Chapter 4.

The Z-Transform

Gao Xinbo
School of E.E., Xidian Univ.
Xbgao@ieee.org
Xbgao@lab202.xidian.edu.cn
http://see.xidian.edu.cn/teach/matlabdsp/
 Review
 The discrete-time Fourier transform approach for representing
discrete signals using complex exponential sequence.
 Advantages for LTI system
 It describes systems in the frequency domain using the frequency respon
se function H.
 The computation of the sinusoidal steady-state response is great facilitat
ed by the use of H.
 Response to any arbitrary absolutely summable sequence x(n) can easily
be computed in the frequency domain by multiplying the transform X an
d the frequency response H.
 Shortcomings to the FT
 1. There are many useful signals in practice, such as u
(n), nu(n), for which the DTFT does not exist.

 2. The transient response of a system due to initial con

ditions or due to changing inputs cannot be computed u
sing the DTFT approach.
 Extension of the DTFT
 To address the above two problems, z-transform is proposed.
 Bilateral (two-sided) version provides another domain in which a
large class of sequence and systems can be analyzed.
 Unilateral (one-sided) version can be used to obtain system
response with initial conditions or changing inputs.
 The bilateral z-transform

X ( z )  Z [ x(n)]   x ( n) z  n
n  

z is a complex variable.
The set of z values for which X(z) exists is called the region of convergence
(ROC) and is given by
Rx  | z | Rx 
For some positive numbers Rx- and Rx+.

1
x(n)  Z 1[ X ( z )]  
n 1
X ( z ) z dz
2j C

C is counterclockwise contour encircling the origin and lying in the ROC.

 Comments
1. The complex variable z is called the complex frequency given
by z | z | e jw , where |z| is the attenuation and w is the real
frequency;
2. Since the ROC is defined in terms of the magnitude |z|, the
shape of the ROC is an open ring. Note that Rx- may be equal to
0 and/or Rx+ could possibly be infinity;
3. If Rx+ <Rx-, then the ROC is a null space and the ZT does
not exist;
 The FT is a special case of the
ZT
The function |z|=1 (or z  e ) is a circle of unit radius in the z-pl
jw

ane and is called the unit circle. If the ROC contains the unit circl
e, then we can evaluate X(z) on the unit circle.

X ( z ) | z e jw  X (e ) 
jw
 x ( n ) e  jwn
 F [ x(n)]
n  

Therefore the discrete-time Fourier transform X() may

be viewed as a special case of the z-transform X(z).
 Examples
 Example 4.1
 Positive-time sequence
 Example 4.1
 Negative-time sequence
 Example 4.1
 Two-sided sequence
 Compare their ROCs, zeros and poles.
 Properties of the ROC
1. The ROC is always bounded by a circle since the convergence
condition is on the magnitude |z|;
2. The ROC for right-sided sequences (n<n0, x(n)=0) is always o
utside of a circle of radius Rx-. (if n0>=0, x(n) is a causal seq.)
3. The ROC for left-sided sequences (n>n0, x(n)=0) is always ins
ide of a circle of radius Rx+. (if n0<=0, x(n) is a anticausal seq
uence)
4. The ROC for two-sided sequences is always an open ring Rx-
<|z|<Rx+ if it exists.
 Properties of the ROC
5. The ROC for finite-duration sequences(n<n1 and n>n2,
x(n)=0) is the entire z-plane. If n1<0, then z=infinity is not in
the ROC. If n2>0, then z=0 is not in the ROC;
6. The ROC cannot include a pole since X(z) converges
uniformly in there;
7. There is at least one pole on the boundary of a ROC of a
rational X(z);
8. The ROC is one contiguous region, the ROC does not come in
pieces.
 Important properties of the z-transform
 1. Linearity
Z [ a1 x1 (n)  a2 x2 (n)]  a1 X 1 ( z )  a2 X 2 ( z ); ROC : ROCx1  ROCx2

 2. Sample shifting
Z [ x(n  n0 )]  z  n0 X ( z ); ROC : ROC x
 3. Frequency shifting
z
Z [a n x(n)]  X  ; ROC : ROC x scaled by | a |
a
 4. Folding
X [ x(n)]  X (1 / z ); ROC : Inverted ROC x
 Important properties of the z-transform
 5. Complex conjugation
Z [ x* (n)]  X * ( z * ); ROC : ROC x
 6. Differentiation in the z-domain
dX ( z )
Z [nx(n)]   z ; ROC : ROC x Multiplication by a ramp
dz
 7. Multiplication
1

1
Z [ x1 (n) x2 (n)]  X 1 ( v ) X 2 ( z / v ) v dv ROC : ROC x1  ROC x2
2j C

 8. Convolution
Z [ x1 (n) * x2 (n)]  X 1 ( z ) X 2 ( z ) ROC : ROCx1  ROCx2
Some common z-transform pairs
Sequence Transform ROC
 ( n) 1 z
u ( n) 1
1 z 1
| z | 1
 u ( n  1) 1
1 z 1
| z | 1
n
a u ( n) 1
1 az 1
| z || a |
 b u ( n  1)
n 1
1 bz 1
| z || b |
 Some common z-transform pairs
Sequence Transform ROC
1
( a sin w ) z
[a n sin w0 n]u (n) 0
1 2 2
| z || a |
1  (2a cos w0 ) z  a z
1
1  ( a cos w ) z
[a n cos w0 n]u (n) 0
1 2 2
| z || a |
1  (2a cos w0 ) z  a z
1
az
n
na u (n) | z || a |
(1  az 1 ) 2
1
bz
 nb u ( n  1)
n
| z || b |
(1  bz 1 ) 2
 Examples:
 Convolution:
 Ex4.4
 Ex4.5
 Ex4.6 Using z-transform properties and the z-
transform table, determine the z-transform of

( n2)  
x(n)  (n  2)(0.5) cos  (n  2)u (n  2)
3 
 Inversion of the z-tranform
 From the definition of the inverse z-transform computation
requires an contour evaluation of a complex integral that, in
general, is a complicated procedure.

 The most practical approach is to use the partial fraction

expansion method. It makes use of the z-transform table. The z-
transform, however, must be a rational function. This
requirement is generally satisfied in digital signal processing.
 Central Idea
 When X(z) is a rational function of z-1, it can be expressed as
a sum of simple (first-order) factors using the partial fraction
expansion. The individual sequences corresponding to these
factors can then be written down using the z-transform table.

 Method: Given

1 M
b0  b1 z    bM z
X ( z)  1 N
, Rx  | z | Rx 
1  a1 z    a N z
 Express it as
~ ~ 1 ~  ( N 1)
b0  b1 z    bN 1 z M N
X ( z)  1
1  a1 z    a N z N
  k c z k

           k0 
Proper rational part polynomial part if M  N

This can be obtained by performing polynomial division if

M>=N using the deconv function
 Perform a partial fraction expansion on the proper
rational part of X(z) to obtain

M N
N
Rk
X ( z)   1
  c k z k

k 1 1  p k z k0 
M N

Where pk is the k-th pole of X(z) and Rk is the residue at pk.

It is assumed that the poles are distinct for which the residu
es are given by
~ ~ 1 ~  ( N 1)
b0  b1 z    bN 1 z 1
Rk  1 N
(1  p k z ) | z  pk
1  a1 z    a N z
 For repeated poles the expansion has a more general
form. If a pole pk has multiplicity r, then its expansion
is given by
r Rk ,l z  (l 1) Rk ,1 Rk , 2 z 1 Rk ,r z  ( r 1)
 (1  p z
l 1
1 l
)

1  pk z 1

(1  pk z )1 2

(1  pk z 1 ) r
k

Write x(n) as
N  1  M N
x(n)   Rk Z  1
1 
  ck  ( n  k )
k 1 1  pk z  k 0    
M N

Finally, use the relation from Table to complete x(n)

1  z   p n
k u ( n) | z k | Rx 
Z   n
 z  pk   pk u (n  1) | z k | Rx 
 Examples
 Ex4.7 Find the inverse z-transform of
z
x( z )  2
3z  4 z  1

Since the ROC is not specified, there are three possible

ROCs.
Right-sided sequence
Left-sided sequence
Two-sided sequence
 1

0.8

0.6

0.4
Imaginary Part

0.2

-0.2

-0.4

-0.6

-0.8

-1

-1 -0.5 0 0.5 1
Real Part
 Matlab Implementation
 A Matlab function residuez is available to compute the residue p
art and the direct (or polynomial) terms of a rational function in
z-1. Let 1 M
b0  b1 z    bM z B( z )
X ( z)  1 N

1  a1 z    a N z A( z )
M N
N
Rk
 1
  c k z k

k 1 1  pk z
k0 
M N

be a rational function in which the numerator and the

denominator polynomials are in ascending powers of z-1.
 Then [R,p,c] = residuez(b,a)
The returned column vector R contains the residues;
Column vector p contains the poles locations;
Row vector c contains the direct term.
If p(k)=…=p(k+r-1) is a pole of multiplicity r, then th
e expansion includes the term of the form
Rk Rk 1 Rk  r 1
1
 1 2

1  pk z (1  pk z ) (1  pk z 1 ) r

Similarly, [b,a]=residuez(R,p,c) can be used to convert t

he partial fraction expansion back to polynomials with co
efficients in row vectors b and a.
 Examples
 Ex4.8 Check Ex4.6
 Ex4.9 Compute the inverse z-transform of

1
X ( z)  1 1
, | z | 0.9
(1  0.9 z )(1  0.9 z )
Polynomial coefficients=Poly([root1,root2,…rootn])
 Ex4.10 Determine the inverse z-transform of
1  0.4 2 z 1
X ( z) 
1  0.8 2 z 1  0.64 z  2
 So that the resulting sequence is causal and contains no complex numbers.
 System representation in the z-domain
 Similar to the frequency response function H(ejw), we can define
the z-domain function, H(z), called the system function. Howeve
r, unlike H(ejw), H(z) exists for systems that may not be BIBO st
able.
 Definition 1: The system function H(z) is given by

H ( z )  Z [h(n)]   h( n ) z  n , Rh  | z | Rh 
n  

Y ( z )  H ( z ) X ( z ) : ROC y  ROCh  ROC x

 System function from the difference
equation representation
N M
y (n)   ak y (n  k )   bl x(n  l )
k 1 l 0
Z-transform N M
Y ( z )   ak z Y ( z )   bl z l X ( z )
k

k 1 l 0

Y ( z ) B( z ) b0 z  M ( z M    bbM0 )
H ( z)    N N
X ( z ) A( z ) z ( z    an )

After factorization, we obtain

M
 ( z  zl )
N M
H ( z )  b0 z l 1
N
 ( z  pk )
k 1
 M
 ( z  zl )
N M
H ( z )  b0 z l 1
N
 ( z  pk )
k 1

Where zl’s are the system zeros and pk’s are the system poles.
Thus H(z) can also be represented in the z-domain using a pole-
zero plot. This fact is useful in designing simple filters by prope
r placement of poles and zeros.
To determine zeros and poles of a rational H(z):
Matlab function: roots: polynomial--root
poly: root-- polynomial
Zplane(b,a): plot the poles and zeros, given the numerator r
ow vector b and the denominator row vector a.
Similarly,
Zplane(z,p): plots the zeros in column vector z and the poles
in the column vector p.
1

0.8

0.6

0.4
Imaginary Part

0.2

-0.2

-0.4

-0.6

-0.8

-1

-1 -0.5 0 0.5 1
Real Part
 Transfer function representation
If the ROC of H(z) includes a unit circle (z=ejw), then we can eva
luate H(z) on the unit circle, resulting in a frequency response fun
ction or transfer function H(ejw).
M
 (e jw  zl )
H (e jw )  b0 e j ( N  M ) w l 1
N
 (e jw  pk )
k 1
| e jw  z1 |  | e jw  z M |
| H (e ) || b0 | jw
jw

| e  p1 |  | e jw  p N | Interpretation, illustration
M N
H (e )  [0 or  ]  [( N  M ) w]   (e  z k )   (e jw  pk )
jw jw
       
constant linear k 1      k1     
nonlinear
 Matlab implementation
[H,w]=freqz(b,a,N)
Returns the N-point frequency vector w and the N-point complex f
requency response vector H of the system.
It is evaluated at N points equally spaced around the upper half of
the unit circle.
[H,w] = freqz(b,a,N,’whole’)
Uses N points around the whole unit circle for computation.
H=freqz(b,a,w)
It returns the frequency response at frequencies designated in vect
or w, normally between 0 and pi.
 Relationships between system
representation
Express H(z) in
1/z cross H(z) Take inverse z-
multiply and transform
take inverse Take z-transform Take
solve for Y/X z-transform
Difference
h(n)
Equation Substitute
z=ejw Take inverse
DTFT

Take DTFT solve H(ejw)

Take Fourier
for Y/X
transform

The dashed paths exist only if the system is stable

 Stability and Causality
 Theorem 2 z-domain LTI stability
 An LTI system is stable if and only if the unit
circle is in the ROC of H(z)
 Theorem 3 z-domain causal LTI stability
 A causal LTI system is stable if and only if the
system function H(z) has all its poles inside the
unit circle.
 Solutions of the difference equations
 Linear constant coefficient difference equations
 Particular and homogeneous solution
 Zero-input (initial condition) and the zero-state
responses
 Z-transform
 Transient and steady-state responses
 The one-sided z-transform
 The one-sided z-transform of a sequence x(n) is given by

Z [ x(n)]  Z [ x(n)u (n)]  X [ z ]   x(n) z  n
 

n0

Then the sample shifting property is given by

Z  [ x(n  k )]  Z [ x(n  k )u (n)]

 
  x(n  k ) z n
  x ( m) z  ( m  k )
n 0 m k
1
 
  x ( m) z ( m  k )
   x ( m) z  m  z  k
m k  m 0 
 Solve difference equations with nonzero
initial conditions or with changing inputs
N M
y ( n )   ak y ( n  k )   bm x(n  m), n0
k 1 m 0

Subject to these initial conditions:

 y (i), i  1,, N  and  x(i), i  1,, M 
Example 4.14 3 1
y (n)  y (n  1)  y (n  2)  x(n), n  0
2 2
n
1
x( n)    u ( n)
4
s.t. y (1)  4, and y (2)  10
 

Y ( z) 1  z  z
3
2
1 1
2
2
  1 1
1
z 1
 (1  2 z 1 )
4

1
1 1 1
 1  z ( 1  2 z )
Y ( z)  3 1
4
1 2

1 2 z  2 z 1  32 z 1  12 z  2
9 1 1 2
 2  z  2 z
Y ( z)  4
(1  12 z 1 )(1  z 1 )(1  14 z 1 )

 1 2 1
Y ( z)  1
 3
1
 3
1 z 1
2 1 z 1  z 1
1
4

 1  n 2 1  1  n 
y (n)        u (n)
 2  3 3  4  
 Forms of the solutions (1)
Homogeneous and particular
parts:
 1  n 2  11
n

y (n)     u (n)    u (n)

 2  3  3 4  

      particular part
homogeneous part

The homogeneous part is due to the system poles

and the particular part is due to the input poles.
 Forms of the solutions (2)
Transient and steady-state response:

1  1 n  1 n  2
y ( n)         u ( n)  u ( n)
 3  4   2   3 

       steady  state response
transient response

The transient response is due to poles that are inside

the unit circle, while the steady-state response is due
to poles that are on the unit circle.
Note that when the poles are outside the unit circle,
the response is termed an unbounded response.
 Forms of the solutions (3)
Zero-input (or initial condition) and zero-state responses:
YZS ( z )  H ( z ) X ( z ) YZI ( z )  H ( z ) X IC ( z )
X IC ( z )  {1,2}
1  1 n  1 n 8    1 n 
y (n)      2   u (n)  3   2u (n)
 3  4   2  3    2  
                
zero  state response zero input response
 Matlab Implementation
 y=filter(b,a,x,xic)
 Xic is an equivalent initial-condition input array
 Xic = filtic(b,a,Y,X)
 b and a are the filter coefficient array
 Y and X are the initial-condition arrays from the initial condi
tions on y(n) and x(n), respectively, in the form

Y  [ y (1), y (2), , y ( N )]
X  [ x(1), x(2), , x( M )]
References and Assignment
 Textbook: pp80 ~110
 Chinese reference book: pp43~63
 Exercises:
 1: p4.1a,b,d; p4.3a,b; p4.10
 2: p4.15;4.18;