C H A P T E R 3
TransformRepresentationofSignals
andLTISystems
As you have seen in your prior studies of signals and systems, and as emphasized
in the review in Chapter 2, transforms play a central role in characterizing and
representing signals and LTI systems in both continuous and discrete time. In
thischapterwediscusssomespecicaspectsoftransformrepresentationsthatwill
play an important role in later chapters. These aspects include the interpreta-
tion of Fourier transform phase through the concept of group delay, and methods
referred to as spectral factorization for obtaining a Fourier representation
(magnitude and phase) when only the Fourier transform magnitude is known.
3.1 FOURIER TRANSFORM MAGNITUDE AND PHASE
The Fourier transform of a signal or the frequency response of an LTI system isin
generalacomplex-valuedfunction. Amagnitude-phaserepresentationofaFourier
transform X(j) takes the form
X(j) =|X(j)|e
jX(j)
. (3.1)
In eq. (3.1), X(j) denotes the (non-negative) magnitude and X(j) denotes | |
the(real-valued)phase. Forexample, ifX(j)isthesincfunction, sin()/,then
|X(j)|istheabsolutevalueofthisfunction,whileX(j)is0infrequencyranges
wherethesincispositive,and infrequencyrangeswherethesincisnegative. An
alternative representation is an amplitude-phase representation
A()e
jAX(j)
(3.2)
in which A() = |X(j)| is real but can be positive for some frequencies and
negativeforothers. Correspondingly,
A
X(j) =X(j)whenA() = + X(j) ,
and
A
X(j) =X(j) whenA() =|X(j)|.
| |
Thisrepresentationisoften
preferred when its use can eliminate discontinuities of radians in the phase as
A()changessign. Inthecaseofthesincfunctionabove,forinstance,wecanpick
A()=sin()/ and
A
=0. Itisgenerallyconvenientinthefollowingdiscussion
for us to assume that the transform under discussion has no zeros on the j-axis,
so that we can take A() =|X(j)| for all (or, if we wish, A() =|X(j)| for
all ). A similar discussion applies also, of course, in discrete-time.
In either a magnitude-phase representation or an amplitude-phase representation,
thephaseisambiguous,asanyintegermultipleof2canbeaddedatanyfrequency
c 47 Alan V. Oppenheim and George C. Verghese, 2010
48 Chapter 3 Transform Representation of Signals and LTI Systems
without changing X(j) in (3.1) or (3.2). A typical phase computation resolves
thisambiguitybygeneratingthephasemodulo2,i.e.,asthephasepassesthrough
+ it wraps around to (or from wraps around to +). In Section 3.2
we will nd it convenient to resolve this ambiguity by choosing the phase to be
a continuous function of frequency. This is referred to as the unwrapped phase,
since the discontinuities at are unwrapped to obtain a continuous phase curve.
TheunwrappedphaseisobtainedfromX(j)byaddingstepsofheightequalto
or 2 wherever needed, in order to produce a continuous function of . The
steps of height are added at points where X(j) passes through 0, to absorb
signchangesasneeded;thestepsofheight2 areaddedwhereverelseisneeded,
invoking the fact that such steps make no dierence to X(j), as is evident from
(3.1). WeshallproceedasthoughX(j)isindeedcontinuous(anddierentiable)
at the points of interest, understanding that continuity can indeed be obtained in
all cases of interest to us by adding in the appropriate steps of height or 2.
Typically,ourintuitionforthetime-domaineectsoffrequencyresponsemagnitude
or amplitude on a signal is rather well-developed. For example, if the Fourier
transformmagnitudeissignicantlyattenuatedathighfrequencies,thenweexpect
the signal to vary slowly and without sharp discontinuities. On the other hand, a
signal in which the low frequencies are attenuated will tend to vary rapidly and
without slowly varying trends.
Incontrast,visualizingtheeectonasignalofthephaseofthefrequencyresponse
ofasystemismoresubtle,butequallyimportant. Webeginthediscussionbyrst
consideringseveralspecicexampleswhicharehelpfulinthenconsideringthemore
general case. Throughout this discussion we will consider the system to be an all-
passsystemwithunitygain,i.e. theamplitudeofthefrequencyresponseA(j) = 1
(continuoustime)orA(e
j
)=1(discretetime)sothatwecanfocusentirelyonthe
eect of the phase. The unwrapped phase associated with the frequency response
will be denoted as
A
H(j) (continuous time) and
A
H(e
j
) (discrete time).
EXAMPLE 3.1 Linear Phase
Consider an all-pass system with frequency response
H(j) =e
j
(3.3)
i.e. in an amplitude/phase representation A(j) = 1 and
A
H(j) = . The
unwrapped phase for this example is linear with respect to , with slope of .
For input x(t) with Fourier transform X(j), the Fourier transform of the output
isY(j) =X(j)e
j
andcorrespondinglytheoutputy(t)isx(t). Inwords,
linearphasewithaslopeofcorrespondstoatimedelayof(oratimeadvance
if is negative).
For a discrete time system with
H(e
j
) =e
j
||< (3.4)
the phase is again linear with slope . When is an integer, the time domain
interpretationoftheeectonaninputsequencex[n]isagainstraightforwardandis
Alan V. Oppenheim and George C. Verghese, 2010 c
Section 3.1 Fourier Transform Magnitude and Phase 49
asimpledelay(positive)oradvance(negative)of . Whenisnotaninteger, | |
the eect is still commonly referred to as a delay of , but the interpretation is
more subtle. If we think of x[n] as being the result of sampling a band-limited,
continuous-time signal x(t) with sampling period T, the output y[n] will be the
result of sampling the signal y(t) =x(tT) with sampling period T. In fact we
saw this result in Example 2.4 of chapter 2 for the specic case of a half-sample
delay, i.e. =
2
1
.
EXAMPLE 3.2 Constant Phase Shift
As a second example, we again consider an all-pass system with A(j) = 1 and
unwrapped phase
for >0
_
A
H(j) =
+
0
for <0
as indicated in Figure 3.1
+
0
-
0
FIGURE 3.1 Phase plot of all-pass system with constant phase shift,
0
.
Note that the phase is required to be an odd function of if we assume that the
system impulse response is real valued. In this example, we consider x(t) to be of
the form
x(t) =s(t) cos(
0
t+) (3.5)
i.e. an amplitude-modulated signal at a carrier frequency of
0
. Consequently,
X(j) can be expressed as
X(j) =
1
S(jj
0
)e
j
+
1
S(j+j
0
)e
j
(3.6)
2 2
where S(j) denotes the Fourier transform of s(t).
For this example, we also assume that S(j) is bandlimited to < , with | |
suciently small so that the term S(jj
0
)e
j
is zero for < 0 and the term
S(j+j
0
)e
j
iszerofor >0,i.e. that(
0
)>0. Theassociatedspectrum
of x(t) is depicted in Figure 3.2.
Alan V. Oppenheim and George C. Verghese, 2010 c
50 Chapter 3 Transform Representation of Signals and LTI Systems
X(j )
0
-
0
0
0
S(j +j )e
-j
S(j -j
0
)e
+j
- +
0 0
FIGURE 3.2 Spectrum of x(t) with s(t) narrowband
With these assumptions on x(t), it is relatively straightforward to determine the
output y(t). Specically, the system frequency response H(j) is
e
j0
_
>0
H(j) =
+j0
(3.7)
e <0
Since the term S(jj
0
)e
j
in eq. (3.6) is non-zero only for >0, it is simply
multiplied by e
j
, and similarly the term S(j+j
0
)e
j
is multiplied only by
e
+j
. Consequently, the output frequency response, Y(j), is given by
Y(j) = X(j)H(j)
=
1
S(jj
0
)e
+j
e
j0
+
1
S(j+j
0
)e
j
e
+j0
(3.8)
2 2
which we recognize as a simple phase shift by
0
of the carrier in eq. (3.5), i.e.
replacing in eq. (3.6) by
0
. Consequently,
y(t) =s(t) cos(
0
t+
0
) (3.9)
This change in phase of the carrier can also be expressed in terms of a time delay
for the carrier by rewriting eq. (3.9) as
_ _
0
_ _
y(t) =s(t) cos
0
t
0
+ (3.10)
3.2 GROUP DELAY AND THE EFFECT OF NONLINEAR PHASE
In Example 3.1, we saw that a phase characteristic that is linear with frequency
corresponds in the time domain to a time shift. In this section we consider the
c Alan V. Oppenheim and George C. Verghese, 2010
Section 3.2 Group Delay and The Eect of Nonlinear Phase 51
eect of a nonlinear phase characteristic. We again assume the system is an all-
pass system with frequency response
H(j) =A(j)e
jA[H(j)]
(3.11)
withA(j)=1. Ageneralnonlinearunwrappedphasecharacteristicisdepictedin
Figure 3.3
A
+
1
-
1
-
0
+
0
FIGURE 3.3 Nonlinear Unwrapped Phase Characteristic
AswedidinExample3.2,weagainassumethatx(t)isnarrowbandoftheformof
equation(3.5)andasdepictedinFigure3.2. WenextassumethatinFigure3.2
is suciently small so that in the vicinity of
0
,
A
H(j) can be approximated
suciently well by the zeroth and rst order terms of a Taylors series expansion,
i.e.
_
d
_
A
H(j)
A
H(j
0
) + (
0
)
A
H(j) (3.12)
d
=0
Dening
g
() as
d
g
() =
A
H(j) (3.13)
d
our approximation to
A
H(j) in a small region around =
0
is expressed as
A
H(j)
A
H(j
0
)(
0
)
g
(
0
) (3.14)
Similarly in a small region around =
0
, we make the approximation
A
H(j)
A
H(j
0
)(+
0
)
g
(
0
) (3.15)
As we will see shortly, the quantity
g
() plays a key role in our interpretation of
the eect on a signal of a nonlinear phase characteristic.
With the Taylors series approximation of eqs. (3.14) and (3.15) and for input
signalswithfrequencycontentforwhichtheapproximationisvalid,wecanreplace
Figure 3.3 with Figure 3.4.
Alan V. Oppenheim and George C. Verghese, 2010 c
52 Chapter 3 Transform Representation of Signals and LTI Systems
0
slope= -
g
(
0
)
+
1
+
0
+
-
0
-
0
-
1
slope= -
g
(
0
)
FIGURE3.4 Taylorsseriesapproximationofnonlinearphaseinthevicinityof
0
where
1
=
A
H(j
0
)
and
0
=
A
H(j
0
) +
0
g
(
0
)
SinceforLTIsystemsincascade,thefrequencyresponsesmultiplyandcorrespond-
ingly the phases add, we can represent the all-pass frequency response H(j) as
the cascade of two all-pass systems, H
I
(j) and H
II
(j), with unwrapped phase
as depicted in Figure 3.5.
A
H
I
(j )
H
I
(j ) H (j )
II
x
I
(t) x(t) x
II
(t)
+
0
-
0
slope = -
g
(
0
)
A
H
II
(j )
FIGURE 3.5 An all-pass system frequency response, H(j), represented as the cas-
cade of two all-pass systems, H
I
(j) and H
II
(j).
Alan V. Oppenheim and George C. Verghese, 2010 c
Section 3.2 Group Delay and The Eect of Nonlinear Phase 53
We recognize H
I
(j) as corresponding to Example 3.2. Consequently, with x(t)
narrowband, we have
x(t) = s(t) cos(
0
t+)
_ _
0
_ _
x
I
(t) = s(t) cos
0
t
0
+ (3.16)
Next we recognize H
II
(j) as corresponding to Example 3.1 with =
g
(
0
).
Consequently,
x
II
(t) =x
I
(t
g
(
0
)) (3.17)
or equivalently
_ _
0
+
0
g
(
0
)
_ _
x
II
(t) =s(t
g
(
0
)) cos
0
t
0
+ (3.18)
Since, from Figure 3.4, we see that
1
=
0
+
0
g
(
0
)
equation (3.18) can be rewritten as
_ _
1
_ _
x
II
(t) =s(t
g
(
0
)) cos
0
t
0
+ (3.19a)
or
x
II
(t) =s(t
g
(
0
)) cos [
0
(t
p
(
0
))+] (3.19b)
where
p
, referred to as the phase delay, is dened as
p
=
1
0
.
In summary, according to eqs. (3.18) and (3.19a), the time-domain eect of the
nonlinearphaseforthenarrowbandgroupoffrequenciesaroundthefrequency
0
is
todelaythenarrowbandsignalbythegroupdelay,
g
(
0
),andapplyanadditional
phase shift of
1
0
to the carrier. An equivalent, alternate interpretation is that the
time-domainenvelopeofthefrequencygroupisdelayedbythegroupdelayandthe
carrier is delayed by the phase delay.
Thediscussionhasbeencarriedoutthusfarfornarrowbandsignals. Toextendthe
discussiontobroadbandsignals,weneedonlyrecognizethatanybroadbandsignal
can be viewed as a superposition of narrowband signals. This representation can
in fact be developed formally by recognizing that the system in Figure 3.6 is an
identity system, i.e. r(t) =x(t) as long as
H
i
(j)=1 (3.20)
i=0
By choosing the lters H
i
(j) to satisfy eq. (3.20) and to be narrowband around
center frequencies
i
, each of the output signals, y
i
(t), is a narrowband signal.
Consequently the time-domain eect of the phase of G(j) is to apply the group
Alan V. Oppenheim and George C. Verghese, 2010 c
54 Chapter 3 Transform Representation of Signals and LTI Systems
G(j )
x(t) r(t)
x(t)
r(t)
H
0
(j ) G(j )
H
i
(j ) G(j )
r
i
(t)
r
0
(t)
g
i
(t)
g
0
(t)
FIGURE 3.6 Continuous-time all-pass system with frequency response amplitude,
phase and group delay as shown in Figure 3.7
FIGURE 3.7 Magnitude, (nonlinear) phase, and group delay of an all-pass lter.
delayandphasedelaytoeachofthenarrowbandcomponents(i.e. frequencygroups)
y
i
(t). Ifthegroupdelayisdierentatthedierentcenter(i.e. carrier)frequencies
Alan V. Oppenheim and George C. Verghese, 2010 c
Section 3.2 Group Delay and The Eect of Nonlinear Phase 55
FIGURE 3.8 Impulse response for all-pass lter shown in Figure 3.7
i
, then the time domain eect is for dierent frequency groups to arrive at the
output at dierent times.
As an illustration of this eect, consider G(j) in Figure 3.6 to be the continuous
time all-pass system with frequency responseamplitude, phase and group delay as
shown in Figure 3.7. The corresponding impulse response is shown in Figure 3.8.
IfthephaseofG(j)werelinearwithfrequency,theimpulseresponsewouldsimply
beadelayedimpulse,i.e. allthenarrowbandcomponentswouldbedelayedbythe
sameamountandcorrespondinglywouldadduptoadelayedimpulse. However,as
weseeinFigure3.7,thegroupdelayisnotconstantsincethephaseisnonlinear. In
particular, frequencies around 1200 Hz are delayed signicantly more than around
other frequencies. Correspondingly, in Figure 3.8 we see that frequency group
appearing late in the impulse response.
A second example is shown in Figure 3.9, in which G(j) is again an all-pass
systemwithnonlinearphaseandconsequentlynon-constantgroupdelay. Withthis
example, we would expect to see dierent delays in the frequency groups around
= 2 50, = 2 100, and = 2 300 with the group at = 2 50 having
the maximum delay and therefore appearing last in the impulse response.
Inbothoftheseexamples,theinputishighlyconcentratedintime(i.e. animpulse)
and the response is dispersed in time because of the non-constant group delay, i.e.
Alan V. Oppenheim and George C. Verghese, 2010 c
56 Chapter 3 Transform Representation of Signals and LTI Systems
FIGURE 3.9 Phase, group delay, and impulse response for an all-pass system: (a)
principalphase;(b)unwrappedphase;(c)groupdelay;(d)impulseresponse. (From
Oppenheim and Willsky, Signals and Systems, Prentice Hall, 1997, Figure 6.5.)
Alan V. Oppenheim and George C. Verghese, 2010 c
4
2
0
-2
-4
0 50 100 150 200 250 300 350 400
Frequency (Hz)
P
h
a
s
e
(
r
a
d
)
0 50 100 150 200 250 300 350 400
0
-5
-10
-15
-20
Frequency (Hz)
P
h
a
s
e
(
r
a
d
)
600
400
200
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-200
-400
-600
Time (sec)
0 50 100 150 200 250 300 350 400
0.10
0.08
0.04
0.06
0.02
0
Frequency (Hz)
G
r
o
u
p
d
e
l
a
y
(
s
e
c
)
(a)
(b)
(c)
(d)
Image by MIT OpenCourseWare, adapted from Signals and Systems, Alan
Oppenheim and Alan Willsky. Prentice Hall, 1996.
Section 3.3 All-Pass and Minimum-Phase Systems 57
thenonlinearphase. Ingeneral,theeectofnonlinearphaseisreferredtoasdisper-
sion. In communication systems and many other application contexts even when
a channel has a relatively constant frequency response magnitude characteristic,
nonlinearphasecanresultinsignicantdistortionandothernegativeconsequences
because of the resulting time dispersion. For this reason, it is often essential to
incorporate phase equalization to compensate for non-constant group-delay.
As a third example, we consider an all-pass system with phase and group delay as
shown in Figure 3.10
1
. The input for this example is the touch-tone digit ve
which consists of two very narrowband tones at center frequencies 770 and 1336
Hz. Thetime-domainsignalanditstwonarrowbandcomponentsignalsareshown
in Figure 3.11.
FIGURE3.10 Phaseandgroupdelayforall-passlterfortouch-tonesignalexample.
Thetouch-tonesignalisprocessedwithmultiplepassesthroughtheall-passsystem
ofFigure3.10. Fromthegroupdelayplot,weexpectthat,inasinglepassthrough
the all-pass lter, the tone at 1336 Hz would be delayed by about 2.5 milliseconds
relativetothetoneat770Hz. After200passes,thiswouldaccumulatetoarelative
delay of about 0.5 seconds.
In Figure 3.12, we show the result of multiple passes through lters and the accu-
mulation of the delays.
3.3 ALL-PASS AND MINIMUM-PHASE SYSTEMS
Two particularly interesting classes of stable LTI systems are all-pass systems and
minimum-phase systems. We dene and discuss them in this section.
1
This example was developed by Prof. Bernard Lesieutre of the University of Wisconsin,
Madison, when he taught the course with us at MIT
c Alan V. Oppenheim and George C. Verghese, 2010
58 Chapter 3 Transform Representation of Signals and LTI Systems
FIGURE 3.11 Touch-tone signal with its two narrowband component signals.
3.3.1 All-Pass Systems
An all-pass system is a stable system for which the magnitude of the frequency
response is a constant, independent of frequency. The frequency response in the
case of a continuous-time all-pass system is thus of the form
H
ap
(j) =Ae
jHap(j)
, (3.21)
where Ais aconstant, not varying with. Assuming the associated transfer func-
tion H(s) is rational in s, it will correspondingly have the form
M
s+a
k
H
ap
(s) =A . (3.22)
sa
k
k=1
Note that for each pole at s= +a
k
this has a zero at the mirror image across the
imaginary axis, namely at s ; and if a
k
is complex and the system impulse = a
responseisreal-valued,everycomplexpoleandzerowilloccurinaconjugatepair,
k
and a zero at s = a
k
. An example of a
pole-zero diagram (in the s-plane) for a continuous-time all-pass system is shown
so there will also be a pole at s +a =
k
inFigure(3.13). ItisstraightforwardtoverifythateachoftheM factorsin(3.22)
has unit magnitude for s=j.
c Alan V. Oppenheim and George C. Verghese, 2010
Section 3.3 All-Pass and Minimum-Phase Systems 59
200
passes
200
passes
200
passes
200
passes
200
passes
FIGURE3.12 Eectofpassingtouchtonesignal(Figure3.11)throughmultiplepasses
of an all-pass lter and the accumulation of delays
.
For a discrete-time all-pass system, the frequency response is of the form
H
ap
(e
j
) =Ae
jHap(e
j
)
. (3.23)
If the associated transfer function H(z) is rational in z, it will have the form
M
H
ap
(z) =A
z
1
b
k
. (3.24)
1b
k
z
1
k=1
The poles and zeros in this case occur at conjugate reciprocal locations: for each
poleatz=b
k
thereisazeroatz= 1/b
k
. Azeroatz=0(andassociatedpoleat)
isobtainedbysettingb
k
=inthecorrespondingfactorabove,afterrstdividing
boththenumeratoranddenominatorbyb
k
;thisresultsinthecorrespondingfactor
in (3.24) being just z. Again, if the impulse response is real-valued then every
complex pole and zeros will occur in a conjugate pair, so there will be a pole at
z = b
k
and a zero at z = 1/b
k
. An example of a pole-zero diagram (in the z
plane)foradiscrete-timeall-passsystemisshowninFigure(3.14). Itisoncemore
Alan V. Oppenheim and George C. Verghese, 2010 c
60 Chapter 3 Transform Representation of Signals and LTI Systems
Im
1
1
2
Re
2 1
1
FIGURE 3.13 Typical pole-zero plot for a continuous-time all-pass system.
straightforward to verify that each of the M factors in (3.24) has unit magnitude
for z=e
j
.
The phase of a continuous-time all-pass system will be the sum of the phases as-
sociated with each of the M factors in (3.22). Assuming the system is causal (in
addition to being stable), then for each of these factors Re{a
k
} < 0. With some
s+a
algebra it can be shown that each factor of the form
k
now has positive group
sa
k
delayatallfrequencies,apropertythatwewillmakereferencetoshortly. Similarly,
assuming causality (in addition to stability) for the discrete-time all-pass system
z
1
b
in (3.24), each factor of the form
k
with b
k
<1 contributes positive group
1b
k
z
1
| |
delay at all frequencies (or zero group delay in the special case of b
k
= 0). Thus,
in both continuous- and discrete-time, the frequency response of a causal all-pass
system has constant magnitude and positive group delay at all frequencies.
3.3.2 Minimum-Phase Systems
Indiscrete-time,astablesystemwitharationaltransferfunctioniscalledminimum-
phase if its poles and zeros are all inside the unit circle, i.e., have magnitude less
than unity. This is equivalent in the DT case to the statement that the system is
stable and causal, and has a stable and causal inverse.
A similar denition applies in the case of a stable continuous-time system with a
rational transfer function. Such a system is called minimum-phase if its poles and
c Alan V. Oppenheim and George C. Verghese, 2010
Section 3.3 All-Pass and Minimum-Phase Systems 61
0.8
Unit circle
3/4
4/3
Im
Re
FIGURE 3.14 Typical pole-zero plot for a discrete-time all-pass system.
nite zeros are in the left-half-plane, i.e., have real parts that are negative. The
systemisthereforenecessarilycausal. Ifthereareasmanynitezerosasthereare
poles, then a CT minimum-phase system can equivalently be characterized by the
statement that both the system and its inverse are stable and causal, just as we
had in the DT case. However, it is quite possible and indeed common for
a CT minimum-phase system to have fewer nite zeros than poles. (Note that a
stable CT system must have all its poles at nite locations in the s-plane, since
poles at innity would imply that the output of the system involves derivatives of
theinput,whichisincompatiblewithstability. Also,whereasintheDTcaseazero
at innity is clearly outside the unit circle, in the CT case there is no way to tell
if a zero at innity is in the left half plane or not, so it should be no surprise that
the CT denition involves only the nite zeros.)
Theuseofthetermminimumphaseishistorical,andthepropertyshouldperhaps
moreappropriatelybetermedminimumgroupdelay,forreasonsthatwewillbring
outnext. Todothis,weneedafactthatweshallshortlyestablish: thatanycausal
andstableCTsystemwitharationaltransferfunctionH
cs
(s)andnozerosonthe
imaginary axis can be represented as the cascade of a minimum-phase system and
an all-pass system,
H
cs
(s) =H
min
(s)H
ap
(s). (3.25)
Similarly, in the DT case, provided the transfer function H
cs
(z) has no zeros on
Alan V. Oppenheim and George C. Verghese, 2010 c
62 Chapter 3 Transform Representation of Signals and LTI Systems
the unit circle, it can be written as
H
cs
(z) =H
min
(z)H
ap
(z). (3.26)
The frequency response magnitude of the all-pass factor is constant, independent
offrequency,andforconvenienceletussetthisconstanttounity. Thenfrom(3.25)
|H
cs
(j)|=|H
min
(j)|, and (3.27a)
grpdelay[H
cs
(j)]=grpdelay[H
min
(j)]+grpdelay[H
ap
(j)] (3.27b)
and similar equations hold in the DT case.
We will see in the next section that the minimum-phase term in (3.25) or (3.26)
canbeuniquelydeterminedfromthemagnitudeofH
cs
(j),respectivelyH
cs
(e
j
).
Consequently all causal, stable systems with the same frequency response magni-
tudedieronlyinthechoiceoftheall-passfactorin(3.25)or(3.26). However,we
have shown previously that all-pass factors must contribute positive group delay.
Thereforeweconcludefrom(3.27b)thatamongallcausal,stablesystemswiththe
same CT frequency response magnitude, the one with no all-pass factors in (3.25)
will have the minimum group delay. The same result holds in the DT case.
Weshallnowdemonstratethevalidityof(3.25); thecorrespondingresultin(3.26)
fordiscretetimefollowsinaverysimilarmanner. Consideracausal,stabletransfer
function H
cs
(s) expressed in the form
M1
(sl
k
)
M2
(sr
i
)
H
cs
(s) =A
k=1 i=1
(3.28)
N
)
n=1
(sd
n
where the d
n
s are the poles of the system, the l
k
s are the zeros in the left-half
plane and the r
i
s are the zeros in the right-half plane. Since H
cs
(s) is stable and
causal, all of the poles are in the left-half plane and would be associated with the
factor H
min
(s) in (3.25), as would be all of the zeros l
k
. We next represent the
right-half-plane zeros as
M2 M2 M2
(sr
i
)
(sr
i
) = (s+r
i
)
(s+r
i
)
(3.29)
i=1 i=1 i=1
Since Re{r
i
} is positive, the rst factor in (3.29) represents left-half-plane zeros.
Thesecondfactorcorrespondstoall-passtermswithleft-half-planepoles,andwith
zeros at mirror image locations to the poles. Thus, combining (3.28) and (3.29),
H
cs
(s) has been decomposed according to (3.25) where
M1
(sl
k
)
M2
(s+r
i
)
H
min
(s) =A
k=1 i=1
(3.30a)
N
(sd
n
)
n=1
M2
H
ap
(s) =
(sr
i
)
(3.30b)
(s+r
i
)
i=1
Alan V. Oppenheim and George C. Verghese, 2010 c
Section 3.4 Spectral Factorization 63
EXAMPLE 3.3 Causal, stable system as cascade of minimum-phase and all-pass
Consider a causal, stable system with transfer function
H
cs
=
(s1)
(3.31)
(s+2)(s+3)
The corresponding minimum-phase and all-pass factors are
(s+1)
H
min
(s)= (3.32)
(s+2)(s+3)
H
ap
(s) =
s1
(3.33)
s+ 1
3.4 SPECTRAL FACTORIZATION
The minimum-phase/all-pass decomposition developed above is useful in a variety
ofcontexts. Onethatisofparticularinteresttousinlaterchaptersariseswhenwe
wearegivenorhavemeasuredthemagnitudeofthefrequencyresponseofastable
system with a rational transfer function H(s) (and real-valued impulse response),
and our objective is to recover H(s) from this information. A similar task may be
posed in the DT case, but we focus on the CT version here. We are thus given
|H(j)|
2
=H(j)H
(j) (3.34)
or, since H
(j) =H(j),
|H(j)|
2
=H(j)H(j). (3.35)
Now H(j) is H(s) for s=j, and therefore
H(j)
2
=H(s)H(s) (3.36) | |
s=j
For any numerator or denominator factor (sa) in H(s), there will be a corre-
sponding factor (sa) in H(s)H(s). Thus H(s)H(s) will consist of factors
in the numerator or denominator of the form (sa)(sa) =s
2
+a
2
, and will
therefore be a rational function of s
2
. Consequently H(j)
2
will be a rational | |
functionof
2
. Thus,ifwearegivenorcanexpress H(j)
2
asarationalfunction | |
2
of
2
,wecanobtaintheproductH(s)H(s)bymakingthesubstitution
2
=s .
TheproductH(s)H(s)willalwayshaveitszerosinpairsthataremirroredacross
theimaginaryaxisofthes-plane,andsimilarlyforitspoles. Foranypoleorzeroof
H(s)H(s)attherealvaluea,therewillbeanotheratthemirrorimagea,while
foranypoleorzeroatthecomplexvalueq,therewillbeothersatq
,q andq
,
c Alan V. Oppenheim and George C. Verghese, 2010
64 Chapter 3 Transform Representation of Signals and LTI Systems
forming a complex conjugate pair (q, q
) and its mirror image (q
,q). We then
need to assign one of each mirrored real pole and zero and one of each mirrored
conjugate pair of poles and zeros to H(s), and the mirror image to H(s).
If we assume (or know) that H(s) is causal, in addition to being stable, then
we would assign the left-half plane poles of each pair to H(s). With no further
knowledgeorassumptionwehavenoguidanceontheassignmentofthezerosother
than the requirement of assigning one of each mirror image pair to H(s) and the
other to H(s). If we further know or assume that the system is minimum-phase,
thentheleft-half-planezerosfromeachmirroredpairareassignedtoH(s),andthe
right-half-plane zeros to H(s). This process of factoring H(s)H(s) to obtain
H(s) is referred to as spectral factorization.
EXAMPLE 3.4 Spectral factorization
Considerafrequencyresponsemagnitudethathasbeenmeasuredorapproximated
as
2
+ 1
2
+ 1
|H(j)|
2
=
4
+13
2
+ 36
=
(
2
+4)(
2
+9)
(3.37)
Making the substitution
2
=s
2
, we obtain
s
2
+ 1
H(s)H(s) =
(s
2
+4)(s
2
+9)
(3.38)
which we further factor as
H(s)H(s) =
(s+1)(s+1)
(3.39)
(s+2)(s+2)(s+3)(s+3)
It now remains to associate appropriate factors with H(s) and H(s). Assuming
the system is causal in addition to being stable, the two left-half plane poles at
s=2 and s=3 must be associated with H(s). With no further assumptions,
eitheroneofthenumeratorfactorscanbeassociatedwithH(s)andtheotherwith
H(s). However, if we know or assume that H(s) is minimum phase, then we
would assign the left-half plane zero to H(s), resulting in the choice
(s+1)
H(s)= (3.40)
(s+2)(s+3)
Inthediscrete-timecase,asimilardevelopmentleadstoanexpressionforH(z)H(1/z)
fromknowledgeof|H(e
j
)|
2
. ThezerosofH(z)H(1/z)occurinconjugaterecipro-
cal pairs, and similarly for the poles. We again have to split such conjugate recip-
rocalpairs,assigningoneofeachtoH(z),theothertoH(1/z),basedonwhatever
additional knowledge we have. For instance, if H(z) is known to be causal in ad-
dition to being stable, then all the poles of H(z)H(1/z) that are in the unit circle
are assigned to H(z); and if H(z) is known to be minimum phase as well, then all
the zeros of H(z)H(1/z) that are in the unit circle are assigned to H(z).
Alan V. Oppenheim and George C. Verghese, 2010 c
MIT OpenCourseWare
[Link]
6.011 Introduction to Communication, Control, and Signal Processing
Spring 2010
For information about citing these materials or our Terms of Use, visit: [Link]