The Z-Transform in DSP

Lecture 10-12
Andreas Spanias
spanias@asu.edu

2006

Copyright 2006 Andreas Spanias

10-12-1

Poles and Zeros of H(z)

In general the transfer function is rational; it has a numerator and a
denominator polynomial.
The roots of the numerator and denominator polynomials are called the
zeros and the poles respectively.
Pole-zero decompositions of H(z) are quite useful and provide intuition in
signal analysis and filter design.
L

( z  ] 1 )( z  ] 2 )...(z  ] L )
H ( z) G
( z  p1 )( z  p2 )...(z  pM )

(z ] )
i

i 1
M

(z  p )
i

i 1

where G is a gain factor

2006

Copyright 2006 Andreas Spanias

10-12-2

Example: Poles and Zeros of H(z)

0.2 z
z  0 .8

H ( z)

the zero (O) of this transfer function is at z=0

the pole (X) of this transfer function is at z=0.8

Imaginary part

1
0.5

-0.5
-1
-1
2006

-0.5

0
0.5
Real part

Copyright 2006 Andreas Spanias

1
10-12-3

Example: Poles-Zeros of a Second Order System

1  1 . 3435 z  1  0 .9025 z  2
1  0 .45 z 1  0 .55 z  2

H (z)

H ( z)

( z .95e
( z .7416e

j 45o

j 72 .34 o

)( z .95e

)( z .7416e

lm
X

 j 45o

 j 72 .34 o

X  poles

0  zeros

Note that the filter

coefficients are
real valued and
therefore poles
and zeros occur in
complex
conjugate pairs.

Re

0
X
2006

Copyright 2006 Andreas Spanias

10-12-4

Poles and Zeros and Stability

The location of the poles determines the stability of the filter.
In fact, for stability all the poles of a causal filter must be inside
the unit circle, that is

for all i = 1, 2, . . . , M

pi 1

IIR filters may be all-pole or pole-zero and stability is always a

concern. FIR or all-zero* filters are always stable.
(* depending on the way the transfer function is normalized the term all-zero for FIR systems may be imprecise because
there may be poles located at zero in an FIR system - similarly an all-pole IIR may have zeros located at zero)

2006

Copyright 2006 Andreas Spanias

10-12-5

The Frequency Response Function

The transfer function is

H (z)

b 0  b 1 z  1  ...  b L z  L
1  a 1 z  1  ...  a M z  M

by evaluating on the unit circle, i.e. for

j:

H (e )
2006

e j:

b0  b1e  j:  ...  bL e  jL :
1  a1e  j:  ...  a M e  jM :
Copyright 2006 Andreas Spanias

10-12-6

The Frequency Response Function (Cont.)

The frequency response function and is a complex and periodic
With period 2R. The normalized frequencies : are associated to the
sampling frequencies fs by
Sampling period

:
rad

ZT

2S

f
fs
Sampling frequency

rad/s

where fs is the sampling frequency and f is any frequency of

interest. In practice, one detemines the frequency response up to
half the sampling frequency (fold-over frequency).
2006

Copyright 2006 Andreas Spanias

10-12-7

The Frequency Response Function (Cont.)

2S

S
Foldover Frequency

fs/2

fs

The frequency response is usually plotted w.r.t. normalized frequencies (:)

The frequency response is periodic with period fs (2 S )
Since frequencies of interest are up to the bandwidth of the analog signal
the spectrum is usually plotted up to fs/2, (S ) the foldover frequency
2006

Copyright 2006 Andreas Spanias

10-12-8

The Frequency Response and Poles and Zeros

The magnitude frequency response function
L

H ( e j: )

e j:  ] i

e j:  p i

i 1
M

i 1

The phase frequency response function

j:

arg(H(e ))

arg(e
i 1

2006

j:

] i )  arg(e j:  pi )
i 1

Copyright 2006 Andreas Spanias

10-12-9

Remarks on Effects of Poles and Zeros on H(ej:)

Poles tend to create peaks in the magnitude frequency response
Zeros tend to create valleys in the magnitude frequency response
Selective filters are designed efficiently by placing poles close to
the unit circle
Sharp notches are achieved efficiently with zeros placed very
close to the unit circle
if restricted to an all-pole filter a sharp notch in the frequency
response will require high order design (many poles).
if restricted to an all-zero (FIR) filter a sharp (high Q) peak in the
frequency response will require many zeros (long impulse
response or high orderCopyright
FIR). 2006 Andreas Spanias
2006

10-12-10

Z plane and Frequency Magnitude Response

lm

0  zeros

Re

2.0
Magnitude H(z)

Magnitude Response

X  poles

1.5

1.0

0.5

0.0

32

64

96

128

Frequency Index (Theta=2*PI*Index/128)

FFT index

2006

Foldover Frequency
fs/2

fs

Copyright 2006 Andreas Spanias

10-12-11

Zero Locations and Frequency Response

o
o

foldover

Moving zeros towards the unit circle creates sharper valleys

2006

Copyright 2006 Andreas Spanias

10-12-12

Pole Locations and Frequency Responses

X
X

X
X

As the poles move towards the unit circle we get sharper peaks
and if the poles are placed on the unit circle we get an oscillator.
2006

Copyright 2006 Andreas Spanias

10-12-13

Computing Filter Responses Using the

Inverse Z-Transform
Partial Fractions:

H (z)

Given a the transfer function

b 0 z L  b1 z L  1  ...  b L
z M  a 1 z M  l  ...  a M
for distinct poles write H(z) as:

H ( z)

2006

c0  c1

z
z
 ...  c M
z  p1
z  pM

Copyright 2006 Andreas Spanias

10-12-14

DSP EEE 407/591

Andreas Spanias

Lecture 11

2006

Copyright 2006 Andreas Spanias

10-12-15

Inverse Z-Transform
The coefficients cj are found using partial fractions, and given the
transform pairs:

z
, n t 0` l
, z ! pi

z  pi

z
n
^ p i , n  0 ` l z  p , z  p i
i

^p

n
i

as well as the ROC, one can find the sequence corresponding to H(z).
2006

Copyright 2006 Andreas Spanias

10-12-16

Partial Fractions
Given a rational function:

H (z)

b 0 z L  b1 z L  1  ...  b L
z M  a 1 z M  1  ...  a M

If M>=L for distinct poles we can write H(z) as:

H ( z)

c0  c1

z
z
 ...  c M
z  p1
z  pM

where ci are constants to be determined

2006

Copyright 2006 Andreas Spanias

10-12-17

Partial Fractions Example

Given the second order function

H z

z2  z
z 2  56 z  16

Use partial fractions to write H(z) as :

H z  9

2006

z
z

8
z - 21
z - 31

Copyright 2006 Andreas Spanias

10-12-18

Partial Fractions Example (Cont.)

and given the ROC

1
2

z !
then

h(n )

1
9

1
 8
3

, n t 0

If the ROC is defined differently then the time domain sequences

will be different.

2006

Copyright 2006 Andreas Spanias

10-12-19

Partial Fractions Example (Cont.)

For example, ROC:

yields

2006

1

3

1
2
n

h(n )

1
 8
3

h(n )

1
 9

Copyright 2006 Andreas Spanias

, n t 0
, n  0

10-12-20

Partial Fractions Example (Cont.)

For the ROC:

1
3

yields

h( n )

2006

1
1
8  9 , n  0
3
2

Copyright 2006 Andreas Spanias

10-12-21

Inverse Z-Transform - Repeated Poles

Given a rational function:

H (z)

z 

then the time domain sequence is For ROC:

z

2006

1
6
1
6

1
,n t 0
6

h(n)

 n  1 

h(n)

1
  n  1 
6

Copyright 2006 Andreas Spanias

,n  0

10-12-22

DSP EEE 407/591

Andreas Spanias

Lecture 12

2006

Copyright 2006 Andreas Spanias

10-12-23

Inverse Z-Transform using the Residue Theorem

We will restrict our discussion to causal sequences. Extension to noncausal is straight-forward. Given the z-transform:
f

X (z)

x(n )z n

Cauchys Integral Theorem:

f

2006

X ( z )z

k 1

dz

x ( n ) z k  n 1 d z

Copyright 2006 Andreas Spanias

10-12-24

Inverse Z-Transform using the Residue Theorem

(Cont.)
If the integration is counterclockwise on a contour which is within
the ROC and includes the unit circle, then
f

X ( z )z

k 1

dz

x ( n ) z

k  n 1

dz

If the path encloses the origin then

therefore

k  n 1

2 S jG ( k  n )

dz

x(n)

2S j

X ( z )z

n 1

dz

Note the similarity with the inverse DTFT which is a special case of the z transform
2006

Copyright 2006 Andreas Spanias

10-12-25

Inverse Z-Transform using the Residue Theorem

(Cont.)
Cauchys residue theorem states that for rational polynomials X(z) the
integral above can be computed as a sum of residues, that is given:

X ( z)

A( z )
( z  p1 )( z  p 2 )...( z  p M )
then
M

x(n )

res [ z n  1 X ( z )] z

pi

i 1

2006

Copyright 2006 Andreas Spanias

10-12-26

Inverse Z-Transform using the Residue Theorem (Cont.)

For poles of multiplicity m

1 d m1
m n1
m1 (z  pi ) z X (z)
(m 1)! dz
z

n1

res[z X (z)]z

pi

pi

For single pole

res [ z n 1 X ( z )] z

[( z  pi ) z n 1 X ( z )] z

pi

pi

Examples: Single pole

X(z)
2006

z
z a

Copyright 2006 Andreas Spanias

10-12-27

Inverse Z-Transform using the Residue Theorem

(Cont.)

res[ z n 1 X ( z )] z

n 1
( z  a ) z z  a
z

an
a

Pole of multiplicity 2

z2
X (z)
(z  a)2
n1

res[z X (z)]z

2006

d
z2
2 n1

(z  a) z
(z  a)2 z
dz
Copyright 2006 Andreas Spanias

(n 1)an
a

10-12-28

Example: Steady-State and Transient Response

of Digital Filters Using the inverse z-transform
y n 
x n  1

z 1

Consider the first order IIR filter

 0 .7

H ( z)

Y ( z)
2006

3z
( z  0 .7 )
H ( z) X ( z)

X ( z)

z
( z  1)

3z 2
( z  1)( z  0.7)

Copyright 2006 Andreas Spanias

10-12-29

Example: Steady-State and Transient Response

(Cont.)
Steady-state
pole of i/p

transient
pole of filter

y(n) res[ z n1Y ( z)]z 1  res[ z n1Y ( z)]z

0.7

( z 1) z n1z 2
( z  0.7) z n1z 2
y(n) 3
 3

(
z

0
.
7
)(
z

1
)
(
z

1
)(
z

0
.
7
)

z 1
z
y(n) 10  7(0.7)n , n t 0
2006

Copyright 2006 Andreas Spanias

10-12-30

0.7

Example:Steady-State and Transient Response

y(n) 107(0.7)n , n t 0

10
9
8
7
6
5
4
3
2
1
0
0

2006

10

12

14

16

Copyright 2006 Andreas Spanias

18

20
10-12-31