ECT303 DIGITAL SIGNAL

PROCESSING
MODULE 3 -PART IV
DESIGN OF ANALOG IIR FILTERS

Ms. Neethu Radha Gopan, Asst. Prof, Dept. of ECE, RSET, Kakkanad
 Introduction to IIR Filters

➢ IIR filters are of recursive type, where the present output sample depends on the present
input, past input & past output samples.
𝑁 𝑀
➢ IIR filters are described by the difference equation
𝑁 𝑀 Y z 1 + ෍ 𝑎𝑘 𝑧 −𝑘 = 𝑋(𝑧) ෍ 𝑏𝑘 𝑧 −𝑘
𝑦 𝑛 = − ෍ 𝑎𝑘 𝑦 𝑛 − 𝑘 + ෍ 𝑏𝑘 𝑥(𝑛 − 𝑘) 𝑘=1 𝑘=0

𝑘=1 𝑘=0
𝑁 𝑀 The system function is given by
𝑦 𝑛 + ෍ 𝑎𝑘 𝑦 𝑛 − 𝑘 = ෍ 𝑏𝑘 𝑥(𝑛 − 𝑘) Y z σ𝑀 𝑏 𝑧 −𝑘
𝑘=0 𝑘
𝑘=1 𝑘=0 𝐻 𝑧 = =
𝑋 𝑧 1 + σ𝑁𝑘=1 𝑎𝑘 𝑧
−𝑘
➢ Taking Z transform, we get,
𝑁 𝑀

Y z + ෍ 𝑎𝑘 𝑧 −𝑘 𝑌(𝑧) = ෍ 𝑏𝑘 𝑧 −𝑘 𝑋(𝑧)
𝑘=1 𝑘=0
Design of Digital Filters from Analog Filters

❑ The design of an IIR filter for the given specifications is to find the filter coefficients
𝑎𝑘 & 𝑏𝑘 .
❑ The most common method of designing IIR digital filters is known as indirect method,
which involves first designing an analog prototype filter and then transforming the
prototype to a digital filter.
❑ For the given specifications of a digital filter , the derivation of the digital filter
transfer function requires 3 steps:
1. Map the desired digital filter specifications into those for an equivalent analog filter.
2. Derive the transfer function for the analog prototype.
3. Transform the transfer function of the analog prototype into an equivalent digital
filter transfer function.
Types of IIR Filters

1. Butterworth Filters
2. Chebyshev Filters
Analog Butterworth Low Pass Filters

Ω𝑝 − pass band frequency (radians/sec)

Ω𝑠 − stop band frequency (radians/sec)
Ω𝑐 − 3dB cutoff frequency (radians/sec)
𝜀 −parameter for allowable passband
𝜆 − parameter for allowable stopband
𝛼𝑝 −pass band attenuation
𝛼𝑠 −stop band attenuation

Butterworth Approximation of Magnitude Response

Butterworth Low Pass Filters

➢ Butterworth low pass filters are all pole filters with a flat passband and is
characterized by the magnitude frequency response
1 1
|𝐻 𝑗Ω | = =
2𝑁 1/2 2𝑁 1/2
Ω Ω
1+ 1 + 𝜀2
Ω𝑐 Ω𝑝

Where N is the order of the filter, Ω𝑐 is the cut off frequency , Ω𝑝 is the pass band
edge frequency.
➢ The magnitude square response of a normalized Butterworth filter
(Ω𝑐 = 1𝑟𝑎𝑑/sec) is
1
|𝐻 𝑗Ω |2 = 2𝑁
1+ Ω

➢ The above function has a total of 2N poles which lie on a circle – N poles on the left
half of the s plane as well as right half of the s plane due to the presence of H(s) &
H(-s).
 Q. Find the expression for order N of an analog low pass Butterworth filter .

Soln: The magnitude response of a Butterworth filter is

1
|𝐻 𝑗Ω | =
2𝑁 1/2
Ω
1+ 𝜀2
Ω𝑝
1
|𝐻 𝑗Ω |2 = 2𝑁
1 + 𝜀 2 Ω/Ω𝑝

Taking logarithm on both sides,

1
10 log |𝐻 𝑗Ω |2 = 10 log 2𝑁
1 + 𝜀 2 Ω/Ω𝑝
2𝑁
20 log |𝐻 𝑗Ω | = 10 log 1 − 10 log 1 + 𝜀 2 Ω/Ω𝑝 −−− −(1)
At, Ω = Ω𝑝 , the attenuation is equal to 𝛼𝑝, therefore eqn (1) becomes

20 log |𝐻 𝑗Ω𝑝 | = −𝛼𝑝 = −10 log 1 + 𝜀 2

𝛼𝑝 = 10 log(1 + 𝜀 2 )

log(1 + 𝜀 2 ) = 0.1𝛼𝑝

1 + 𝜀 2 = 100.1𝛼𝑝

𝜀 2 = 100.1𝛼𝑝 − 1

𝜀= 100.1𝛼𝑝 − 1 −−−− −(2)

At, Ω = Ω𝑠 , the stop band attenuation is equal to 𝛼𝑠 , Therefore eqn (1) becomes
2𝑁
20 log |𝐻 𝑗Ω𝑠 | = −𝛼𝑠 = −10 log 1 + 𝜀 2 Ω𝑠 /Ω𝑝
 2𝑁
𝛼𝑠 = 10 log 1 + 𝜀2 Ω𝑠 /Ω𝑝
Taking logarithm at both sides,
2𝑁
0.1 𝛼𝑠 = log 1 + 𝜀 2 Ω𝑠 /Ω𝑝
𝑁 100.1𝛼𝑠 − 1
log Ω𝑠 /Ω𝑝 = log
2𝑁 100.1𝛼𝑝 − 1
1 + 𝜀 2 Ω𝑠 /Ω𝑝 = 100.1𝛼𝑠
2𝑁
𝜀 2 Ω𝑠 /Ω𝑝 = 100.1𝛼𝑠 − 1 100.1𝛼𝑠 − 1
𝑁 log Ω𝑠 /Ω𝑝 = log
100.1𝛼𝑝 − 1
2𝑁 100.1𝛼𝑠 − 1
Ω𝑠 /Ω𝑝 = (100.1𝛼𝑠 −1)/𝜀 2 = 0.1𝛼𝑝
10 −1
100.1𝛼𝑠 − 1
log 0.1𝛼𝑝
𝑁 100.1𝛼𝑠 − 1 𝑁= 10 −1
Ω𝑠 /Ω𝑝 = log Ω𝑠 /Ω𝑝
100.1𝛼𝑝 − 1
Q. Find the expression for cut-off frequency of an analog low pass
Butterworth filter .

Soln: The magnitude square response of a Butterworth filter is

2
1
|𝐻 𝑗Ω | = 2𝑁
1 + Ω/Ω𝑐 2𝑁
Ω𝑝 /Ω𝑐 = 𝜀2 = 10 0.1𝛼𝑝 − 1

2
1
𝐴𝑙𝑠𝑜 |𝐻 𝑗Ω | =
1+ 𝜀2 Ω/Ω𝑝
2𝑁 Ω𝑝 /Ω𝑐 = (10 0.1𝛼𝑝 − 1 )1/2𝑁

Comparing both equations, we get

Ω𝑝 Ω𝑝
2𝑁 2𝑁 Ω𝑐 = = −−− −(1)
1 + Ω/Ω𝑐 = 1 + 𝜀 2 Ω/Ω𝑝 1 𝜀 1/𝑁
0.1𝛼𝑝 2𝑁
10 −1
2𝑁 2𝑁
Ω/Ω𝑐 = 𝜀2 Ω/Ω𝑝
From the previous derivation, we know that,
2𝑁 100.1𝛼𝑠 − 1
Ω𝑠 /Ω𝑝 = 0.1𝛼𝑝
10 −1
Ω𝑠 Ω𝑝
1/2𝑁 ∴ Ω𝑐 = =
100.1𝛼𝑠
−1 1 1
Ω𝑠 /Ω𝑝 = 10 0.1𝛼𝑠 − 1 2𝑁 10 0.1𝛼𝑝 − 1 2𝑁
100.1𝛼𝑝 − 1
1/2𝑁
100.1𝛼𝑠 − 1
Ω𝑠 = Ω𝑝 −− −(2)
100.1𝛼𝑝 − 1

Ω𝑝
From equation (1) , we have Ω𝑐 = 1
0.1𝛼𝑝 2𝑁
10 −1
Substituting this in equation 2, we get

Ω𝑠 = Ω𝑐 100.1𝛼𝑠 − 1 1/2𝑁
Steps to Design Analog Butterworth Low Pass Filter

1. From the given specifications, find the order of the filter N.

2. Round it off to the next highest integer.
3. Find the transfer function H(s) for Ω𝑐 = 1 rad/sec for the value of N.
4. Calculate the value of cut off frequency Ω𝑐 .
𝑠
5. Find the transfer function 𝐻𝑎 𝑠 for the above value of Ω𝑐 by substituting s→ in
Ω𝑐
H(s).
Necessary Equations for Butterworth LPF Design

10 0.1𝛼𝑠 − 1 𝜆
log 0.1𝛼𝑝 log
𝑁= 10 −1= 𝜀
Ω𝑠 Ω𝑠
log log
Ω𝑝 Ω𝑝

∴𝜆= 10 0.1𝛼𝑠 − 1 & 𝜀= 10 0.1𝛼𝑝 − 1

The poles on the left half of the s plane is given by 𝑠𝑘 = 𝑒 𝑗𝜙𝑘 , k =1,2,….N
𝜋 2𝑘 − 1 𝜋
𝜙𝑘 = +
2 2𝑁
Ω𝑝
Ω𝑐 =
( 10 0.1𝛼𝑝 − 1)1/𝑁
Q. Design an analog Butterworth filter that has a -2dB passband attenuation at a
frequency of 20 rad/sec and at least -10 dB stopband attenuation at 30 rad/sec.

Soln: Given 𝛼𝑝 = 2 𝑑𝐵, 𝛼𝑠 = 10 𝑑𝐵, Ω𝑝 = 20 𝑟𝑎𝑑/ sec , Ω𝑠 = 30 𝑟𝑎𝑑/𝑠𝑒𝑐

10 0.1𝛼𝑠 − 1 10 − 1
log 0.1𝛼𝑝 log log(3.92)
10 −1= 100.2 − 1 = = 3.37
𝑁= log(1.5)
Ω 30
log 𝑠 log
Ω𝑝 20
𝑁≈4
In order to find the transfer function, we need to find the poles. There are 2N=8 poles. We
only consider the 4 poles on the left half of the s plane.
𝑠𝑘 = Ω𝑐 𝑒 𝑗𝜙𝑘 , k =1,2,…. N (Take Ω𝑐 = 1𝑟𝑎𝑑/sec)
𝜋 2𝑘 − 1 𝜋
𝜙𝑘 = +
2 2𝑁
 𝜋 𝜋 5𝜋
k=1, 𝜙1 = + = , 𝑠1 = 𝑒 𝑗5𝜋/8 = cos 5𝜋/8 + 𝑗𝑠𝑖𝑛 5𝜋/8= −0.383 + 𝑗 0.924
2 8 8
𝜋 3𝜋 7𝜋
k=2, 𝜙2 = + = , 𝑠2 = 𝑒 𝑗7𝜋/8 = −0.924 + 𝑗 0.383
2 8 8
𝜋 5𝜋 9𝜋
k=3, 𝜙3 = + = , 𝑠3 = 𝑒 𝑗9𝜋/8 = −0.924 − 𝑗 0.383
2 8 8
𝜋 7𝜋 11𝜋
k=4, 𝜙4 = + = , 𝑠4 = 𝑒 𝑗11𝜋/8 = −0.383 − 𝑗 0.924
2 8 8

Hence the denominator terms becomes

(𝑠 + 0.383 − 𝑗 0.924)(𝑠 + 0.924 − 𝑗 0.383)(𝑠 + 0.924 + 𝑗 0.383) (𝑠 + 0.383 + 𝑗 0.924)
1
∴ 𝐻 𝑠 =
( 𝑠 + 0.383 2 + 0.924 2 )( 𝑠 + 0.924 2 + 0.383 2 )
1
𝐻 𝑠 = 2
(𝑠 + 0.765𝑠 + 1)(𝑠 2 + 1.847𝑠 + 1)
 Ω𝑝 20
Ω𝑐 = = 1/4
= 21.385
( 10 0.1𝛼𝑝
− 1)1/𝑁 (0.765)

𝑠
The transfer function for Ω𝑐 = 21.385 can be obtained by substituting s→ in
21.385
H(s).
1
∴ 𝐻 𝑠 = 2 2
𝑠 𝑠 𝑠 𝑠
+ 0.765 +1 + 1.847 +1
21.385 21.385 21.385 21.385

0.209 × 106
𝐻 𝑠 = 2
𝑠 + 16.36 𝑠 + 457.3 (𝑠 2 + 39.5 𝑠 + 457.3)
Q. For the given specifications, design an analog Butterworth filter.
0.9≤ |H(jΩ)| ≤ 1, for 0 ≤ Ω ≤ 0.2𝜋 & |H(jΩ)| ≤ 0.2, for 0.4 𝜋 ≤ Ω ≤ 𝜋

Soln: Given
1 1
Ω𝑝 = 0.2𝜋, Ω𝑠 = 0.4𝜋, = 0.9, = 0.2
1+ 𝜀2 1+ 𝜆2
𝜆
log
𝑁= 𝜀
Ω𝑠
log
Ω𝑝
1 𝜀 = 0.484
= 0.9
1+𝜀 2

1
= 0.2 𝜆 = 4.898
1 + 𝜆2

𝑁 = 3.34 ≈ 4
From the previous question , we have found the poles of the transfer function of
a normalized Low pass Butterworth filter for N = 4
1
∴ 𝐻(𝑠) =
(𝑠 2 + 0.765𝑠 + 1)(𝑠 2 + 1.847𝑠 + 1)
Ω𝑝 Ω𝑝 0.2𝜋
Ω𝑐 = = 1/𝑁 = 1/4
= 0.24 𝜋
( 10 0.1𝛼𝑝
− 1)1/𝑁 𝜀 (0.484 )
𝑠
The transfer function for Ω𝑐 = 0.24 𝜋 can be obtained by substituting s→ in H(s).
0.24 𝜋
1
𝐻 𝑠 = 2 2
𝑠 𝑠 𝑠 𝑠
+ 0.765 +1 + 1.847 +1
0.24 𝜋 0.24 𝜋 0.24 𝜋 0.24 𝜋
0.323
𝐻 𝑠 =
(𝑠 2 + 0.577 𝑠 + 0.568)(𝑠 2 + 1. 39𝑠 + 0.568)
Frequency Transformations in Analog Domain

1. Low Pass to Low Pass Filter

Given a normalized low pass filter, it is desirable to have a low pass filter with a cut off
𝑠
frequency Ω𝑐 .This is done by s→ .
Ω𝑐

2. Low Pass to High Pass Filter

Given a normalized low pass filter, it is desirable to have a high pass filter with a cut off
Ω𝑐
frequency Ω𝑐 .This is done by s→ .
𝑠
3. Low Pass to Band Pass Filter
The transformation for converting a normalized low pass filter to a band pass filter
with cut off frequencies, Ω𝑙 & Ω𝑢 can be done by
𝑠 2 + Ω 𝑙 Ω𝑢
𝑠→
𝑠(Ω𝑢 − Ω𝑙 )
−Ω1 2 + Ω𝑙 Ω𝑢
𝐴=
Ω1 (Ω𝑢 − Ω𝑙 )

Ω2 2 + Ω𝑙 Ω𝑢
𝐵=
Ω2 (Ω𝑢 − Ω𝑙 )
Ω𝑠
Ω𝑟 = = min{ 𝐴 , |𝐵|}
Ω𝑝
4. Low Pass to Band Stop Filter
The transformation to convert a normalized low pass filter to a band stop filter is

𝑠(Ω𝑢 − Ω𝑙 )
𝑠→ 2
𝑠 + Ω𝑙 Ω𝑢
Ω1 (Ω𝑢 − Ω𝑙 )
𝐴=
−Ω1 2 + Ω𝑙 Ω𝑢

Ω2 (Ω𝑢 − Ω𝑙 )
𝐵=
−Ω2 2 + Ω𝑙 Ω𝑢
Ω𝑠
Ω𝑟 = = min{ 𝐴 , |𝐵|}
Ω𝑝
Q. For the given specifications, 𝛼𝑝 = 3 𝑑𝐵, 𝛼𝑠 = 15 𝑑𝐵, 𝑓𝑝 = 159.24 𝐻𝑧,
𝑓𝑠 = 79.62 𝐻𝑧, design a highpass filter.

Soln: First we design a normalized low pass filter and then transform to a high pass filter.
Corresponding Low pass filter specifications:
𝛼𝑝 = 3 𝑑𝐵, 𝛼𝑠 = 15 𝑑𝐵,
𝑓𝑝 = 79.62 𝐻𝑧, 𝑓𝑠 = 159.24 𝐻𝑧

Ω𝑝 = 2𝜋𝑓𝑝 = 500 rad/sec

Ω𝑠 = 2𝜋𝑓𝑠 = 1000 rad/sec

10 0.1𝛼𝑠 − 1
log 0.1𝛼𝑝
𝑁= 10 − 1 = 2.47 ≈3
Ω
log 𝑠
Ω𝑝
 𝑠𝑘 = Ω𝑐 𝑒 𝑗𝜙𝑘 , k =1,2,…. N (Take Ω𝑐 = 1𝑟𝑎𝑑/sec)
𝜋 2𝑘 − 1 𝜋
𝜙𝑘 = +
2 2𝑁
𝜋 𝜋 2𝜋
k=1, 𝜙1 = + = , 𝑠1 = 𝑒 𝑗2𝜋/3 = cos 2𝜋/3 + 𝑗𝑠𝑖𝑛 2𝜋/3 = −0.5 + 𝑗 0.866
2 6 3
𝜋 3𝜋
k=2, 𝜙2 = + = 𝜋, 𝑠2 = 𝑒 𝑗𝜋 = −1
2 6

𝜋 5𝜋 4𝜋
k=3, 𝜙3 = + = , 𝑠3 = 𝑒 𝑗4𝜋/3 = −0.5 − 𝑗 0.866
2 6 3

1 1
∴ 𝐻 𝑠 = =
(𝑠 + 1)(𝑠 + 0.5 − 𝑗0.866)(𝑠 + 0.5 + 𝑗0.866) (𝑠 + 1)( 𝑠 + 0.5 2 + 0.866 2 )
1
The normalized Butterworth LPF is given by 𝐻 𝑠 =
(𝑠 + 1)(𝑠 2 + 𝑠 + 1)
The transfer function for the highpass filter is obtained by substituting
Ω𝑐
s→ in H(s).
𝑠

Ω𝑝
Ω𝑐 = 0.1𝛼𝑝
= 1000 𝑟𝑎𝑑/ sec ( Ω𝑝 should be that of HPF)
( 10 −1)1/𝑁
1
∴ 𝐻(𝑠) =
2
1000 1000 1000
+1 + +1
𝑠 𝑠 𝑠

𝑠3
𝐻(𝑠) =
(𝑠 + 1000)(𝑠 2 + 1000𝑠 + 10002 )
Q. Design a bandstop Butterworth filter : Stopband 100 to 600 Hz, 20 dB
attenuation at 200 & 400 Hz, gain at w=0 is unity and passband attenuation is 3 dB

Soln:
Given 𝑓𝑙 = 100 𝐻𝑧, 𝑓1 = 200 𝐻𝑧
𝑓2 = 400 𝐻𝑧, 𝑓𝑢 = 600 𝐻𝑧
𝛼𝑠 = 20 𝑑𝐵, 𝛼𝑝 = 3𝑑𝐵

Ω𝑙 = 2𝜋𝑓𝑙 = 200𝜋 rad/sec

Ω1 = 2𝜋𝑓1 = 400𝜋 rad/sec
Ω2 = 2𝜋𝑓2 = 800𝜋 rad/sec
Ω𝑢 = 2𝜋𝑓𝑢 = 1200𝜋 rad/sec
 Ω1 (Ω𝑢 − Ω𝑙 ) 400𝜋 (1200𝜋 − 200𝜋)
𝐴= = =5
2
−Ω1 + Ω𝑙 Ω𝑢 −(400𝜋)2 + (200𝜋) 1200𝜋

Ω2 (Ω𝑢 − Ω𝑙 ) 800𝜋 (1200𝜋 − 200𝜋)

𝐵= = = −2
−Ω2 2 + Ω𝑙 Ω𝑢 −(800𝜋)2 + (200𝜋) 1200𝜋

Ω𝑠
Ω𝑟 = = min 𝐴 , 𝐵 =2
Ω𝑝
The order for a normalized Butterworth LPF is
10 0.1𝛼𝑠 − 1
log 0.1𝛼𝑝
𝑁= 10 −1 = 3.32 ≈ 4
Ω
log 𝑠
Ω𝑝
1
For N=4, 𝐻 𝑠 = (𝑠 2 + 0.765𝑠 + 1)(𝑠 2 + 1.847𝑠 + 1)
 The transfer function for the bandstop filter is obtained by substituting
𝑠(Ω𝑢 −Ω𝑙 )
s→ in H(s).
𝑠2 +Ω𝑙 Ω𝑢

1000𝜋𝑠
ie, s→
𝑠2 +24×104 𝜋2

1
𝐻 𝑠 =
2 2
1000𝜋𝑠 1000𝜋𝑠 1000𝜋𝑠 1000𝜋𝑠
+ 0.765 2 +1 + 1.847 +1
𝑠 2 + 24 × 104 𝜋 2 𝑠 + 24 × 104 𝜋 2 𝑠 2 + 24 × 104 𝜋 2 𝑠 2 + 24 × 104 𝜋 2
 References
29

1. Proakis J. G. and Manolakis D. G., Digital Signal Processing, 4/e, Pearson

Education, 2007.
2. P. Ramesh Babu, Digital Signal Processing, Scitech Publications (India) Pvt
Ltd.
30 END of PART -IV

THANK YOU!