107

Chapter 4

The z-Transform

4.1 Introduction

We have seen in the chapter on sampling that the Fourier transform plays a key role in
representing and analyzing discrete-time signals and systems. Here, we will consider a
generalization of the Fourier transform referred to as the z-transform. The z-transform for
discrete-time signals is a counterpart of the Laplace transform for continuous-time signals. A
principal motivation for introducing this generalization is that the Fourier transform does not
converge for all sequences and it is useful to have a generalization of the Fourier transform that
encompasses a broader class of signals. A second advantage is that in analytical problems the z-
transform notation is often more convenient than the Fourier transform notation. Still another
advantage is that the z-transform allows us to bring the power of complex variable theory to bear
on problems of discrete-time signals and systems.

Here, we define the z-transform representation of a sequence. Its properties and some of the
important theorems associated with the z-transform will be presented. We basically use the z-
transform as a tool for the analysis, design, and implementation of digital filters. The inverse z-
transform will be given towards the end of this chapter. In discussing the z-transform we use a
number of results from the theory of complex variables.

4.2 The z-Transform

The z-transform of a sequence x(n) is defined as an infinite sum or infinite power series

( )= ( ) 4.1

where z is a continuous complex variable. The z-transform may be considered as an operator that
transforms a sequence into a function. This z-transform operator z is indicated as

Z
x (n)  X ( z)   x ( n) z n
.
n  

The z-transform as defined in Equation (4.1) is often referred to as the two-sided or bilateral z-
transform, to differentiate it from the one-sided or unilateral z-transform, which is defined as
108 Introduction to Discrete-time Signal Processing

X  ( z)   x ( n) z n

n0

The bilateral and unilateral z-transforms are the same only if ( ) = 0 for ≤ −1. Most of the
discussion in this chapter will be on the bilateral z-transform. However, the unilateral z-
transform will be treated at the end of the chapter, where its significance for solving linear
constant coefficient difference equations with nonzero auxiliary conditions will be emphasized.

The first time we encountered the z-transform in this text was in connection with the spectrum of
sampled signals where the spectrum (DTFT) was related to Equation (4.1) for particular values
of z, i.e. |z| = 1, the unit circle in the z-plane. This shows that the Fourier transform is a special
case of the z-transform. Thus, the z-transform evaluated on the unit circle corresponds to the
Fourier transform (DTFT) of the sequence.

Since the z-transform is an infinite power series, it exists only for those values of z for which this
series converges. The region of convergence (ROC) of
X(z) is the set of all values of z for which ( ) attains a Im z-plane
finite value. Thus, the z-transform together with its ROC
gives a unique representation of a sequence.

The complex variable z may be expressed, in polar form,

as z  re j  and the z-transform becomes -1 0 1 Re

X ( z)   x ( n) z
n  
n


  x ( n) r
n  
 n  j n
e unit circle

Uniform convergence of the z-transform requires that

the sequence be absolutely summable. Applying this to Equation (4.1) leads to


 x ( n) r  n  
n  

for absolute convergence of the z-transform.

The infinite sum X(z) may be expressed in closed form, i.e., it can be expressed as a simple
mathematical function. An important form of X(z) is the one where X(z) is a rational function,
inside the region of convergence, i.e.,
N ( z)
X (z) 
D( z )
where N(z) and D(z) are polynomials in z. The values of z for which ( ) = 0 are called zeros of
X(z) and the values of z for which X(z) is infinite are called the poles of X(z). Note that the finite
 The z-transform 109

poles of X(z) are the roots of the denominator polynomial D(z). Poles may also exist at z = 0 and
= ∞. The region of convergence of the z-transform and the location of the poles are very
closely related to the nature of the sequence x(n).

Example 4.2.1: Consider the exponential sequence x (n)  a n u(n) . The z-transform is given
by

X ( z)  a
n  
n
u( n) z  n

  (az
n 0
1 n
)

Convergence of ( ) imposes the requirement

| | < ∞.

The above sum will converge for values of z for which | | < 1, i.e., | | > | |. These
values of z for which the infinite sum converges define the region of convergence. In this
region of convergence the infinite series converges to

( ) = ( )

1 −
= , ( ) = , | |<1
1− 1−

= , | |>| |
−
This shows that the z-transform converges for any finite | | and for some values of z. The
Fourier transform, however, converges only if | | < 1. Note that = 1 corresponds to
( ) = ( ), the unit step sequence, with z-transform
1
X ( z)  , z  1.
1  z 1
Notice that ( ) has a pole at = and a zero at = 0. The ROC is outside the circle
defined by | | = | |. The pole-zero plot of ( ) is shown in Figure 4.2. For the unit step
sequence ( ) = ( ), the pole is at z = 1 and the region of convergence is outside the
unit circle.
In general, the ROC can be the whole z-plane, outside of the circle defined by | | = | |,
inside of a circle defined by | | = | |, or an annular ring defined by | | < | | < | | for
some numbers ‘a’ and ‘b’. We will also see that the numbers ‘a’ and ‘b’ are associated with the
110 Introduction to Discrete-time Signal Processing

poles of ( ). We will also see that the z-transform ( ) is not unique unless its region of
convergence is also given. The following example illustrates this point.

Im
1 x(n) = anu(n)
z-plane
ROC

0
O a× 1 Re

•••

0 5 n unit circle

Fig. 4.1. Time domain plot of the sequence in Fig. 4.2. Pole-zero plot and region of
Example 4.2.1 convergence for Example 4.2.1

Example 4.2.2: Consider the sequence ( ) = − (− − 1). This exponential sequence

exists for negative n only, i.e., −∞ < < −1, and the exponential sequence is
inverted. The z-transform is
 1
X ( z)   
n  
a n u(  n  1) z  n   a
n  
n
z n
 
  a
n 1
n n
z  1  (a
n 0
1
z) n

If | a-1z | < 1, or equivalently, | z | < | a|, the above sum converges to

1 1 z
X ( z)  1   , z a
1  a 1 z 1  az 1 z a
The sequence and the pole-zero plots are shown in Figures 4.3 & 4.4, respectively.
Im
x(n) z-plane
0
n
ROC

0
o ×a 1 Re

unit circle
Fig. 4.3. Time domain plot of the sequence in Fig. 4.4. Pole-zero plot and ROC of
Example 4.2.2. Example 4.2.2.
 The z-transform 111

Note that in Examples 4.2.1 & 4.2.2 the sequences are different, but their z-transforms and the
pole-zero plots are the same. However, the regions of convergence are completely disjoint. Also,
( ) will be rational if ( ) is a linear combination of real or complex exponential sequences.

Example 4.2.3: Consider the linear combination of exponential sequences

n n
 1  1
x ( n)    u(n)     u(n).
 2  3
The z-transform is
   1 n  1
n

X ( z)      u(n)
n    2
    u( n)  z  n
 3 
 n  n
 1 n  1  n
    z
 
n 0 2
 
n0
  z
 3
1 1
1 1 2(1  z )
12
  
1 1 1 1
1  z 1 1  z 1 (1  z 1 )(1  z 1 )
2 3 2 3
1
2 ( − 12)
=
1 1
− 2 ( + 3)
.
Convergence of ( ) requires that
1 1
<1 − <1
2 3
i.e.,
| | > 1/2 | | > 1/3.
This gives a region of convergence | | > 1/2. This example reveals the fact that the z-
transform of a linear combination of two sequences is the same linear combination of the
individual z-transforms. In general, we state the fact that the z-transform is a linear
operator, as can be verified from the defining equation. On the other hand, the region of
convergence of the combination is the intersection of the individual regions of
convergences.

Example 4.2.4: Consider the sequence ( ) given by

n n
 1  1
x ( n)     u( n)    u(  n  1).
 3  2
112 Introduction to Discrete-time Signal Processing

Using the general result of Example 4.2.1,

1 1 1
− ( ) ⎯⎯⎯ , | |> .
3 1 3
1+3
Using the result of Example 4.2.2, for the second term,
1 1 1
− (− − 1) ⎯⎯⎯, | |< .
2 1 2
1−2
z-plane
Using the linearity property of the z-transform,
ROC
1 1 1 1
X ( z)   , z  and z 
1 1 1 1 3 2
1 z 1 z
3 2
1 1 × oo ×1/2
2(1  z ) 1 1 -1/3 0 1/12
12
 ,  z 
1 1 3 2
(1  z 1 )(1  z 1 )
3 2

Note that the ROC for this example is the annular

region 1/3 < | z | < 1/2 . The pole-zero plot and the
ROC is shown in Figure (4.5).
Fig. 4.5. Pole-zero plot and ROC for
Example 4.2.4.

The z-transform of a sequence may be expressed as a ratio of polynomials in positive powers of z

or polynomials in negative powers of z. However, from the definition of the z-transform, we see
that sequences that are zero for n < 0 have z-transforms involving only powers of rather than
z. Nevertheless, even when x(n) is nonzero for n < 0, X(z) can still be expressed as a rational
function in powers of .

A finite-length sequence is a sequence that is nonzero on some finite interval ≤ ≤ . The

z-transform of such a sequence becomes, from the definition,
N2

X ( z)   x (n) z
n  N1
n

Such finite-duration sequences have no problems of convergence as long as each term in the
summation is finite. The following examples illustrates the above fact.

Example 4.2.5: Compute the z-transform of the unit sample sequence

 1, if n  0
x( n)   (n)  
 0, if n  0
 The z-transform 113

The z-transform may be obtained using the definition by setting = = 0. This gives
( ) = 1, for all z.
The ROC is the entire z-plane.

Example 4.2.6: Consider the finite-length sequence ( ), where N is finite,

, 0≤ ≤ −1
( )=
0, ℎ
Its z-transform is
N 1 N 1
X ( z)   a n z n   (az 1 n
)
n0 n 0

1−( )
=
1−
1 −
=
−
o z-plane
Zeros: N=8
The above sum will be finite as long o o
as | | < ∞ and ≠ 0. Thus, the pole-zero
cancellation
ROC contains the entire z-plane 7
except the origin. Note that for real o -1 0
x 1 a
‘a’ there will be a pole of multiplicity
N-1 at the origin and N-1 zeros
distributed around a circle of radius
7-poles o o
‘a’. Also there is a pole/zero o
cancellation at z = a.

4.3 Properties of the Region of Convergence of the z-Transform

As may be seen from the previous examples the region of convergence is one property of a se-
quence. Sequences having the same general behavior have regions of convergence having the
same general behavior. These properties are summarized below.
1. A region of convergence has the geometric shape of a disk in the z-plane centered at the
origin, i.e., 0 ≤ ≤ | | ≤ ≤ ∞, where ri is the inner radius of the disk which may be
zero and ro is the outer radius of the disk which may extend to infinity.
2. A ROC cannot contain any poles.
114 Introduction to Discrete-time Signal Processing

3. If ( ) is a finite-duration sequence, then the ROC is the entire z-plane except possibly
= 0 or = ∞.
4. If ( ) is a right-sided sequence, i.e., a sequence that is zero for −∞ < < , then the
ROC extends outward from the outermost finite pole in ( ) to ( and possibly including)
= ∞.
5. If ( ) is a left-sided sequence, i.e., a sequence that is zero for > > −∞, the ROC
extends inward from the innermost nonzero pole of ( ) to (and possibly including)
= 0.
6. If ( ) is a two-sided infinite-duration sequence, then the ROC will consist of a ring in
the z-plane, bounded on the interior and exterior by a pole.

Some of these general properties of the ROC of the z-transform of sequences are indicated in the
figures below for causal, anticausal, and two-sided finite-duration and infinite-duration
sequences, respectively.

Finite duration sequences and their regions of convergence

Sequence ROC

causal sequence entire z-plane z-plane

except z = 0

-1 0 1 2 3 4 5 n
0

anticausal sequence entire z-plane z-plane

except z = ∞

-5 -4 -3 -2 -1 0 1 2 n 0
 The z-transform 115

two-sided sequence z-plane

entire z-plane
except z = 0
and z = ∞

-5 -4 -3 -2 -1 0 1 2 n 0

Infinite duration sequences and their regions of convergence

Sequence ROC

causal sequence
z-plane

. . . . . .
r1
-1 0 1 2 3 4 5 n |z| > r1

z-plane
anticausal sequence

. . . . . . r1
-5 -4 -3 -2 -1 0 1 2 n
|z| < r1

two-sided sequence
z-plane

. . . . . . r1
-5 -4 -3 -2 -1 0 1 2 n r2
r1< |z| < r2
116 Introduction to Discrete-time Signal Processing

4.4 z-Transform Properties

The z-transform is often employed for the analysis of discrete-time signals and systems. The
power of this transform is a consequence of some very important properties that are often used in
conjunction with analysis techniques. We will consider some of the most frequently used
properties of the z-transform. Many of these properties may be verified using the definition of
the z-transform. In the following discussion, ( ) denotes the z-transform of ( ).

1. Linearity: The linearity property states that if

Z Z
( )  ( ) and ( )  ( ),
then
Z
( ) = ( ) + ( )  ( ) + ( ).
for some constants a and b. This property may be proved easily by using the
definition of the z-transform. The linearity property states that the z-transform of
a linear combination of signals is the same linear combination of the z-transforms.
The linearity property has already been used in the previous examples.

Example 4.4.1: Determine the z-transform of ( ) = cos( ) ( ).

Using Euler’s identity, x(n) may be expressed as
1 j0 n 1
x( n)  cos( 0 n)u (n)  e u ( n)  e  j 0 n u (n)
2 2
and using the linearity property of the z-transform
 
1 j 0n  n 1
X ( z)  2 e z   2 e  j 0n z  n
n 0 n 0
1 1 1 1
 j 0  1
  j 0  1
2 1 e z 2 1 e z
1
1  z cos( 0 )

1  2 z 1 cos( 0 )  z  2
The region of convergence of ( ) may be obtained as the intersection of the
individual regions of convergence. For this sinusoid, the ROC is | | > 1.

Z
2. Time Shifting: If ( )  ( ),
then
( − ) 
Z
( ),
where D is assumed to be an integer.
 The z-transform 117

The ROC of ( ) is the same as that of ( ), except at = 0 if > 0 and

= ∞ if < 0. Note that if D is positive, the original sequence ( ) is shifted
right, and if D is negative, ( ) is shifted left. The ROC is affected by the factor
which introduces poles at = 0 or at = ∞. This property follows
directly from the definition, i.e., if ( ) = ( − ), then

n
Y ( z)   x ( n  D) z
n  

Making change of variables = − ,

( )= ( ) ( )

( ) = ( )

= ( )
The time shifting property and the linearity property are extremely useful in the
analysis of discrete-time LTI systems.

Example 4.4.2: Consider the z-transform

1 1
X ( z)  , z  .
1 2
z
2
Note that from the ROC, the sequence ( ) is right-sided. ( ) may be
expressed as
1 1
( ) = =, | |> .
1 1 2
1−2 1−2
From the time-shifting property and the previous examples, we recognize this
transform as being associated with a time shift of one sample to the right of the
sequence
n
1
x( n )    u ( n )
 2
i.e.,
n 1
1
x( n )    u ( n  1)
 2
118 Introduction to Discrete-time Signal Processing

3. Multiplication by an Exponential Sequence (i.e., Scaling in the z-Domain):

Z
If ( )  ( ), ROC: <| | <
Z
then ( )  ( ), ROC: | | < | | < | |

for any real or complex constant a. The reader is advised to prove the above
property and verify that the region of convergence is indeed as given.

As a consequence of this property, all pole-zero locations are scaled by a factor a

since if ( ) has a pole at z = z1 , then ( ) will have a pole at z = az1 . This
may be interpreted as shrinking or expanding of the z-plane if a is real and
positive, i.e., pole and zero locations change along radial lines in the z-plane. If,
however, a is complex with unity magnitude, i.e., a  e j 0 , the scaling
corresponds to a rotation in the z-plane by an angle of 0 (counterclockwise if
0 > 0, i.e., poles and zeros are rotated along circles centered at the origin. This
may be interpreted as a frequency shift or frequency translation, associated with
modulation. This corresponds to multiplication of a signal in the time domain by
the complex exponential sequence a n u ( n)  e j0 n u ( n) .

Example 4.4.3: We know that

1
u(n) Z  , | z|  1.
1  z 1
We wish to determine the z-transform of
x( n)  a n cos( 0 n)u (n)
Note that x(n) may be expressed as
1 1
x ( n)  (ae j 0 ) n u (n)  ( ae  j 0 ) n u (n)
2 2
Using the multiplication by an exponential sequence property, we notice that
1
1 2
(ae j 0 ) n Z
 , | z||a|,
2 1  ae j 0 z 1
and
1
1 2
( ) ⎯⎯⎯⎯ , | | < | |.
2 1−
Using the linearity property, we get
 The z-transform 119

1 1 1 1
X ( z)  j 0  1
 , | z|  | a|
2 1  ae z 2 1  ae  j 0 z 1
1  a cos( 0 ) z 1
 , |z| > |a|
1  2a cos( 0 ) z 1  a 2 z  2

4. Differentiation in the z-Domain:

Z
If ( )  ( ),
dX ( z)
then nx(n) Z   z
dz
This property may easily be proved by differentiating the z-transform, i.e.,

n
X ( z)   x (n) z
n  

Differentiate both sides

d d  
d
dz
X ( z)   x ( n) z  n
dz n  
  x(n) dz z n

n  
 
 n1 1 n
  (n) x(n)z  z  nx(n) z
n   n  

i.e.,
d
X ( z)   z 1 Z {nx(n)}
dz
i.e.,
d
z X ( z) Z  nx (n)
dz
The ROC is the same as that of x(n) except possibly = 0 or = ∞.

Example 4.4.4: i) Determine the z-transform of

x(n)  na n u (n)
Using the differentiation property, it follows that
d 1
X ( z)   z , | z|  | a|
dz 1  az 1
az 1
 , | z|  | a |
(1  az 1 ) 2
Thus, we get
120 Introduction to Discrete-time Signal Processing

az 1
na n u(n) Z  , | z|  | a |
(1  az 1 ) 2

ii) Use the differentiation and time-shifting property to determine the inverse z-
transform of X(z) given as
X ( z)  log(1  az 1 ), | z|  | a |
Note that this form of X(z) has never been encountered so far. However, try to do
something to it to bring it to forms that have been encountered so far. The log
function will be a rational function when differentiated. Hence, use the
differentiation property.
d  az  2
X (z) 
dz 1  az 1
To make use of the differentiation property, this may be written as
d  az  2
X ( z) 
dz 1  az 1
1 az 1
 
z 1  az 1
i.e.,
d  az 1
z X ( z) 
dz 1  az 1
Note that the problem has been reduced to finding the sequence whose z-
transform is
az 1 1
 az 1
1  az 1 1  az 1
which is recognized as the z-transform of ( a ) n u(n) multiplied by ‘a’ and
delayed by one sample interval, i.e., a (  a ) n 1 u(n  1) . Since this is the
differentiated transform,
nx(n)  a( a ) n 1 u ( n  1).
Therefore,
a n 1 Z
x( n)  ( 1) n 1 u ( n  1)  log(1  az 1 ), | z |  | a |.
n

Z
5. Time Reversal: If x(n)  X(z), ROC: < | | <
then Z
x(-n)  X(1/z), ROC: <| |<
 The z-transform 121

This property follows easily from the definition of the z-transform and the details
are left to the student as an exercise.

Example 4.4.5: Consider the sequence

x(n)  a  n u ( n).

which is a time-reversed version of a n u(n) . From the time-reversed property the

z-transform follows as
1 − 1
( )= = , | |< .
1− 1−

Z
6. Convolution property: If ( )  ( )
Z
and ( )  ( ),
then
Z
( ) = ( ) ∗ ( )  ( ) ( ).

Proof: The convolution of x1(n) and x2(n) is defined as


x( n)   x (k )x (n  k ).
k  
1 2

The z-transform of x(n) is

  
X ( z)   x(n)z n
n  
  [  x1 (k ) x2 (n  k )]z n
n  k  

Assuming we can interchange the order of summations,

 
X ( z)   x1 (k )[  x2 (n  k )z  n ]
k  n 

Changing the index of summation in the second sum from n to m = n - k, we get

 
X (z)   x1 (k )[  x2 (m)z  m ]z  k
k   m  
 
k m
  x (k )z  x (m)z
k  
1
m  
2

 X1 ( z) X 2 ( z)
The region of convergence is, at least, the intersection of that for X1(z) and X2(z).
If a pole that borders on the region of convergence of one of the z-transforms is
canceled by a zero, then the region of convergence of the result may be larger.
122 Introduction to Discrete-time Signal Processing

As we shall see, in a later chapter, the convolution property plays a particularly

important role in the analysis of LTI systems.

Example 4.4.6: Let ( ) = ( ) and ( ) = ( ). The corresponding z-

transforms are

1
X 1 ( z)  a n
z n  , | z||a|
n 0 1  a z 1
and

n 1
X 2 ( z)  z
n 0

1  z 1
, | z | 1

If | | < 1, then the z-transform of the convolution of x1(n) and x2(n) is

X ( z)  X 1 ( n) X 2 ( n )
1

(1  a z )(1  z 1 )
1

1  1 a 
   , | z | 1
1 a 1 z 1
1  a z 1 

Therefore, inverse z-transforming X(z), we get

1
x(n)  x1 (n) * x2 (n)  [1  a n1 ]u (n).
1 a

7. Correlation Property:
Z
If ( )  ( )
Z
( )  ( )
then the correlation of the two sequences is defined as

= ( )∗ (− ) = ( ) ( + )

Its z-transform is
Z
rx1x2 (k )  Rx1x2 ( z )  X 1 ( z ) X 2 ( z 1 ).

Proof: Recall that the correlation of two sequences may be expressed as the
convolution
rx1 x2 (k )  x1 (k ) * x2 ( k ).
 The z-transform 123

Using the convolution and time-reversal properties of the z-transform, we get

Rx1 x2 ( z )  X 1 ( z ) X 2 ( z 1 ).

The region of convergence of Rx1 x2 ( z ) is at least the intersection of that of X1(z)

and X2(z -1). When the sequences to be correlated are the same we get the auto-
correlation of the sequence. Correlation, like convolution, has very important
applications in signal processing.

8. Initial Value Theorem: If x(n) is causal, then

x (0)  lim X ( z )
z

This may be shown, for ( ) causal, by considering the limit of each term in the
definition of the z-transform.

lim X ( z )  lim  x(n) z  n
z  z 
n 0

 lim [ x(0)  x(1) z 1  x(2) z  2  x (3) z 3   ]

z 

 x(0).

9. Complex Conjugation: If ( ) is a complex sequence with z-transform ( ), then the z-

transform of its complex conjugate is
* Z * *
x (n)  X (z )
with a region of convergence the same as that of ( ).This property follows from
the definition of the z-transform.

We have so far derived the z-transforms of most of the frequently encountered sequences in
applications and the general properties of the z-transform. For easy reference the z-transform of
some of the sequences is tabulated below.

Sequence Transform ROC

1. ( ) 1 All z
1
2. u(n) |z| > 1
1  z 1
1
3. -u(-n-1) |z| < 1
1  z 1
124 Introduction to Discrete-time Signal Processing

All z except
4. ( − ) z-m
0 or ∞
1
5. ( ) |z| > |a|
1  az 1
1
6. − (− − 1) | z| < |a|
1  az 1
a z 1
7. ( ) | z| > |a|
(1  a z 1 ) 2
a z 1
8. − (− − 1) | z| < |a|
(1  a z 1 ) 2
1  cos( 0 ) z 1
9. cos( ). ( ) | z| > 1
1  2 cos( 0 ) z 1  z  2
1  sin( 0 ) z 1
10. sin( ). ( ) | z| > 1
1  2 sin( 0 ) z 1  z  2
1  r cos( 0 ) z 1
11. cos( ). ( ) | z| > r
1  2r cos( 0 ) z 1  r 2 z 2
1  r sin(  0 ) z 1
12. ( ). ( ) | z| > r
1  2r sin(  0 ) z 1  r 2 z  2
a n , 0  n  N  1 1 aN zN
13.  | z| > 0
0 , otherwise 1  az 1

4.5 The Inverse z-Transform

The z-transform is a powerful tool for discrete-time LTI system analysis. This analysis involves
finding the z-transform of sequences, manipulations of these transforms, and finding the inverse
z-transform of the final expressions.

Several formal and informal methods have been developed for determining the inverse z-
transform given an algebraic expression and some region of convergence. The z-transform is not
unique unless an associated region of convergence is attached to it. In most cases informal
methods can be employed to find the inverse z-transform. Some of the less formal inversion
methods are the partial fraction expansion method and expansion of the z-transform into a
series. In both cases we try to bring the z-transform expression into a simpler form which is
easily recognizable as the z-transform of a known sequence.
 The z-transform 125

4.5.1. Partial Fraction Expansion Method

In most cases the z-transform is expressed as a rational function in powers of z-1 as

k
b z
k 0
k
X ( z)  
k
a z
k 0
k

This form is not recognizable as a known transform. Thus, it has to be changed to a simpler form
or rational function whose inverse is easily recognizable. Hence, the z-transform may be
expressed as
M
b0  (1  ck z 1 )
k 1
X ( z)  N
a0  (1  d k z 1 )
k 1

where the ck ' s are the nonzero zeros and the d k ' s are the nonzero poles of ( ), respectively.
For < , X(z) is strictly proper, and if all poles are first order, then ( ) may be expressed
as a sum of simpler fractions as
N
Ak
X ( z)  1 d
k 1 z 1
k

Each coefficient Ak may be determined by multiplying both sides of the equation by (1 - dk z-1)
and evaluating the resulting expression at = as
Ak  (1  d k z 1 ) X ( z ) | z d k
If, however, ≥ , then the rational function has to be made proper through long division until
the remainder is proper rational. Hence, for ≥ , the partial fraction expansion would have
the form
M N N
Ak
X ( z )   Br z  r   1
.
r 0 k 1 1  d k z

The first part is a polynomial whose inverse is simple and the Ak‘s are obtained as before. It is to
be noted that the assumption so far is that ( ) has distinct poles. If, however, ( ) has multiple
poles and ≥ , then the partial fraction expansion has to be modified further. If ( ) has a
pole of order s at z = di , (i.e., k = i,) then the expansion becomes
M N N s
r Ak Cm
X ( z)  B z
r 0
r  1 d
k 1 z 1
  (1  d z
m 1
1 m
)
.
k i
k i

The coefficients Ak and Br are obtained as before. The coefficients Cm are obtained from
126 Introduction to Discrete-time Signal Processing

1 d sm
Cm  [(1  d i z 1 ) s X ( z )]z  di , m  1, 2, 3,  , s
( s  m)!( d i ) s m d ( z 1 ) s  m

The above expression may be derived by a similar argument as in the distinct poles case, but
differentiating the resulting expression until a constant remains in the numerator of one of the
terms Cm in the partial sum. This is repeated until all coefficients Cm are obtained. The student is
advised to verify the expression for Cm.

Once the partial fraction expansion is obtained finding the inverse z-transform, x(n), can be
obtained by using the linearity property of the z-transform and finding the inverse transform of
each partial fraction term in the sum. The terms Br z - r, for ≥ , corresponds to shifted and
scaled impulse sequences, i.e., terms of the form ( − ). The fractional terms, for distinct
poles case, correspond to exponential sequences. The case of multiple poles can be handled using
z-transform properties. The region of convergence is used to determine uniquely the sequence.

Example 4.5.1: Consider the z-transform

2 z 1
X ( z) 
1  2 z 1  0.75 z  2
The transform ( ) is a proper rational function with M =1 and N = 2. The denominator
polynomial may be represented as
1  2 z 1  0.75 z 2  (1  0.5 z 1 )(1  1.5 z 1 )
The poles of ( ) are at z = 0.5 and at z = 1.5. The partial fraction representation of
( ) is
2
Ak A1 A2
X ( z)  1 d
k 1 z 1

1  0. 5 z 1

1  1.5 z 1
k
where
1 2 z 1
A1  (1  0.5 z ) X ( z ) | z  0.5   2
1  1.5 z 1 | z  0.5
2 z 1
A2  (1  1.5 z 1 ) X ( z ) | z  1.5   2
1  0.5 z 1 | z  1.5
Thus,
2 2
X ( z)   1

1  0.5 z 1  1.5 z 1
In order to determine a unique inverse a region of convergence has to be specified. Note
that X(z) has two real and distinct poles at = 0.5 and = 1.5. These poles define
three different regions of convergence for X(z). These ROCs each give rise to a unique
inverse. The corresponding inverses will be i) causal, ii) two-sided, and ii) anti-causal.
 The z-transform 127

i) If we consider the causal inverse, i.e., ROC is | | > 1.5, then the above partial
fractions are recognized to have the following inverses:
2 Z
   2(0.5) n u (n) z-plane
1  0.5 z 1
2 Z
 2(1.5) n u (n)
1  1.5 z 1
Then
-1 III
x 1
x
0.5
x(n)   2(0.5) n u (n)  2(1.5) n u (n)
II

ii) Consider the anti-causal case, i.e., region III, I C

then
x( n)  2(0.5) n u ( n  1)  2(1.5) n u ( n  1)
iii) Consider region II, where x(n) is a two-sided sequence, i.e., x(n) has nonzero
components for both negative and positive n, we get
x( n)   2(0.5) n u (n)  2(1.5) n u ( n  1)
Note that this last case (ROC: region II) results in a stable but non-causal sequence. The
second case (ROC: region III) results in a non-causal and unstable sequence. Finally,
the first case (ROC: region I) results in a causal but unstable sequence. In simple terms
the stability of the sequences is checked by examining the sequences for boundedness for
all n.

Example 4.5.2: Consider the transform

1
X (z)  .
(1  z )(1  z 1 ) 2
1

Note that the pole at z = 1 is a multiple pole of multiplicity 2 and that at z = -1 is a

simple pole. Hence, we have to use the appropriate partial fraction expansion for X(z),
i.e.,
1 A1 C1 C2
X ( z)     .
(1  z )(1  z 1 ) 2
1
1 z 1
1 z 1
(1  z 1 ) 2
where A1 is readily obtained from
 C1 C2 
(1  z 1 ) X ( z )| z  1   A1  1
 1 2 
.
 1 z (1  z ) |
z  1

which gives
128 Introduction to Discrete-time Signal Processing

1 1
(1  z 1 ) X ( z )|  A1   .
z  1 (1  z 1 ) 2 |z  1 4

Similarly, C2 is obtained from

 (1  z 1 ) 2 
(1  z 1 ) 2 X ( z )|   1
A1  (1  z 1 )C1  C2  .
z 1
 1 z |z  1

which gives,
1 1
(1  z 1 ) 2 X ( z )|  C2  1
 .
z 1 1  z |z  1 2

Finally, C1 is obtained by differentiating with respect to , before evaluating at =

1, the following expression
 (1  z 1 ) 2 
(1  z 1 ) 2 X ( z )   1
A1  (1  z 1 )C1  C 2 
 1 z 
i.e., with the derivative of the first term removed, since it will eventually be zero when
evaluated at z = 1, we get
d d
dz 1

(1  z 1 ) 2 X ( z ) |z  1  
dz 1
 
(1  z 1 )C1  C 2 |z  1 .

i.e.,
d d 1
d z 1

(1  z 1 ) 2 X ( z ) |z  1  
d z 1  z 1 |z  1
1
  C1.

and
d 1 1
C1    .
d z 1  z 1 |z  1
1
4

Thus, the partial fraction expansion of X(z) becomes

1 1 1 1 1 1 1
X ( z)     .
(1  z )(1  z 1 ) 2
1
4 1 z 1
4 1 z 1
2 (1  z 1 ) 2
The inverse z-transform, assuming the inverse is a causal sequence, i.e., ROC: | | > 1,
becomes
1 1 1
x(n)  (1) n u (n)  u (n)  nu( n).
4 4 2
Note that X(z) has two possible regions of convergence and hence two inverses. The
student is advised to find the anti-causal inverse.
 The z-transform 129

We have so far considered cases where poles are distinct and real and repeated and real. Let us
now consider the case where all poles are distinct but some of them are complex. The complex
poles result in complex exponential sequences. However, for ( ) real ( ) will have real
coefficients and if there is a complex pole , there will also be a complex conjugate pole ∗ .
The corresponding coefficients in the partial fraction expansion are also complex conjugates.
Thus, the contribution of these complex conjugate poles to x(n) can be reduced to the real term,
i.e.,

( ) = [ + ∗ ( ∗) ] ( )
= | | | | +| | | | ( )
| || | ( ) ( )
= + ( )
= 2| || | cos( + ) ( )
,
if the ROC is | | > | | with  and  being the phase components of and .

Example 4.5.3: Consider the following z-transform

1
X ( z)  .
1
1  z  z 2
1

2
The partial fraction expansion of X(z) will have the form
A1 A2
X ( z)  1

1  p1 z (1  p2 z 1 ) 2
where
1 1 1 1
p1  j , p2  p1* and A1  j , A2  A1*
2 2 2 2
i.e.,
1 j 4 1  j 4 1 j 4 1  j 4
p1  e , p2  e and A1  e , A2  e .
2 2 2 2
Thus, the inverse, assuming x(n) is causal, is
n
 1   
x( n )  2  cos( n  )u (n).
 2 4 4

Example 4.5.4: Consider the transform

1  2 z 1  z 2 1  2 z 1  z 2
X ( z)  3 1
 1
, ROC : | z |  1.
1  2 z 1  2 z  2 (1  2 z 1 )(1  z 1 )
130 Introduction to Discrete-time Signal Processing

Since M = N = 2 and the poles are distinct, X(z) can be represented as

A1 A2  1  5 z 1
X ( z)  A0    2
1
1  2 z 1 1  z 1 1
(1  2 z 1 )(1  z 1 )

where A0 = 2, obtained by long division, and

 1  5 z 1
A1   9
1  z 1 | z 1  2
 1  5 z 1
A2   8
1  12 z 1 1
|z 1
Thus,
9 8
X ( z)  2  
1 z 1
2
1
1  z 1
and
n
1
x(n)  2 (n)  9  u (n)  8u (n).
2

4.5.2. Power Series Expansion

As noted earlier, the defining expression for the z-transform is a power series where the sequence
values x(n) are the coefficients of z - n. Thus, if the z-transform is given as a power series of the
form

n
X ( z)   x( n ) z
n  

   x( 2) z 2  x (1) z  x(0)  x (1) z 1  x (2) z  2  

we can determine any particular value of the sequence by finding the coefficient of the
appropriate power of z -1. We have already used this approach in finding the inverse transform of
the polynomial part of the partial fraction expansion when ≥ . This approach is also very
useful for finite-length sequences where X(z) may have no simpler form than a polynomial in
z -1, as we shall see in the case of finite impulse response digital filters.

Example 4.5.5: i) Consider the z-transform

X ( z)  z 2 (1  12 z 1 )(1  z 1 )(1  z 1 )
 z 2  12 z  1  12 z 1

which has an inverse z-transform

 The z-transform 131

x(n)   (n  2)  12  (n  1)   (n)  12  (n  1)
 1, n  2
 1
 , n   1
 2
  1, n  0
 1
 2, n 1
 0, otherwise

ii) Consider the z-transform
X ( z )  log(1  az 1 ), | z | | a | .

the power series expansion of the logarithmic function log(1 + ), with | x | < 1, is
1 2 1 3 1 4 (1) n 1 n
log(1  x )  x  x  x  x   x  , n  0
2 3 4 n

(1) n 1 n
  x
n 1 n
Using the above power series representation, we can write X(z) as

(1) n 1
X ( z )  log(1  az 1 )   (az 1 ) n
n 1 n
This expression has the form of the z-transform definition with the sequence x(n) being
(−1)
( )= , ≥1
0, ≤0
Note:- The inverse of X(z) may also be obtained using the differentiation property of the
z-transform. This will be left as an exercise to the student.

4.5.3. Inverse z-Transform Using Contour Integration

This is a formal method of obtaining the inverse z-transform. It is based on the Cauchy Integral
Theorem of complex theory. We will use this theorem to evaluate the inverse using contour
integration and residue theory.

The Cauchy integral theorem states that

1 1, k 1
 z k dz  
2 j c
0, k 1
132 Introduction to Discrete-time Signal Processing

where C is a counterclockwise contour in the z-plane that encircles the origin. Using the z-
transform definition

n
X (z)   x (n ) z
n  

Multiply both sides of this equation by z k -1 and integrate with a contour integral for which the
contour of integration C encloses the origin and lies entirely in the region of convergence of X(z),
i.e.,

1 1
 X ( z ) z k 1dz    x (n)z  n  k 1dz ,
2 j c 2 j n 
c

Assuming that the order of integration and summation can be interchanged (valid if the series is
convergent), we get

1 1
 X ( z ) z k 1dz   x(n) 2 j  z  n  k 1
dz ,
2 j c
n  
c

Using the Cauchy integral theorem, the contour integral on the right-hand side reduces to 1 if n-
k+1=1 , i.e., if n = k. Under this condition the infinite sum reduces to a single term (for n = k )
and the above equation becomes
1
 X ( z ) z k 1dz  x( k )
2 j c

Therefore, the formal inverse z-transform relation is given by the contour integral
1
x( n)   X ( z ) z n 1dz.
2 j c

Contour integrals of the above form are often conveniently evaluated using Cauchy’s residue
theorem, which when applied to the above, gives
1
x( n)  X ( z ) z n 1dz
2 j c
  [residues of X ( z ) z n 1 at the poles inside C ]
  n 1
  [residues of X ( z ) z at the poles outside C ]

This expression is valid for any proper z-transform X(z). For ( ) rational, it may be
expressed as
R( z )
X ( z ) z n 1  ,
( z  d0 )s
where ( ) has s poles at z = do and R(z) has no poles at z = do . The residue of
( ) at z = do is given by
 The z-transform 133

1  d s 1 
Residue of X ( z ) z n 1 [at z  d 0 ]   s 1 R( z )
( s  1)!  dz  zd 0

In particular, if there is only a first-order pole at z = do, i.e., s = 1, then

Residue of X ( z ) z n 1 [at z  d 0 ]  R( d 0 )
This is, in general, a powerful inverse z-transforming method and has applications in later
chapters.

Example 4.5.6: Let us use the contour integration method to find the inverse of
1
X (z)  1 1
, | z | 12 . z-plane
1 z 2
Using the inversion integral, we get
-1
x 1
n 1 0.5
1 1 z
x (n)   X ( z ) z n 1dz   dz
2 j c 2 j c 1  12 z 1
C
1 zn
 dz
2 j c z  12

where the contour of integration, C, is a circle of radius greater than 0.5, i.e., inside the
region of convergence. Note that
z n 1 zn
X ( z ) z n 1  1
 1
1  z 1
2
z2

and has a pole at z = 0.5. The contour of integration, C, will be a circle of radius greater
than 0.5. This pole will be encircled by the contour C. For ≥ 0, poles inside the
contour of integration contribute to x(n), i.e., the residue of the pole at z = 0.5 is
n
1
x( n)    , n  0.
2
For < 0, ( ) has a multiple-order pole at z = 0 whose order depends on n. For
n = -1, the pole is first order with a residue
1 1
n   1, Residue[ at z  0]    2
z ( z  0.5) 0.5
1 1
Residue[ at z  0.5]   2
z ( z  0.5) 0.5
134 Introduction to Discrete-time Signal Processing

1 1
n   2, Residue[ 2
at z  0]    4
z ( z  0.5) 0.25
1 1
Residue[ 2 at z  0.5]   4
z ( z  0.5) 0.5
etc.

Thus, x(-1) = -2+2 = 0 and x(-2) = -4+4 = 0, etc. We can safely generalize that x(n) is
zero for negative values of n.

4.6 Causality and stability

We have seen that the z-transform can be employed to characterize discrete-time signals and
systems. Two important characterizations of signals and systems are causality and stability. A
causal signal of the form
x ( n)  A1 p1n u (n)  A2 p2n u (n)  A3 p3n u ( n)  
will have z-transform
A1 A2 A3
X ( z)  1
 1
 
1  p1 z 1  p2 z 1  p3 z 1
with the conditions, for each of the terms, | z | > p1 , | z | > p2 , | z | > p3 , and so on,. This gives a
common region of convergence
| z |  max| p i |
i

that is, outside the circle defined by the pole of maximum magnitude. If the signal is completely
anticausal, i.e.,
x( n)   A1 p1n u ( n  1)  A2 p2n u ( n  1)  A3 p3n u ( n  1)  
with
A1 A2 A3
X ( z)  1
 1
 
1  p1 z 1  p2 z 1  p3 z 1
then the ROC conditions for each term will be | z | < p1 , | z | < p2 , | z | < p3, and so on. Thus the
common ROC in this case is
| z |  min| pi |
i

i.e., the inside of the circle defined by the pole of minimum magnitude.

In summary, causal signals are characterized by ROCs that are outside the maximum pole circle.
Anticausal signals have ROCs that are inside the minimum pole circle. Mixed signals have ROCs
 The z-transform 135

that are the annular region between two circles - with the poles that lie inside the inner circle
contributing causally and the poles that lie outside the outer circle contributing anticausally.

Stability can also be characterized in the z-domain in terms of the choice of the ROC. It can be
shown that a necessary and sufficient condition for stability of a signal x(n) is that the ROC of
the corresponding z-transform contain the unit circle. For a signal or system to be simultaneously
stable and causal it is necessary that all its poles lie strictly inside the unit circle.

An important class of signals is the so-called marginally-stable signals, which neither diverge nor
converge for large n. Rather, they remain bounded. The unit-step, alternating-step, and more
general sinusoidal signals fall in this class. Such signals have poles that lie on the unit circle.
Consider a complex sinusoid of frequency 0
x( n)  e j 0 n u (n), causal
j 0 n
x ( n)   e u ( n  1), anticausal

This is a special case of the exponential signal with = . Note that the plain unit-step u(n)
and alternating step (−1) ( ) are special cases of this with = 0 and = . The
corresponding z-transform is
1
X ( z) 
1  e j0 z 1
with the single pole on the unit circle and an ROC either | | > 1 for the causal case or
| | < 1 for the anticausal case.

The stability of a linear time-invariant system can also be expressed in terms of the system
function. A necessary and sufficient condition for a linear time-invariant system to be bounded-
input bounded-output (BIBO) stable is


| h(n) |
n  
 

where h(n) is the impulse response of the system. This implies that H(z) must contain the unit
circle. Hence, if the system is BIBO stable, the unit circle is contained in the ROC of H(z), and a
causal system is BIBO stable if and only if all the poles of H(z) are inside the unit circle.

4.7 Frequency Spectrum

In Chapter 1 we looked at the spectrum of sampled signals in terms of the signal sample values,
i.e., the discrete-time Fourier transform (DTFT). It was indicated that the DTFT is a special case
of the more general z-transform, with z values restricted on to the unit circle. The DTFT of a
discrete-time signal x(n) is defined as
136 Introduction to Discrete-time Signal Processing

 jn
X ( )   x (n )e
n  

This is recognized as the z-transform evaluated at z = e jω, i.e.,

 
 jn
X ( z)|
z  e j
  x( n) z  n
n  
  x ( n )e
n  
 X ( )
|
z  e j

The frequency response ( ) of a LTI system, defined by h(n) and transfer function H(z), is
defined in the same way, namely,

 jn
H ( )   h ( n )e
n  
 H ( z )|
z  e j

Recall that the Fourier spectrum of a sampled signal x(nT), given as

 
~
X(f )   x( nT )e  j 2fnT   x(nT )e  jn

n   n 

was a periodic replication of the original analog spectrum, at multiples of the sampling rate fs . In
units of , the period of the replicated spectrum becomes 2 . Thus, ( ) may be considered
over one period, say the Nyquist interval, − ≤ < .
The inverse DTFT recovers the discrete-time signal or sequence x(n) from its spectrum ( )
over the Nyquist interval, i.e.,
1 
j n
x( n ) 
2   X ( )e

d

4.8 The Unilateral z-Transform

The one-sided or unilateral z-transform of x(n) is defined as

X  ( z)   x(n) z n

n 0

Note that the lower limit of the sum is zero regardless of the values of x(n) for n < 0. The
unilateral z-transform is identical to the bilateral z-transform only if x(n) is causal. The ROC for
the case of the unilateral transform is always the exterior of some circle defined by the pole
furthest from the origin. As a result we don’t need to specify the ROC for unilateral transforms.

Example 4.8.1: Consider the unit impulse sequence

x( n )   (n)

The bilateral z-transform is

 The z-transform 137

 
X ( z)   x( n) z  n    (n ) z n
 1.
n   n 0

The unilateral z-transform is

 
X  ( z)   x( n) z  n    (n ) z n
 1.
n 0 n 0

This result is to be expected since x(n) is zero for n < 0.

Example 4.8.2: Consider the shifted impulse

x(n)   ( n  1)
Note that the two transforms are

n
X ( z)    (n  1) z
n  
 z.

X  ( z)    (n  1) z n
 0.
n 0

Note that almost all properties of the bilateral z-transform carry over to the unilateral z-transform
with the exception of the shifting property which becomes:
Z
Shifting Property: i) Time Delay: If ( )  ( ), then
k
Z
x(n  k )  z  k [ X  ( z )   x ( n ) z n ]
n 1
This may be verified using the definition for the unilateral z-transform.

Proof: Let y ( n )  x (n  k ) .
 
Y  ( z)   y ( n) z n   x( n  k ) z n

n 0 n 0

m
  x(m) z z k , with m  n  k
mk
1 
m k
  x( m) z z   x( m) z  m z  k
m  k m 0
k 
k
 z  x(n) z n
 z  k  x( m) z  m
n 1 m 0
i.e.,
k
( ) = z k [ x(n) z n  X  ( z )]
n 1
138 Introduction to Discrete-time Signal Processing

For x(n) causal, the unilateral and bilateral transforms are the same.


ii) Time Advance: Z
If x(n)  X+(z), then
k

Z
x(n  k )  z k [ X  ( z)   x( n) z n
], k 0
n 0

Show or prove the above property.


Final Value Theorem: Z
If x(n)  X+(z), then
lim ( ) = lim( − 1) ( )
→ →

The limit exists if the ROC of (z - 1)X+(z) includes the unit circle.

The principal use of the unilateral z-transform is in analyzing systems described by linear
constant-coefficient difference equations with non-initial rest conditions. In applying the
unilateral z-transform to problems of this type, the linearity property and the time-invariance
property are particularly important. The linearity property of the unilateral z-transform is
identical to the linearity property of the bilateral z-transform.

4.9 Solution of Difference Equations

We have seen the time domain solution of difference equations earlier in Chapter 1. Here we
employ the unilateral z-transform to solve difference equations. The unilateral z-transform is a
very efficient tool for solving difference equations with nonzero initial conditions. With the
unilateral transform, the difference equation is transformed into an equivalent algebraic equation.
This algebraic equation can be solved easily to obtain the transform of the desired solution. The
desired discrete-time signal is obtained by inverting the resulting z-transform.

Example 4.9.1: Consider a system for which the input and output satisfy the linear
constant-coefficient difference equation
1
y ( n)  y (n  1)  x( n)
2
The input for ≥ 0 is
x ( n )  1, n  0,
and the initial condition at n = -1 is y(-1) = 1.
 The z-transform 139

Applying the unilateral z-transform to the difference equation and using the linearity
property as well as the time-shifting property, we have
1 1
Y  ( z )  [ z 1Y  ( z )  y (1)]  X  ( z) 
2 1  z 1
From which it follows that
 1  1 1
Y  ( z )1  z 1   1
 y ( 1)
 2  1 z 2
1  1 1 
Y  ( z)   y (1)
1 1 1  z 1 2 
1 z
2
1 1 1
 1 1

2 1 2 z (1  z )(1  12 z 1 )
1

Performing a partial fraction expansion of the second term gives

1 1 1 2
Y  ( z)   
1 1
2 1 2 z 1 1
1 2 z 1  z 1
2 1 1
  .
1 z 1
2 1  12 z 1
Since the unilateral z-transform of the sequence
x ( n)  an , n  0,
is
1
X  (z)  ,
1  az 1
y(n) is seen to be
n
1 1
y (n)  2    , n  0.
2 2
140 Introduction to Discrete-time Signal Processing

Problems

4.1. Determine the z-transform of each of the following sequences. Include with your answer
the region of convergence in the z-plane and a sketch of the pole-zero plot. Express all
sums in closed form; can be complex.
(a) x a (n)  a n , 0  1

1, 0  n  N 1

(b) xb (n)  0, n N
0, n 0

 n, 0n N
 2 N  n, N 1  n  2N

(c) xc (n)  
 0, nN
 0, n0

4.2. Let ( ) denote a causal sequence; i.e., ( ) = 0, < 0. Furthermore, assume

that (0) ≠ 0.
(a) Show that there are no poles or zeros of X(z) at = ∞ i.e., that limit as z goes to
infinity of ( ) is nonzero and finite.
(b) Show that the number of poles in the finite z-plane equals the number of zeros in the
finite z-plane. The finite z-plane excludes = ∞.

4.3. Consider the z-transform X(z) whose pole-zero plot is shown.

z-plane

o x x x
-1 0 1 2 3

(a) Determine the region of convergence of X(z) if it is known that the Fourier transform
exists. For this case, determine whether the corresponding sequence is right-sided,
left-sided, or two-sided.
(b) How many possible two-sided sequences has the pole-zero plot shown?
(b) Is it possible for the pole-zero shown to be associated with the sequence that is both
stable and causal? If so, give the appropriate region of convergence.
 The z-transform 141

4.4. Determine the sequence x(n) with z-transform

X ( z )  (1  2 z )(1  3z 1 )(1  z 1 )

4.5. Determine the inverse z-transform of each of the following. In parts (a) - (c), use the
methods specified. In part (d), use any method you prefer.
1  13 z 1
(a) Long division: X ( z)  , x(n) is right  sided
1  13 z 1
3
(b) Partial fraction: X ( z)  1 1
, x (n) is stable
z   8 z 1
4
1
(c) Power series: X ( z )  ln(1  4 z ), | z| 4
1
(d) X ( z )  , | z |  (3) 1 / 3
1 1
1 z
3

4.6. Determine the inverse z-transform for each of the following using any method.
1
(a) X ( z )  , stable sequence
1
(1  z 1 )(1  2 z 1 )(1  3 z 1 )
3
1
(b) X ( z )  e z
c) X ( z )  sin( z ), ROC includes | z |  1
3 z 3
(d) X ( z )  , left  sided sequence
1 1 2
(1  z )
4

4.7. Determine the inverse z-transform of

1 1
X ( z)  log [2( 2  z )], | z |
2
a) by using the power series

xi
log (1  x)    , | x | 1
i 1 i
b) by first differentiating X(z) and then using this to recover x(n).

4.8. Let x(n) be a causal stable sequence with z-transform X(z). Define the complex cepstrum
xˆ ( n ) as the inverse transform of the logarithm of X(z), i.e.,
142 Introduction to Discrete-time Signal Processing

Xˆ ( z )  log X ( z ) Z
 xˆ(n)

where the ROC of Xˆ ( z ) includes the unit circle. (Strictly speaking the logarithm of a
complex number requires some careful considerations. Furthermore, the logarithm of a
valid z-transform may not be a valid z-transform. However, for now we will assume it is
valid). Determine the complex cepstrum for the following sequence.
x(n)   ( n)  a ( n  N ), | a | 1.

4.9. The pole-zero diagram shown corresponds to the z-transform of X(z) of a causal sequence
x(n). Sketch the pole-zero diagram of Y(z), where y(n) = x(-n+3). Also specify the region
of convergence for Y(z).
1 1
Poles : z  j ,
2 2
1 1
z  j ,
2 2
3
z   ,
4
Zeros : z  0.

4.10. The system function of a causal linear time-invariant system is

1  z 1
H (z) 
3
1  z 1
4
The input to the system is
x( n )    u (n)  u(n  1).
1 n
2
a) Find the impulse response of the system for all values of n.
b) Find the output y(n) for all values of n.
c) Is the system stable, i.e., is h(n) absolutely summable?

4.11. For each of the following difference equations and associated input and initial conditions,
determine the response y(n) by using the unilateral z-transform.
1
a) y (n)  3 y (n  1)  x (n), x (n)  ( ) n u (n), y (1)  1.
2
1 1
b) y (n)  y (n  1)  x(n)  x(n  1), x(n)  u (n), y (1)  0.
2 2
1 1
c) y (n)  y (n  1)  x (n)  x (n  1), x(n)  u (n), y (1)  1.
2 2
The z-transform 143