EENG 5610: Digital Signal Processing

Class 3: The z-Transform and Its Applications

Dr. Xinrong Li
Department of Electrical Engineering
University of North Texas

Outline
The z-Transform
Properties of z-Transform
Rational z-Transforms
The Inverse z-Transform
Analysis of LTI Systems in the z-Domain
The One-Sided z-Transform

Dr. Xinrong Li

EENG 5610, UNT

The Direct z-Transform

The z-transform of a discrete-time signal x(n)
The z-transform plays the same role in the analysis of discrete-time
signals and LTI systems as the Laplace transform does in the
n
analysis of continuous-time
signals
and
LTI
systems:
X
(
z
)

x
(
n
)
z

z
n
X(z) = Z {x(n)}, or x(n) X ( z )
z is a complex variable.

From the mathematical point of view, the z-transform is just an

alternative representation of a signal.

Region of convergence (ROC) of X(z)
The ROC of X(z) is the set of all values of z for which X(z) attains a

finite value. The z-transform only exists in the ROC.

Anytime we cite a z-transform we should also indicate its ROC.
The ROC of a finite-duration signal is the entire complex z-plane,
except possibly the points z = 0, and/or z = .
3

Dr. Xinrong Li

EENG 5610, UNT

 Example 3.1.1:
X 1 ( z ) 1 2 z 1 5 z 2 7 z 3 z 5 , ROC : entire z - plane except z 0.
X 2 ( z ) z 2 2 z 5 7 z 1 z 3 , ROC : entire z - plane except z 0 and z .
X 3 ( z ) Z { (n)} 1, ROC : entire z - plane.
X 4 ( z ) Z { (n k )} z k , k 0, ROC : entire z - plane except z 0.
X 5 ( z ) Z { (n k )} z k , k 0, ROC : entire z - plane except z .

Example 3.1.2:
1
x ( n ) ( ) n u ( n)
2

1
1
1
X ( z ) Z {x(n)} ( z 1 ) n
,
ROC
:
|
z
|

1 0.5 z 1
2
n 0 2
N 1

1 a
1
Sum
a nof
infinite, geometric series
a n (a very
if often
| a | 1used formula):
N

n 0

Dr. Xinrong Li

1 a

n0

1 a

EENG 5610, UNT

 The ROC is generally an annular region in the z-plane:

Define : z re j , then
X ( z)
X ( z)

x ( n) z

x(n)r n e jn

x(n)r n e jn

n jn

x ( n) r e

n
1

x(n)r

x ( n) r n
n 0

x ( n) r
n

n 1

n 0

x ( n)
rn

The problem of finding ROC

for X(z) is equivalent to
determining the range of
values for r for which the
sequence x(n)r-n is absolutely
summable.
5

x ( n) r n

Dr. Xinrong Li

If r2 > r1,
X(z) does not exist.

EENG 5610, UNT

 Example 3.1.3
x ( n) n u ( n )
X ( z)

x ( n) z

( z 1 ) n
n 0

1
, ROC : | z | | |
1 z 1

Example 3.1.4
x(n) nu (n 1)
X ( z)

x ( n) z

1 z

( 1 z ) n
n 1

1
1 1 z

1
, ROC : | z | | |
1 z 1

Dr. Xinrong Li

EENG 5610, UNT

 Two observations from the examples:

A discrete-time signal x(n) is uniquely determined by its z-transform

and the ROC of X(z).

The ROC of a causal signal is the exterior of a circle of some radius
r2, while the ROC of an anti-causal signal is the interior of a circle of
some radius r1.
If there is a ROC for an infinite-duration two-sided signal, it is a ring

(annular region) in the z-plane.

The ROC of a signal depends both on its duration (finite or infinite) and
on whether it is causal, anti-causal, or two sided.

The Inverse z-Transform

Inversion by contour integration:

x (n)

(based on Cauchys integral theorem)

1
n 1
X
(
z
)
z
dz
2 j C

The integral is taken over a closed contour C within the ROC of X(z),

enclosing the origin in a counterclockwise direction.

We will mainly deal with signals and systems in the z-domain which
have rational z-transforms (i.e., a ratio of two polynomials). We will
develop a much simpler method for inverse z-transform for such signals
& systems.
7

Dr. Xinrong Li

EENG 5610, UNT

Properties of the z-Transform

Linearity
If Z{x1(n)} = X1(z) and Z{x2(n)} = X2(z),

then Z{a1x1(n) + a2x2(n)} = a1X1(z) + a2X2(z).

The linearity property helps us to find the z-transform of a signal by

expressing the signal as a sum of elementary signals, for each of which,

the z-transform is already known.
When several z-transforms are combined, the ROC of the overall
transform is, at least, the intersection of the ROC of the individual ones.
If the linear combination of several signals has finite duration, the ROC
of its z-transform is exclusively dictated by the finite-duration nature of
this signal, not by the ROC of the individual transforms.

Time shifting
If Z{x(n)} = X(z), then Z{x(n-k)} = z-kX(z).
The ROC of z-kX(z) is the same as X(z) except for z = 0 if k > 0 and
8

the definition:
z = if k < 0. This is easily verified from

Dr. Xinrong Li

X ( z)

x(n) z EENG 5610, UNT

n

 Scaling in the z-domain

If Z{x(n)} = X(z), ROC: r1 < | z | < r2,

then Z{anx(n)} = X(a-1z), ROC: | a | r1 < | z | < | a | r2,

for any constant a, real or complex.

Time reversal
If Z{x(n)} = X(z), ROC: r1 < | z | < r2,

then Z{x(-n)} = X(z-1), ROC: 1/r2 < | z | < 1/r1,

Reflection in the time domain corresponds to inversion in the z-domain.

Differentiation in the z-domain

d X ( z)
If Z{x(n)} = X(z), then

dz

z 1Z {nx(n)}

Both transforms have the same ROC.

Convolution of two sequences

If Z{x1(n)} = X1(z) and Z{x2(n)} = X2(z),

then Z{x1(n) * x2(n)} = X1(z) X2(z).

The ROC is at least the intersection of that of X1(z) and X2(z).
This is one of the most powerful properties of the z-transform because it

Dr. Xinrong
canLibe

UNT
used to efficiently compute convolution ofEENG
two 5610,
signals.

x (n) * x (n) = Z-1{X (z) X (z)} = Z-1{Z{x (n)} Z{x (n)}}.

 Correlation of two sequences

rx1 x2 (l )

x ( n) x ( n l )

Rx1 x2 ( z ) X 1 ( z ) X 2 ( z 1 )

The ROC of Rx1x2(z) is at least the intersection of that of X1(z) and X2(z-1).
z
Multiplication of two
sequences
1

x(n) x1 (n) x2 (n) X ( z )

2 j C

z 1
v dv
v

X 1 (v ) X 2

In the integral, C is a closed counterclockwise contour, enclosing the

origin, in the *ROC common

to both
z
1 X1(v) and *X 2(1/v).
z * 1
x(n) x1 (n) x2 (n) X ( z )
X (v) X 2 * v dv
For complex-valued
sequences,2 j C 1
v

The initial value theorem

x(0) lim X ( z )
z
If the signal x(n)
is causal, (i.e., if x(n) = 0 for n < 0)

then

10

Dr. Xinrong Li

EENG 5610, UNT

 Some common z-transform pairs

11

Dr. Xinrong Li

EENG 5610, UNT

Rational z-Transforms
Rational z-Transforms, and poles and zeros
Rational z-transform, an important family of z-transforms, is a

rational function of two polynomials in z-1 (or z).

B ( z ) b0 b1 z 1 ... bM z M
X ( z)

A( z ) a0 a1 z 1 ... a N z N

b0 N M ( z z1 )( z z 2 )...( z z M )
z
a0
( z p1 )( z p2 )...( z p N )

(if a0 0 and b0 0)

G z N M

(z z
k 1
N

(z p )
l

l 1

X(z) has M finite zeros, N finite poles, and | N - M | zeros (if N > M) or
12

Dr. Xinrong
Li
poles

EENG 5610, UNT

(if N < M) at the origin z = 0.
If a polynomial has real coefficients, its roots are either real or occur in

 Pole-Zero Plot
X(z) can be represented graphically in the complex plane by showing

the location of the poles by crosses () and the zeros by circles (o).
The multiplicity of the poles and zeros is indicated by a number close to

the corresponding cross or circle.

Example 3.3.2

an ,

Determine the pole-zero plot of the signal (a > 0): x(n)

0,
Solution:

X ( z)

x( n) z

0 n M 1
elsewhere

1 (az 1 ) M
zM aM

M 1
1 az 1
z ( z a)

Since a 0, z M a M has M roots :

zk ae j 2k / M

0 k M 1.

The zero z0 a cancels the pole at z a.

Thus,
X ( z)

( z z1 )( z z 2 )...( z z M 1 )
z M 1

13

Dr. Xinrong Li

EENG 5610, UNT

 Magnitude response of z-transform as surface

Since z is a point in the complex plane, | X(z) | is a two-dimensional

function and describes a surface.

The surface plot of | X(z) | shows high peaks at the poles and deep valley

close to the zeros of the z-transform X(z).

Example:
X ( z)

14

Dr. Xinrong Li

( z 1)
( z 0.9e j / 4 )( z 0.9e j / 4 )

EENG 5610, UNT

Pole Location and Time-Domain

Behavior for Causal Signals
The time behavior of a signal depends strongly on the

location of its poles relative to the unit circle | z | = 1.

Here we deal exclusively with real, causal signals.

If a real signal has a z-transform

with one pole, this pole has to
be real. The only such signal is:
x ( n) a n u ( n)
1
, ROC : | z | | a |
1 az 1
z1 0, p1 a.
X ( z)

15

Dr. Xinrong Li

EENG 5610, UNT

 A causal signal with a double real pole has the form:

x(n) na nu ( n),
X ( z)

16

az
,
1 2
(1 az )

Dr. Xinrong Li

ROC : | z | | a |

z1 0, p1 p2 a.

EENG 5610, UNT

 A causal signal with a pair of complex-conjugate poles:

x(n) a n cos( w0 n)u (n),
1 az 1 cos w0
X ( z)
,
1 2az 1 cos w0 a 2 z 2
ROC : | z | | a |

17

Dr. Xinrong Li

EENG 5610, UNT

 Causal signal with a double pair of complex conjugate poles on the unit

circle:

Summary:
Causal signals with simple real poles or simple complex conjugate pairs of

poles that are inside or on the unit circle always have bounded amplitude.
A signal with a pole (or a complex conjugate pair of poles) near the origin
decays more rapidly than the one with a pole near (but inside) the unit
circle.
Properties of causal signals also applies to causal LTI systems, since their
impulse response is a causal signal.
If a pole of a system is outside the unit circle, the impulse response of the system

becomes unbounded and thus the system is unstable.

18

Dr. Xinrong Li

EENG 5610, UNT

The System Function of LTI System

Output of a relaxed LTI system: Y(z) = H(z)X(z), y(n) = x(n)*h(n).
H(z) is the z-transform of the unit sample response h(n).

System function:

H(z) = Y(z) / X(z),

H ( z ) n h(n) z n

LTI system described by linear constant-coefficient difference equation:

N
M
M
k
y ( n) k 1 ak y (n k ) k 0 bk x( n k )
b
z
Y ( z)

k
k 0

H ( z)
N
M
k
k
X ( z ) 1 N ak z k
Y ( z ) k 1 ak Y ( z ) z k 0 bk X ( z ) z
k 1

H ( z ) orkFIR
b z , (ak 0average
, 1 k (MA)
N ) system
All-zero
0 k or moving

H ( z)

b0
N

Pole-zero system

(bk 0, 1 k M )

k 1 a(belongs
All-pole 1system
kz
to IIR system)
19

Dr. Xinrong Li

EENG 5610, UNT

Inversion of the z-Transform

Three methods for the evaluation of inverse z-transform
The inverse z-transform by contour integration

x (n)

1
n 1
X
(
z
)
z
dz
2 j C

The integral is taken over a closed counterclockwise contour C in the

ROC of X(z), enclosing the origin. For simplicity, C can be taken as a

circle in the ROC.
The inverse z-transform by power series expansion
Expand X(z) into a series of terms, in the variables z and z-1.

The inverse z-transform by partial-fraction expansion

Factorize X(z) into a sum of simple terms and do table lookup.

20

Dr. Xinrong Li

EENG 5610, UNT

 The inverse z-transform by partial-fraction expansion

The basic idea of such method is to decompose a function X(z):
X ( z ) 1 X 1 ( z ) 2 X 2 ( z ) ... K X K ( z )
where xk(n) = Z-1{Xk(z)}, 1 k K, are available in the table of
transform pairs. Then, x(n) can be obtained as (linearity property):
x(n) 1 x1 (n) 2 x2 (n) ... K xK ( n)
This approach is particularly useful if X(z) if a rational function:

B( z ) b0 b1 z 1 ... bM z M
X ( z)

A( z ) 1 a1 z 1 ... a N z N

A rational function is called proper if aN 0 and M < N.

An improper rational function (M N) can always be written as the sum

of a polynomial and a proper rational function:

B ( z)
B( z )
X ( z)
c0 c1 z 1 ... cM N z ( M N ) 1
A( z )
A( z )
The inverse z-transform of the polynomial can be found by inspection.

An alternative form of a proper rational transform function X(z):

X ( z ) b0 z N 1 b1 z N 2 ... bM z N M 1

z
z N a1 z N 1 ... a N

21

OurLijob
Dr. Xinrong

is to perform partial-fraction expansion to

express
EENG
5610, this
UNT as a sum
of simple fractions. We will distinguish two cases.

 Case I in partial-fraction expansion: Distinct poles

Suppose the poles of X(z), p1, p2, , pN, are all different (distinct).

Then,
where

AN
X ( z)
A1
A2

...
z
z p1 z p2
z pN
( z pk ) X ( z )
Ak
, 1 k N
z
z pk

This expansion formula holds for both real and complex poles.
Complex conjugate poles result in complex-conjugate coefficients in

the partial-fraction expansion.

Case II in partial-fraction expansion: Multiple-order poles

If X(z) has an mth-order pole (z - pk)m, the partial-fraction expansion
A1k
A2 k
Amk

must contain the terms: z p ( z p ) 2 ... ( z p ) m

k
k
k
1 d m i ( z pk ) m X ( z )
where Aik

(m i )! dz
z

, 1 i m
z pk

Once we have performed partial-fraction expansion, we are

ready to perform the inversion of the z-transform X(z).

22

Dr. Xinrong Li

EENG 5610, UNT

 Case I in inverse z-transform: Distinct poles

With the partial-fraction expansion, we can have

X ( z ) A1

1
1
1

...

A
2
N
1 p1 z 1
1 p 2 z 1
1 p N z 1

The inverse z-transform can be obtained by inverting each of the

n
terms
and
taking
corresponding
linear
combination:

(
p
)
u ( n),
if ROC
: | z | | pk | (causal signals)

k
1

1 pk z

( pk ) n u (n 1), if ROC : | z | | pk | (anti - causal signals)

If the signal x(n) is causal, the ROC is | z | > pmax.

Suppose all poles are distinct but some of them are complex. If x(n)

is real,
the polynomials
real coefficients and complex
A
Ak* in X(z) have
1
n
k
Z will appear

| rk cos( k npair.
k )uThen,
( n), ROC : | z | | pk | rk ,
2 | Ak conjugate
poles
in complex
1 p z 1 1 p * z 1
k
k

where Ak | Ak | e j k , pk rk e j k .

23

pz
Dr.
Z
np nu (nMultiple
), ROC : | poles
z | |EENG
p | . 5610, UNT

Xinrong
Case
II Liin inverse
z-transform:
1 2
(
1

pz
)

Use multiple differentiation formula. For example,

1

 Example 3.4.9

1 z 1
Determine the causal signal x(n), for which X ( z )
1 z 1 0.5 z 2
Solution:

1 z 1
A1
A2
X ( z)

1 z 1 0.5 z 2 1 p1 z 1 1 p2 z 1

10 j 71.565
1 j / 4
e
, p1 p2* 0.5 j 0.5
e .
2
2
1

Hence, x(n) 10 ( ) n cos( n 71.565)u (n).

4
2
where A1 A2* 0.5 j1.5

Example 3.4.10

1
Determine the causal signal x(n), for which X ( z )
(1 z 1 )(1 z 1 ) 2
Solution:
1 1
3 1
1 z 1
X ( z)

4 1 z 1 4 1 z 1 2 (1 z 1 ) 2
3 n
1
(1) n u (n).
4 2
4

Hence, x(n)

24

Dr. Xinrong Li

EENG 5610, UNT

Analysis of LTI Systems in the z-Domain

Response of Systems with Rational System Functions
Systems and signals as a ratio of two polynomials (if the system is

initially relaxed, i.e., y(k) = 0, k < 0, or zero-state response):

H ( z)

B( z )
,
A( z )

X ( z)

N ( z)
Q( z )

Y ( z) H ( z) X ( z)

B( z ) N ( z )
A( z )Q ( z )

Assume H(z) contains simple poles pi, 1 i N, and X(z) contains

simple poles qj, 1 j L, where pi qj. In addition, assume there is

N
L
N
L
Q j Then:
Ai
n
noYpole-zero
cancellation.
( z)

, y (n) Ai ( pi ) u (n) Q j (q j ) n u (n)

i 1 1

pi z 1

j 1 1

q j z 1

i 1

j 1

The first part of y(n), ynr(n), is called the natural response of the system.
The 2nd part of y(n), yfr(n), is called forced response of the system. The

influence of the system on this response is through the scale factors Qj.
25

TheLiscale
Dr. Xinrong

factors Ai and Qj are functions of bothEENG

sets of
poles.
5610, UNT

When X(z) and H(z) have one or more poles in common, the Y(z) will

 Transient and Steady-State Response

If the poles of the system H(z), | pk | < 1 for all k, then ynr(n) decays

to zero as n tends to infinity. In such a case, the natural response of

the system is referred to as the transient response.
The rate at which ynr(n) decays toward zero depends on the magnitude of

the pole positions. If all poles have small magnitude, the decay is rapid.
If one or more poles are located near the unit circle, the corresponding
terms in ynr(n) will decay slowly toward zero.
If the poles of the input X(z), | qk | < 1 for all k, then yfr(n) will decay

toward zero as n tends to infinity.

When the causal input is a sinusoid, the poles fall on the unit circle and

the forced response is also a sinusoid that persists for all n 0. In this
case, the forced response is called the steady-state response of the
system.
For the system to sustain a steady-state output for n 0, the input signal
must persist for all n 0.

26

Dr. Xinrong Li

EENG 5610, UNT

 Causality and Stability

An LTI system is causal if and only if the ROC of the system H(z) is
the exterior of a circle of radius r < , including the point z = .
A causal LTI system is the one whose unit sample response h(n) satisfies

the condition h(n) = 0, n < 0.

An LTI system is BIBO stable if and only if the ROC of the system

function H(z) includes the unit circle.

A necessary and sufficient condition for an LTI system to be BIBO stable

is: n | h(n) |
This condition implies that the system function H(z) must contain the
unit circle within its ROC (the converse is also true) because:

| H ( z ) | n | h(n) z n | n | h(n) | | z n | ,

when evaluate on the unit circle (i.e., | z | 1) : | H ( z ) | n | h(n) | .

The conditions for causality and stability are different and that one

does not imply the other.

Causal system can be stable or unstable, and vice versa.

However, for a causal system, the condition on stability can be

narrowed to some extent.

27

A causal
Dr. Xinrong
Li

LTI system is BIBO stable if and only ifEENG

all the
poles of H(z) are
5610, UNT
inside the unit circle.

 Example 3.5.2
Determine the ROC and h(n) for the LTI system:

3 4 z 1
1
2
H ( z)

1 3.5 z 1 1.5 z 2 1 0.5 z 1 1 3z 1

The system has two poles: z1 = 0.5, z2 = 3.

(a) If the system is stable:
Since the system is stable,
h(n) its
(0ROC
.5) n u (must
n) 2include
(3) n u (nthe
1)unit circle and hence the ROC
is 0.5 < | z | < 3, and
(b) If the system is causal:
h(n)
0.5) n u ( nits) ROC
2(3) nis
u (|nz) | > 3 (the system is unstable)
Since the system
is (causal,

and hence:
(c) If the system is anti-causal:
h(n) [(0.5) n 2(3) n ]u (n 1)
If the system is anti-causal, its ROC is | z | < 0.5 (the system is unstable)
and hence:

28

Dr. Xinrong Li

EENG 5610, UNT

 Pole-Zero Cancellation
When a z-transform has a pole that is at the same location as the

zero, the pole is cancelled by the zero and the term containing the
pole in the inverse z-transform vanishes.
The order of the system is reduced by one if pole-zero cancellation

occurs in the system function itself.

The pole of the system is suppressed by the zero in the input signal (or
vice versa) if pole-zero cancellation happens in the product of system
function and the input signal.
Thus, it is possible to suppress one or more modes in the system

response by carefully selecting poles and zeros in the input and the
system function.
When the zero is located very close to the pole but not exactly at the

same location, the term in the response has a very small amplitude.
Non-exact pole-zero cancellation can happen in practice due to
insufficient numerical precision used in representing system coefficients.
Thus, one should not attempt to stabilize an inherently unstable system
by placing a zero in the input signal at the location of the pole.

29

Dr. Xinrong Li

EENG 5610, UNT

The One-Sided/Unilateral z-Transform

Why one-sided z-transform?
The (two-sided) z-transform is not useful for the non-relaxed
systems (i.e., systems with initial values) since it requires that the
signal be specified for the entire time range -< n < .
The one-sided z-transform is developed to solve such problems, i.e.,
the difference equations with initial conditions.
Definition of the one-sided z-transform:

X ( z ) Z {x(n)} x(n) z n Z {x(n)u (n)}

n 0

X+(z) does not contain information about x(n) for n < 0.

X+(z) is unique only for causal signals, because only causal signals

are zero for n < 0.

The ROC of X+(z) is always the exterior of a circle. When dealing
with one-sided z-transforms, its not necessary to refer to their ROC.
30

Dr. Xinrong Li

EENG 5610, UNT

 Properties of one-sided z-transform

Almost all properties for the two-sided z-transform carry over to the

one-sided z-transform, except the shifting property.

Shifting property
Case 1 time delay:

If Z {x(n)} X ( z ), then Z {x(n k )} z X ( z ) x(n) z n ,

n 1

The
shifting property can be intuitively explained:
Z {x(n k )} [ x(k ) x(k 1) z 1 ... x(1) z k 1 z k X ( z )],

If Z2 {xtime
(n)} advance:
X ( z ), then Z {x(n k )} z X ( z ) x(n) z n ,
Case
n0

k 1

for k 0.

for k 0.

for k 0.

If Z {x(n)} X ( z ), then lim x(n) lim ( z 1) X ( z )

Final value theorem:

31

TheLilimit
Dr. Xinrong

z 1

exists if the ROC of (z - 1)X+(z) includes

the5610,
unit UNT
circle.
EENG

 Solution of difference equations

The one-sided z-transform is a very efficient tool for the solution of

difference equation with nonzero initial conditions.

Example 3.6.5:
y(n) is the nth term of the Fibonacci sequence (initial condition y(-1) = 0,

y(-2) = 1): y(n) = y(n - 1) + y(n - 2). Determine a close-form expression

of y(n).
Solution:
y (n) y (n 1) y (n 2),

Y ( z ) z 1Y ( z ) y (1) z 2Y ( z ) y (2) y (1) z 1 .

Thus, Y ( z )

1
1

1 z z

1 5 1 5

y ( n)

2
2 5

32

Dr. Xinrong Li

1 5
, A1, 2 p1, 2 / 5 (partial - fraction expansion) ,
2
n
1 5
u (n).
2

p1, 2

1 5

2 5

EENG 5610, UNT