Chapter 4

Pole zero placement method of

design of filters
 Contents
• Ideal filter characteristics

• IIR and FIR filters

• Notch filter

• Comb filter

• All pass filter

• Digital resonator

• Digital sinusoidal oscillator

 Filter ?
• Frequency selective network

• Time domain representation of filter

• Frequency domain representation of filter

– Explanation is easier in frequency domain for a filter

• To pass desirable range of frequencies

• To stop / attenuate undesirable range of

frequencies / single frequency / narrow band
 Types of filters & characteristics
• Low pass Filter, HPF, BPF, BSF, all pass filter

• Frequency domain representation of filter

– Plot of frequency response

• Why on either side of frequency = 0?

• ‘all pass filter’ ?

 Ideal filter characteristics
• For all filters desirable characteristics are

• Constant frequency response in pass band

• Zero frequency response in stop band

• Linear phase characteristics (response) in pass

band

• All frequency components of input signal falling

within pass band experience same delay
 Ideal filter characteristics
• Example Ce  jn0 1  2
H ( ) 
0 otherwise

• If LPF, lower cutoff is zero.

• Verify magnitude frequency response

• It is unity

• Verify phase response

• It is linear phase response
 Ideal filter characteristics
• It is linear phase response = n0

• Determine output: y(n)=Cx(n-n0) in pass band & zero

in stop band

• This leads to
– All frequency components of input signal falling within
pass band experience same delay
• If linear phase characteristics is not satisfied, signal in
pass band gets distorted.

• Pulse dispersion is an example for the same

 Ideal filter characteristics
• Ideal characteristics of a filter usually not
realizable.

• It is
– Leading to its impulse response infinite no. of samples
– & impulse response non causal
– Sum of all samples not finite and hence not stable.
 Pole zero placement method of design
• An approximate approach

• Easy method of design of filters

– Place poles within unit circle
– Zeros can be any where in z plane
– All complex zeros and poles must occur in conjugate pairs.

• Exhibits good performance for

– LPF with small cutoff frequency, close to zero
– HPF with very high cutoff frequency, close to Nyquist limit
– BPF & BSF with narrow band width
 Significance of zero
 j
• Zero at z e
j  j
• Lead to numerator factor ( z  e )( z  e )

• Effect on frequency response: (e j  e j )(e j  e  j )

• Becomes zero for frequency ??

• Thus placing zero reduces the magnitude
response
 Significance of pole
 j
• Pole at z re within unit circle (r<1)
j  j
• Leads to denominator factor ( z  re )( z  re )
j j j  j
• Effect on frequency response: ( e  re )( e  re )

  (e j  re j )(e j  re  j
)
• At ,
(1  r )e j (e j  re  j )

• The frequency response increases as the pole moves

closer and closer to unit circle
• Thus placing pole increases the magnitude response
 LPF design
• Example 1: design a single pole LPF:
• Solution: Because it is single pole, it has to be real
pole.
• Let it occur at z=a
1
• Thus system transfer function is  H1 ( z ) 1  az  1 (1)

• What are the frequency components in pass band?

Refer to the plot of frequency response…
• What is the value of frequency response in pass
band?
• What is maximum gain? 1 a
 H1 ( z )  1
( 2)
• To make max gain as unity, final: 1  az
 LPF design
1 a
H1 ( z )  ( 2)
1  az  1
• Verify gain at frequency = 0 and gain at maximum
frequency pi.
• Example 2: Design a single pole, single zero LPF:
• Solution: start from previous example.
• Why zero is required?
• To reduce gain further at frequencies away from
pass band
• Where to place this zero?
• Insert zero into the system at maximum frequency
pi that is equivalent to z = -1
 LPF design
(1  a )(1  z  1 )
 H 2 ( z)  (3)
1  az  1
• Verify gain at frequency = 0
• To make max gain as unity, final:
(1  a )(1  z  1 )
 H 2 ( z)  (3)
2(1  az  1 )

• Verify gain at maximum frequency pi to confirm

correctness or closer to ideal filter design
 HPF design
• Example 3: Design a HPF:
• Solution: design low pass filter and fold the pole-zero
locations
• There was zero at frequency
j =pi. Now place pole at
that frequency. z ae  a
• There was
z pole
1 at frequency =0. Now place zero there.
(1  a )(1  z  1 )
 H 3 ( z)  (4)
2(1  az  1 )
• Verify gain at frequency = 0
• Verify gain at maximum frequency pi to confirm
correctness or closer to ideal filter design
 LPF design
• Example 4: Given system transfer function of a LPF
with 2 poles. b0
H ( z) 
(1  pz  1 ) 2
• Determine b0 and p that satisfies conditions:
2

H (0) 1 & H ( ) 0.5
4
• Solution: At frequency =0,
b0
H ( 0)  2
1
(1  p )
 b0 (1  p ) 2
 LPF design
• Determine p from second condition
2
2 
  b0 
H ( )    0.5
4  (1  pe  j 4 ) 2 
 
2
 
 (1  p )  2

   0.5
 (1  pe  j 4 ) 2 
 

• It leads to p=0.32.
• Substitute these values in H(z)
BPF design
 Digital resonator
• Special two pole band pass filter.
• Determine system transfer function for zeros at
origin, and for zeros at z=1 & z=-1
• Ensure magnitude frequency response of unity at
frequency corresponding to pole position
 Digital notch filter
• Either one or more dips / notches in frequency
response. Zeros are kept at those frequencies for
the purpose
• Leads to FIR filter (all zero system)
• Determine system transfer function
• Frequencies closer to the ones which are to be
suppressed, also experience attenuation.
• To overcome this, poles are kept …near these
zeros
 Digital comb filter
• Many dips / notches in frequency response.
• Shape of frequency response?
• Why name comb?
1 M
• Consider y (n)   x(n  k )
M  1 k 0

• Represents moving average filter

• Determine system transfer function
1 M k 1 1  z  ( M 1)
H ( z)  
M  1 k 0
z 
M 1 1  z  1
 Digital comb filter
1 1  z  ( M 1)
H ( z) 
M 1 1  z  1
• Where are zeros and poles?

• Zeros at z e  j 2k /( M 1) 1 k M

• What about k=0?
• The zero at k=0 cancels with pole at z=1. Thus the
system has M zeros
• Is it FIR / IIR?
• All zero filter.. FIR
 Digital all pass filter
• Where are zeros and poles?
• From the definition of filter response in pass band
k
determine system transfer function H ( z )  z
• Purely a ‘delay introducing’ system
– Phase characteristics is important
• Another way of writing system transfer function is
a N  a N  1 z  1  a N  2 z  2 ...  a1 z  ( N  1)  z  N
H ( z) 
1  a1 z  1  a2 z  2 ...  a N  1 z  ( N  1)  a N z  N
N

 k
a z  (N k)

 k 0N & a0 1
 k
a z k

k 0
 Digital all pass filter
z  N A( z  1 )
 H ( z) 
A( z )

• Or if all pass filter has a pole at z0 then it has a zero

at its reciprocal 1/z0
• This is so for every pole and zero
• Plot the same in z plane…a complex pole and
corresponding zero
 Digital all pass filter
• Instead of considering many poles and zeros,
consider an all pass filter with poles at
j  j
z re , re
j j
• Then there will be zeros at z e  / r, e / r
 Digital all pass filter
• pole at z re j
1  j
z e /r
1 j
z re 1
j
• zero at z e  /r
1 j
z re
• Thus system transfer function is

1 j 1  j
z  re z  re
H ( z)  j  1  j  1
1  re z 1  re z
 Digital all pass filter
z  1  re j z  1  re  j
H ( z) 
1  re j z  1 1  re  j z  1
 j j  j  j
e  re e  re
H ( )  j  j  j  j
1  re e 1  re e

 j  j  j j
e  re e  re
 j  j  j  j
1  re e 1  re e

e  j (1  re j (   ) ) e  j (1  re j (  ) )
  j (   )  j (  )
1  re 1  re
 Digital all pass filter

 j j (   )  j j (  )
e (1  re ) e (1  re )
  j (   )  j (  )
1  re 1  re
  ( )  2
1 r sin(   )
 2 tan
1  r cos(   )
1 r sin(   )
 2 tan
1  r cos(   )
• The filter is used to ensure linear phase characteristics for
a system which does not exhibit it …
 Digital oscillator
• Pole on unit circle…
• System transfer function is
b0
H ( z)  j  1  j  1
(1  e z )(1  e z )
• Impulse response ….
• Output y(n)=…
• Implement….
 Effect of Pole position on bandwidth & center
frequency
 j
• Plot pole at z re

• Mark band width:

• Band width: Frequencies covered by arc obtained

by drawing two lines from origin around the pole

• Effect of pole position on bandwidth?

• Effect of pole position on center frequency?
 Effect of Pole position on band width & center frequency
 j
z re

Unit circle

• Band width: Frequencies covered by arc on circle obtained by drawing

two lines from origin around the pole on to circle of radius r

• Effect of pole position on bandwidth: increase r increases BW


• Effect of pole position on center frequency: depends on angular
displacement of the pole from real axis, i.e.,
 Discrete Cosine Transform (DCT)

• Forward DCT: DCT

N1
 (2n  1)k
v(k )  (k ) x(n) cos( ); 0 k  N  1
n 0 2N
1 2
 ( 0)  &  (k )  ; 1 k  N  1
N N

• Inverse DCT:
N1
 (2n  1)k
x(n)   (k )v(k ) cos( ); 0 n  N  1
k 0 2N
 DCT
• Thus the kernel matrix for DCT is
 1
 k 0 & 0 n  N  1
 N
c(k , n) 
 2 cos(  (2n  1)k ); 1 k  N  1 & 0 n  N  1
 N 2N
&
v(k ) c(k , n) x(n)
x(n) c(k , n)  1 v(k )
 0.5000 0.5000 0.5000 0.5000
 0.6533 0.2706 - 0.2706 - 0.6533 
• For N=4, c ( k , n )  
 0.5000 - 0.5000 - 0.5000 0.5000 
 
 0.2706 - 0.6533 0.6533 - 0.2706 
 Narrow BPF
H () BW
r 1 
fs
f0
 x 3600
fs

0 f0 fS /2=fmax
BW

0 1800
• Zeros are kept at z=1 (to make DC component zero) and z=-1
(to make response at max digital frequency =0). Thus total 2
zeros  j
• Poles are at z re . Thus 2 poles
 Narrow BPF
BW
r 1 
• Construct H(z) fs
f0
 x 3600
fs

K ( z  1)( z  1)
H ( z)  j  j
( z  re )( z  re )

(1  r ) 1  2r cos 2  r 2
K
2 sin 
 Notch filter
• Narrow band stop filter
• Place poles same way as in BPF
• Place zeros with same angles, but on unit circle
• Same equation for r and angle
j  j
K ( z  e )( z  e )
H ( z)  j  j
( z  re )( z  re )

1  2r cos   r 2
K
2  2 cos 
 First order LPF
• When cutoff frequency is less than fs/4, pole is
placed at z=a.
fc
a 1 2
fs
• When cutoff frequency is greater than fs/4, pole is
placed at z=-a.
fc
a  (1    2 )
fs
• In both the cases, zero is placed at z=-1 to ensure
highest possible frequency is suppressed.
LPF with cutoff frequency is less than fs/4

Frequency
response

Unit 0 fC fS /2
circle =fmax
LPF with cutoff frequency is greater than fs/4

Frequency
response

fC
Unit 0 fS /2
circle =fmax
 First order HPF
• When cutoff frequency is less than fs/4, pole is
placed at z=a.
fc
a 1 2
fs
• When cutoff frequency is greater than fs/4, pole is
placed at z=-a.
fc
a  (1    2 )
fs
• In both the cases, zero is placed at z=1 to ensure
zero frequency (DC component) is suppressed.
HPF with cutoff frequency is less than fs/4

Frequency
response

Unit 0 fC fS /2=fmax
circle
HPF with cutoff frequency is greater than fs/4

Frequency
response

0 fC fS /2
=fmax
Unit
circle
Digital sinusoidal oscillator8.47 352
Comb filter 8.49
 Effect of Pole position on band width & center frequency

BW

r z re j

Unit
circle

• Band width: Frequencies covered by arc on circle obtained by drawing

two lines from origin around the pole on to circle of radius r

• Effect of pole position on bandwidth: increase r increases BW


• Effect of pole position on center frequency: depends on angular
displacement of the pole from real axis, i.e.,