Chapter 10

THE z-TRANSFORM
 Chapter 10 THE z-TRANSFORM
Sec. 10.1 The z-Transform
Sec. 10.2 The Region
g of Convergence
g for The z-Transform
Sec. 10.3 The Inverse z-Transform
Sec 10.4
Sec. 10 4 Geometric Evaluation of the Fourier
Transform from The Pole-Zero Plot
Sec 10.5
Sec. 10 5 Properties of The z-Transform
z Transform
Sec. 10.6 Some Common z-Transform Pairs
Sec. 10.7 Analysis
l i andd Characterization
h i i off LTI Systems Usingi
z-Transforms
 Chapter 10 THE z-TRANSFORM
Sec. 10.8 Discrete All-Pass System, Minimum Phase
System,and
y , Spectral
p Factorization
Sec. 10.9 System Function Algebra and Block Diagram
Representations
Sec. 10.10 The Unilateral z-Transform
Sec 10.11
Sec. 10 11 Summary
 Sec. 10.1 THE z-TRANSFORM
z
for a discrete-time linear time-invariant system with impulse response h[n], the response
y[n] of the system to a complex exponential input of the form zn is

(10 1)
(10.1)
where
(10 2)
(10.2)

Key concepts
(i) definition of the z-transform;
((ii)) the discrete-time Fourier transform is a special
p case of the z-transform where z = e jω,
i.e., z is on the unit circle;
((iii)the
) region
g of convergence
g ((ROC)) of the z-transform;
(iv)zeros, poles, and the pole-zero plot of the z-transform;
(v) The z-transform is linear.
If x[n] = α x1[n] +β x2[n], then X(z) =αX1(z) + βX2(z)
P.831
Definition 10.1
10 1 z-Transform
z Transform

The z-transform of a general discrete-time signal x[n] is defined as

(10 3)
(10.3)
where z is a complex variable. For convenience, the z-transform of x[n] will
sometimes
so et es be denoted
de oted as Z{x[n]}
{x[n]} and
a d the
t e relationship
e at o s p between
betwee x[n] aandd its
ts z-transform
ta so
is indicated as
(
(10.4)
)

we considered a number of important

p relationships
p between the Laplace
p
transform and the Fourier transform for continuous-time signals.

P.832
 To explore
p these relationships,
p we express
p the complex
p variable z in the
polar form as

(10.5)
In terms of r and ω, eq. (10.3) becomes

or equivalently,

(10 6)
(10.6)

P.832
 we see that X(re jω) is the Fourier transform of the sequence x[n] multiplied by a real
exponential
p g r–n; that is,,
signal

(10.7)
for r = 1, or equivalently, | z | = 1, eq. (10.3) reduces to the Fourier transform; that is,

(10.8)

the z-transform reduces to the Fourier transform on the contour in the complex z-
plane
l corresponding
di to a circle
i l withi h a radius
di off unity,
i as indicated
i di d in
i Figure
i 10.1.
10 1
This circle in the z-plane is referred to as the unit circle.

P.832
P.833
 [[Example
p 10.1]]

Consider the signal x[n] = anu[n]. Then, from eq. (10.3),

For convergence of X(z),

X(z) we require that < ∞.
∞ Consequently,
Consequently the region of
convergence is the range of values of z for which | az–1 | < 1, or equivalently, | z | > | a |.
Then

P.833
 (10 9)
(10.9)

Thus, the z-transform

Thus z transform for this signal is well-defined
well defined for any value of a,
a with an ROC
determined by the magnitude of a according to eq. (10.9). For example, for a = 1,
x[n] is the unit step sequence with z-transform
z transform

We see that the zz-transform

transform in eq. (10.9) is a rational function. Consequently,just as
with rational Laplace transforms, the z-transform can be characterized by its zeros
((the roots of the numerator ppolynomial)
y ) and its ppoles ((the roots of the denominator
polynomial). For this example, there is one zero, at z = 0, and one pole, at z = a. The
ppole-zero pplot and the region
g of convergence
g for Example p 10.1 are shown in Figure
g
10.2 for a value of a between 0 and 1. For | a | > 1, the ROC does not include the
unit circle, consistent with the fact that, for these values of a, the Fourier transform
of anu[n] does not converge.
P.834
P.834
 [[Example
p 10.2]]

[ ] = –anu[–n
N llett x[n]
Now [ – 1].
1] Th
Then

(10 10)
(10.10)

If | a–1z | < 1, or equivalently, | z | < | a |, the sum in eq. (10.10) converges and

(10.11)

The pole-zero plot and the region of convergence for this example are shown in
g
Fig-ure 10.3 for a value of a between 0 and 1.
P.834
P.835
 [[Example
p 10.3]]

L t us consider
Let id a signal
i l that
th t is
i the
th sum off two
t reall exponential
ti l signals:
i l

(10 12)
(10.12)

The z-transform
z transform is then

(10.13)

P.835
 (10.14)

(10.15)

For convergence of X(z), both sums in eq. (10.13) must converge, which requires
that both | (1/3)z– 1 | < 1 and | (1/2)z–1 | < 1, or equivalently, | z | > 1/3 and | z |
> 1/2.
1/2 Thus,
Th the h regioni off convergence is i | z | > 1/2
1/2.
The z-transform for this example can also be obtained using the results of Example
10 1 S
10.1. Specifically,
ifi ll from
f the
th definition
d fi iti off the
th z-transform
t f in
i eq. (10.3),
(10 3) we see
that the z-transform is linear; that is, if x[n] is the sum of two terms, then X(z) will be
the sum of the z-transforms
transforms of the individual terms and will converge when both z-
transforms converge. From Example 10.1,

(10.16)

P.836
 and

(10.17)

and consequently,

(10 18)
(10.18)

as we determined before. The pole-zero plot and ROC for the z-transform of each of
the individual terms and for the combined signal are shown in Figure 10.4.
10 4

P.836
P.837
 [[Example
p 10.4]]

L t us consider
Let id the
th signal
i l

The z-transform of this signal is

(10.19)

P.836
 or equivalently,
i l tl

(10 20)
(10.20)

For convergence of X(z), X(z) both sums in eq.eq (10.19)

(10 19) must converge,
converge which requires
that | (1/3)e jπ/4z–1 | < 1 and | (1/3)e–jπ/4z–1 | < 1 or equivalently, | z | > 1/3. The
pole-zero
pole zero plot and ROC for this example are shown in Figure 10.5.

P.837
P.838
 Sec. 10.2 THE REGION OF CONVERGENCE FOR THE z-
TRANSFORM
Key concepts
9 pproperties
p of the ROC of the z-transform,, especially
p y pproperties
p 2,, 3,, 8,, and 9 and Fig.
g
10.8

Property 1: The ROC of X(z) consists of a ring in the z-plane centered about the origin.

This property is illustrated in Figure 10.6 and follows from the fact that the ROC
consists of those values of z = re jω for which x[n]r–n has a Fourier transform that converges.
the ROC of the z-transform of x[n] consists of the values of z for which x[n]r–n is
absolutely summable:2

(10.21)

P.838
P.839
 P
Property 2 The
t 2: Th ROC ddoes nott contain
t i any poles.
l

Property 3: If x[n] is of finite duration,

duration then the ROC is the entire z-plane,
z plane except
possibly z = 0 and/or z = ∞.

A finite-duration sequence has only a finite number of nonzero values, extending, say,
from n = N1 to n = N2 where N1 and N2 are finite
finite. Thus,
Thus the zz-transform
transform is the sum of a
finite number of terms; that is,

(
(10.22)
)

P.839
 [[Example
p 10.5]]

C id the
Consider th unit
it impulse
i l signal
i l δ[n].
δ[ ] Its
It z-transform
t f is
i given
i by
b

(10 23)
(10.23)

with an ROC consisting of the entire z-plane,

z plane including z = 0 and z = ∞.
∞ On the other
hand, consider the delayed unit impulse δ [n – 1], for which

(10.24)

This z-transform is well defined exceptp at z = 0,, where there is a ppole. Thus,, the ROC
consists of the entire z-plane, including z = ∞ but excluding z = 0. Similarly, consider
p
an impulse advanced in time,, namely,y, δ [[n + 1].
] In this case,,
P.840
 (
(10.25)
)

which is well defined for all finite values of z. Thus, the ROC consists of the entire
finite z-plane (including z = 0), but there is a pole at infinity.

Property 4: If x[n] is a right-sided sequence, and if the circle | z | = r0 is in the ROC,

then all finite values of z for which | z | > r0 will also be in the ROC.

(10.26)

Property
P 5 If x[n]
t 5: [ ] is
i a left-sided
l f id d sequence, andd if the h circle
i l | z | = r0 is
i in
i the
h ROC,
ROC then
h
all values of z for which 0 < | z | < r0 will also be in the ROC.

P.840
P.841
 from eq. (10.3), the summation for the z-transform will be of the form

(
(10.27)
)

where N2 may be positive or negative.

Property 6: If x[n] is two sided, and if the circle | z | = r0 is in the ROC, then the ROC
will consist of a ring in the z-plane that includes the circle | z | = r0.

P.841
P.842
 [[Example
p 10.6]]

C id the
Consider th signal
i l

Then

(10.28)

P.843
 [[Example
p 10.6]]

Since x[n] is of finite length,

length it follows from Property 3 that the ROC includes the
entire z-plane except possibly the origin and/or infinity. In fact, from our discussion
of Property 3,
3 since x[n] is zero for n < 0,
0 the ROC will extend to infinity.
infinity However,
However
since x[n] is nonzero for some positive values of n, the ROC will not include the origin.
This is evident from eq.
eq (10.28),
(10 28) from which we see that there is a pole of order
N – 1 at z = 0. The N roots of the numerator polynomial are at

(10.29)
The root for k = 0 cancels the ppole at z = a. Consequently,
q y, there are no ppoles other
than at the origin. The remaining zeros are at

(10.30)
The pole-zero pattern is shown in Figure 10.9.
P.843
P.843
 [[Example
p 10.7]]

Let
(10.31)

This two-sided sequence is illustrated in Figure 10.10, for both b < 1 and b > 1. The z-
transform for the sequence can be obtained by expressing it as the sum of a rightsided
and a left-sided sequence. We have

(10.32)
From Example10.1,
p ,

(
(10.33)
)
and from Example 10.2,
(10.34)
P.844
P.844
 In Figures 10.11(a)–(d) we show the pole-zero pattern and ROC for eqs. (10.33) and
((10.34),
), for values of b > 1 and 0 < b < 1. For b > 1,, there is no common ROC,, and thus
the sequence in eq. (10.31) will not have a z-transform, even though the rightsided and
left-sided components
p do individually.
y For b < 1, the ROCs in eqs. q ((10.33)) and ((10.34))
overlap, and thus the z-transform for the composite sequence is

(10.35)

or equivalently,

(10.36)

Th corresponding
The di pole-zero
l pattern andd ROC are shown
h iin Figure
Fi 10
10.11(e).
11( )

P.845
 Property 7: If the z-transform X(z) of x[n] is rational, then its ROC is bounded by poles
or extends to infinity.
y

p y 8: If the z-transform X(z)

Property ( ) of x[n] is rational, and if x[n] is right
g sided, then the
ROC is the region in the z-plane outside the outermost pole–i.e., outside the circle of
radius equal to the largest magnitude of the poles of X(z). Furthermore, if x[n] is causal
(i.e., if it is right sided and equal to 0 for n < 0), then the ROC also includes z = ∞.

Property 9: If the z-transform X(z) of x[n] is rational, and if x[n] is left sided, then the
ROC is the region in the z-plane inside the innermost nonzero pole–i.e., inside the circle
of radius equal to the smallest magnitude of the poles of X(z) other than any at z = 0 and
extending inward to and possibly including z = 0. In particular, if x[n] is anticausal (i.e.,
if it
i is
i left
l f sided
id d andd equall to 0 for
f n > 0),
0) then
h the
h ROC alsol iincludes
l d z = 0. 0

P.845
P.846
 [[Example
p 10.8]]

Let us consider all of the possible ROCs that can be connected with the function

(10 37)
(10.37)

The associated pole

pole-zero
zero pattern is shown in Figure 10.12(a).
10 12(a) Based on our discussion in
this section, there are three possible ROCs that can be associated with this algebraic
expression for the zz-transform.
transform. These ROCs are indicated in Figure 10.12(b) – (d). Each
corresponds to a different sequence. Figure 10.12(b) is associated with a right-sided
sequence,
q , Figure
g 10.12(c)
( ) with a left-sided sequence,
q , and Figure
g 10.12(d)
( ) with a two-
sided sequence. Since Figure 10.12(d) is the only one for which the ROC includes the
unit circle,, the sequence
q corresponding
p g to this choice of ROC is the onlyy one of the three
for which the Fourier transform converges.

P.847
P.848
 Sec. 10.3 THE INVERSE z-TRANSFORM
z
Key concepts
(i) method 1 of the inverse z-transform: integration [see equat ions (10.40) and (10.41)];
(ii) Method 2 of the inverse z-transform: partial fraction expansion;
(iii) method 3 of the inverse z-transform: power series;
(iv) method 3-1: with the aid of long division to expand X(z) as a power series;
( ) method
(v) h d 3-2: with
i h the
h aid
id off Taylor’s
l series
i to expandd X(z)
( ) as a power series;
i
(vi) The inverse z-transform of Ai /(1 – ai z–1);(10.21)

as expressed in eq. (10.7),

(10 38)
(10.38)

for any value of r so that z = re jω is inside the ROC.

ROC Applying the inverse Fourier
transform to both sides of eq. (10.38) yields

P.847
 Or

(10.39)

or, moving the exponential factor r n inside the integral and combining it with the term
e jωn,

(
(10.40)
)

we can recover x[n][ ] from

f it z-transform
its t f evaluated
l t d along
l a contour
t z = re jω in
i the
th ROC,
ROC
with r fixed and ω varying over a 2π interval.
i terms
in t off an integration
i t ti in
i the
th z-plane,
l eq. (10.40)
(10 40) can bbe rewritten
itt as

(10 41)
(10.41)
P.848
 [[Example
p 10.9]]

Consider the z-transform

z transform

(10 42)
(10.42)

There are two poles,

poles one at z = 1/3 and one at z = 1/4,
1/4 and the ROC lies outside the
outermost pole. That is, the ROC consists of all points with magnitude greater than
that of the pole with the larger magnitude, namely the pole at z = 1/3. From Property
4 in Section 10.2, we then know that the inverse transform is a right-sided sequence.
As described in Appendix
pp I,, X(z)
( ) can be expanded
p by
y the method of ppartial fractions.
For this example, the partial-fraction expansion, expressed in polynomials in z–1, is

(10.43)

P.849
 Thus,, x[n]
[ ] is the sum of two terms,, one with z-transform 1/[1 ) –1] and the other
[ – ((1/4)z
with z-transform 2/[1 – (1/3)z–1]. In order to determine the inverse z-transform of each of
these individual terms, we must specify the ROC associated with each. Since the ROC
for X(z) is outside the outermost pole, the ROC for each individual term in eq. (10.43)
must also be outside the pole associated with that term. That is, the ROC for each term
consists of all points with magnitude greater than the magnitude of the corresponding
pole. Thus,

(10.44)
where
h

(10 45)
(10.45)

(10.46)

P.849
 From Example 10.1, we can identify by inspection that and

(10.47)

andd

(10 48)
(10.48)

and thus
thus,

(10 49)
(10.49)

P.850
 [[Example
p 10.10]]

Now let us consider the same algebraic expression for X(z) as in eq. eq (10.42),
(10 42) but with the
ROC for X(z) as 1/4 < | z | < 1/3. Equation (10.43) is still a valid partial-fraction
expansion of the algebraic expression for X(z),
X(z) but the ROC associated with the
individual terms will change. In particular, since the ROC for X(z) is outside the pole at z
= 1/4,
1/4 the ROC corresponding to this term in eq. eq (10.43)
(10 43) is also outside the pole and
consists of all points with magnitude greater than 1/4, as it did in the previous example.
However, since in this example the ROC for X(z) is inside the pole at z = 1/3, that is,
since the points in the ROC all have magnitude less than 1/3, the ROC corresponding to
this term must also lie inside this ppole. Thus,, the z-transform ppairs for the individual
components in eq. (10.44) are

(10.50)

P.850
 and

(10.51)

Th signal
The i l x1[n]
[ ] remains
i as in
i eq. (10.47),
(10 47) while
hil from
f Example
E l 10.2,
10 2 we can identify
id if

(10 52)
(10.52)

so that

(10 53)
(10.53)

P.850
 [[Example
p 10.11]]
Finally, consider X(z) as in eq. (10.42), but now with the ROC | z | < 1/4. In this case the
ROC iis iinside
id bboth
th poles,
l i.e.,
i the
th points
i t in
i the
th ROC all ll have
h magnitude
it d smaller
ll than
th
either of the poles at z = 1/3 or z = 1/4. Consequently the ROC for each term in the
partial fraction expansion in eq.
partial-fraction eq (10.43)
(10 43) must also lie inside the corresponding pole.
pole As a
result, the z-transform pair for x1[n] is given by

(10.54)

while the z-transform pair for x2[n] is given by eq. (10.51). Applying the result of
Example 10.2
10 2 to eq.
eq (10.54),
(10 54) we find that

so that

P.851
 Sec. 10.5 PROPERTIES OF THE z-TRANSFORM
z

Key concepts
(1) Learning the properties of (i) linearity, (ii) time shifting, (iii) scaling in the z-domain,
(i ) time
(iv) i reversal, l (v)
( ) time
i expansion,i (vi)
( i) conjugation,
j i (vii)
( ii) convolution,
l i (viii)
( iii)
multiplication of n (i.e., differentiation in the z-domain), and (ix) the initial value theorem.
Th
These properties
ti are summarized i d in
i Table
T bl 10.1.
10 1 (2) In
I fact,
f t many (but (b t nott all)
ll) properties
ti off
the z-transform are similar to those of the discrete-time Fourier transform (summarized in
Table 5.1).
5 1) Remember that, hen z = e jω, the z-transform
that when transform is reduced
red ced to the discrete-time
discrete time
Fourier transform.

10.5.1 Linearity

(10.71)

P.858
 10 5 2 Time Shifting
10.5.2

(10.72)

10.5.3 Scaling in the z-Domain

If

Then

(10.73)

where
h | z0 | R is
i the
h scaled
l d version
i off R.
R
P.859
 An important special case of eq. (10 73) is when z0 = e jω0. In this case,
eq (10.73) case | z0 | R = R and

(10 74)
(10.74)

10.5.4 Time Reversal

(
(10.75)
)

P.861
P.861
 10 5 5 Time Expansion
10.5.5
The sequence x(k)[n], introduced in Section 5.3.7 and defined as

(
(10.76)
)

If

Then

(10.77)

we see that the coefficient of the term z–n equals the value of the signal at time n. That is,
with

P.862
 it follows that

(10.78)

10 5 6 Conjugation
10.5.6
If

(
(10.79)
)
then

(10.80)

P.862
 10 5 7 The Convolution Property
10.5.7

(10.81)

P.863
 [E
[Example
l 10.15]
10 15]
Consider an LTI system for which

(
(10.82)
)

Where

Note that

(10.83)
with ROC equal to the entire z-plane except the origin. Also, the z-transform in eq.
(10.83) has a zero at z = 1. From eq. (10.81), we see that if

P.864
 Then

(10.84)

with ROC equal to R, with the possible deletion of z = 0 and/or addition of z = 1.

Note that for this system

That is,, y[
y[n]] is the first difference of the sequence
q x[n].
[ ] Since the first-difference
operation is commonly thought of as a discrete-time counterpart to differentiation, eq.
((10.83)) can be thoughtg of as the z-transform counterpart
p of the Laplace
p transform
differentiation property presented in Section 9.5.7.

P.864
 [E
[Example
l 10.16]
10 16]
Suppose that we now consider the inverse of first differencing, namely,
accumulation or summation. Specifically, let w[n] be the running sum of x[n]:

(10.85)

Then, using eq. (10.81) together with the z-transform of the unit step in Example
10.1, we see that

(10.86)

with ROC including at least the intersection of R with | z | > 1. Eq. (10.86) is the
discrete-time z-transform counterpart of the integration property in Section 9.5.9.

P.864
 10 5 8 Differentiation in the z-Domain
10.5.8 z Domain

(10.87)

P.865
 [E
[Example
l 10.17]
10 17]
If

(
(10.88)
)

Then

(10.89)

By differentiating, we have converted the z-transform to a rational expression. The

inverse z-transform of the right-hand side of eq. (10.89) can be obtained by using
Example 10.1 together with the time-shifting property, eq. (10.72), set forth in
Section 10.5.2. Specifically, from Example 10.1 and the linearity property,

P.865
 (10 90)
(10.90)

Combining this with the time-shifting property yields

Consequently,

(10.91)

P.865
 [E
[Example
l 10.18]
10 18]
As another example of the use of the differentiation property, consider determining
the inverse z-transform for

(10.92)

From Example 10.1,

(10.93)

and hence,

(10.94)

P.866
 10 5 9 The Initial-Value
10.5.9 Initial Value Theorem
If x[n] = 0,
0 n < 0,
0 then

(10 95)
(10.95)

This property follows by considering the limit of each term individually in the expression
for the z-transform, with x[n] zero for n < 0.

P.866
 Sec. 10.6 SOME COMMON zz-TRANSFORM PAIRS

Key concept
the z-transform pairs summarized in Table 10.2

P.867
 Sec. 10.7 ANALYSIS AND CHARACTERIZATION OF LTI
SYSTEMS USING z-TRANSFORMS
Key concepts
(i) system functions (transfer functions) for discrete-time LTI systems;
(ii) Properties for causal signals and stable signals (summarized in the blocks in Sections
10.7.1 and 10.7.2);
(iii) using the z-transform to analyze the linear constantcoefficient difference equations;
(iv) Deriving and analyzing the system function, as in Examples 10.26 and 10.27;
( ) Characteristic
(v) h i i off the
h z-transform
f f a di
for discrete-time
i allpass
ll system

F
From th
the convolution
l ti property
t presented
t d in
i Section
S ti 10.5.7,
10 5 7

(10 96)
(10.96)

P.867
P.868
P.869
 10 7 1 Causality
10.7.1
A discrete-time LTI system is causal if and only if the ROC of its system function is the
exterior of a circle, including infinity.

A discrete-time LTI system with rational system function H(z) is causal if and only if: (a)
the ROC is the exterior of a circle outside the outermost pole; and (b) with H(z)
expressed as a ratio of polynomials in z, the order of the numerator cannot be greater
than the order of the denominator.

P.869
 [E
[Example
l 10.21]
10 21]
Consider a system with system function

(
(10.97)
)

Since the ROC for this system

y function is the exterior of a circle outside the
outermost pole, we know that the impulse response is right-sided. To determine if
the system is causal, we then need only check the other condition required for
causality, namely that H(z), when expressed as a ratio of polynomials in z, has
numerator degree no larger than the denominator. For this example,

(10.98)

P.870
 so that the numerator and denominator of H(z) are both of degree two, and consequently
we can conclude that the system is causal.
causal This can also be verified by calculating the
inverse transform of H(z). In particular, using transform pair 5 in Table10.2, we find that
the impulse response of this system is

(10.99)

Since h[n]
[ ] = 0 for n < 0,, we can confirm that the system
y is causal.

P.870
 10 7 2 Stability
10.7.2
An LTI system is stable if and only if the ROC of its system function H(z) includes the
unit circle, | z | = 1.

P.871
 [E
[Example
l 10.22]
10 22]
Consider again the system function in eq. (10.97). Since the associated ROC is the
region | z | > 2, which does not include the unit circle, the system is not stable. This
can also be seen byy noting g that the impulse
p response
p in eq.
q ((10.99)) is not
absolutely summable. If, however, we consider a system whose system function
has the same algebraic
g expression
p as in eq.
q ((10.97)) but whose ROC is the region
g
1/2 < | z | < 2, then the ROC does contain the unit circle, so that the corresponding
system is noncausal but stable. In this case, using transform pairs 5 and 6 from
Table 10.2, we find that the corresponding impulse response is

(10.100)

which is absolutely summable.

P.871
 Also, for the third possible choice of ROC associated with the algebraic expression for
H(z) in eq.
eq (10.97),
(10 97) namely,
namely | z | < 1/2
1/2, the corresponding system is neither causal (since
the ROC is not outside the outermost pole) nor stable (since the ROC does not include
the unit circle).
circle) This can also be seen from the impulse response,
response which (using transform
pair 6 in Table 10.2) is

A causal LTI system

y with rational system
y function H(z)
( ) is stable if and only
y if all of the
poles of H(z) lie inside the unit circle–i.e., they must all have magnitude smaller than 1.

P.871
 [E
[Example
l 10.24]
10 24]

The system function for a second

second-order
order system with complex poles was given in
eq.(10.69), specifically,

(10.101)

with poles located at z1 = re jθ and z2 = re–jθ. Assuming causality, we see that the
ROC is outside the outermost ppole ((i.e.,, | z | > | r |)
|). The ppole-zero pplot and ROC
for this system are shown in Figure 10.16 for r < 1 and r > 1. For r < 1, the poles
are inside the unit circle, the ROC includes the unit circle, and therefore, the
system is stable. For r > 1, the poles are outside the unit circle, the ROC does not
include the unit circle, and the system is unstable.

P.872
 [E
[Example
l 10.25]
10 25]

Consider an LTI system for which the input x[n] and output y[n] satisfy the
linearconstant-coefficient difference equation

(10.102)

Applying the z-transform to both sides of eq. (10.102), and using the linearity
pproperty
p y set forth in Section 10.5.1 and the time-shiftingg pproperty
p y ppresented in
Section 10.5.2, we obtain

Or

(10.103)

P.872
P.873
 From eq. (10.96), then,

(10.104)

This provides the algebraic expression for H(z), but not the region of convergence.
In fact,
fact there are two distinct impulse responses that are consistent with the
difference equation (10.102), one is right sided and the other is left sided.
Correspondingly, there are two different choices for the ROC associated with the
algebraic expression (10.104). One, | z | > 1/2, is associated with the assumption that h[n]
is right
g sided,, and the other,, | z | < 1/2,, is associated with the assumption
p that h[n]
[ ] is left
sided.
Consider first the choice of ROC equal q to | z | > 1/2. Writing g

P.873
 we can use transform ppair 5 in Table 10.2, together
g with the linearityy and time-shiftingg
properties, to find the corresponding impulse response

For the other choice of ROC, namely, | z | < 1/2, we can use transform pair 6 in
Table 10.2 and the linearity and time-shifting properties, yielding

In this case, the system is anticausal (h[n] = 0 for n > 0) and unstable.

P.874
 consider an LTI system
y for which the input
p and output
p satisfyy a linear constant-
coefficient difference equation of the form

(10.105)

(10 106)
(10.106)

P.874
 [E
[Example
l 10.26]
10 26]

Suppose that we are given the following information about an LTI system:
1. If the input to the system is x1[n] = (1/6)n u[n], then the output is

where a is a real number.

2. If x2[[n]] = ((–1))n, then the output
p is y2[[n]] = . As we now show,, from
these two pieces of information, we can determine the system function H(z) for
this system, including the value of the number a, and can also immediately
deduce a number of other properties of the system.
The z-transforms of the signals specified in the first piece of information are

(10.107)

P.875
 (10.108)

From eq. (10.96), it follows that the algebraic expression for the system function is

(10.109)

Furthermore, we know that the response to x2[n] = (–1)n must equal (–1)n multiplied
by the system function H(z) evaluated at z = –11. Thus from the second piece of
information given, we see that

(10.110)

P.875
 Note that the convolution of cn and h[n] can be expressed in term of H(z), which is the z-
transform of h[n]:

After substituted c by –1, we have

which leads to eq. (10.110).

Solving eq. (10.110), we find that a = –9, so that

(10.111)

P.876
 Or

(10 112)
(10.112)

or, finally,
fi ll

(10 113)
(10.113)

Also, from the convolution property,

Also property we know that the ROC of Y1(z) must include at least
the intersections of the ROCs of X1(z) and H(z). Examining the three possible ROCs for
H(z) (namely,
(namely | z | < 1/3,
1/3 1/3 < | z | < 1/2
1/2, and | z | > 1/2)
1/2), we find that the only choice that
is consistent with the ROCs of X1(z) and Y1(z) is | z | > 1/2.

P.876
 Since the ROC for the system includes the unit circle, we know that the system is stable.
F h
Furthermore, ffrom eq. (10
(10.113)
113) with
i h H(z)
H( ) viewed
i d as a ratio
i off polynomials
l i l in
i z, the
h order
d
of the numerator does not exceed that of the denominator, and thus we can conclude that
th LTI system
the t is i causal.
l Also,
Al using
i eqs. (10.112)
(10 112) andd (10.106),
(10 106) we can writeit the
th
difference equation that, together with the condition of initial rest, characterizes the
system:

P.876
 Sec. 10.9 SYSTEM FUNCTION ALGEBRA AND BLOCK
DIAGRAM REPRESENTATIONS
Key concepts
learning how to plot block diagrams of a discrete system using the z-transform, including
• the direct form,
• the cascade form, and
(iii) The parallel form, see Fig. 10.21. Also note the alternative way for direct-form
representation, see Figs. 10.20 and 10.22.

10.9.1 System Functions for Interconnections of LTI Systems

The system function algebra for analyzing discrete-time block diagrams such as series,
parallel,
ll l andd feedback
f db k interconnections
i t ti is
i exactly
tl the
th same as that
th t for
f the
th corresponding
di
continuous-time systems in Section 9.8.1.

P.880
P.881
 [[Example
p 10.28]]
Consider the causal LTI system with system function

(10.124)

Using the results in Section 10.7.3, we find that this system can also be described
by the difference equation

together with the condition of initial rest. In Section 2.4.3 we constructed a block
diagram representation for a first
first-order
order system of this form, and an equivalent
block diagram (corresponding to Figure 2.28 with a = –1/4 and b = 1) is shown
in Figure
gu e 10.19(a). e e, z–1 iss thee sys
0. 9(a). Here, system
e function
u c o of o a unit
u delay.
de ay. That
a is,
s, from
o
the time-shifting property, the input and output of this system are related by

P.881
 The block diagram in Figure 10.19(a) contains a feedback loop much as for the system
considered in the previous subsection and pictured in Figure 10.18.
10 18 In fact,
fact with some
minor modifications, we can obtain the equivalent block diagram shown in Figure
10 19(b) which is exactly in the form shown in Figure 10.18,
10.19(b), 10 18 with H1(z) = 1 and H2(z)
= –1/4z–1. Then, applying eq. (10.123), we can verify that the system function of the
system in Figure 10.19
10 19 is given by eq.
eq (10.124).
(10 124)

P.882
P.882
 [[Example
p 10.29]]
Suppose that we now consider the causal LTI system with system function

(10.125)

As eq. (10.125) suggests, we can think of this system as the cascade of a system
with system function 1/[1 – (1/4)z–1] and one with system function 1 – 2z–1. We
have illustrated the cascade in Figure 10.20(a), in which we have used the block
10 19(a) to represent 1/[1 – (1/4)z–1].
diagram in Figure 10.19(a) ] We have also represented
1 – 2z–1 using a unit delay, an adder, and a coefficient multiplier. Using the time-
shifting property, we then see that the input v[n] and output y[n] of the system with
system function 1 – 2z–1 are related by

P.882
 While the block diagram in Figure 10.20(a) is certainly a valid representation of the
system in eq
eq. (10
(10.125),
125) it has an inefficiency whose elimination leads to an alternative
block-diagram representation. To see this, note that the input to both unit delay elements
in Figure 10.20(a)
10 20(a) is v[n],
v[n] so that the outputs of these elements are identical; i.e.,
ie

Consequently, we need not keep both of these delay elements, and we can simply use
the output of one of them as the signal to be fed to both coefficient multipliers.
multipliers The
result is the block diagram representation in Figure 10.20(b). Since each unit delay
element requires a memory register to store the preceding value of its input, the
representation in Figure 10.20(b) requires less memory than that in Figure 10.20(a).

P.883
P.883
 [[Example
p 10.30]]
Next, consider the second-order system function

(10.126)

which is also described by the difference equation

(10.127)

Using the same ideas as in Example 10.28, we obtain the block-diagram

p
representation for this system
y shown in Figure
g 10.21(a).
( ) Specifically,
p y, since the two
system function blocks in this figure with system function z–1 are unit delays, we
have

P.883
P.884
 so that eq. (10.127) can be rewritten as

Or

which is exactly what the figure represents.

The block diagram in Figure 10.21(a) is commonly referred to as a direct-form
representation,
i since
i the
h coefficients
ffi i appearing
i in
i the
h diagram
di can be
b determined
d i d by
b
inspection from the coefficients appearing in the difference equation or, equivalently,
th system
the t function.
f ti Alternatively,
Alt ti l as ini continuous
ti time,
ti we can obtain
bt i both
b th
cascadeform and parallel-form block diagrams with the aid of a bit of system function
algebra Specifically,
algebra. Specifically we can rewrite eq.
eq (10.126)
(10 126) as

(10 128)
(10.128)
P.884
 which suggests the cascade-form representation depicted in Figure 10.21(b) in which
the system is represented as the cascade of two systems corresponding to the two
factors in eq. (10.128).
Also, by performing a partial
partial-fraction
fraction expansion, we obtain

which leads to the parallel-form representation depicted in Figure 10.21(c).

P.885
 [Example 10.31]
Finally,
i ll consider
id the
h system function
f i

(10 129)
(10.129)

W iti
Writing

(10 130)
(10.130)

suggests representing the system as the cascade of the system in Figure 10.21(a)
10 21(a) and
the system with system function . However, as in Example 10.29, the
unit delay elements needed to implement the first term in eq.
eq (10.130)
(10 130) also produce the
delayed signals needed in computing the output of the second system. The result is the
direct-form
direct form block diagram shown in Figure 10.22, the details of the construction of
which are examined in Problem 10.42. The coefficients in the direct-form
p
representation can be determined byy inspection
p from the coefficients in the system
y
P.885
function of eq. (10.129).
 We can also write H(z) in the forms

(10.131)

And

(10.132)

Eq. (10.131) suggests a cascade-form representation, while eq. (10.132) leads to a

parallel-form block diagram. These are also considered in Problem 10.42.

P.885
P.886
 Sec. 10.10 THE UNILATERAL z-TRANSFORM
Key concepts
(i) definition of the unilateral z-transform;
(ii) the properties of the unilateral z-transform (Table 10.3);
(iii) using
i theh unilateral
il l z-transform
f to solve
l the
h linear
li difference
diff equations
i with
i h initial
i ii l
values
Definition 10.2
Unilateral z-Transform
Transform
The unilateral z-transform of a sequence x[n] is defined as

(10.133)
As in previous chapters,
chapters we adopt a convenient shorthand notation for a signal and its
unilateral z-transform:
(10 134)
(10.134)
P.886
 [[Example
p 10.32]]
Consider the signal

(10.135)

Since x[n] = 0, n < 0, the unilateral and bilateral transforms are equal for this
example and thus,
example, thus in particular,
particular

(10.136)

P.887
 [Example 10.33]
Let
((10.137))
In this case the unilateral and bilateral transforms are not equal, since x[–1] = 1 ≠ 0.
The bilateral transform is obtained from Example 10.1 and the time-shifting
property set forth in Section 10.5.2. Specifically,

(10.138)
In contrast, the unilateral transform is

Or

(10 139)
(10.139)
P.888
 [[Example
p 10.34]]
Consider the unilateral z-transform

(10.140)

In Example 10.9, we considered the inverse transform for a bilateral z-transform

X(z) of the same form as in eq.
eq (10.140)
(10 140) and for several different ROCs.
ROCs In the case
of the unilateral transform, the ROC must be the exterior of the circle of radius
equal to the largest magnitude of the poles of (z)(z)–in
in this instance, all points z with
| z | > 1/3. We can then invert the unilateral transform exactly as in Example 10.9,
y
yielding g

(
(10.141)
)

In eq. (10.141), we have emphasized the fact that inverse unilateral z-transforms
provide us with information about x[n] only for n ≥ 0.
P.888
 10.10.2 Properties of the Unilateral z-Transform

The unilateral z-transform has many important properties, some of which are identical to
their bilateral counterparts and several of which differ in significant ways.
ways Table 10.3
10 3
summarizes these properties.

Table 10.3 states that if x1[n] = x2[n] = 0 for all n < 0, then

(10.146)

An example of such application is to the accumulation or summation property in Table

10.3.
0.3. Specifically,
Spec ca y, if x[n] = 0 for
o n < 0, then
e

((10.147))
P.890
 [[Example
p 10.36]]
Consider the causal LTI system described by the difference equation
(10 148)
(10.148)

together with the condition of initial rest.

rest The system function for this system is
(10.149)

Suppose that the input to the system is x[n] = αu[n], where α is a given constant. In
this case, the unilateral (and bilateral) zz-transform
transform of the output y[n] is

(10.150)

Applying Example 10.32 to each term of eq. (10.150) yields

(10.151)
P.891
 To develop the shifting property for the unilateral transform, consider the signal

(10.152)

Then

P.892
 Or

(10.153)

so that

(10.154)

By repeated application of eq. (10.154), the unilateral transform of

(10.155)

Is

(10.156)
P.892
 Teacher Note
In general,
general if k is a positive integer,
integer then

There is also a time advance property for unilateral transforms that relates the transform
of an advanced version of x[n]
[ ] to X(z).
( ) Specifically,
p y as shown in Problem 10.66,

(10.157)

P.892
 [[Example
p 10.37]]
Consider again the difference equation (10.148) with x[n] = αu[n] and with the
initial condition
(10.158)
Applying the unilateral transform to both sides of eq.
eq (10.148)
(10 148) and using the
linearity and time delay properties, we obtain

(10.159)

Solving for y(z) yields

(10.160)

P.893
Referring to Example 10.36 and, in particular, eq. (10.150), we see that the second term on the
right hand side of eq.
right-hand eq (10
(10.160)
160) equals the unilateral z-transform
z transform of the response of the system
when the initial condition in eq. (10.158) is zero (β = 0). That is, this term represents the
response of the causal LTI system described by eq eq. (10
(10.148),
148) together with the condition of
initial rest. As in continuous-time, this response is frequently referred to as the zero-state
response, i.e., the response when the initial condition or state is zero.
The first term on the right-hand side of eq. (10.160) is interpreted as the unilateral transform
of the zero-inputp response–i.e.,
p , the response
p of the system
y when the input
p is zero (α( = 0).
) The
zero-input response is a linear function of the value β of the initial condition. Moreover, eq.
((10.160)) illustrates the fact that the solution of a linear constant-coefficient difference equation
q
with nonzero initial state is the superposition
of the zero-state and zero-input responses. The zero-state response, obtained by setting the
initial condition to zero, corresponds to the response of the causal LTI system defined by the
difference equation and the condition of initial rest. The zero-input response is the response to
the initial condition alone with the input set to zero. Problems 10.21 and 10.46 provide other
examples illustrating the use of unilateral transforms to solve difference equations with
nonzero initial conditions.
P.893
 Finally, for any values of α and β, we can expand y(z) in eq. (10.160) by the method of
partial fractions and invert the result to obtain y[n] . For example,
example if α = 8 and β = 1,
1

(10.161)

and applying
pp y g the unilateral transform ppair in Example
p 10.32 to each term yields
y

(
(10.162)
)

P.893