minimized. Hence for any PWM scheme, a primary objective can be identified which is to calculate the converter switch ON times which create the desired (low-frequency) target output voltage or current. Having satisfied this primary objective, the secondary objective for a PWM strategy is to determine the most effective way of arranging the switching processes to minimize unwanted harmonic distortion , switching losses, or any other specified performance criterion. It is interesting to note that despite the wealth of material that has been published relating to PWM, only three significantly different alternatives for determining the converter switch ON times have ever been usefully proposed for fixed-frequency modulation systems. These alternatives are: 1. Switching at the intersection of a target reference waveform and a highfrequency carrier (naturally sampled PWM). 2. Switching at the intersection between a regularly sampled reference waveform and a high-frequency carrier (regular sampled PWM). 3. Switching so that the integrated area of the target reference waveform over the carrier interval is the same as the integrated area of the converter switched output (direct PWM).

Before proceeding with any investigation into PWM strategies it is necessary to establish a common basis for comparing the performance of various algo-

rithms, and this has been the subject of considerable research work over the years. Some researchers argue the merits of a particular PWM implementation on the basis of diminished harmonics [1], others propose first- or second-order filtered distortion performance factors [2], and others evaluate the RMS harmonic ripple current in a typical load such as an induction motor [3]. Many variations of these approaches have been published, all with particular arguments as to why the performance index proposed is preferable. Unfortunately, one result of all this work is some confusion, since it becomes difficult to precisely compare the performance of PWM strategies presented by different researchers. With recent PWM innovations claiming performance improvements of fractions of a percent over previous approaches, it is important to be sure how much of the improvement claimed is because of more careful physical implementation, rather than any intrinsic advantages of the new scheme. In this text, particular care has been taken to compare all PWM variations on exactly the same basis. The approach used is threefold: First, where they are available, analytical solutions to the PWM strategies that are to be compared are used to determine the magnitude of the various harmonic components. This ensures that harmonics caused by simulation roundoff errors or practical implementation effects such as dead time, switch ON-state voltages, DC bus voltage ripple, etc., are not confused with intrinsic harmonic differences between the PWM strategies being compared. Second, PWM strategies are compared at exactly the same phase leg switching frequencies. Where necessary (e.g., with discontinuous modulation strategies) the carrier frequency of a PWM implementation is varied to ensure that harmonic comparisons are done at precisely the same average device switching frequency per fundamental cycle. - Third, the first-order weighted total harmonic distortion factor WTHDO \ (WTHD normalized to the fundamental component at a modulation index of M = I) is used as a single-term performance indicator for more rapid comparison of PWM alternatives. As explained in Section 2.2, WTHD can be interpreted as the normalized current ripple expected into an inductive load when fed from the switched waveform, and is a widely applicable performance indicator which has a useful physical significance.

T
I I
I I

Double Fourier Integral Analysis ofa Two-Level Pulse Width-Modulated

Waveform

3.3

Double Fourier Integral Analysis of a Two-Level Pulse Width-Modulated Waveform

The essential concept of a two-level pulse width-modulated converter system is that a low-frequency target reference waveform is compared against a highfrequency carrier waveform, and the comparison output is used to control the state of a switched phase leg. When the reference waveform is greater than the carrier waveform, the phase leg is switched to the upper DC rail. When. the reference waveform is less than the carrier waveform, the phase leg is switched to the lower DC rail. The resultant phase leg output is a stream of pulses switching between the upper and lower DC rails which (hopefully) have, as a fundamental component, the target reference waveform, but also incorporate a series of unwanted harmonics generated as a consequence of the switching process. Determination of the harmonic frequency components of a PWM switched phase leg output is quite complex and is often done by using a Fast Fourier Transform analysis of a simulated time-varying switched waveform. This approach offers the benefits of expediency and reduced mathematical effort but requires considerable computing capacity (especially when analyzing PWM systems with a higher carrier frequency) and always leaves the uncertainty as to whether a subtle simulation roundoff or error may have slightly tamished the results obtained. In contrast, an analytical solution which exactly identifies the harmonic components of a PWM waveform ensures that precisely the correct harmonics are being considered when various PWM strategies are compared against each other. The most well-known analytical method of determining the harmonic components of a PWM switched phase leg was first developed by Bowes and Bird [4], who adapted to modulated converter systems an analysis approach originally developed for communication systems by Bennet [5] and Black [6].

The analysis process assumes the existence of two time variables x(t)

= CDct+ec andy(t)

= CDot+eo,where

Tc

= =

carrier interval arbitrary phase offset angle for carrier waveform

ec

Double Fourier Integral Analysis of a Two-Level Pulse Width-Modulated

Waveform

CDo

2n/To To

= fundamental

(sinusoid) angular frequency,

CDo

< CDc (3.2)

periodic in both x and y. That is, identical unit cells are assumed to exist infinitely in both the x andy directions as shown in Figure 3.2. From Fourier transform theory [7J, any time-varying function f(t) expressed as a summation of harmonic components f(t)
=

with
= period =

of fundamental waveform

can be

eo

arbitrary phase offset angle for fundamental waveform ~o +

These variables x(t) and yet) can be thought of as representing the time variation of the high-frequency modulating wave and low-frequency modulated wave, respectively, with each of these variables being considered as independently periodic. The problem of solving for thelmodulated wavefo_rr.::l can then be addressed by exploring the existence of a unit cell which identifies contours within which f(t) is constant for cyclic variations of x(t) and y(t), as depicted in Figure 3.1.

L [amcosmCDt
m = 1

<X!

+ bmsinmCDtJ

iCt)

~f
bm
= ~

f(t)cosmCDt

dCDt

The value of the function f(t) = f[x(t),y(t)J within each contour region over the unit cell represents the output phase leg voltage and can be represented in topolQgical form as shown by allowing the z axis to take on the values of f(x,y) so that the functional dependence can be considered as three dimensional. The x axis and y axis of the cell are scaled in radians corresponding to the frequency of the modulating carrier and low-frequency fundamental reference waveforms, respectively, and span from -n to n. (Note that a span of 0 to 2n is often alternatively used by many texts, but this makes no difference to the solution process which only requires cyclic variation of x and y over 2 n intervals. The range of -n to n is used in this book to simplify the form of the analytical solutions which are developed during the text.) In general, the value of the function f[x(t), yet)] at any point in the unit cell can be represented as a Fourier Series if it is assumed that the function is

f(t) sinmCDt dCDt

In Appendix 1, it is shown that a Fourier harmonic component form similar to Eq. (3.3) can be developed for a double variable controlled waveform f(x,y),as

~WV~

~~lZ

A
j(x,y)
=

00

00

~O

+ L
n = I

[AOncosny

+ Bonsinny]

+ L[AmOcosmx

+ Bmosinmx]

It is useful to examine Eq. (3.10) carefully. The carrier index variable m and the baseband (angular) frequency of each harmonic component index variable n define the of the switched phase leg

output voltage, as (mcoc + ncoo)' This means, for example, values of m = 2 and n = 4 define the fourth sideband harmonic in the group of harmonics that + L L [Amncos(mx

+ ny) + Bmnsin(mx

+ ny)J

are located around the second carrier harmonic sideband (2 co c group). This sideband harmonic

(i.e., within the second carrier frequency of

m=ln=-oo (n"* 0)

will have an absolute

+ 4 coo) rad/s, where co c is the carrier frequency and co 0 is the target fun= 0,

damental frequency. Amn

2~2

f f
-IT IT

IT

IT

j(x,y)cos(mx

+ ny) dx dy

Special groups of these harmonics occur for m quencies are defined by n alone, and for n harmonic components,
= 0,

where the harmonic

fre-

where the harmonic frequencies and the carrier the fun-

-IT IT

are defined by m alone. These groups are termed the baseband damental component component, of the switched waveform

Bmn

2~2

f f
-IT

respectively. Note also that with this definition,

j(x,y)sin(mx

+ ny) dx dy

is the first baseband harmonic

although it is often useful to maintain it as a separate component. of the harmonic components defined in Eq. (3.10) are the for particular values Equation (3.9) gives a genthese coeffiof the in that it outto take account

-IT

or in complex form

The magnitudes

A mn and the B mn coefficients, Cmn

which must be evaluated

= Amn +jBmn = 2~2

-IT

f f

j(mx+ny) j(x,y)e dxdy(3.9)

ofm and n for each PWM scheme to be considered. cient values, particular once the integrals

eral form of the Fourier integral that must be solved to determine have been adapted switching pattern that is to be used. This adaption

Replacing x by coct + 8c andy by coot + 80, Eq. (3.6) can be alternatively expressed in time-varying form as A jet)
00

is discussed

detail in this and succeeding ate magnitudes

chapters, but for now it is simply assumed that define the switched

is possible to evaluate Eq. (3.10) for any particular

values ofm and n, to gener-

=:2 +

OO

L
n
=

[AOncos(n[coot+
I

80J) + Bonsin(n[coot+
Fundamental Component & Baseband Harmonics

for all the harmonic components of the inverter phase leg. where m

80])]

put waveform

I--_.

DC Offsetl

The first term ofEq. (3.10), Aoo/2 DC offset component

0, corresponds (if any).

to the

of the pulse width-modulated

waveform

+ L
m = 1

[AmOcos(m[coct

+ 8cJ) + Bmosin(m[coct + 8cJ)J

I Carrier

Harmonics

The first summation

term,

(.~=
00

where m

0, defines the output

m=l

n=-oo (n"* 0)

rAmncOs(m[coct

+ 8cJ + n[coot + 80J)


Harmonics

n = 1

fundamental fundamental

low-frequency

synthesized

waveform

and its baseband harmonics harmonics around the or (preferably) eliminated

+ Bmnsin(m[coct + 8cJ + n[coot + 80J)

(if they exist). This term includes low-order undesirable output which should be minimized process (except for the n
=

I Sideband

with the modulation

1 element which, of course, and must be retained).

defines the desired fundamental

output waveform

Naturally Sampled Pulse Width Modulation

The second summation term, carrier wave harmonics, which

(:z= ,:

3.4
= I ... )

Naturally Sampled Pulse Width Modulation


Sine-Sawtooth Modulation
modulation strategy is termed naturally target reference carrier waveform. waveform A simple one PWM, which compares a sinusoid) against a low-frequency is illustrated

where n

0, corresponds

to the

are relatively

high-frequency

components,

3.4.1
sampled

since the lowest frequency term is the modulating

carrier frequency. where m, n*"O,

The earliest and most straightforward (usually example

The final double summation is the ensemble waveform

term,

(:z=: = :z=: =
I

-00

... )

a high-frequency

(n '" 0)

of this type of modulation

in Figure 3.3, showing

of all possible frequencies

formed by taking the sum and difand the reference are gen-

phase leg of an inverter driven by a sawtooth wave carrier. It can be noted that, as described earlier, the phase leg is switched to the upper DC rail when the To reference waveform obtain a sinusoidal is greater than the sawtooth wave carrier and to the lower is greater than the reference waveform. strategy, the reference output using this modulation wave-

ference between the modulating

carrier waveform harmonics

and its associated baseband harmonics. These combinations

erally referred to as sideband harmonics, harmonic frequencies.

and exist as groups around the carrier

DC rail when the carrier waveform form has the form

In general the two angular frequencies However, if t~e Fourier coefficient the integration

COo

and co c will not be an integer over very

ratio, and the switched pulse train created by the phase leg will not be periodic. integral Eq. (3.9) is evaluated reference waveform where many periodiC cycles of the fundamental one cycle, eventually periodic for the same overall integration waveform instead of just solution

v;z = Mcos(coot+
M co 0
80
=

80) = Mcosy

(3.11)
output volt-

of the carrier frequency cycles will also be period. Hence, the harmonic period To' for

modulation index or modulation depth (i.e., normalized age magnitude) with range 0 < M < 1 target output frequency arbitrary output phase

= =

for f( t) defined by Eq. (3.10) is valid even for the case where the switched is not periodic over one fundamental

and the use of the asterisk denotes the "commanded" that the phase leg reference because the sawtooth this type of -;;'wto;th waveform carrier ramps from normalized

or "target" value. Note about zero, values of -1 to 1. With na!!:-

There does exist a possible noninteger

source of error if the Fourier coefficients

can be made symmetrical

carrier ratios are evaluated over less than the very many fundamencarrier and fundabased on the harOf course, such an error would affect only the predicted component not its frequency, defined in Eq. (3.10). However, it has been found that if

tal cycles which may be required to achieve both periodic mental waveforms. magnitude monic components all reasonable of each harmonic

carrier, o;ly the trailing edge of the pulse varies as M

changes value, and hence this type of modulation rally sampled modulation.

[_.-

_.

is termed trailing edge ----_._---

The unit cell for trailing edge sawtooth modulation where rather than f(x,y) the function for the case of PWM switching

is shown in Figure 3.4,

such errors exist in the theoretical ing roundoff errors of simulation ,~


I

solutions (and in most cases they do not), for studies and can be neglected. method of determining a Fast Fouto both

carrier frequency ratios they are much smaller than the comput-

having many areas of constant value in the x, y plane, simply has two values of

+ 2 Vdc
to the DC bus

This is in sharp co;trast the harmonic rier Transform components

to the more conventional of a switched waveform

l
if

and 0 as a result of the circuit switching action. These values correspond voltage potential of point a in Figure 3.3 with respect to the negative

by computing

point n for the two possible switching states of the phase leg. Note th.at defining the switched output with respect to the negative DC bus in this way rather than with respect to the midpoint z of the DC bus, considerably mathematics simplifies the a DC of the Fourier solution at the trivial expense of introducing

(FFT) of the waveform. of the simulation

This method is quite sensitive and the periodicity

the time resolution

of the overall wave-

form. In general, harmonic investigations the small magnitude

of PWM strategies using time simuare the focus of study.

lation studies are only effective for exact integer carrier ratios, particularly sideband or baseband harmonics

offset of

+ Vdc into the final solution.

+ Vdc + Vdc

TI4-L
Load

;, ~ r
vst

MCOSClO,t+ 60)

1.WW

-1.0

Naturally sampled trailing edge PWM with halfbridge (one phase leg) voltage source inverter. The boundary between the two switched levels represents the locus of all possible solutions to the switching problem defined as the intersection of the sawtooth wave,. defined along the x axis, and the sine wave reference command, defined along the y axis. Over multiple carrier and fundamental cycles the sine wave waveform replicates over the x, y plane as shown in Figure 3.5. As time t unfolds, x = 0) c t and y = 0)ot define a straight line in the x, y plane with slope y / x = 0)0/0) c = carrier ratio (note that both the carrier and fundamental angles are assumed to be zero at time t = 0, or in other words both e c and e 0 are zero for this discussion). Intersections of this line with the boundary locus between the two switching voltage levels correspond to the actual switching instant for particular values of 0) 0 and 0) c as shown in Figure 3.6. Hence, if p is the number of

Figure 3.5 The x,y plane for trailing edge sawtooth modulation sh.owing solution for particular values of 0)0 and O)c.

the cell along the x axis counting from the origin, then for naturally sampled trailing edge PWM the switching instants are defined such that
j(x, y) changes from 0 to 2 Vdc when

The function j(x,y) plotted versus O)ct in Figure 3.6(b) is the switched output of the phase leg. Note that when van' the voltage from phase leg output to the negative DC bus, is plotted versus O)ot (projected onto the y axis instead of the x axis) the waveform is the same as j(x, y), except it is scaled differently with respect to time (in radian measures of the modulated reference frequency rather than the modulating carrier frequency). Equations (3.12) and (3.13) specifY the limits of the integration regions defined in Eq. (3.9) for which j(x,y) is nonzero (and constant). Using these limit definitions, Eq. (3.9) can then be adapted to become

For m plifies to

0, n > 0 [the baseband harmonics in Eq. (3.10)], Eq. (3.14) simn

V~c

rc

e
-n

jny

dxdy

-n
n

-2

Vdc I

(rcMcosy + rc)e

jny

dy

rc

2VdCi
!(x,y) (b)

-n

nnni
x
n nMcosy

rc~c

1 rt
n

jnn

rcMcosy( cosny

+jsinny)dy

}] +

rc( e

;n

-jnn

(3.17)
=

wet

Figure 3.6 ~ Half-bridge switching for trailing edge PWM: (a) x, y plane showing intersection of reference with unit cells and (b) x, y plane showing resulting PWM voltage.

The integral within the curly brackets is zero except for the case where n = 1, while the second term is zero for all nonzero values of n, so that Eq. (3.17) evaluates to

2~2

-n

I I

2Vdce

j(mx + ny)

dx dy

-n

Equation (3.18) defines the magnitude of the fundamental harmonic of the modulated waveform, which equals as anticipated the target reference defined in Eq. (3.11), multiplied by Vdc . For m> 0, n fies to
=

which can be evaluated for particular values of the index variables m and n. For m
=

0 [the carrier harmonics in Eq. (3.10)], Eq. (3.14) simplin 1tMcosy

0 [the DC offset in Eq. (3.10)], Eq. (3.14) simplifies to

V~c

rc

I I
1t

n
~c

I I
-n
-1t

/mxdx

dy

-n

-n

-n

Y V In (jmnMCOS ....!!.E. e

-e -jmn)

Vdc -2 rc

(rcMcosy

+ rc) dy

Vdc 2 = -2 (2rc ) = rc

2 Vdc

rc2

jm

-1t

When substituted back into Eq. (3.10), this gives a DC component of

Aoo

2 =

Vdc

with

Boo

-.-2

Vdc

I
-n

1t

jm1tMcosYd

y-e

-jmn In -n

dy

2 =

(3.16)

Jmrc

The first integral within the square brackets ofEq. (3.19) can be evaluated using Eq. (A2.16), so that Eq. (3.19) becomes
. AmO+JBmO
=

2 Vdc --[Jo(mnN1)-(-l) Jmn

]
=

(3.20) as
(3.21)

2V

CD

+ ~"

which can be reexpressed using (-1


AmO + jBmo
=

~m
m
=

l.[cosmn-Jo(mnM)]sin(m[coct+
l

e c ])

cosmn

2Vdc j--[

mn

cosmn-Jo(mnM)]

+-n

~ dCI
?V

00

For m> 0, n:;t:. 0 [the sideband harmonics in Eq. (3.10)], the inner integral ofEq. (3.14) evaluates to give
A

m = 1

mn J mn

+B

Vdc . 2 Jmn

It

jny( jmltMcosy -:jmlt e -e) dy

-It

Amn

+jIJ~n

= -

'V J d~[2n/Jn(mnM) mn

- e-

jmlt

jnlt (e

-jnlt ~e In

)J

(3.23)

From Figure 3.3, it is clear that the phase leg switched output voltage measured with respect to the DC bus midpoint z is the same as the switched output voltage measured with respect to the negative DC rail n, but with a - Vdc offset. Hence Eq. (3.26) also defines the harmonic components of vaz( t) , if the first term which defines the DC offset is discarded. Figure 3.7 presents the voltage spectrum for trailing edge naturally sampled PWM for the conditions of a carrier ratio of 21 and a modulation index M of 0.9. This plot shows the single fundamental low-frequency component produced by the modulation process, and the groups of sideband harmonics arranged around the carrier and double carrier harmonics. Some sideband harmonics from the triple carrier group can also be seen on the right-hand side of the plot. All harmonic components including the fundamental have magnitudes which have been normalized with respect to Vdc' since this is the magnitude of the synthesized output fundamental for a modulation index of 1.0 (this convention is followed throughout this book to allow exact comparisons to be made of the various harmonic plots that are presented). The WTHDO for this modulation strategy under these conditions is 4.03%, but this of course varies with both the carrier ratio and the modulation index M Figure 3.8 shows how the WTHDO changes with these parameters, where it can be seen how the harmonic significance reduces with increasing modulation index and an increasing carrier ratio. Note that the precise carrier ratios used in this figure have no particular significance, particularly in terms of being odd and/or triplen. The reader is referred to Section 3.8 for further discussion on the issue of carrier pulse ratio.

The second term in the square brackets of this expression is zero for all nonzero values of n, and since
e
J-n
2
.It

A mn +JB mn

j2 Vdc ---e mn

jnT!:

2J(mnM) n

which can be reexpressed as


. Amn + JB mn
= dc 2V --In(mnM) (.

mn

smn-- Jcosn-

n.

Equation (3.25) is valid for all nonzero values of m and n, i.e., for m = 1,2, ... ,00 andn = -00, ... ,-2,-1, 1,2, ... ,00. The complete harmonic solution for trailing edge naturally sampled modulation of a half-bridge phase leg can now be formed by substituting the results

of Eqs. 3.16, (3.18), (3.21), and (3.25) back into Eq. (3.10), so that the timevarying switched phase leg voltage van(t) can be expressed in terms of its harmonic components as

F7="_-=_-=_-=::-:::-_="!_ :=-:-=:-=:-=::-:::-:::-=:=-: 7: -=:-=::-:::-:::-~:r=-: 7: -=:-==-=_:-:::-~!="" . =-:r IW~T;:;;H;:;D:;=::;;0=-='4='.


- - - 1- -1- -

0;:;3;:;;'o/c7H;, ~
-

12
I I I

------------

------------------------I I ------------I
----l------r
I ___ I L -f-

- - -

-l -

f- -

1- I

I - - - - - - I" - - - - - -

10

- - - - - - - I - - I

I" - - - I" - - I I I

-I

-------:::~::::

I 1'" I I -"t

I -

----t---

to

~
0

- - - r - - I I I I

-r - - -...,
I I I

- - -.,
I

- - -..,
I

- - -..,I

--

::f? '-'
0

Q
------1----I -

::r: f-<
~
----1---------1--

___

.1

J
I

J
I

..J
I

I
I I

I I I

I I I I I I I I I I I

----------------1--------------------1------------I

- 1-

__ __ - -

__.!.
I

.:

.2
I

-'

_1_1-

= _

=J=_ J _
2

I T T I -

I T I -

I ., I I I -

I ..., I I I I .,

I I I I I -

I .., I -I

I I I -

- - - r - - I \

- - - - -1-

______

1_-

I I

~'===::I::
------1--

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Modulation Index M WTHDO for one inverter phase leg with trailing edge naturally sampled PWM for varying modulation indices and carrier ratios. Note that the WTHD results presented in Figure 3.8 (and in all other similar plots in this book) have been calculated using harmonic components taken directly from Eq. (3.26) and normalized with respect to Vdc (i.e., effectively the WTHDO expression developed in Section 2.9). This means that no adjustment is made for changes to the magnitude of the fundamental component as the modulation index varies. Such an approach provides a convenient comparison of the relative levels of harmonics produced by various modulation strategies and operating conditions, but does not give an indication of the significance of the harmonics compared to the fundamental target component. If WTHD relative to the fundamental synthesized component is required, the results presented in Figure 3.8 can be scaled by dividing their magnitudes by the modulation index. However, this approach gives a relative WTHD of infinity at a modulation index of M = 0, which while mathematically valid, is less useful as a practical measure of the distortion level than an absolute value ofWTHDO based on the normalized harmonic components.

20

30

40

Harmonic Number Figure 3.7 Harmonic components for one inverter phase leg with trailing edge naturally sampled PWM, M = 0.9, fJ fo = 21.

The reduction of WTHD with modulation index occurs primarily because the magnitude of the first carrier harmonic reduces with increasing modulation, and this harmonic dominates the WTHD calculation. (It will be seen later that the main carrier harmonics usually have no influence when phase legs are paralleled to make complete inverter systems since they become common mode terms and cancel. Consequently the way in which WTHD varies with increasing modulation index for complete inverter systems is quite different to that of one phase leg.) The reduction ofWTHD with an increasing carrier ratio is easily identified as the reduced influence of the carrier and sideband harmonics on the WTHD calculation as they move toward higher frequencies.

Using the same sinusoidal The more common form of naturally sampled PWM uses a triangular against the reference carrier as for j(x, y)

reference

waveform

as given in Eq. (3.11), the such that

switching instants for double-edge


X

modulation

can be expressed

instead of a sawtooth

carrier to compare

waveform

2ITp - 2(1 + MCOSffiot)

IT

shown in Figure 3.9. With this type of carrier, both sides of the switched output pulse from the phase leg are modulated, monic performance ble-edge naturally sampled modulation. for j(x,y)
p

which considerably

improves the haris termed dou-

changing from 0 to 2 Vde, and


X

of the pulse train. This type of modulation

2ITp

+ 2(1 + Mcoswot)

IT

changing from 2 Vde to O. voltage

+ Vde + Vde Load

Dj
+

The switched

van(t)

measured

from the phase leg output

to the

negative DC bus and plotted versus wet then becomes the waveform Figure 3.11. Under the integration (3.9) becomes

shown in

limits defined by Eqs. (3.27) and (3.28), Eq.

D2

2~2

f f
-]1; _

j(mx +ny)

2 Vdee

dxdy

~(1 + Mcosy)

Double-edge naturally sampled PWM with half-bridge phase leg) voltage source inverter.

(one

which can now be evaluated for the various possible values of m and n. y = wot

A plot of the corresponding val

unit cell for sine-triangle

single-phase

modula-

tion is shown in Figure 3.10. Once again, the unit cell is defined over the inter-IT

< wet < IT rather than from 0 to

2IT

for simplicity of result.

nnn
Figure 3.11 Half-bridge (a) x, y plane switching showing for intersection double-edge of reference

x
with

wet

PWM: unit

cells and (b) x, y plane showing resulting PWM voltage.

For m > 0, n
1t

0 [the carrier harmonics


1t

in Eq. (3.10)], Eq. (3.29) simpli-

dc V 2

rr

1t

2:(1 + Mcosy)
dxdy
1t

fies to

f
- 2:( 1 + Mcosy)
1t

-1t

AmO

+jEmo

Vd2c
rr

1t

2:(1 + Mcosy)

f
1t

ejmxdxdy

-1t

-::(1 + Mcosy) 2

dc V -2

rr

[rr(1 + Mcosy)]

dy

Vdc --

-1t

}mrr

f [ ejm::2(1 + Mcosy) -e -:im::(l 2


-TC

+ MCOSY)]

dy

Thus Aoo = 2 Vdc, Eoo = 0, which gives the expected Vdc offset in the final solution when the negative DC bus is taken as the switched reference voltage. For m fies to
=

0, n> 0 [the baseband harmonics

in Eq.(3.l0)],

Eq. (3.29) simpli-

AmO

+jEmo
=

~:~[/m~Jo(

m~~

- e-jm~Jo(

-m~~

and since Jo( -I;)

Jo( 1;) from Eq. (A2.22), 4 Vdc


=

AOn

+jEon

Vd2c
rr

f f f
TC

TC

jny e dxdy

AmO

+}EmO

mrr Jo

m"2.~ smm"2.

rr

~.

rr

-TC

For m > 0, n [rr(l

=F

0 [the sideband harmonics

in Eq. (3.10)], the inner integral

Vdc -2
rr

+ Mcosy)e

jny

ofEq. (3.29) evaluates to ] dy Amn

+.

}Emn

Vdc

1t

jny [ jm~(l + Mcosy) -jm~(1 + MCOSY)] e - e dy

TC

}mrr
Vdc --

-TC

Vdc rr

[jny e

+ -(e

M 2

j[n+l]y

+e

j[n-l]YJ

) dy

f
2

TC

}mrr
Since

[jm-2 jny jm::2Mcosy -jm:: jny -jrn::MCOSY] e e e -e 2e e 2 dy

TC

f /n

-TC

dy

0 for any nonzero value of n, Eq. (3.31) reduces to

for the case of n

= 1 only. For all other n> 1, AOn +jEon

= 0 .Once again

this equals the target reference as defined in Eq. (3.11), times Vdc

(m- M jmrr n 2

2V dc --J

rr

J[

'

TC.1t InJm2 e 2-

. TC -In.TC] -Jm2 e 2

which simplifies to Amn + jBmn


=

4:~cJn( m~M) sin([m + n]~ of the modulated wave. Note that since n is

- - - - - - - - - - -

- - - -

- 1- 1- 1- 1 - 1 1

- - - - -

- - 1 - - -

1 - - - -I - -1

These terms define the sidebands both positive and negative,

- - - - - -

,- - - - - - -,
- - - - - - - - - - - - - - - -

- ~ - - - -- - - - - - - -< - - - - - - - - - - - - ~ - - - - - - -

- - -

-,

- - - - - - -

- - - -

- - - - -

-,

- - - - - -

~ - WTHDO=4. 00% I- - - - -, - - - - - - - - - - - ,- - - - - - - 1- - - - - ~ - - - - - - 1- - - - - 1 1 r - - - - - ,- - - - - 1

the voltage side bands are equally spread above and

1 _I - -I - -I - -1 - 1 - -

below the central carrier wave harmonics. The complete harmonic solution for double-edge naturally sampled moduthe results

----III "0 C

10-1

~ 0..

'--'

- - - - - - - - - - - - - - - - - - -

lation of a half-bridge phase leg can now be formed by substituting


varying switched phase leg voltage van(t) monic components as can be expressed

1- 1- 1- 1- - - - - - 11

- - - - - J - - - - - - - - - - - ~- - - - - -< - - - - - - - - - - - - - - - ~-

-'
l

- - -

- - -

t- !. !. L I1

- ,- - - - 1- - - - - 1- - - - - - 11 - - - 1

.~
Of) c<:l

- - - - - - 1- - - - - 1

-,
-I -I

- - - - - - - - - - - - -

- - - - -

- - - -

r 1

- - -

,- - - - 1 1

of Eqs. (3.30), (3.32), (3.35), and (3.38) back into Eq. (3.10), so that the timein terms of its har-

1 _I

~
()

10-2

= - = - = - = - 1- = - = - = - = - = - = - =
-

'2
0 c<:l

- - - - 1- - - - - - - -- - - -

= - = - = - = - = - - - - - - - -

= ,=- = - = - - - - - - - J

- - - - 1- - - - - - -1 - - - - - - - - - - - - - - 1 - - - - - - -1
1 1 1 1

~
10-3

~' 4 Vdc

+ -;-

LJ
OO

;/0

n ~ n m2~ sinm2 cos(m[coct

+ ec])

,- - - - - - -, - - - - ,- - - - - - - - - - - - - - - -, - - - - - - - - _I 1-

-'
l

- - - - - - - - -

L - I- - t- !. !. 1

= - = - = = - = - = - = - 1-= = - - - - - - - - - - - 1- - - - - 1

- - - 1- - - - - - - -- - - - - - - - 1- - - - - -

- - - - -

- - - 1

r 1 L I-

,- - 1 1111 1 1

- -

m = 1

- - - - - - 1- - - - - - 1- - - - - - - - - - - - 11

- - - - - -I - - - - - -

- - - - - - - - - - -1

+-n

4V

dCL

- - - - - - 1- -

;;/n(
n=-oo (n '" 0)

-r

- -

-,

J - - - - - - - - - - - - - - - - - - - ~- - - - -' - -

, ,
1

: -

- -

- I- !. !.
1

- -

m~j~ x

sin([m + n]~

-r

- -

- - - - - -

- -

- - -

l
1

- - - -

r 1

- -

,- - 1

, 40

20

30

m = 1

cos(m[coct+ec]+n[coot+eo])
components of vaz( t), if the

Harmonic Number

As before, Eq. (3.39) also defines the harmonic first term which defines the DC offset is discarded.

Figure 3.12

Harmonic

components

for one inverter phase leg with double-

edge naturally sampledPWM,


naturally sam-

0.9,

1/1

21 .

Figure 3.12 presents the voltage spectrum for double-edge pled PWM for the conditions ponent produced monics arranged of 0.9. Once again this plot shows the single fundamental by the modulation around the carrier and double carrier

of a carrier ratio of 21 and a modulation

index M comharSome sampled of

It should be noted that while the WTHDO for this modulation is because the major harmonic carrier fundamental contributor

strategy

of

low-frequency harmonics. naturally

4.00% is almost identical to that of trailing edge naturally sampled PWM, this to WTHD in this case is the main at 21fo' As will be seen later this carrier is eliminated for all multiple phase leg converter topologies since it is a common mode component, and the performance of the two PWM strategies then differ markedly. and since the carrier fundamentals naturally sampled PWM, naturally index for double-edge are the

process, and the groups of sideband

sideband harmonics

from the triple carrier group can also be seen on the right feature of double-edge sideband components around odd multiples eliminated

of the plot. However, the significant PWM is that the odd harmonic the carrier fundamental, multiples double-edge to single-edge sin[(m + n)nl2] of the carrier

and even harmonic sideband components fundamental, are completely

around even by the of

Because of this carrier dominance, variation ofWTHDO with modulation here.

very similar for trailing edge and double-edge

expression

in Eq. (3.39). This is an intrinsic advantage using a triangular

sampled

naturally sampled modulation modulation

carrier compared

PWM is virtually identical to that shown in Figure 3.8 and is not shown again

using a sawtooth carrier.

~ f vanU) sinmx dx

An alternative approach to calculate the harmonic spectrum of naturally sampled PWM is to modulate the duty cycle of a simple buck converter with a lowfrequency sinusoidal reference [8]. This approach can be developed by considering the switching process of one inverter phase leg during a few arbitrary cycles of the sawtooth carrier waveform, as shown in Figure 3.13. If the reference waveform is assumed to be constant within each carrier cycle (i.e., Ie fo) with a duty cycle of d = 1 + Mcosy, from Eqs. (AU), (Al.2), and (Al.3), the Fourier series of the resultant switched waveform is
00

-n

2Vde --[

mn

cosmn - cos(mnMcosy)]

Note that when m = 0, ao = 2 Vde(l + Mcosy) From Eqs. (Al.2) and (Al.3)

and bo

cos(mnMcosy)

Jo(mnM)

+2

I
n=!

cosn~Jn(mnM)cosny

van(t)

ao

"2+ "" L..J


m = I

(amcosmx+bmsinmx)

2
n nMcosy

I
n = 1

sinn~Jn(mnM)cosny

am

~ fVan(t)cosmx
-n

dx

2Vde n f
-n

cosmx

dx

2Vde --[sin(mnMcosy) mn

Substituting these relationships into Eqs. (3.41) and (3.42) allows Eq. (3.40), the Fourier form of the switched output voltage with respect to the negative bus, to be written as (3.41 )
van(t)
=

+ sinmn]

m*O

Vde(l

+ Mcosy) + --nde L..J "" m [cosmn -Jo(mnM)]


m= I

2V

00

sinmx

(3.45)

+ 4Vde~

L..J m

1.

I
n=l

sinn~Jn(mnM)cosnycosmx

m =!

_ 4Vde
n

L..J
m
=

1.
m
1

I
n=!

cosn~Jn(mnM)cosnysinmx

wet

Production of PWM waveforms by comparison of sawtooth carrier wave and low-frequency control input wave.

For the first term, the result would be the same for negative n, since only odd n terms can exist because sin(nn/2) = 0 for even n, and from Eq. (A2.18), J_n(~) = -In(~) but also sin(nnl2) = -sine -nn/2), for odd n.

For the second term, the result would also be the same for negative n, since because cos (nn/2) = 0 for odd n, only even n terms can exist, and from Eq. (A2.l8), J_n(~) = In(~) for even n. Hence, the summation terms produce the same result if they are summed from -00 to -1 rather than from 1 to 00. Alternatively, the summation can be taken over both the ranges -00:::; n :::; -1 and 1 :::; n :::; 00 provided the result is divided by two. Thus Eq. (3.45) can be rewritten as

Similarly
sinmxcosNy

+ sinmxcos(-Ny)

sin(mx + Ny) + sin(mx-Ny)

(3.50)

van(t)

Vdc(1 + Mcosy)

2 Vdc + -;-

L ;;;
00

[cosmn

-Jo(mnM)]
er'

sinmx
J

(3.51)

an

(t)

Vd (1 +Mcosy)+-c

2Vdc

Lm
00

1 -[cosmn-Jo(mnM)]sinmx

(3.46)

+--

2Vdc n

m = !

m=l

2 ~dC ,,~ ~;;;1 ~ ~


m=l n=-oo (n '" 0)

n sinn'2Jn(mnM)cosnycosmx
_ 2:

dC

m=

L L
00

m=!

;;;In(mnM)Sinn~COS(mx
'~(\

+ ny)

n=-oo

(n'" 0)
00

,
sin(mx + ny) N-"'
+

I'

,(

~/n(mnM)cosn~

! n=-oo (n '" 0)

1-1'

-r~'

/.I

lr

Nn

00

00

_ 2 ~dC ,,~

"-m1" n ~ cosn'2Jn(mnM)cosnysinmx
m=l n=-oo
(n '" 0)

Replacing x by (J) ct + e c and y by (J) ot + 80' it is evident that this result is the same as Eq. (3.26) obtained in Section 3.4.1 using the double Fourier integral technique.

",c,'

Now, for each value of m, two terms can be taken from the inner summation expressions with the positive and negative values of n that have the same absolute magnitude, say, for example, n = N. Consequently, using the trigonometric identities

2:[
sinmxcosny
=

cos(mx

+ ny) + cos(mx - ny)]


sin(mx-ny)]

A similar approach can be used to develop the harmonic spectrum of a sinusoidal reference modulated by a triangular carrier. For a triangular carrier, the reference waveform has a duty cycle of d = (1 + Mcosy)/2, and once again is initially assumed to be constant within the carrier interval. For this case, the Fourier series corresponding to the resultant switched waveform becomes
ao

~[sin(mx+ny)+

van(t)

-+ 2

L
m=
1

00

(amcosmx

+ bmsinmx)

(3.52)

where
cosmxcosNy

+ cosmxcos( -Ny)

2:[ cos(mx + Ny) + cos(mx


+
1 2:[ cos(mx

-Ny)]
1t

1t

2(1 + Mcosy)

- Ny) + cos (mx - (-Ny] (3.49)

am

~ f van(t)cosmx
-1t

dx

2Vdc n

f
-~(l+Mcosy) 2

cosmx

dx

(cos(mx

+ Ny) + cos(mx-Ny

dC 4V mTC[ sin (TC m (1 + Mcosy) 2

)J

Equation

(3.56) can be readily rearranged

by changing

the inner summation

limits to become

Vde + VdeMcosy + ---;-

4 Vde

L...J

1 (TC ~ TC ;/0 m2 Mj sinm2

m =!

- ~(l + Mcosy) 2

4: f f ;;J
de
m=l n=-oo (n '" 0)

m~M) sin([m + n]~) cos(mx

+ ny)

Note that when m

= 0, ao = 2Vdc(l + Mcosy).
which is identical

Equation. (3.53) can be rewritten as

to Eq. (3.39) when x and yare The simplicity

replaced by wet + 8e and

am

4:~C[sinm~cos(

m~MCOSY)

+ cosm~sin(

m~MCOSy)J

mol-O
(3.55)

wot + 80, respectively.

of this approach compared to the double

integral method should be apparent.

One major

limitation

with naturally

sampled

PWM is the difficulty because

of its

implementation

in a digital equation

modulation

system,

the intersection To overcome this sysrefercarrier carrier of the

between the reference sinusoid and the triangular or sawtooth carrier is defined

+ 8VdC~
TC

L...J m

by a transcendental limitation

and is complex

to calculate.

the modem popular alternative are sampled values

is to implement

the modulation during each

m=l

tem using a "regular sampled" PWM strategy, where the low-frequency ence waveforms and then held constant are compared

Vdc(l + Mcosy)

4: f ~/o(
dC
m = I

interval. cosmx waveform

These sampled

against the triangular

m~M) sinm~

to control the switching varying reference.

process of each phase leg, instead

sinusoidally

8Vdc +-TC

L L ~/J m~M)
00 00

The sampled reference waveform sin([m + n]~

must change value at either the positive depending on the sampling changing the referswitch

cosny

cosmx

or positive/negative

peaks of the carrier waveform,

m=l

n=!

strategy. This change is required to avoid instantaneously transitions if it was allowed to occur.

Vdc(l + Mcosy)

4: f ;/0(
de
m =!
n

ence during the ramping period of the carrier, which may cause multiple

m~M) sinm~ cosmx


For a sawtooth carrier, sampling occurs as the carrier waveform end of the ramping period. For a triangular falls at the carrier, sampling can be symmetri-

deL +-4V
TC

OO

~ 1; (mr.!:.Jv?sin([m
L...J m
2)
n=!

+ n]r.!:.l[cos(mx + ny) 2J + cos(mx - ny)J

cal, where the sampled reference is taken at either the positive or negative peak of the carrier and held constant for the entire carrier interval, or asymmetrical,
where the reference is resampled every half carrier interval at both the positive

and the negative carrier peak. These alternatives are shown in Figure 3.14. Note that there is no concept of symmetrical and asymmetrical single-edge PWM, because the phase leg switching transition is calculated every (sawtooth) carrier cycle. Figure 3.14 shows that the sampling process produces a stepped reference waveform which is phase delayed with respect to the original reference wave-

form. For a sawtooth carrier and symmetrical sampling, this phase delay is one half the carrier interval, while for asymmetrical sampling the phase delay is one quarter the carrier interval. This phase delay can be compensated by phase advancing the reference waveform to produce the result shown in Figure 3.15.

Phase Advanced Sampled Reference

Phase Advanced Symmetrically Sampled Reference Triangular Carrier

)I

Sinusoidal Reference Sinusoidal Reference (c) Asymmetrically Sampled Reference

Phase Advanced Asymmetrically Sampled Reference

Figure 3.14

Regular sampling for (a) sawtooth carrier, (b) symmetrical sampling with triangular carrier (positive peak sampled), and (c) asymmetrical sampling with triangular carrier.

Regular sampling with phase advanced reference for (a) sawtooth carrier - l/2 carrier advance, (b) symmetrically sampled triangular carrier - l/2 carrier advance, and (c) asymmetrically sampled triangular carrier - l/4 carrier advance.

Since the same result can be obtained by phase delaying the sampling point on the original reference waveform rather than phase advancing the reference waveform itself, it is clear that this adjustment only affects the phase of the harmonics of the resultant PWM switched waveform, not their amplitudes. Figure 3.16 shows how the width and the placement of the switching pulse created by a phase leg varies depending on the carrier type and the sampling process used. (Note that the change in the target reference within one carrier interval is grossly exaggerated from normal in Figure 3.16 for purposes of illustration). As can be seen from this figure, the effect of regular sampling is to change the position of the switching instants within each carrier interval. This effect can be reflected into the double integral harmonic integration of Eq. (3.9) by changing the limits of the inner integral over which the function j(x, y) is nonzero for each type of modulation. In general-'terms, the solution to Eq. (3.9) for two-level modulation can be expressed as
1

Naturally Sampled

Symmetrica Regular Sampled

-2

f f
1t -1t

xf

2 Vdce
r

j(mx + ny)

dx dy NOTE: All sampling points are phase delay compensated

2rr
Xr

Asymmetrical Regular Sampled

= =

instant at which phase leg switches to 2 Vdc instant at which phase leg switches back to 0 V

xf

For naturally sampled PWM with a sawtooth carrier, these inner integral limits have been previously identified as
xr

-rr

xf

rrMcosffiot

(=

rrMcosy)

(3.59)

where the 2rr rollover included in Eqs. (3.12) and (3.13) has been deleted for convenience since it has no effect on the integration. For naturally sampled PWM with a triangular carrier, the inner integral limits have been previously identified as
x
r = --

Switched pulse from one inverter phase leg for (a) sawtooth carrier and (b) triangular carrier, for naturally and regular sampled PWM alternatives (phase delayed sampling).

rr

2 (1 + M cos ffi 0 t)

Using these limits, the harmonic components of a naturally sampled PWM waveform under a sawtooth or a triangular carrier have been developed previously as Eqs. (3.26) and (3.39), respectively. In order to similarly calculate the harmonic components of a regular sampled PWM waveform, it is now necessary to modify these inner integral limits to take account of the effects of the sampling process.

where again the 2rr rollover included in Eqs. (3.27) and (3.28) has been deleted for convenience.

The intersection trajectory shown in Figure 3.l7(b) is a staircase variable Figure 3.17(a) shows how the switching instants for regular sampling can be determined by the intersection between the sampled sinusoidal reference waveform and the solution trajectory line y = y'+(coo/coc)x. Figure 3.17(b) shows how the same switching instants can be determined by the intersection between the continuous sinusoidal reference waveform and a sampled solution trajectory where y is held constant within each carrier interval. Both approaches create the same phase leg switched output, but the approach of Figure 3.17(b) is more mathematically tractable to evaluate since the sinusoidal reference waveform remains smooth.
y' , which has a constant value frozen at x
=

0 within each carrier interval. For


=

example, during the first carrier interval, y' interval, y'


COo
= -

0; during the second carrier

2rc ; and so on. In general the value of y' within each carrier

COc

interval can be expressed as


COo

y'

-2rcp
COc

(3.61)

where p represents the pth carrier interval in a fundamental cycle. The staircase variable y' can also be expressed in terms of the continuous variables x and y as
y'
=

y--(x-2rcp)
COc

COo

The inner integral limits of Eq. (3.58) for the case of regular sampled PWM with a sawtooth carrier can now be defined using this new staircase variable as
xr

= -rc

xf

= rcMcosy'

(3.63)

so that Eq. (3.58) becomes

V~c

rc

11

e
-11

j(mx + ny)

dxdy

-11

Equation (3.64) can be evaluated by changing the integration variable y to = y' + (co/coc)x using Eq. (3.62), where the 2rcp rollover term associated with x can once again be discarded since it has no effect on the integration, to become

I
Figure 3.17

f(x,y)

V~c

rc

f rOO'! J'++ =:'JJ


11
-11 -11

<U dy'

Half-bridge switching for trailing edge PWM: (a) sampled sinusoidal reference waveform and (b) solution trajectory held constant within each carrier interval.

V~c

rc

11

llMCOSY'

j([m + n: o]x + n
e
c

y')

d" dy'

-11

Equation (3.65) can be evaluated for particular values ofm and n as follows: For m
=

0 , the solution process for Eq. (3.15) remains valid, so that

For m = 0, n> 0, the solution process leading to Eq. (3.25) can be used, with m replaced by n(wo/wJ. This gives baseband harmonic coefficients of

+ --n

2 Vdc

00

1
c

~ -[cosmn-Jo(mnM)]sin(m[w L..Jm
m = I

t+ 8 c ])

Equation (3.67) reflects a significant difference for regular sampled PWM compared to naturally sampled PWM, since the baseband harmonic coefficients are no longer zero for n > 1 . However, in practice the magnitude of the additional harmonics usually roll off fairly rapidly with n for any reasonable carrier ratios .. For m> 0, n so that AmO + jEmo For m> 0, n
=f:. = =

2Vdc +--

oo

m= 1

0, the solution process leading to Eq. (3.21) remains valid, 2Vdc mn

As before, Eq. (3.70) also defines the harmonic components of vaz(t), first term which defines the DC offset is discarded.

if the

j--[

cosmn-Jo(mnM)]

0, the solution process leading to Eq. (3.25) can be used,

with m replaced by m + n( w/

wJ.

This gives sideband coefficients of

Figure 3.18 presents the voltage spectrum for trailing edge regularly sampled PWM for the conditions of a carrier ratio of 21 and a modulation index M of 0.9. Comparison of this plot with the spectrum for naturally sampled PWM shown in Figure 3.7 shows the anticipated difference between the two modulation strategies - the presence of low-order baseband harmonics above the desired fundamental for regular sampled PWM. These harmonics are a consequence of the regular sampling process and occur for any regular sampled PWM strategy. However, the roll-off of their magnitude with n is affected by the carrier ratio and the modulation strategy. Higher carrier ratios (and/or sinetriangle sampled PWM as shown in the next section of this chapter) achieve a considerable improvement in the rate of roll-off of the baseband harmonics, and are therefore generally preferable. The other subtle variation in the harmonics created by regular sampling is a slight shift of carrier sideband energy between the lower and the higher sideband harmonics. This can be seen by careful comparison of Figure 3.18 versus Figure 3.7, where it can be seen how regular sampling has distorted the symmetry of the sidebands compared to the naturally sampled case. This effect is essentially a property of regular sampled PWM, although the exact distortion does vary a little depending on the particular modulation strategy used.

As before, Eq. (3.69) is valid for all nonzero values of m and n, i.e., for m = 1, ... , 00 and n = -00, ... , -1, 1, ... , 00 . The complete harmonic solution for trailing edge regular sampled modulation of a half-bridge phase leg can now be formed by substituting the results of

Eqs. (3.66), (3.67), (3.68), and (3.69) back into Eq. (3.10), so that the timevarying switched phase leg voltage von( t) can be expressed in terms of its harmonic components as

: : = = = _ ~ __ :
-------;------f---------------

~ _ =1 WTHDO=5.03%~
-

y'

y-

COo

-(x-211:p)

coe

====

I ------1-------

I ------------I ----i------r

= J

====

= = [= I
I

==

-----------

=:(

= == = =

I -----1------

~ ~ ci. ~
.-;:
-0 ::l
<l)

10-1

:::

'" ~

OIl

10-2

'a 0
'" ~

()

s ,..,
10-3

I ----1----I

I 1 - I J __

-< -< -

nn
Figure 3.19
Xr

Half-bridge switching for symmetrical regular sampled double-edge PWM showing (a) staircase solution trajectory and (b) resultant switched PWM voltage.

The inner integral limits for regular sampled PWM with a triangular carrier can be defined similar to Eq. (3.60) using the same staircase variable y' as Figure 3.18
edgeregularsampledPWM,

Harmonic components for one inverter phase leg with trailing M = 0.9, fe/fa = 21.

-~(1

+ Mcosy')

xf

11:

2(1

+ Mcosy')

so that Eq. (3.58) becomes The WTHDO for regular sampled modulation using a sawtooth carner under the conditions listed in Figure 3.18 is 5.03%, which is higher than for naturally sampled PWM and primarily reflects the influence of the additional baseband harmonic components.
1t

V~e
11:

-1t

f f
2
1t 1t

1t

2(1 + Mcosy') j(mx

+ ny)

dxdy

- !:E( 1 + Mcosy')

Once again this expression y = y' + (coo/coe)x to become A similar approach can be adopted for symmetrical regular sampled PWM with a triangular carrier, except that the rising edge inner integral limit xr must now also be defined using the staircase variable y'. Figure 3.19 shows how the intersection occurs between the sinusoidal reference and the staircase solution trajectory for symmetrical regular sampled PWM.

can be evaluated

by replacing

y with

V~e
11:

2( 1 + Mcosy')
e

ID

j([m

n---':!.]x + nyJ
IDe

dx dy'

-1t

Equation (3.73) can be evaluated for particular values of m and n as follows: For m
=

0, the solution process for Eq. (3.30) remains valid, so that

For m

0, n > 0, the solution process leading to Eq. (3.38) can be used,


.

with m replaced by n( OJol OJc)

This gives baseband harmonic


C_

coefficients

of

A
On

+J'B
On

_4_V_d_ J (n OJo!!:.M\sin([n OJo+ nJ!!:.) n OJ c 2) OJ c 2 o

[ro ] nroc

11:

+
PWM coeffi-

4: f
dC m= I

Equation

(3.75) again shows the difference

between

regular

sampled harmonic

and naturally

sampled PWM, with the existence of baseband

cients for n > 1 for the case of regular sampled PWM.

As before, the DC offset voltage ence point for the switching generality. Equation

Vdc simply reflects the selection of the referand can be discarded without loss of

waveform

For m> 0, n so that

0, the solution process leading to Eq. (3.35) remains valid,

(3.78) can be expressed

more compactly

for the switched

output

voltage measured with respect to the DC bus midpoint,


v
(t)

vaz(t), as

az

= --

4V

dc

11:

I I
m=OHn=l m>OHn=oo

oo

oo

~ In( q~M) sin([q + n]~


x

(3.79)

For m > 0, n"* 0, the solution process leading to Eq. (3.38) can again be used, with m replaced by m + n(OJoIOJc)' This gives sideband coefficients of

cos(m[OJct +

ecJ

+ n[OJot

+ eoJ)

where q = m + n(OJo/roJ (and is not necessarily integer). Note the change in the lower summation limit for n depending on whether m = 0, which is required because baseband harmonics are only defined for positive values of n. Figure 3.20 shows the voltage harmonic spectrum for symmetrical sampled PWM with a triangular carrier for the conditions scheme compared regular

of a carrier ratio of natu-

Once again, Eq. (3.77) is valid for all nonzero m


=

values of m and n, i.e., for


,00.

21 and a modulation differences

index M of 0.9. This figure shows a number of significant to both double-edge

1,2, ...

,00

andn

= -00, ...

,-2,-1,

1,2, ...

for this modulation

rally sampled PWM and regular sampled PWM with a sawtooth carrier. The complete harmonic solution for symmetrical regular sampled modulation of a half-bridge phase leg can now be formed by substituting the results of Eqs. (3.74), (3.75), (3.76), and (3.77) back into Eq. (3.10), so that the timevarying switched phase leg voltage v an( t) can be expressed monic components as in terms of its harFirst, the carrier sideband cancellations that were achieved by double-edge reduction in the magnitude the cancellation of

naturally sampled PWM have not been fully achieved for the case of symmetric regular sampling. While there is considerable the odd sideband harmonics band harmonics around the even carrier multiples, around the odd carrier multiples and the even sideis nowhere

- - - - - -1- - - - - - - 1- ------------I

- - - - -I - - - - -I I I -I I I : : -I

------1------t---

---

-------------

------

= = = = =

I I
I

= = = == = r I: = == == I = = = = = = I ------------- ---I I ----1-----I I

I - 1- I

-----l------rI

~ ci.. '--"
<l) "0

10-1

- - - -----: : : : - - -

_1----: 1- : : : -

- -1-

- - - - -\

= = = = = = I: = = = = = =

For asymmetrical regular sampling, the inner integral limits ofEq. (3.58) must now be modified to account for the two sampling points that are established in each carrier interval. This can be achieved by creating two staircase variables, one frozen at x = -n/2 within each carrier interval to determine the rising edge of the switched pulse, and the other frozen at x = n/2 within each carrier interval to determine the falling edge of the switched pulse. These variables can be expressed as

.2
Of)

~
(.)

10-2
_____ 1 _

E ....
CI:l

_____

___ ___

J J
I

_ _

- 1I 1------

::r:

I - 1- -

---l---I

y;

CO y- -

o(

coe

x - 2 np - ~ -

----:---T 10 20 30 40

The inner integral limits for asymmetrical regular sampled PWM with a triangular carrier can be defined using these new staircase variables as
Xr

= -

~(1 + Mcosy;)

Harmonic Number Figure 3.20 Harmonic components for one inverter phase leg with triangular carrier and symmetrical regular sampled PWM, M = 0.9,

so that Eq. (3.58) becomes


1t

V~e

1/1

= 0

21.

-1t

f f

1t

2:(1

+ Mcosyj) j(mx +ny)

dxdy

near complete. This has significant implications for the harmonics produced by phase legs operated together as single- or three-phase bridge inverters, as will be considered in later chapters. Second, the baseband harmonic components produced by the regular sampling process still exist, but their magnitude has rolled off much more rapidly than for regular sampling with a sawtooth carrier. For the conditions presented in this chapter, the second harmonic in the baseband for sawtooth carrier regular sampled modulation has a magnitude of 6%, while for symmetric regular sampled modulation with a triangular carrier, the magnitude of the second harmonic is below 0.5%. This clearly illustrates the benefits of using a triangular carrier in a regular sampled modulation system.

To solve Eq. (3.83), the switched waveform in each carrier interval is split into two sections for analysis, with the results added by superposition. The first section has a modulated rising edge in the first half carrier interval and a falling edge in the center of the carrier interval. The second section has a rising edge at the center of the carrier interval and a modulated falling edge in the second half carrier interval. Each waveform section is evaluated separately, and their harmonics are then combined to determine the harmonics of the overall switched waveform.

fr(x, y) steps from

Oto2Vdcat 2Vdc to 0 at

x-2np x-2np x-2np x-2np

xr

and from

0 0
=

fJx, y) steps from


and from Equation (3.83) then becomes

o to 2Vdc

at

~;"[ i

~(1 + McosY,'l

dy,' +

~(1 + McosY)l

dy)] ~

2 Vd, (3.87)

2Vdc to 0 at

xf

This result has the expected Vdc offset as before. For m > 0, n
=

0, Eq. (3.86) simplifies to

f f
1t

/(mx

+ ny) dx

dy

f f

jmx

dxdy;

1t

f f
-1t

1t

2(1 + Mcosyj) ej(mx

2(1 + Mcosyj)

+ ny) dx dy

jmx

dxdy)

The two sections of Eq. (3.85) can be evaluated by replacing y with


y

= y/+(Ul/UlJ[x+(n12)]

and y = y)+(Ulo/Ulc)[x-(n12)]

for each

section, respectively, derived from Eq. (3.81), to become


.

f
Vdc
}mn
2

1t

[l_e-jm~(1+McOSY;)]

dy;

-1t

f f
1t

1t

[ejm~(1

+McOSYj)_l]

dy)

-1t

Vdc

-1t

f[

1- e

-jm~ 2

-jm~Mcosy 2

,] r

dy;

}mn (3.86) Equation (3.86) can be evaluated for particular values of m and n as follows: For m
=

1t

f[
-1t

jm~

jm- MCosy ' 2 f -1 ]

1t

dy)

0, Eq. (3.86) simplifies to

.h +jq'!:.Mcosy On Iy th e t erms Wlt an e- 2 component need be solved for the outer Equation (3.90) identifies that the harmonics of the carrier wave for asymmetrical regular sampled PWM are the same as for double-edge naturally sampled PWM and for symmetrical regular sampled PWM, as per Eqs. (3.35) and (3.76). integration, since

Ai

ny

dy

O. Hence Eq. (3.92) can be simplified to

TC

Ol

A mn +;'B mn _- Vde . 2
jqn

f
-TC

jn......!!.'!:. jnYr - jq'!:.2 -jq'!:.Mcosy' - e Ole2 e e e 2

Yr

'

f f
A mn +"B ;
mn--2

Ol ,OTC

-:ln

Ole2

ejnYj

IT

IT

ejq2

jq-Mcosy' e 2

dy)

de
IT

n
~(l+MCOSY;) ([ j

f
m + n( co/co

Ol]0 m + n Ol
e

x + ny; -

[Ol]) 0 n Ol
e

dxdy)

= Vde
;qn
. 2

-TC TC

jny ere

- jm'!:. 2

- jq'!:.Mcosy , 2 r dy;

o
(3.91) and solving the inner integration gives

jnYj e e

jm'!:. jq'!:.Mcosy' 2 e 2 j

dy)

-TC

TC

j(ny'

rOle

+ [n Ol o]'!:.) [ 2

1- e

- jq~(1

+ Mcosy
r

')]

dy; Amn

+;Bmn=

"

2n Vde, jqn2)

n[

-In

n ~ _ 'm'!:. -q2~ e j 2 +In

q~~ejm2

~"TC]

-TC

f
.

TC

j(ny;_[n:o]~) e

[ jq~(1

+ McosYj)

-1

dYj

and using Eq. (A2.21) to A mn +;B mn = 2V ;qn

de[J ( q2~~(.n ;
n

jrn~

-;

,-n 2 J]
TC

-jrn-

Vde

f
2

TC

j n-

[Ol]

0 7t

Ole

. 1t

. 1t

jny

1-

- jq2

- jq2MCOSYr]

dy;

2Vde
. ;qn

[J ( q2~~(
n

j[m+n]'!:.

2 -

-j[m+n]'!:.J] 2

-TC

;qn

J e +::]l
-IT

and which finally simplifies to


jny; e
[

jq2 e

" TC

jq2 e

. TCM

COSY '

-1 dy)

Am" +iBm"

= [

m + -n
COe

4::] 4[
n

m + :"n
e

JiM)

sin([m + n]~

For m

0, n

7=

0, Eq. (3.96) can be used with m set to 0, to give

AOn

+JEan

= [

4 Vdc
CO

JI1 -n;/vJ)
CO c

(COo

n ~. TC smn2:

-nn
COc

Figure 3.21 shows the voltage harmonic spectrum for symmetrical regular sampled PWM with a triangular carrier for the conditions of a carrier ratio of 21 and a modulation index M of 0.9. This figure shows a significant harmonic improvement compared to symmetrical regular sampled PWM with a triangular carrier. In particular, the odd harmonic sideband components around odd multiples of the carrier fundamental, and even harmonic sideband components around
even multiples of the carrier fundamental, are once again completely elimi-

which once again contains the baseband harmonic components expected from regular sampled PWM. The complete harmonic solution for asymmetrical
regular sampled modulation of a half-bridge phase leg can now be formed by substituting the results

of Eqs. (3.87), (3.90), (3.96), and (3.97) back into Eq. (3.10), so that the timevarying switched phase leg voltage vanU) can be expressed in terms of its harmonic components as
van(t) Vdc + ----;4Vdc
~

nated by the sin([m + n]~ eliminated by the sin ( n ~


100
- - - - - - - - - - -

expression in Eq. (3.98), despite the regular samexpression in the first summation term.
- -< - - - " - - - - - -< - - - - - - -; - - - - - - - - - - - - ~- I

pling process. In addition, the even low-order baseband harmonics have been

L..

n=1

- - - -,I 1- - - - - - - - 1- - - - - - -I - I- - - - - - - I - - - - - - - - - - - - I I - - - - - - -- --- - I

- - - - - -

,- - - - - - -,

- - -

_ c _ - t- -

- WTHDO=3 - t- - - - - - - - - - - - - - - - - I - '- - - - - - - - - I I - - - - - - - - - - I I I

.98%
- - - - -

- - - - -

- - - --

1 I

- - - - - - r -

+ ----;-

4 Vdc

L ;/0
00

( n ~ n m2:M) sinm2: cosCm[coct + 8J)

m = 1

- - - - - - - - - - - - - - - -

- 1- 1- I- I - I I

,- -- - - - -

I _I -I

- - - - - - -

- - - - - - - -I
- - - - -

-,

- - - - -

+-TC

4V

dCL

oo

- - - - - - I- - - - - -

, -,
I I I -I

- - -- - - -< - - - - -; - - -- - -- J

- - J - - - - - - - - -

, ,

- - - - - - - - - -

t-

.. -

;- -

- - -

,- - - - I

'I I

- 1- 1- 1- I- I - I I

- - - - - - - - -

- - - - - - - - - - - - - - - - - - - - -

- - - - - 1- - - - I

r tt-

- - - I- - - - - -

sin( [m + n ]~)
x

= - = - = - = - = - = - = - = - = - = - I= - =
- - - - -

m = 1

coSCm[coct + 8c] + n[coot + 80])

- - - - - - - - - - I- - - - - - - I- - - - - - - - - - - - - - - - I- - - - I

- -I - - - -I

= - = -

,
I

- - - -- - - -

-< - - - - - - J - - - - - -J - I

= - = - = - = - = - = - - ~- - - - - - - -

= == == - 1- = - - = - = - = - - - - - - - 1- - - - - - - - -I - - - - - - -

c
'I I L t-

,
I

- -

As before the DC offset voltage Vdc can be discarded from Eq. (3.98) without loss of generality, and Eq. (3.98) can be expressed more compactly for the switched output voltage with respect to the DC bus midpoint, vaz( t) , as

- - - - - - I- - - - - -

-,
-I

- - - - - 1- - - I

r -

- -

,- - - - 11-

I
-

~- -

az

(t)

= 4V --

dc

TC

L L
m=OBn=\ m>O
B
n=OO

oo

oo

- - - - I- - - - - - - - - - I- I I

- - - - I- - - - I- - - - I-

- - -

- -I - - -I
- -I
I

_I

- - -

- - - -

- - - - - - - -< - - - - - - - - ~- - - - J - - - I

J -<

- ccI

.. -

- - - - - - - - - I
I

,- - - - - - - - -

- - - - - - I - I I

~ In( q~M) sin([m +


x

n]

(3.99)

- - - - I - - - - - - -I
I

- - -

1
I

- - - -

f
I

- -

- - - - - -

coscm[coct+8J+n[coot+80])

20

30

40

Harmonic Number Figure 3.21 Harmonic components for one inverter phase leg with triangular carrier and asymmetrical regular sampled PWM, M = 0.9, 1/10 = 21.

where again q = m+n(coo/coJ (and is not necessarily integer). Note the change in the lower summation limit for n depending whether m = 0, which is once more required because baseband harmonics are only defined for positive values ofn.

These benefits are an intrinsic advantage of asymmetrical sampling compared to symmetrical sampling for triangular carrier (or equivalent) regular sampled modulation systems and have significant consequences when the harmonic canceling that occurs between phase legs is considered for the various converter topologies discussed in later chapters.

Another method of determining the width of the switched pulses is occasionally proposed as an alternative to regular sampled PWM. The concept is to switch the inverter to create an active pulse interval for each carrier interval that exactly achieves the same volt-second average as the original target waveform. The method is not usually practical to implement since it requires integration over the carrier interval, but it has been presented in the literature [9] as a more accurate reference against which other strategies can be compared. In genera! terms, the required pulse width is given by

interval and a falling edge in the second half-carrier interval. For an asymmetrical system, the pulses in each successive half-carrier intervals are placed alternatively at the start and the end of each interval (i.e., back to back across a complete carrier interval). Since every two pulses therefore run together, no intermediate phase leg switching is required between every second pulse, and there is no increase in switching frequency despite the two pulse width calculations in each carrier interval. The pulses produced by this strategy are very similar to those produced by symmetrical and asymmetrical regular sampling, respectively, as shown in Figure 3.16(b), except that the pulse widths vary slightly because of the alternative technique for pulse width calculation. Figure 3.22 shows the voltage harmonic spectrum for asymmetrically calculated direct PWM with a triangular carrier for the conditions of a carrier ratio of 21 and a modulation index M of 0.9. As could be anticipated, the performance of this approach is almost exactly the same as for that of asymmetrical regular sampled PWM shown in Figure 3.21. A similar result is obtained when symmetrically evaluated direct modulation is compared against symmetrical regular sampled PWM. A simple variation of direct modulation is to approximate the reference waveform as a straight line across the evaluation interval. Equation (3.100) is then replaced with
_ COS[CDotJ

= __ dulyeycle

(ti

+ 1-

1__

t)

li+ 1 MCOS(CD
0

t) dt

I.

This pulse width can be evaluated either on a symmetrical basis over a complete carrier interval or on an asymmetrical basis at half carrier intervals. Mathematically, these two alternatives can be expressed as
1.+t>T
I

Wdulyeyele

COS[CDoti+

(3.103)

t.,1 T

f
I.
I

M cos ( CD 0 t) dt

which is very close to the value of M cos [CD oti + 0.5] used in regular sampled systems for all reasonable carrier ratios and consequently produces much the same harmonic performance. In reality, direction modulation offers virtually no harmonic benefit at any reasonable carrier ratio. For very low carrier ratios, there is a slightly better resolution of the fundamental magnitude compared to the reference (remember that regular sampling does introduce some error in the baseband fundamental

I.

+ t>T
2 MCOS(CDot)

~T

dt

4 Vde 1 because of the -------J


n:

(
n

[n(CDo/CDe)]

n--M
CD

CDon:

2 e

term

III

Eqs. (3.78) and (3.98)

'

for asymmetrical evaluation. For a symmetrically calculated system, the resultant pulse is placed at the center of the carrier interval, to create a rising edge in the first half-carrier

and this error increases with a decreasing carrier ratio). However, very low carrier ratios also lead to significant carrier sideband harmonics intruding into the spectrum below the fundamental (these harmonics are often referred to as subharmonics but are really very low-frequency sidebands from the first carrier

- - - - - - - - - - - -

- 1- 1- 1 - - - - -- 1 - - - - - - 1

- - -

- -I - - - - - - - -I - - - - - - - - - - -1 - - - - 1 - - - - - 1

- - - ~- - - ~- - WTHDO - - - - - - +- - - - - - - - - - ,- - - - - - - - -, - - - - - - - - - - - - - ~- - - - - - ~- - - - - - - -j

1 1

- - - - - - -

4 .OO%~ - - - - -

l
I
I

- - - - - - - - - - - -- - - - - - - - - - - - -- -

,- - - - - - -,
,_
1

- - - - - - - - - - - -

11

- - - - - r 1

- - - -

,- - - - - 1

;:i

P..
'-'
Q)

- 1- 1- - 11

- - - - - _I - - - - - -1 - - - - - -

'0

- - - - - - - - - - - - -1 - - 1 - - - - - - -1

- - - - -

-I

- - - - - J - - - -j - - - - - - -j - - - - - - - - J - - - - -- - -

~
1 1

- - - - - - -

L
j.

- - +-

,- -

- - -

'. 1 1

- 1- - - - - - - 1- - - - - - - 1- - - - - - - - 1- - - 1 1

I I
I

spective of the switching frequency ratio, the PWM process produces a fundamental component, baseband harmonic components when any form of regular sampling is involved, and sideband harmonics grouped around multiples of the carrier frequency, at frequencies of mfc nlo' In some literature, non-integer frequency ratios have been identified as undesirable because they are claimed to introduce subharmonics below the fundamental. For example, a carrier frequency ratio of 29/3 has sometimes been said to create subharmonics of one-third and two-third times the fundamental component. However, the PWM solutions that have been presented in this chapter identify that any harmonics that might exist in this region are really only low-order carrier sideband components. In the case above, for example, the so-called sub-harmonics would really correspond to the 9th and loth lower sidebands grouped around the first carrier frequency harmonic. But as shown in Figures 3.7 and 3.12, the amplitudes of these sideband harmonics are negligible after the first few components. For applications using very low carrier frequency ratios (say < 11), there is an argument for using an integer frequency ratio to ensure that all significant lower sideband harmonics from the first carrier group have a frequency above (or at worst equal to) the fundamental frequency. For natural and asymmetrical regular sampled PWM with a triangular carrier, there is further justification in using an odd frequency ratio to ensure that the lowest significant sideband harmonic will be at least three times the fundamental frequency (recall from Sections 3.4.2 and 3.6.2 that these modulation strategies eliminate odd sideband components around odd carrier multiples, so that for m = 1, Ie nlo must always be odd for an integer carrier frequency ratio and the lowest possible sideband component is therefore a third harmonic of the fundamental). Finally, it should be noted that the selection of an appropriate carrier frequency ratio can become a little more complex for the more sophisticated PWM strategies discussed in the next three chapters of this text. Essentially, the choice depends on how rapidly the sideband harmonics from the first carrier group roll off in magnitude away from the carrier, and this must be considered individually for the various PWM strategies presented. The issue will be revisited as appropriate in later chapters.

- - - - - - - -

.~
ro 10-2

- - - - - -

,- - - - - - -,
1 1 1

- - - - - 1- - - - - - - -

r L j. - r - -

- - -

,- - - - - 1 1

= -= = == - - = - = - = - 1- - - --- - - - -

- - - - - - - - 1- - - - - -

- - - - -1 - - - - -I
-I

- - - - - - - - - - - - 1 - - - - - - -1 - - - 1 - - - - - - -1

- - -

- - - - - - - - - -

= -

= -

J
-j

- 1 - J J 1

= = - = - = - - -

- - - - - - -

'. '. 1

= = 1- = - - - - - - - 1- - - 1- - - - - - 11

= = - = - = - - - - - - - - - - - - - - - -

= -

- - - - - 3

,- - - - - - -,
1 1

- - 1- - - 1

r 1

- -

- - - - - - - - 1 1 - - -- - - -- -- - - 1 - - - - - - -1
1 1

- - - 1- - - - 1- - - - - - - - - - 1- - - -

- - -

_I -I -I

- - J - - - - -j - - - - - - -j - - - - - - - - - - - - - - - - - - J - - - -

- +- j. L

11-

,- - - - - ,- - - - - - 1

~
1 1

'. '.
1 1

- -

- -

- - -

1 1 1 1

- - - - - -

- - -

- - - - 11

Tf

- -

-,
1

- 1

-1

- -

r -

- - -

- -

Harmonic components for one inverter phase leg ,:ith triangular carrier and direct calculation of pulse penod, M = 0.9, 1/10 = 21.

harmonic group), and hence strategies such as programmed PWM discussed in Chapter 9 are often preferable. As a result of these observations the direct modulation approach will not be considered further in this text.

3.8

Integer versus Non-Integer Frequency Ratios

It should be noted at this point that nothing in the detailed mathematical development presented in this chapter constrains the carrier ratio to integer values, and even irrational values of switching frequency are also permissible. Irre-

3.9

Review of PWM Variations

Table 3.1 summarizes the magnitudes of the significant harmonics for the three approaches to the determination of switched pulse width presented in this chapter, normalized with respect to the naturally sampled PWM fundamental component. From this table it can be seen that there is negligible difference between the fundamental components for the carrier ratio of 21 that was used, and it can therefore be argued that there is virtually no fundamental component differentiation between the techniques at the carrier ratios typically used in modem inverter systems. In terms of the secondary objective of harmonic minimization, it can be seen that the harmonic performance of direct modulation is almost identical to regular sampled modulation. This is only to be expected, since regular sampled
Table 3.1 Harmonic Components for Single-Phase Leg PWM Strategies, fe/fa = 21, All Harmonics Normalized with respect to Vde (%) Naturally Sampled PWM (%) (Sawtooth Carrier) 90.0 0.0 0.0 0.0 0.0 2.1 7.0 17.7 30.5 25.5 51.2 25.5 30.5 17.7 7.0 2.1 3.88 Regularly Sampled PWM(%) (Sawtooth Carrier) 89.8 6.0 0.61 0.07 0.01 0.8 4.3 15.0 31.9 27.9 51.2 21.5 28.3 19.4 9.8 4.1 4.91 Naturally Sampled PWM(%) (Triangular Carrier) 90.0 0.0 0.0 0.0 0.0 0.0 1.2 0 26.8 0.0 71.2 0.0 26.8 0.0 1.2 0.0 3.86 SymmetricalRegular Sampled PWM(%) (Triangular Carrier) 89.7 0.5 0.1 0.0 0.0 0.0 0.6 1.1 24.8 5.3 71.2 5.0 28.1 1.8 1.9 0.1 3.86 Asymmetrical Regular Sampled PWM(%) (Triangular Carrier) 89.9 0.0 0.2 0.0 0.0 0.0 0.7 0.0 25.1 0.0 71.2 0.0 28.4 0.0 1.9 0.0 3.84 M = 0.9,

PWM simply linearizes the sinusoidal reference about its average for each carrier or half-carrier interval. For carrier ratios above about 15, little difference would be expected between the linearized volt-second average and the exact volt-second average calculated by direct modulation. As a consequence, it would be anticipated that regular sampled PWM and direct PWM would have the same harmonic consequences, since they essentially produce the same switched pulse widths, and these pulses are placed in the center of each halfcarrier interval. In terms of regular sampled PWM compared to naturally sampled PWM, there are a number of specific differences that can be seen from the results presented in Table 3.1. First, as predicted by the analytical results of Section 3.6, regular sampling creates low-order baseband harmonics just above the fundamental component. These low-order baseband harmonic multiples are usually quite small because of the rapid reduction in the Bessel function in the first term of Eqs. (3.78) and (3.98) and so are often overlooked by researchers, especially when carrier ratios are greater than about 15. However, they are an intrinsic consequence of regular sampled modulation, and their rate of attenuation varies with different modulation strategies and carrier ratios. For example, single ended modulation has a much slower decay rate for these tenns, so that for this implementation they can make a nontrivial contribution to the WTHD even for the typical pulse ratios of at least 30 which are used in modern inverters. Second, comparing natural and regular sampled PWM, it is found that the regular sampled modulation process attenuates the lower side sideband harmonics and increases the high-side sideband harmonics around the carrier frequency. This can also be seen by careful comparison of Figures 3.12 and 3.20, and is an intrinsic property of regular sampled PWM [10]. This attenuation preferentially reduces the lower order harmonic components of the switched waveform and can significantly improve the WTHD for modulation implementations with a low pulse ratio. But as a consequence, claims for an improved "new" PWM algorithm compared to naturally sampled PWM may simply reflect these effects of regular sampling rather than any intrinsic advantage of the new scheme, and this may not be appreciated by the proponents. Third, the major difference between symmetrical and asymmetrical PWM is that symmetrical regular sampled PWM causes additional sideband harmonic components to be present in the output voltage spectra, while asymmet-

Harmonic Number

Direct Modulation Asymmetrical Calculation (%) 89.9 0.0 0.2 0.0 0.0 0.0 0.7 0.0 25.0 0.0 71.3 0.0 28.3 0.0 2.0 0.0 3.85

1 2 3 4 5 16 17 18 19 20 21 22 23 24 25 26 WTHDO

rical regular sampled PWM and naturally sampled PWM do not create these components. This is an intrinsic limitation of symmetrical sampled modulation compared to asymmetrical sampled modulation [10].

[1]

S.R. Bowes a~d R. Bullough, "PWM switching strategies for current-fed inverter dnves, lEE Proceedings (London), vo!' 131 Pt. B, no. 5., Sept. 1984 pp. 195-202. ' M.A. Boost and P.D. Ziogas, "State-of-the-art carrier PWM techniques: a critical evaluation," IEEE Trans. on Industry Applications, vo!' 24, no. 2, March 1988, pp. 271-290. H.W. Van der Broeck and H.C. Skudelny, "Analytical analysis of the hannonic effects of a PWM ac drive," IEEE Trans. on Power Electronics, vol. 3, no. 2, March/April, 1988, pp. 216-223. S. Bowes and B.M. Bird, "Novel approach to the analysis and synthesis of modulation processes in power converters," lEE Proceedings (London), vol. 122, no. 5, May 1975, pp. 507-513. W.R. Bennett, "New results in the calculation of modulation products," Bell System Technical Journal, vol. 12, Apri11933, pp. 228-243. H.S. Black, Modulation Theory, Van Nostrand, New York, 1953. E.O. Brigham, The Fast Fourier Transform, Prentice-Hall, Englewood Cliffs NJ,1974. ' T.H. Barton, "Pulse-width modulation wavefonn - the bessel approximation," Con! Rec. IEEE Industry Applications Society Annual Mtg., Toronto, 1978, pp. 1125-1130.
III

[2]

This chapter has presented the basic concepts of determining switched pulse width by modulation, and has determined from these concepts that: Low-order harmonic multiples of the fundamental are produced by the regular sampling process. The lower side sideband harmonic components are attenuated slightly, and the high-side sideband harmonic components are increased slightly for regular sampled PWM. Symmetrical regular sampled PWM leads to additional (farrier sideband ha~onics in the phase leg output voltage and is therefore inferior to asymmetrical PWM. Direct modulation produces an almost identical harmonic response to regular sampled PWM. For full inverter systems with multiple phase legs operating together, both the absence of harmonics by virtue of the modulation process and the elimination of harmonics by cancellation between the phase legs of an inverter, playa significant part in determining the harmonics seen in the I-I output voltages of the inverter. Hence it is important to have a precise realization of the phase leg switched output when investigating the harmonic performance of various inverter topologies and modulation implementations. The theoretical understanding presented in this chapter, and the exact match between theoretical analysis and numerical simulation that has been achieved (not elaborated on here, see [11]), provide the basis to evaluate the performance of the more complex switched inverter structures investigated in the following chapters of this book.

[3]

[4]

[5] [6] [7] [8]

The

[9]

Y. Kim and M. Ehsani, "An algebraic algorithm for microcomputer-based (direct) inverter pulse width modulation," IEEE Trans. on Industry Applications, vo!' 23, no. 4, July/Aug. 1987, pp. 654-660. J.T. Boys and P.G. Handley, "Hannonic analysis of space vector modulated PWM wavefonns," lEE Proceedings (London), Pt. B, vo!. 137, no. 4, July 1990, pp. 197-204. D.G. Holmes, "A generalised approach to the modulation and control of hard switched converters," Ph.D. Thesis, Monash University, Australia, 1997.

[10]

[11]

Você também pode gostar