Chapter 4

z-Transform

4.1 z-Transform

The z-transform is a powerful tool to deal with discrete-time signals, both from a conceptual
and practical standpoint. It converts a discrete-time domain signal into a discrete-frequency
domain signal in the complex plane. In practice, it is widely used to analyze and process digital
data such as JPEG images, MP3 and MP4 songs, and ZIP files. Specifically, in DSP it is
applied to system design, analysis of the stability of a system, finding the frequency response,
designing digital filters, or analysis of linear discrete system. The main use we will be making
of the z-transform here is solving linear di↵erence equations associated to linear time-invariant
discrete-time systems (LTIDT systems). The z-transform provides a tractable way to solve
those linear di↵erence equations. Conceptually speaking, the z-transform is a useful tool to
detect the presence of exponentially increasing or decreasing oscillations in an input signal.
Furthermore, it is possible to characterize a system from an impulse response by computing the
z-transform. In particular, it provides a way to identify whether a system is stable or instable
(see Definition 3.4.8). The z-transform can also detect the presence of signals that increase
or decrease over time. Finally, it is an e↵ective tool to determine the frequency response of a
system.
Formally speaking, the z-transform is a weighted infinite series of the input signal and the
geometric series z n (a formal definition will arrive soon). It converts the di↵erence equations
associated to a LTIDT system into a rational function (a ratio of polynomials) in the complex
domain. By studying the properties of that rational function is possible to find the solution
to the di↵erence equation. Moreover, the inverse of the z-transform, under certain conditions,
exists, and then it is possible to go back and forth from the time domain to the frequency

49
50 z-Transform

domain. When we take a signal and change from the time domain to the frequency domain we
gain insights as to the characteristics of that signal. Figure 6.1 shows a diagram illustrating
the way of operating of the z-transform when applied to solving di↵erence equations.

Figure 4.1: Using the z-transform to solve di↵erence equations

Review 4.1.1 Review (again) Section 7.3.2 on numerical series, in particular, convergence of
geometric series, the necessary condition for convergence of series, and the ratio test.

Definition 4.1.1 The z-Transform. The z-transform of an arbitrary discrete time signal
x(n) is defined as
1
X
X(z) = Z[x(n)] = x(n)z n (4.1)
n= 1

where z is a complex variable. The z-transform is conceived as an operator; hence, the notation
Z[x(n)]. If x(n) is right-sided or left-sided, the z-transform is said to be unilateral. A usual
z-transform is called bilateral.

Problem 4.1.2 Use the ratio test for convergence of series to determine where Equation 4.1
makes sense. Notice that X(z) can be written as
1
X 0
X 1
X
n n n
X(z) = x(n)z = x(n)z + x(n)z
n= 1 n= 1 n=1

Definition 4.1.3 Region of convergence The region of the complex plane where X(z) exists
is called the region of convergence (ROC, for short). In general, the ROC will have the form
of a ring r1 < |z| < r2 in the complex plane, where r1 , r2 0 (see Problem 4.1.2).

Exercise 4.1.4 Find the z-transform of the unit sample sequence and the unit step sequence.
Find their ROC too.

Exercise 4.1.5 Find the z-transform and specify their ROC for the following sequences.

(1) The right-sided sequence x(n) = an · u(n), where a 2 C.

4.2. Properties of the z-Transform 51

(2) The finite sequence given by the table below.

n -2 1 0 1 2
x(n) 2 3 -1 2 1

(3) x(n) = 2 (n + 2) + (n + 1) + 3 (n) (n 1) + (n 2).

(4) x(n) where 8 ✓ ◆n

>
> 1
< if n 0
x(n) = ✓5 ◆n
>
> 1
: if n < 0
3

Problem 4.1.6 Write the z-transform when z is given in exponential form z = r · ei✓ .

Exercise 4.1.7 Given x(n), n 2 Z, find the z-transform of the time shifting sequence x(n
k), k 2 Z.

4.2 Properties of the z-Transform

Theorem 4.2.1 Properties of the z-transform. Let x1 (n) and x2 (n) be two sequences
whose z-transforms have ROC R1 = {r1 < |z| < s1 } and R2 = {r2 < |z| < s2 }, respectively.
Further assume that R1 \ R2 is not the empty set. Then the z-transform holds the following
properties.

(1) Linearity. Let a1 , a2 be two arbitrary complex numbers. The equality

Z[a1 x1 (n) + a2 x2 (n)] = a1 Z[x1 (n)] + a2 Z[x2 (n)] (4.2)

holds on R1 \ R2 = {max(r1 , r2 ) < |z| < min(s1 , s2 )}.

(2) Time reversal. The z transform of x1 ( n) is

Z[x1 ( n)] = X1 (z 1 ) (4.3)

⇢
1 1 1
with ROC = < |z| < .
R1 s1 r1
(3) Time shifting. The z transform of x1 (n k) for k 2 Z is
k
Z[x1 (n k)] = z · Z[x1 (n)] (4.4)

with the same ROC as x1 (n) except for z = 0 if k > 0 and z = 1 if k < 0.

(4) Scaling in the z-domain. Assume now that a1 6= 0. The z transform of an1 x1 (n) is
✓ ◆
n z
Z[a1 x1 (n)] = X1 (4.5)
a1

with ROC |a1 |R = {|a1 |r1 < |z| < |a1 |s1 }
52 z-Transform

(5) Di↵erentiation in the z-domain. The derivative of X1 (z) is given by the formula

dX1 (z)
Z[nx1 (n)] = z· (4.6)
dz
with the same ROC as x1 (n).

(6) Convolution. The convolution of two sequences is given by

Z[x1 (n) ⇤ x2 (n)] = X1 (z) · X2 (z) (4.7)

where the formula is valid on R1 \ R2 = {max(r1 , r2 ) < |z| < min(s1 , s2 )}.

(7) Cross-correlation. The cross-correlation of two sequences is given by

Z[rx1 x2 (n)] = X1 (z) · X2 (z 1 ) (4.8)

with ROC R1 \ (1/R2 ) = {max(r1 , 1/r2 ) < |z| < min(s1 , 1/s2 )}.

(8) Conjugate of a complex sequence

Z[x1 (n)] = X 1 (z) (4.9)

with the same ROC as x1 (n).

Problem 4.2.2 From Theorem 4.2, prove statements (2), (4), (5), (6), and (7).

Table 4.1 on next page summarizes the results in Theorem 4.2.

Exercise 4.2.3 Referring to Program 3.7.3, define in MATLAB the signals x1 (n) =
[1, 1, 0, 1, 1, 0, 1, 1] and x2 (n) = [ 1, 2, 1, 0, 1/2, 1/2, 1, 1] on n = [ 3, 4], and signal
x3 (n) = [1, 2, 3, 4] on n = [ 1, 2]. Compute the z-transform of the sequences below. Compute
also the energy of x1 (n) and x1 (n) ⇤ x3 (n).
1 Pn
(1) n · x1 (n) · x2 (n) (5) the accumulator system k= 1 x3 (k)
n
(2) x (n) delayed by 5 samples (6) the moving average for M1 = M2 = 1
2

(3) the time-reversed sequence of x1 (n) (7) the cross-correlation of x1 (n) and x2 (n)

(4) the compressor x2 (2n) system (8) the autocorrelation of x1 (n)

4.2. Properties of the z-Transform 53

Property Sequence z-transform ROC

Linearity a1 x1 (n) + a2 x2 (n) a1 Z[x1 (n)] + a2 Z[x2 (n)] R1 \ R2

Time reversal x1 ( n) X1 (z 1 ) 1/R1

k
Time shifting x1 (n k) z · Z[x1 (n)] with the same ROC
as x1 (n) except
for z = 0 if k > 0
and z = 1 if k < 0
✓ ◆
z
Scaling in an1 x1 (n) X1 |a1 |R1
a1
z-domain
dX1 (z)
Di↵erentiation nx1 (n) z· R1
dz
in the z-domain
Convolution x1 (n) ⇤ x2 (n) X1 (z) · X2 (z) R1 \ R2
Correlation rx1 x2 (n) X1 (z) · X2 (z 1 ) R1 \ (1/R2 )

Conjugate of a x1 (n) X 1 (z) R1

complex sequence
Table 4.1: Properties of the z-transform

Problem 4.2.4 Let x(n) be a causal system. Prove that x(0) = limz!1 Z[x(n)].

(1) x(n) = u( n 1)

(2) x(n) = nu(n)

(3) x(n) = nu(n) if n > 0 and 0 elsewhere

(4) x(n) = sin(an), where a 2 R

(5) x(n) = u(n) u(n 5)

Theorem 4.2.5 Let x(n) be a finite signal defined on n = [0, N ].

(1) Delay property #1. Let a1 , a2 be two arbitrary complex numbers. Assume that R1 \ R2
is not the empty set. The equality
k
Z[y(n k)u(n k)] = z Y (z) (4.10)

(2) Representation property. Assume that value y[ 1], y[ 2], . . . , y[ k] are defined and
known for some integer k > 0. Then, y(n k) can be written as
k
X
y(n k) = y(n k)u(n k) + y( m) (n m + k) (4.11)
m=1
54 z-Transform

(3) Delay property #2. Assume that values y[ 1], y[ 2], . . . , y[ k] are defined and known
for some integer k > 0. Then

k
!
X
k
Z[y(n k)] = z Y (z) + y( m)z m (4.12)
m=1

4.3 Solution of Di↵erence Equations by Using the

z-Transform

In Section 3.6.2 we examined higher-order linear constant-coefficient di↵erence equations. In

these types of equations y(n) depended solely on past samples of the dependent variable y(n
1), . . . , y(n k), for some integer k > 0. Thanks to the z-transform we will now be able to solve
more complex di↵erence equations where y(n) will depend not only on y(n 1), . . . , y(n k) but
also on past samples of the input signal x(n), . . . , x(n m), for some positive number m > 0.
For the rest of this section, we will assume that the system is causal and the input signal x(n)
is of finite duration, say defined on n = [N1 , N2 ], N1 < 0 < N2 . The type of di↵erence equation
we are trying to solve is

N2
X nXN1
y(n) = bm x(n m) am y(n m) (4.13)
m=N1 m=1

The system is causal on the dependent variable as y(n) depends on y(N1 ), y(N1 +1), . . . , y(n 1).
However, it is not causal on the independent variable. This kind of equation is more general
than the ones studied in Section 3.6.2 (which were recursive, homogeneous LCCDE). We start
o↵ by taking the z-transform in the previous equation.

" N2 nXN1
#
X
Z[y(n)] = Y (z) =Z bm x(n m) am y(n m)
m=N1 m=1
N2
X nXN1
= bm Z[x(n m)] am Z[y(n m)] (by linearity)
m=N1 m=1
N2
X nXN1
m m
= bm z Z[x(n)] am z Z[y(n)] (time shifting)
m=N1 m=1
N2
X nXN1
m m
=X(z) bm z Y (z) am z
m=N1 m=1
4.3. Solution of Di↵erence Equations by Using the z-Transform 55

Let’s now work Y (z) out.

N2
X nXN1
m m
Y (z) =X(z) bm z Y (z) am z
m=N1 m=1
nXN1 N2
X
m m
Y (z) + Y (z) am z = X(z) bm z
m=1 m=N1
nXN1
! N2
X
m m
Y (z) 1 + am z = X(z) bm z
m=1 m=N1
P N2 m
m=N1 bm z
Y (z) = X(z) · Pn N1
1+ m
m=1 am z

Notice that the term multiplying X(z) in the left-hand side of the last equation is a complex
rational function, that is, the ratio of two polynomials in z. That function is known by the
name of transfer function H(z).
P N2 m
m=N1 bm z bN1 z N1 + bN1 +1 z (N1 +1) + . . . + bN2 z N2
H(z) = P N1 = (4.14)
1 + nm=1 am z m 1 + a1 z 1 + . . . + an N1 z (n N1 )
Finally, we have the nice equation
Y (z) = X(z) · H(z) (4.15)
This equation will be key to solving the di↵erence equation as we will see later on in this
chapter.
Returning to Equation 4.14, the expression of H(z), let’s do some algebra manipulations.
We start by multiplying the numerator and denominator by z N2 .
N1
bN1 z + bN1 +1 z (N1 +1) + . . . + bN2 z N2
H(z) =
1 + a1 z 1 + . . . + an N1 z (n N1 )
z N2 bN z N1 + bN1 +1 z (N1 +1) + . . . + bN2 z N2
= N2 · 1
z 1 + a1 z 1 + . . . + an N1 z (n N1 )
bN z N2 N1 + bN1 +1 z N2 (N1 +1) + . . . + bN2 z N2 N2 PH (z)
= 1 N2 N 1 N (n N )
=
z + a1 z 2 + . . . + an N 1 z 2 1 QH (z)
where polynomials PH (z) and QH (z) are defined as
PH (z) = bN1 z N2 N1
+ bN1 +1 z N2 (N1 +1)
+ . . . + bN2 z N2 N2

QH (z) = z N2 + a1 z N2 1
+ . . . + an N1 z
N2 (n N1 )

By the fundamental theorem of algebra, every n-th order polynomial can be factored into a
product of n first-order polynomials. Therefore, H(z) can be written in factored form as
(z pN1 )(z pN1 +1 ) · . . . · (z pN2 )
H(z) = bN1 (4.16)
(z qN2 )(z qN2 1 ) · . . . · (z qN2 (n N1 ) )
The numerator roots {pN1 , . . . , qN2 } are called the zeros of the transfer function, and the
denominator roots {qN2 , . . . , pN2 (n N1 ) } are called the poles of the filter (or just poles for
short). The poles are called simple if all qi are di↵erent; otherwise, they are said to be multiple
poles.
56 z-Transform

Exercise 4.3.1 Identify the zeros and the poles of the following transfer functions.
z2 + 1 z 2 + 2z 15
(1) H(z) = (2) H(z) =
z 2 3z + 1 (z 2 2z + 1)(z 2 + 2z + 5)
Problem 4.3.2 Properties of the ROC. Prove the following properties about the ROC.

(1) ROC does not contain poles.

(2) The ROC for a finite duration causal sequence is the entire z-plane except for z = 0.

(3) The ROC for a non-causal finite duration sequence is the entire z-plane except for z = 1.

(4) The ROC for a finite duration two-sided sequence is the entire z-plane except for z = 0
and z = 1.

(5) The ROC for an infinite duration right-sided sequence is the exterior of a circle which
may or may not include z = 1.

(6) The ROC for an infinite duration left-sided sequence is the interior of a circle which may
or may not include z = 1.

(7) The ROC of an infinite duration two-sided sequence is a ring in the z-plane.

4.4 The Inverse z-Transform

Definition 4.4.1 Inverse of the z-transform. Given a complex function X(z), its inverse
z-transform is a sequence x(n) such that Z[x(n)] = X(z). We also notate it by writing x(n) =
Z 1 [X(z)].

Computing the inverse of a z-transform is, in general, a hard problem. However, for certain
important cases there are specific methods. The three more common methods are: Cauchy’s
residue theorem, partial fraction expansion, and power series expansion. Cauchy’s residue
theorem is beyond the scope of this course. We will mainly focus on partial fraction expansion
and to a lesser extent on power series expansion.

4.4.1 Partial Fraction Expansion Method

Assume that the transfer function is the ratio of two polynomials,

(z p1 )(z p2 ) · . . . · (z pN ) PH (z)
H(z) = b1 =
(z q1 )(z q2 ) · . . . · (z qM ) QH (z)

If the degree of the numerator is greater or equal than that of denominator, the fraction is
said to be an improper fraction. When N < M , then the fraction is a proper fraction. If
H(z) is not a proper fraction, it is enough to carry out the division between PH (z) and QH (z).

PH (z) P1 (z)
H(z) = = C(z) +
QH (z) Q(z)
4.4. The Inverse z-Transform 57

P1 (z)
where P1 (z) is a polynomial of degree less than N , which implies that is already a proper
Q(z)
fraction. Notice that C(z) has no poles as all its powers on z are positive. The partial fraction
expansion of a proper rational fraction is an expression that can take two forms, depending on
whether the fraction has simple poles or multiple poles. If all poles are unique, the expansion
is
XM
P1 (z) ck
H(z) = C(z) + = C(z) + (4.17)
Q(z) k=1
z qk
where ck are complex constants. These constants can be determined by using the following
formula
ck = (z qk ) · H(z)|z=qk (4.18)
that is, by evaluating at z = qk the expression (z qk )·H(z) (this is the meaning of the notation
|z=qk ).
Before carrying on with te second form of the partial fraction expansion, let’s give an
example of the first form.
Example. Let X(z) be the z-transform
z
X(z) =
(z a)(z b)

where a, b 2 C and a 6= b and a, b 2 ROC. It is proper rational fraction with two distinct poles,
and therefore can be written as
z c1 c2
= +
(z a)(z b) z a z b
z a
To obtain c1 , we evaluate (z a)X(z) = at z = a. That returns . Analogously, c2
z b a b
z b
is computed from the function (z b)X(z) = evaluated at z = b, which yields .
z a b a
Plugging the constants into the equality, we obtain
z a 1 b 1
= · + ·
(z a)(z b) a b z a b a z b

If we want to find the inverse z-transform, how will we proceed? From Exercise 4.1.5, we know
1
that the z-transform of right-sided sequence x(n) = an · u(n), n 2 N is Z[x(n)] = . The
1 az 1
following algebraic manipulations shows how to obtain the inverse of X(z) in terms of x(n).
1
Let’s take the term ; computing the inverse of the other term is similar and will leave it
z a
to the reader.
1 1 1 z 1 1
= 1
= = =z · Z[x(n)]
z a z a·z·z z(1 a · z 1) 1 a·z 1

1
By the time shifting property (see Table 4.1), z ·Z[x(n)] is just x(n 1), that is, the right-sided
sequence x(n 1) = an 1 · u(n 1).
58 z-Transform

Let’s now describe the second form of the partial fraction expansion, where the poles are
not unique. Assume for the sake of simplicity that there is just one pole q1 whose multiplicity
is r whereas the rest of the poles are unique. The expansion is written as

X r XM
P1 (z) ck ck
H(z) = C(z) + = C(z) + r
+ (4.19)
Q(z) (z q1 ) z qk
|k=1 {z } k=r+1
| {z }
multiple poles distinct poles

where constants ck are given by

• for constants c1 , . . . , cr ,

1 dr k
ck = ((z q1 )r · H(z)) |z=qk , k = 1, . . . , r (4.20)
(r k)! dr k z

dr k
The expression means the (r k)-th derivative.
dr k z
• for the remaining constants cr+1 . . . , cM , it still holds ck = (z qk ) · H(z)|z=qk .
Pr ck
If there are more poles with multiplicities, similar terms to k=1 have to be inserted
(z q1 )r
in Equation 4.19.
Example. Let’s illustrate the above description with an example. Let X(z) be the following
z-transform
z(z + 1)
X(z) =
(z 1)3
When decomposed in partial fractions, we write

z(z + 1) c1 c2 c3
X(z) = 3
= + 2
+
(z 1) z 1 (z 1) (z 1)3

To determine constants c1 , c2 , c3 , we need to find the following derivatives.

d2 3 00
For c1 : (z 1) · X(z) = (z(z + 1)) =2
d2 z
d 0
For c2 : (z 1)3 · X(z) = (z(z + 1)) = 2z + 1
dz
For c3 : (z 1)3 · X(z) = z(z + 1)

We finally find c1 , c2 , c3 by evaluating the above derivatives at z = 1 and apply Equation 4.20
1 1 1
c1 = · 2 = 1, c2 = · 3 = 3 , c3 = · 2 = 2
2! 1! 0!
Therefore,
z(z + 1) 1 3 2
X(z) = 3
= + 2
+
(z 1) z 1 (z 1) (z 1)3
To determine the inverse z-transform, we need to find the inverse of the z-transform of the
1 1 1
terms , 2
, and . We will do it separately
z 1 (z 1) (z 1)3
4.4. The Inverse z-Transform 59

1
• Inverse z-transform of . In Exercise 4.1.5, the reader was asked to find the z-
z 1
z
transform of the right-sided sequence x(n) = an ·u(n). The answer was Z[an u(n)] =
z a
1
for |z| > |a|. In order to obtain , we set a = 1 and compute the z-transform of the
z 1
time-shift u(n 1) (assuming |z| > 1).
1
X 1
X 1
n n z 1
Z[u(n 1)] = u(n 1)z = z = 1
=
n=0 n=1
1 z z 1


1 1
Therefore, Z is x(n) = u(n 1).
z 1
1 z
• Inverse z-transform of 2
. We know that the z-transform of
is x(n) =
(z 1) (z 1)2
nu(n). Assuming still that |z| > 1, with the following algebraic manipulations we obtain
the desired inverse transform.
z
z 1+1 1 1
= = +
(z 1)2 (z 1)2 z 1 (z 1)2
  
1 z 1 1 1 1
Z 2
=Z +Z
(z 1) z 1 (z 1)2
  
1 1 1 z 1 1
Z 2
=Z 2
Z
(z 1) (z 1) z 1

1
Z 1 = nu(n) u(n 1)
(z 1)2

By examining the values taken by x(n) = nu(n) u(n 1) when n 2 N and n > 0, we
conclude that it can be rewritten as x(n) = (n 1)u(n 1).
1
• Inverse z-transform of . For this term, we will use power series expansion. We
(z 1)3
know that, for |z| > 1, the following equality holds
1
X 1
X
n 1 1 n 1 1 1
z =z z =z 1
=
n=0 n=0
1 z z 1
60 z-Transform

By di↵erentiating twice this equality, we obtain

✓ ◆0 1
!0 1
1 1 X X
n 1 n 2
= = z = ( n 1)z
z 1 (z 1)2 n=0 n=0
1
X
1
= ( n 1)z n 2
(z 1)2 n=0
✓ ◆0 1
!0 1
1 2 X X
n 2 n 3
= = ( n 1)z = ( n 1)( n 2)z
(z 1)2 (z 1)3 n=0 n=0
1
X
1 1 n 3
= (n + 1)(n + 2)z
(z 1)3 2 n=0
X1 1
(1) 1 n (2) 1X n
= (n 2)(n 1)z = (n 2)(n 1)u(n 3)z
2 n=3
2 n=0
1 ✓ ◆
(3) X n 1 n
= u(n 3)z
n=3
2

where: in (1) a changed of index variable was made; in (2) the unit step function was
introduced in the summation; and in (3) we used combinatorial notation to make the
expression more concise. Finally, the inverse z-transform we were looking for turns out
to be  ✓ ◆
1 1 n 1
Z = u(n 3)
(z 1)3 2

z(z + 1)
Putting all the pieces together, we arrive at the final expression for X(z) = .
(z 1)3
✓ ◆
1 n 1
Z [X(z)] = 2 u(n 3) + 3 · (nu(n) u(n 1)) + u(n 1).
2

which again is only valid for |z > |1|.

For a long list of inverse z-transform, check out the Wikipedia article z-transform [Wik19].
From the examples of inverse z-transform we have examined so far, it is noticeable that
the inverse z-transform is easier to compute when the transfer function is written in terms of
1
z 1 rather in terms of z. For example, the inverse z-transform of is just the unit step
 1 z 1
1
1
sequence u(n). However, computing Z requires more e↵ort. Because of this, in many
1 z
textbooks the above theoretical development is done in terms of z 1 . To us, it makes more
sense to show the reader the true reason to change variables from z to z 1 . Let’s describe how
to find the inverse z-transform in terms of z 1 . We first rewrite the transfer function H(z) as
follows.
4.4. The Inverse z-Transform 61

(z p1 )(z p2 ) · . . . · (z pN )
H(z) = b1
(z q1 )(z q2 ) · . . . · (z qM )
1
(z p1 z · z )(z p2 z · z 1 ) · . . . · (z pN z · z 1 )
= b1 1 )(z
(z q1 z · z q2 z · z 1 ) · . . . · (z qM z · z 1 )
z N (1 p1 z 1
)(1 p2 z 1 ) · . . . · (1 pN z 1 )
= b1 M 1 )(1
z (1 q1 z q2 z 1 ) · . . . · (1 qM z 1 )
(1 p1 z 1 )(1 p2 z 1 ) · . . . · (1 pN z 1 )
= z N M · b1 = zN M
· H1 (z)
(1 q1 z 1 )(1 q2 z 1 ) · . . . · (1 qM z 1 )
| {z }
H1 (z)

Assume now that we know the inverse z-transform of H1 (z), call it S(n). Since Z[S(n)] = H1 (z)
and by using the time shifting property,
Z[S(n (N M )] = z N M
Z[S(n)] = z N M
H1 (z) = H(z)
Therefore, the inverse of H(z) is S(n (N M )), just a time shifting of H1 (z) the transfer
function written in terms of z 1 .
Problem 4.4.2 Fill in the table below with the inverse z-transforms. Use the properties listed
in Table 4.1. Specify the ROC where the inverse is valid. Use MATLAB as appropriate for
your computations.

Sequence z-transform ROC Sequence z-transform ROC

1 1
1
|z| > 1 1
|z| > |a|
1 z 1 az

1 1
1
|z| < 1 1
|z| < |a|
1 z 1 az
1
z az 1
|z| < 1 |z| > |a|
(1 z 1 )2 (1 az 1 )2
1
z az 1
|z| < 1 |z| < |a|
(1 z 1 )2 (1 az 1 )2

z 1 (1 + z 1 ) az 1 (1 + az 1 )
|z| < 1 |z| > |a|
(1 z 1 )3 (1 az 1 )3

z 1 (1 + z 1 ) az 1 (1 + az 1 )
|z| < 1 |z| < |a|
(1 z 1 )3 (1 az 1 )3

Table 4.2: Inverse z-transforms

Problem 4.4.3 Determine the impulse response of the system described by the di↵erence
equation:
y(n) 3y(n 1) 4y(n 2) = x(n) + 2x(n 1)
Assume that the system is relaxed initially (all initial values are zero).
 62 z-Transform

4.4.2 Power Series Expansion Method

Review 4.4.1 Review Section 7.4 on power series.

The power series expansion method uses the theory on power series in a very straightforward
manner. The main idea is to use functions for which their power series is known and written
them in terms of z 1 or z, and then identify the coefficients of the power series with inverse
z-transform. As a matter of fact, this is not a general method (power series of every function
are not always known), but it is a useful method because the power series of some important
functions are actually known. Let’s illustrate the method by giving a few examples.
z
Our first example will be computing the inverse z-transform of the function , where
z a
|z| > |a|. From the power series theory we know that, when |x| < 1, the equality below holds
(just summing a geometric series).
X1
1
xn =
n=0
1 x
If we now make the change of variables x = az 1 , then we have
1
X 1
X
1 n 1
(az ) = an z n
= 1
n=0 n=0
1 az

1
The new ROC is |az 1 | < 1, or when working z out, |z| > |a|. Finally, the function
1 az 1
z
is just . By identifying the coefficients in the power series with x(n) we conclude that
z a
x(n) = an · u(n).

Problem 4.4.4 Find the inverse z-transform of X(z) = log(1 + az 1 ), |z| > |a|, by using
power series.

Problem 4.4.5 Make the mind map of this chapter

4.5 MATLAB Exercises and Problems

Program 4.5.1 Complete the definition of the MATLAB function FiniteZtr below. This
function receives a finite signal (n, x) as input and computes its z-transform. Apply your
program to the signal in Figure 4.2.
1 f u n c t i o n [X] = F i n i t e Z t r ( n , x )
2 %The program f i r s t draws t h e f i n i t e s i g n a l x i n t h e r an g e n
3 %Afterwards , i t computes i t s z t r a n s f o r m by u s i n g t h e d e f i n i t i o n
4 stem ( n , x )
5 syms z
6 X=
7 end

Program 4.5.2 Try to find the z-transforms of the sequences below by using the MATLAB
function ztrans. Specify the ROC too.
 4.5. MATLAB Exercises and Problems 63

Figure 4.2: The z-transform of a finite signal

(1) x(n) = (1/10)n · u(n 1) (3) x(n) = 2nu(n)

(2) x(n) = u(n) 2n u(n 1) (4) x(n) = cos(n⇡/2n)u(n)

Program 4.5.3 The MATLAB script below can be used to obtain the inverse z-transform in
di↵erent ways. By using the output of this script, fill in the first row of the table below. Then,
change the script for each of the given X(z) and complete all the rows.
1 %F i r s t , we d e f i n e z and n a s s y m b o l i c v a r i a b l e s
2 syms n
3 syms z
4 X( z )= 1 / ( z ˆ2 3∗ z +2)
5 %We compute , i f p o s s i b l e , t h e i n v e r s e z t r a n s f o r m d i r e c t l y
6 x= i z t r a n s (X( z ) , z , n )
7 %We o b t a i n t h e f i r s t few terms o f x
8 n=0:5;
9 S i x F i r s t T e r m s=i z t r a n s (X( z ) , z , n )
10 %To f i n d t h e p a r t i a l f r a c t i o n expansion , we use r e s i d u e z
11 %C o e f f i c i e n t s a r e i n t r o d u c e d i n X a s power o f z ˆ( 1)
12 [ R, p , c ]= r e s i d u e z ( [ 0 , 0 , 1 ] , [ 1 , 3 , 2 ] )
 64 z-Transform

z-transform Expression Zeros, poles Partial frac- Inverse by First

X(z) in z 1 residues tion expansion look-up table values x(n)
1
z2 3z + 2

1 1 z
2
+
z z z 2
z 1
8z 3 6z 2 + z
1
z2 2z + 1
1
z2 +4

Table 4.3: Finding inverse z-transforms

2z 1
Program 4.5.4 If the z-transform of a causal signal x(n) is X(z) = , compute:
z 1
(1) The value of x(0)
(2) The z-transform of y(n) = 3x(n 1) + 2x(n)
(3) The z-transform of y(n) = n · x(n)
(4) The convolution of x(n) and u(n)

Program 4.5.5 MATLAB have a function to compute the inverse z-transform, impz. For
example, if we write [h,t] = impz(b,a), the function returns the impulse response of the
digital filter with numerator coefficients b and denominator coefficients a; see the program
below.
1 >> a =[1 2 1 ] ;
2 >> b=[1/2 1, 0 . 5 ] ;
3 >> impz ( b , a ) %This s t a t e m e n t draws t h e f i g u r e below .
4 >> h=impz ( b , a )
5 h =
6 0.5000
7 2.0000
8 4.0000
9 6.0000
10 8.0000
11 10.0000
12 12.0000
13 14.0000
14 16.0000
15 18.0000
4.5. MATLAB Exercises and Problems 65

Figure 4.3: Computing the impulse response with impz

z
1
Program 4.5.6 Let H(z) be the transfer function . Use the function tf2zp to
8z 2 6z + 1
find the poles and zeros and draw them in the complex plane. Use the MATLAB help to deal
with this function.