AVR 1

UNIT III - IIR FILTER DESIGN

Syllabus:
Structure of IIR – System design of discrete time IIR filter from continuous time filter
– IIR filter design by impulse invariance – Bilinear transformation – Approximation
derivatives – Design of IIR filter in the frequency domain.
Phase I
1. Structure of IIR –
2. System design of discrete time IIR filter from continuous time filter
Approximation derivatives –
3. IIR filter design by impulse invariance –
4. Bilinear transformation –
5. Design of IIR filter in the frequency domain.
TEXT BOOK
1. John G Proakis and Dimtris G Manolakis, “Digital Signal Processing Principles
- Algorithms and Application”, 3rd Edition, PHI/Pearson Education, 2000.
REFERENCES
1. Alan V Oppenheim, Ronald W Schafer and John R Buck, “Discrete Time
Signal Processing”, 2nd Edition, PHI/Pearson Education, 2000.
AVR 2

NOMENCLATURE
F s : Sampling Rate
Relation between Discrete Time signal and Analog
1 Time signal Frequencies:
T = : Sampling Time
Fs
Ω=2 πF ( Analog signal)
F pass : Analog passband edge frequency
ω=2 πf (Digital Signal)
F stop : Analog stopband edge frequency F
f=
f p : Digital passband edge frequency Fs

f s : Digital stopband edge frequency 2 πF

ω= =ΩT
Fs
AVR 3

IIR FILTER DESIGN TECHNIQUE

System design of discrete time IIR filter from continuous time filter
– IIR filter design by Approximation Derivative
H ( z ) for IIR filter is a rational function .
❑
∑ ❑❑ ❑❑ω ¿
() ❑

(∑ ❑ ❑ )
❑
❑
❑
❑
H ( z )−→ H ( ω)

( ❑❑ ❑❑ ❑❑ )
ω H ( ω ) drawn ¿ Pole zero plot () ω ¿ Analog Filter is converted to Digital Filter by
❑❑ ❑❑ ❑❑
1. Approximation Derivatives
2. Impulse Invariance
3. Bilinear Transformation
If the conversion technique is to be correct, it should possess the following desirable properties:
(Condition to satisfy for Correct Conversion)
1.The jΩ axis in the s-plane should map into the unit circle in the z-plane.
2. A stable analog filter should be converted to a stable digital filter.
Method1: Approximation Derivative:

L {dydt }=s . Y ( s )
Backward difference

Z ( ( y ( n ) − y ( n−1 ) )
T ) =
1−z−1
T
Y (z )

1−z −1
s= ( Mapping¿ s¿ z plane )
T
Problems:
1.Convert the analog filter given below to digital filter using backward difference method.
Assume T = 0.1sec.
1
H ( s )=H a ( s )=
( s +0.1 )2 +9
Solution:
Simplifying the H(s):

1 1
H ( s )= 2
= 2
s + 0.01+0.2 s+ 9 s + 0.2 s+ 9.01
−1
1−z
Substituting s=
T
1
H ( z )=
( ) ( )
−1 2
1−z 1−z−1
+0.2 +9.01
T T
1
H ( z )=
( 1+ z −2 z )
( )+9.01
−2 −1 −1
1−z
2
+ 0.2
T T
AVR 4

2
T
¿
1+ z −2 z +0.2 T ( 1−z ) +9.01T
−2 −1 −1 2

2
T
¿
1+ z −2 z +0.2 T −0.2 T z−1 +9.01 T 2
−2 −1

Rearranging like :Constant + z−1 terms + z−2 terms in the denominator,

T2
¿
1+ 0.2T +9.01 T 2−2 z−1−0.2T z −1 + z−2
2
T
¿ 2 −1 −2
1+ 0.2T +9.01 T −(2+0.2 T )z + z
Usually form is:1+ z−1 terms+ z −2 terms:
2
T
2
1+0.2 T + 9.01T
¿
2+ 0.2T 1
1− 2
z−1 + 2
z −2
1+ 0.2T +9.01 T 1+ 0.2T +9.01 T
Find the poles if T = 0.1.
0.009
H ( z )=
1−1.8196 z−1+ 0.9008 z −2
0.009 0.009
¿ =
( 1− p1 z )(1− p2 z−2) ( 1−(0.91+ j0.27) z−1 ) ( 1−( 0.91− j 0.27 ) z −2)
−1

p1,2=0.91± j 0.27

So H(z) in this problem, act like a Low Pass Filter.

Disadvantage of Approximation Derivative Method:

1
z=
1−sT

Fig: 3.2 s-plane to z-plane mapping

 jΩ axis is mapped into unit circle (small region in unit circle)
 LHP of s-plane (stability of s-plane) is mapped inside the unit circle of z-plane (stability
of z-plane)
Both the conditions are satisfied. But it has a problem in designing high pass filter.
For digital high pass filter, pole should be located near z = -1 point. But any points in the s-plane
(even for Analog high pass filter), is mapped into z= 0 to 1. So high pass filter can't be designed
using Approximation Derivative method. That’s the disadvandage of this method.
System design of discrete time IIR filter from continuous time filter
AVR 5

– IIR filter design by Impulse Invariance

Method2: Impulse Invariance Method
h ( n ) =ha ( nT )
Impulse response of continuous filter is sampled to get identical impulse response in digital
filter. (Impulse response is not different(invariant) from analog filtering. That's why the name:
Impulse Invariance)
Mapping of s to z plane:
Sampling Z
L-1
Ha(s) ha(t) h(n)=ha(nT) H(z)

Mapping s and z

Fig:3.3 Block Diagram of Impulse Invariance Method

N
ck
H a ( s ) =∑ −→(1)
k =1 s− p k

Applying Inverse Laplace Transform: L ( s−a

1
)=e at

N
h a ( t )=∑ c k e p t , t ≥ 0 k

k=1

Applying Impulse Invariance idea:

N N
h ( n ) =ha ( nT )=∑ c k e pk nT
=∑ c k e p Tn , n ≥0
k

k=1 k=1

Applying Z Transform:
∞
H ( z )= ∑ h ( n ) z
−n

n=−∞

∞ N N ∞ N ∞
H ( z )=∑ ∑ c k e p Tn z −n=∑ c k ∑ e p Tn z−n=∑ c k ∑ ( e p T z−1 )
n
k k k

n=0 k=1 k=1 n=0 k=1 n=0

∞
1
Since, ∑ a =
n
, then ,
n=0 1−a

( )
N
ck
H ( z ) =∑ −→ ( 2 )
1− ( e z )
pk T −1
k=1

Comparing equation 1∧2 ,then ,

Pole, pk in the s-plane is plotted as pole, e p T in z-plane. k

So if we have:
c1 c c ❑ c2 c1
H a ( s )= + 2 + …+ N −→ H ( z )= ❑ ❑ ❑ + + …+
s− p1 s− p2 s− pN ❑ ❑ 1−e zp T −1 p T −1
1−e z ❑ 2 N

Equation simplification Useful to analyze problems:

2 2 2 2 2
s +2 as+ a +b = ( s+ a ) +b =( s+ a+ jb ) ( s+ a− jb)
Problem:
2.Convert the analog filter with system function
s+0.1
H a ( s )=
( s +0.1 )2 +9
AVR 6

into a digital IIR filter by means of the impulse invariance method. Assume T = 0.1 second.
Solution:
To convert the poles in s-plane to z-plane. So first need to find pole in s-plane.
Pole in Analog Filter is, pk is found using equating the denominator to zero.
( s+0.1 )2 +9=0
s=−0.1± j3=poles , pk
s+0.1
H a ( s )=
( s +0.1+ j 3 ) ( s +0.1− j3 )
Only summation of poles in s-plane can be replaced by summation of poles in z-plane.
So use partial fraction expansion, to convert from product form to summation of poles form.
Hence,
s+0.1 c c'
H a ( s )= = +
( s +0.1+ j 3 ) ( s +0.1− j3 ) ( s+0.1+ j3 ) ( s+ 0.1− j3 )
If Residue , c=a+ jb, then Residue, c '=a− jb
To find c using Residue Method:
Numerator
Residue=
Differentiation of Denominator
s+0.1
Residue=
d s+ 0.1 s +0.1 1
(s ¿ ¿ 2+0.2 s +9.01)= = = ¿
ds 2 s+0.2 2 ( s+ 0.1 ) 2
1 ' 1
c= ∧c =
2 2
1 1
2 2
H a ( s )= +
( s +0.1+ j 3 ) ( s+ 0.1− j 3 )

Now ∈∑ of poles form . Apply Impulse invariance method mapping , ( pk → e )

pk T

Pole : s1=−0.1− j3∧s 2=−0.1+ j3

1 1 1 1
2 2 2 2
H ( z )= + = +
( 1−e (−0.1− j 3) T z−1 ) ( 1−e(−0.1 + j 3) T z−1 ) ( 1−e−0.1T e− j 3 T z−1 ) ( 1−e−0.1T e+ j 3T z −1 )
Convert the system function to implementable form:
1
( 1−e−0.1 T e j 3 T z−1 +1−e−0.1 T e− j 3T z −1)
2
H ( z )=
( 1−e−0.1T e j 3 T z−1 ) ( 1−e−0.1T e− j 3 T z−1)

1
H ( z )= ¿ ¿
2

1 Basic Maths to help you:

H ( z )= ¿ ¿
2
e jθ =cosθ+ jsinθ
jθ − jθ
e +e =cosθ+ j sinθ +cosθ− jsi
0
e =1
AVR 7

1
( 2−e−0.1T ( 2 cos 3 T ) z−1 )
2
H ( z )=
1−e−0.1 T ¿ ¿
−0.1T −1
1−e c os 3 T z
H ( z )= −0.1T
1−e ¿¿
Given T = 0.1,

cos ( radians )=cos ( radians∗180

π )
−1
1−0.946 z
H ( z )=
1−1.892 z−1 +0.98 z−2

Note on Butterworth Filter:

Find the poles Using Butterworth Circle

Any of the three IIR Design Technique can be

Given Digital Analog Analog Digital

Butterworth Filter Butterworth Filter Butterworth Filter Butterworth
Specification specification Design Filter Design
H(z)

Fig: 3.4 Butterworth Filter Design

Digital Filter specification can be given either directly or indirectly. Using the specification,
draw the frequency response that gives you an idea about the filter.
Indirect Digital Specification (We need 5 parameters):
Passband Ripple: 20 log ⁡(1+ δ¿¿ 1)¿ (dB) at Passband edge frequency: F pass
Stopband Attenuation: 20 log δ 2 (dB) at Stopband edge frequency: F stop
Sampling Rate, F s
From this,
1−δ 1 ≤ H ( e jω ) ≤1 , 0 ≤ ω ≤ω p wher e ω p =2 π f p , f p=(F ¿ ¿ pass /F s )¿

H ( e jω ) ≤ δ 2 ,ω s ≤ ω ≤2 π whereω s=2 π f s , f s=(F ¿ ¿ stop/ F s) ¿

As shown in Fig: 3.5
AVR 8

¿∨H ¿ )|

Fig: 3.5 Practical Filters

Butterworth Filter is givenby :
2 1
|H a ( j Ω )| =H a ( s ) H a (−s )=
( )
2N
jΩ
1+
j Ωc

, where N is the order of the Butterworth Filter , Ω c isthe cutoff freque ncy where
1
the magnitude becomes of the maximum value .
√2

3. Design a digital Butterworth filter satisfying the constraints.

a)
π
0.707 ≤|H ( e )|≤1 for 0≤ ω ≤
jω
2

|H ( e jω )|≤ 0.2 f ∨ 34π ≤ ω ≤ π

With T=1 sec using impulse invariance method.
(OR)
b)
Passband Ripple: 4.64 dB , Passband Edge Frequency¿ 0.25 Hz ,
Stopband Ripple−14 dB , Stopband Edge Frequency=0.375 Hz
Sampling Rate: 1Hz
Solution:

Convert the Given digital filter specification, H ( e jω ) to analog filter specification, H ( j Ω)

Analog Filter specification:

In impulse invariance, ω=Ω T
π
0.707 ≤|H a ( j Ω )|≤1 for 0 ≤ ΩT ≤
2
AVR 9

3π
|H a ( jΩ )|≤ 0.2 for 4
≤ ΩT ≤ π

Substituting , ΩT =Ω ,(Since T =1 is given)

0.707 ≤|H a ( j Ω )|≤1 for 0 ≤ Ω≤ 0.5 π Analog Filter Specification is
found. Next need to find Ha(s):
|H a ( jΩ )|≤ 0.2 for 0.75 π ≤ Ω≤ π
Rearranging the condition:
At Ω=0.5 π , H a ( j0.5 π ) ≥0.707 (1)

At Ω=0.75 π , H a ( j0.7 5 π ) ≤0.2

For a Butterworth Filter,

2 1
|H a ( j Ω )| =H a ( s ) H a (−s )= ( 2)
( )
2N
jΩ
1+
j Ωc

Substituting (1) in (2), ≥∧≤ both are modified ¿ equality , ¿ find N∧Ωc
2 1 2 1
|H a ( j 0.5 π )| =|0.707| = 2N | a
H ( j 0.75 π )| =|0.2| =
2 2

( ) ( )
2N
j 0.5 π j0.75 π
1+ 1+
j Ωc j Ωc

( )
2N
j0.5 π 1
1+ = =2
j Ωc 0.70 72 Taking Log on both sides,
2Nlog

( )
2N
0.5 π
=1
Ωc

Taking Log on both sides,

2Nlog ( 0.5Ω π )=0

c

2 Nlog ( 0.5 π ) −2 N log ( Ω c ) =0

2 N log ( 0.5 π )=2 N log ( Ωc )

log ( 0.5 π )=log ( Ωc )

Taking anti-log on both sides,Ωc =0.5 π

To find Ha (s), we need to find the poles for the Butterworth filter. Poles for Butterworth
filter is found using N and Ωc :

 Draw the circle of radius, Ωc in the s-plane

 Divide the circle by 2N poles equally. (Left Half Plane and Right Half Plane, N poles
each)
a b c
 From this, find the values of the poles using the formula: = = .
sinA sinB sinC

N
Ωc
H a ( s )= ( obtained¿ ( 2 ) )
( s− p 1) ( s− p2 ) ( s− p3 ) ( s−p 4 )
AVR 10

Im(s)

Re(s)

Fig:3.6 Butterworth Circle to find the poles

p1∧ p2 are conjugate pairs . Similarly , p 3∧ p4 are conjugate pairs .

ℑ ( p1 ) Ωc ℜ ( p 1)
= = ℑ ( p1 ) =Ωc sin22.5=0.6011 ; ℜ ( p1 )=1.4512
sin(22.5) sin 90 sin ( 180−90−22.5 )
p1=−1.4512+ j0.6011 , p 2=−1.4512− j0.6011 (Negative because left half of the s-plane)

ℑ ( p3 ) Ωc ℜ ( p3 )
= =
sin ( 45+22.5 ) sin 90 sin ( 180−90−67.5 )

ℑ ( p 3) =Ωc sin 67.5=1.4512 ; ℜ ( p3 )=0.6011

p3,4 =−0.6011 ± j 1.4512 (Negative because left half of the s-plane)

FINDING THE POLES CAN ALSO BE DONE USING FORMULA AS DID IN THE FINAL
PROBLEM (USING BILINEAR TRANSFORMATION) IN THIS UNIT.

Therefore, Ha (s) can be written as:

( 0.5 π )N
H a ( s )=
( s +1.4512− j0.6011 ) ( s +1.4512+ j 0.6011 ) (s +0.6011−1.4512)(s +0.6011+1.4512)

( s+a+ jb ) ( s+ a− jb )=( s+a )2−( jb )2=s2 +2 as+ a2 +b 2

For implementing Impulse Invariance, sum of poles form is needed. To do it, we need to find the
partial fraction expansion using Residue method.

c1 c1 ' c2 c2 '
H a ( s )= + + +
s +1.4512− j0.6011 s+1.4512+ j0.6011 s+ 0.6011− j 1.4512 s+0.6011+ j1.4512
To find c 1∧c 2 using Residue Method :
Numerator
Residue=
Differentiation of Denominator
( 0.5 π )4
Residue=
d
( ( s+ 1.4512− j 0.6011 ) ( s+1.4512+ j0.6011 ) ( s +0.6011−1.4512 ) ( s +0.6011+1.4512 ) )
ds
6.09
Residue=
d
¿¿
ds
d
¿
ds

6.09
Residue=
( s +2.9 s +2.47 ) ( 2 s+ 1.2 )+ ( s 2+ 1.2 s+2.47 ) (2 s+2.9 )
2
 AVR 11
Basic Maths to help you:

Residue , c1 at s=−1.4512+ j 0.6011 , e jθ =cosθ+ jsinθ

jθ − jθ
then using calculator ∈Complex mode , e +e =cosθ+ jsinθ+cosθ− jsinθ
jθ − jθ
6.09 e −e =2 jsinθ
c 1= =0.71− j1.75
1.22+ j 2.99 0
e =1
then ,c '1=0.71+ j1.75

Residue , c2 at s=−0.6011+ j 1.4512 ,

thenusing calculator∈Complex mode ,
6.09
c 2= =−0.73+ j0.3
−7.13− j 2.95
'
then ,c 2=−0.73− j1.75

0.71− j1.75 0.71+ j1.75 −0.73+ j 0.3 −0.73− j0.3

H a ( s )= + + +
s +1.4512− j0.6011 s+1.4512+ j0.6011 s+ 0.6011− j 1.4512 s+0.6011+ j1.4512
Applying Impulse Invariance Technique, pk → e p T k

0.71− j1.75 0.71+ j 1.75 −0.73+ j 0.3 −0.73− j0.3

H (z )= −1.45 T + j 0.6 T −1
+ −1.45 T− j 0.6 T −1
+ −0.6 T + j1.45 T −1
+ −0.6 T − j 1.45 T −1
1−e z 1−e z 1−e z 1−e z

Substituting, T = 1 and writing, e a+ b=e a +e b

H1(z) H2(z)

0.71− j 1.75 0.71+ j1.75 −0.73+ j 0.3 −0.73− j 0.3

H (z )= −1.45 j 0.6 −1
+ −1.45 − j 0.6 −1
+ −0.6 j1.45 −1
+
1−e e z 1−e e z 1−e e z 1−e−0.6 e− j 1.45 z−1

¿ Convert the equation∈implementable form , combine the complex conjugate terms:

( 0.71− j 1.75 ) ( 1−e−1.45 e− j 0.6 z−1 ) + ( 0.71+ j 1.75 ) ( 1−e−1.45 e j 0.6 z−1)
H 1 (z )=
( 1−e−1.45 e j 0.6 z−1 ) ( 1−e−1.45 e− j 0.6 z−1 )
0.71−0.71 e−1.45 e− j 0.6 z −1 − j 1.75+ j 1.75 e−1.45 e− j 0.6 z−1 +0.71−0.71 e−1.45 e j 0.6 z−1 + j 1.75− j1.75 e−1.45 e j 0.6 z −1
H1 (z )=
1−e−1.45 e− j 0.6 z−1−e−1.45 e j 0.6 z−1 +e−2.9 z−2
−1.45
H 1 ( z ) =1.42−0.71 e ¿¿¿

−1.45
1.42−0.71e ( 2 cos 0.6 ) z−1+ j 1.75 e−1.45 (−2 jsin 0.6 ) z−1
H1 (z )=
1−e−1.45 ( 2 cos 0.6 ) z−1 +e−2.9 z−2

cos ( radian )=cos ( radian∗180

pi ) 2
j =−1

H1 (z )=
( 0.6∗180
1.42−1.42 e−1.45 cos
π ) −1
z +3.5 e sin (
0.6∗180
π )z
−1.45 −1

cos ( ) z +e z
−1.450.6∗180 −1 −2.9 −2
1−2 e
π
Using Calculator,
−1 −1 −1
1.42−0.27 z + 0.46 z 1.42+ 0.19 z
H 1 (z )= −1 −2
= −1 −2
1−0.39 z + 0.055 z 1−0.39 z +0.055 z
Similarly proceeding with H 2 ( z ) ,
AVR 12

(−0.73+ j 0.3 ) ( 1−e−0.6 e− j1.45 z−1 ) + (−0.73− j 0.3 ) ( 1−e−0.6 e j 1.45 z−1 )
H 2 (z )=
( 1−e−0.6 e j 1.45 z −1) ( 1−e−0.6 e− j 1.45 z−1)
−0.6 − j 1.45 −1 −0.6 − j 1.45 −1 −0.6 j 1.45 −1 −0.6 j1.45 −1
−0.73+ 0.73 e e z + j0.3− j 0.3 e e z −0.73+0.73 e e z − j0.3+ j0.3 e e z
H2 (z )= −0.6 − j1.45 −1 −0.6 j1.45 −1 −1.2 −2
1−e e z −e e z +e z

−0.6
H 2 ( z ) =−1.46+0.73 e ¿¿

−0.6
−1.46+ 0.73 e ( 2cos 1.45 ) z−1− j0.3 e−0.6 (−2 jsin 1.45 ) z−1
H 2 (z )=
1−e−0.6 ( 2cos 1.45 ) z−1 +e−1.2 z−2
Aft er converting radians¿ degree∧using calculator ,
−1.46+ 0.1 z −1 −0.33 z −1 −1.46−0.23 z−1
H 2 (z )= =
1−0.13 z−1+ 0.3 z −2 1−0.13 z−1 +0.3 z−2
H ( z )=H 1 ( z ) + H 2 ( z )
−1 −1
1.42+0.19 z −1.46−0.23 z
H ( z )= +
1−0.39 z−1+ 0.055 z −2 1−0.13 z−1+ 0.3 z −2

Mapping and Disadvantage of Impulse Invariance Method:

Frequency Domain Representation:
Since in Impulse Invariance technique, Impulse response of analog filter is sampled to
find the corresponding digital filter, frequency response of the digital filter follows the discrete
time signal property. (i.e., response repeats every 2 π cycle Ω) highest frequency is at π .)
H a ( jand

Aliasing Effect Ω

Fig:3.7 Aliasing Effect in Impulse Invariance Technique

( )
+∞
1 jω j 2 πk
H ( e )= ∑ H a
jω
+
T k=−∞ T T
p T
Since mapping corresponds ¿ poles , s= p k =¿ z=e k

z=esT
z=ℜ jω ; s=σ + j Ω
r e jω =e σT e j Ω T
σT jω j ΩT
r =e ; e =e
ω=Ω T ,
Mapping:
For a stable analog fitler , σ <0 ,then r< 1. Hence stable digital filter .
¿
digital filter specification ¿ analog filter specification ¿
Aliasing Effect:
−π <ω< π
−π <Ω T < π
−π π
< Ω<
T T
❑
Since Ω varies from −∞ ¿+∞ , other values of Ω ❑ are also mapped to same−π ¿ π region,
resulting in aliasing effect. (More than one frequency in continuous domain has the same
frequency in digital domain. This is called aliasing.)
Advantage of Impulse Invariance:
AVR 13

Filter Shape is preserved if aliasing effect is ignored.

Disadvantage of Impulse Invariance:
High Pass Filter can’t be designed due to aliasing effect because of more cross over
compared to low pass filter.
System design of discrete time IIR filter from continuous time filter
– IIR filter design by Bilinear Transformation
Approximate Derivative and Impulse Invariance is not useful to design high pass filter. But,
Bilinear Transformation is used to design any filter.
Features of Bilinear Transformation:
 maps the j Ω axis into the unit circle only once (unlike impulse invariance approach),
hence no aliasing effect
 Left half of s-plane is mapped into the unit circle (hence stable analog filter to stable
digital filter)
 Right half of s-plane is mapped outside the unit circle.
Derivation:
Convert to Input and Convert to Digital
Output terms, (X(s) Convert to Time
dd domain (x(t) and y(t)) Domain by means of
and Y(s)) sampling (t = nT)

Apply Trapezoidal
Represent Derivative
formula and convert
in Integral form By Substituting, H(z) is found
to digital domain (t
= nT, t0 = nT-T) by taking Z transform and is
T) compared with Ha (s) to find the
mapping between s and z.

Fig:3.8 Steps to Derive Bilinear Transformation

Let assume,
b
H a ( s )=
s +a
Y ( s)
Since, Y(s) = Ha(s) X(s) ==> Ha(s) =
X (s )
Y ( s) b
= → Y ( s )( s+a )=bX ( s ) → sY ( s )+ aY ( s )=bX ( s )
X ( s ) s+ a
Taking Inverse ℒ Transform ,
1
Laplace Recall: L(dy/dt) =dysY(s)
(t)
+ay ( t )=bx(Lt)( e )=
at

dt s−a

(
y ' ( t ) +ay ( t ) =bx ( t ) since : y ' ( t )=
dy (t)
dt )
C onverting ¿ digital domain ,t=nT ,

y ' ( nT ) +ay ( nT )=bx ( nT )

y ' ( nT )=−ay ( nT ) +bx ( nT )

Since, T is understood implicitly,( i.e. ) y(nT) = y(n), then,

y ' ( n )=−ay ( n ) +bx ( n )−(1)

Represent derivative in integral form:
dy ( t )
∫ dt
dt= y ( t )
AVR 14

∫ y ' ( t ) dt+ y ( t0 ) = y ( t )
t0

Trapezoidal Rule:

t
( y' ( t ) + y' (t 0 ))
∫ y ( t ) dt=( t −t 0 )
'
2
t0

( y ' ( t )+ y ' ( t 0 ) )
( t−t 0 ) 2
+ y ( t 0 )= y ( t )

Converting to Digital Domain, t = nT and previous time, t0 = nT −¿ T,

( y ' ( nT ) + y ' ( nT −T ) )
( nT −( nT −T ) ) + y ( nT −T ) = y ( nT )
2
T
. ( y ( nT )+ y ( nT −T ) ) + y ( nT −T ) = y ( nT )
' '
2
Since T is understood implicitly,
T
. ( y ( n ) + y ( n−1 ) ) + y ( n−1 )= y ( n ) → ( 2 )
' '
2

Substitute (1) in (2), ( y ' ( n )=−ay ( n )+ bx ( n )∧¿ y ' ( n−1 )=−ay ( n−1 )+ bx ( n−1 ) since LTI system )

T
( −ay ( n ) +bx ( n )−ay ( n−1 )+ bx ( n−1 ) ) + y ( n−1 ) = y ( n )
2
Rearranging the above equation:
bT
2
( x ( n )+ x ( n−1 )) + 1−
aT
2 (
y ( n−1 )= y ( n )+
aT
2 )
y (n)

(1+ aT2 ) y ( n )−(1− aT2 ) y ( n−1)= bT2 ( x ( n) + x ( n−1) )

Taking Z Transform¿ find H ( z ) :

(1+ aT2 )Y (z)−(1− aT2 ) z −1

Y (z )=
bT
2
( X ( z)+ z−1 X ( z ) )

(
Y ( z ) 1+
aT
2
− 1−
aT −1
2 (
z = X(z)
bT
2 ) )
( 1+ z−1 ) ( )
bT bT
( 1+ z−1 ) ( 1+ z −1)
Y ( z) 2 2
H ( z )= = =

( ) ) ( )
X (z )
1+
aT
2
− 1−
a T −1
2
z ( 1−z−1 +
aT
2
( 1+ z−1 )

By comparing with Ha(s), only b is in the Numerator, Divide the numerator and denominator by:
T
( 1+ z −1)
2
AVR 15

bT
2 (( 1+ z −1) )
T
( 1+ z −1 )
2 b b
H ( z )= = =
( −1 aT ( 1−z )
) 2 ( 1−z )
−1 −1
1−z + ( 1+ z−1 ) +a +a
2 T( T ( 1+ z
−1
)
1+ z )
−1
T 2
( 1+ z )
−1
2
b
Comparing with H a ( s )= , then
s +a

2 ( 1−z )
−1
s=
T ( 1+ z )
−1

This mapping of s and z is called Bilinear Transformation.

To find relation between ω∧Ω :

2 ( 1−e )
− jω
jω −1 − jω
Since z=e , then z =e , s=
T ( 1+ e )
− jω

jω
Multiply thenumerator ∧denominator by e 2 :

) = 2 (2 jsin ( 2 )) = 2 j
( )( ω2 ) = 2 j tan ω
ω
2(e ) = 2(e
jω jω jω jω − jω
sin
2 ( 1−e )∗e
(2)
− jω 2 − jω 2 2 2
2
−e e −e
s= =
T (e ) T (e ) T ( 2 cos( ω2 )) T cos ( )
jω jω jω jω − jω
ω T
T ( 1+ e )∗e
− jω 2 2 − jω 2 2 2
+e e +e 2

s=σ + jΩ=
2j
T
tan
ω
2 ( )
Equatingimaginary parts ,

Ω=
2
T
tan
ω
2 ( )
¿ ω=2 tan −1 ( ΩT2 )
Mapping of ω∧Ω and Frequency Warping:

−π <ω< π

( ΩT2 )< π
−π <2 tan −1

< tan (
2 ) 2
−π ΩT π −1
<
2

−tan ( )< ( )< tan ( ) →−∞ < (

2 )
π ΩT π ΩT
<+∞
2 2 2
−∞< Ω<+ ∞
So entire Ω is mapped into the unit circle only once, unlike impulse invariance technique.
Hence no aliasing effect.
AVR 16

Frequency
Warped

ω=2 tan −1 ( ΩT2 )

Fig: 3.9 Frequency Warping

From Fig:3.9, compression of frequency response in digital domain has been obtained after
certain frequency (ω s ¿which is called as Frequency Warping. It is due to the non-linear relation
between ω∧Ω . To meet the desired discrete time filter, band edge frequencies are prewarped using
2
Ω= tan
T ( )
ω
2

Advantage of Bilinear Transformation:

 Used to design all filters
 No aliasing effect as in Impulse Invariance Method
Disadvantage of Bilinear Transformation:
 Frequency Warping. (To avoid this, pre-warping is done)

Problems:
( s +0.1 )
1. Convert the analog filter: H a ( s )= into digital IIR filter by means of bilinear
( s +0.1 )2 +16
π
Transformation. Given: Digital should have resonant frequency, ω r=
2
Solution:
Before applying the mapping directly, from the given data, T can be obtained.

( s+a )2+ b2=¿> b iscalled resonant frequency , Ωr

From the given H a ( s )=¿> Ωr=4 Resonant Frequency: Frequency at which Filter
frequency response is exactly unity (Maximum)
Ω=
2
T
tan
ω
2 ( )
()
π
2 2 1
4= tan −→ T =
T 2 2

( ) ( )
−1 −1
2 1−z 1−z
Bilinear Transformation, s= =¿> , s=4
T 1+ z −1 1+ z
−1

H ( z )=[ H a ( s ) ]
( )
−1
1−z
s=4 −1
1+z

( 0.128+0.006 z−1−0.122 z−2 )

H ( z )= −1 −2
1+0.0006 z + 0.975 z
AVR 17

2. Design a single-pole low pass digital filter with a 3-dB bandwidth of 0.2 π , using the
Bilinear transformation applied to the analog filter
Ωc
H a ( s )=
s +Ωc

Where, Ωc is the 3-dB bandwidth of the analog filter.

Solution:
Before applying the mapping directly, from the given data, Ωc can be obtained.
3−dB bandwidth∨Cutoff frequency for digital filter , ω c =0.2 π

Since Bilinear Transformation , Ω=

2
T ( )
tan
ω
2

Ωc =
2
T ( )
tan
ωc 2
2
= tan
T
0.2 π
2 ( )
2
= tan ( 0.1 π )=
T
0.65
T

Value of T is unknown, let's proceed with it,

0.65
T
H a ( s )=
0.65
s+
T
H ( z )=[ H a ( s ) ]
( )
−1
2 1−z
s=
T 1+z −1

0.245 ( 1+ z )
−1
H ( z )=
1−0.509 z−1
(If T is not cancelled, assume T = 1 in case T is not given)
3. Design a digital filter from

❑❑ () ❑❑
()
Solution:

❑❑ () ❑❑ ❑❑ ❑❑
() ❑ ❑

() [ ❑❑ () ] ❑
( )
❑
❑
❑ ❑❑

()
❑ ❑❑
❑ ❑❑( )
( ( )) ( ( ))
❑
❑ ❑❑ ❑ ❑❑
❑ ❑❑ ❑ ❑❑

( ( ❑❑ ) (❑❑) )
( ❑❑)
()
( ❑❑ )❑ ❑ ❑❑
❑ (❑ ) ❑ ❑
❑ ❑❑ ❑ ( )
( ( ❑❑ ) (❑❑) )
( ❑)
() ❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑
(❑ ) ( ❑ ) (❑ ) ❑ ( ❑ )
❑ ❑❑
❑ (❑ )
AVR 18

( (❑❑ ) (❑❑) ) ( ( ❑❑ ) )
()
( ❑❑ )❑ ( ❑❑) ( ❑❑) ❑❑ ( ❑❑ )❑

Since T is not given either directly or indirectly, assume T = 1, then,

( (❑❑ ) ( ❑❑ ) ) ( ❑❑)
()
( ❑❑ )❑ ( ❑❑) ( ❑❑) ( ❑❑ )❑

( ( ❑❑ ) (❑❑❑❑) )
()
( ❑❑ ❑❑ ) ( ❑❑ ) ( ❑❑❑❑ )

( ❑❑ ❑❑ )
() ❑ ❑
❑ ❑

( ❑❑ ❑❑ )
() ❑ ❑
❑ ❑
Try to implement atleast in Direct form I. (Since asked for 16 marks)
( ) (❑❑ ❑❑)
()
( ) ❑❑ ❑❑

() ( ❑❑ ❑❑ ) () ( ❑❑ ❑❑)
()❑❑( )❑❑ ()= ( )❑❑ ()❑❑ ()

()()( )❑❑ ()
( )()( )( )
Now you can implement this in direct form I easily (as we did before).

4. Design a digital Butterworth filter satisfying the following constraints:

|( ❑❑)|
|( ❑❑)|
using Bilinear Transformation.
Solution:
Digital to Analog Specification:

|()|❑❑( ❑
❑)

|()|❑❑( ❑
❑)
Taking only the Passband limit:
❑ ❑❑ ❑ ❑ ❑ () ❑ ( )
❑ ( )
❑ ❑❑ ❑
()
Taking only the stopband limit:
❑ ❑❑ ❑ ❑ () ❑ ❑ ❑
❑ ( )
❑ ❑ ❑ ❑ ❑ ( )
()
Substituting the limit,

|()|
AVR 19

|()|

()
()
Analog Butterworth Filter Design:
❑ ❑
|()| ❑

( )
❑
❑❑

|()| ❑❑ ❑ ❑ |()| ❑❑ ❑ ❑
❑ ❑

( ❑❑ ) ❑
( ❑❑ ) ❑

❑ ❑

( )
❑
❑❑
❑
❑
❑ ( )
❑
❑❑
❑
❑
❑

❑ ❑

( )
❑
❑❑ ( )
❑
❑❑
Taking Log on both sides, Taking Log on both sides,
❑❑ ❑❑
❑❑ ❑❑ – (2)

To solve (1) and (2), substitute, ( )❑❑

❑
❑
Assume N=2 ( always greater ∨equal integer than what we obt ained )
Substitute N = 2 in (1) (satisfying passband criterion),
❑❑
❑❑
❑❑

Since N = 2, so two poles, then H a ( s )

❑
❑❑ ❑
❑
❑
❑❑ ()
( ❑❑ ) ( ❑❑ ) ( ❑❑) ( ❑❑ ) ( ❑❑ )( ❑❑)

Poles for Butterworth filter is given by:

❑❑ ❑❑ ❑ ❑ ❑
( () ❑ )

( () ❑)
❑❑ ❑ ❑ ❑
( ❑ ❑ ) ( ❑)
❑❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑
❑ ❑ (( )( ))
()
( ❑ ❑ ) ( ❑)
❑❑ ❑ ❑ ❑ ❑ ❑ ❑ ❑
❑ ❑ (( )( ))
AVR 20

❑❑ () ❑ ❑❑
()() ❑
❑❑ ❑❑ ❑❑()❑❑❑ ()
Analog filter design is over. Now analog filter can be converted to digital filter using

Bilinear Transformation using s=

2 1−z −1
T 1+ z −1(=2
1−z−1
1+ z
−1 ) ( )
, (Since T is not given, assume T

=1)

() ❑

(( )) (( ❑❑ ))
❑ ❑ ❑
❑
❑❑ ❑

Similarly proceeding (as we did with previous problems):

( ❑❑ ❑❑ )
()
❑❑ ❑❑

( ❑❑ ❑❑) ❑❑❑❑
()
❑❑ ❑❑ ❑❑❑❑
Thus digital filter design is over.
Other commonly used Analog filters:

Structures of IIR (Partially covered)

Realization of IIR is possible through recursive structures. Recursion means output depends on previous
output. Recursion is also called as FEEDBACK.
Realization of y(n) = −¿ a1 y(n-1) +box(n) +b1 x(n-1)

Direct Form I
AVR 21

Conversion from Form I to II

Direct Form II
AVR 22

Useful Equations To remember during Exam - UNIT III

Useful to analyze problems:
❑ ❑ ❑ ❑ ❑
❑ ❑ ❑ () ❑ ( )
−1
1−z
Approximation Derivative: s=
T
Impulse Invariance, Partial fraction expansion can be found using:
❑
❑
❑ ❑
❑❑ ❑❑ ❑❑ ❑

Bilinear Transformation:

❑ ❑❑ ( ) ( )
❑ ❑❑ ❑❑ ❑
❑
Formulas to Solve Butterworth Filter Design:

|()| ()() ❑ ❑
❑

( ❑❑ )
❑
If Impulse Invariance:

If Bilinear Transformation: ω=2 tan

−1
( ΩT2 )
❑❑|( )| ()( )
❑ ❑
❑

( )
❑
❑❑
❑❑
❑❑ () ❑
()

❑ ❑ ❑ () ❑ ( ) Poles , s =Ω e j( ) , k=0,1,2 , … , N −1
( 2 k 1) π π
❑ ❑ ❑ ❑ ❑ +
2N 2
k c

Hint :
Only if find half of the poles is sufficient , others can be found usingcomplex conjugate .

❑❑ () ❑ ❑❑ () ❑❑ ❑❑ ❑❑ ❑❑ ❑❑ ❑
()() ❑ ❑ ❑ ❑ ( )()() [ ]❑
❑❑ ❑ Laplace Transform to Know:
❑ ❑❑ ❑❑ ❑❑ ❑

()() ( ❑( ) )() (❑ ) ❑❑ ❑

Z Transform to Know:
{ ()} () (() ) ❑❑()
AVR 23

Transforming Z plane to frequency domain: (needed for good understanding)

It gives the understanding of what kind of filter we are designing from its H(z).
H ( z )−→ H ( ω) H(ω ¿

0 π

Fig:3.1a Z to ω Conversion 3.1b H ( ω ) drawn ¿ Pole zero plot

( ❑❑ ❑❑❑❑ )
()
❑❑ ❑❑❑❑
Fig:3.1a is pole zero plot of high-pass Filter which can be found from drawing the
corresponding H(ω ¿ .
AVR 24

Other Ways to find Partial Fraction:

❑❑ () ❑
()()

❑❑ ❑❑ ❑❑ ❑❑
❑❑ ()
❑ ❑ ❑ ❑

❑❑
[ ( )(❑)() ]
❑