1 Design Consideration
When a filter produces a unit-sample response that has an infinite duration, it is called an infinite-impulse response
(IIR) filter. The system functions for an analog filter is denoted by H(s), where s=+jΩ, where Ω is angular frequency.
The frequency transfer function of an analog filter, denoted by H(jΩ) is obtained by evaluating the system function H(s)
in the s-plane along the frequency axis, or
H(j)=H(s)|s=jΩ (1)
The squared magnitude response |H (jΩ)|2 = power transfer characteristic of the filter.
The analytic from of the identity given by equation (2) indicates that the singularities of H(s) H(-s) is s-plane are
symmetric about the j axis.
To obtain a stable system function from H(s) H(-s), the poles that lie in the left half of the s-plane are assigned to H(s).
The assignment of zeros is arbitrary.
1
H ( s) H ( s)
1 ( s / jc ) 2 N (3)
1 1
| H ( j) |2 (4)
1 (1) 2N
2
1
| H ( jc ) |dB
2
10 log10 3.0103 (5)
2
N=1
N=6
2N
s
1 (6)
j c
s 1
1 2N e
j (2 k 1) /2 N
, k 0,1, 2,....N 1
Therefore, c
j (7)
Let the maximum passband attenuation in db is αp at passband frequency Ωp and αs is the minimum stopband
attenuation in dB at stopband frequency Ωs. Now the magnitude function can be rewritten as
1
| H ( j) |
1 ( / jc ) 2 N
1/2 (9)
s
p 10log(1 2 )
(11)
0.1 p log(1 2 )
Taking antilog on both side of eqn. (11)
0.1 p
(10 1)1/2 (12)
2
s
2N
0.1 s log 1
p
After simplification
2 2N
0.1 s
s 10 1 (14)
p
Sub. Eqn(14) into eqn.(12),
2N
s 10 s 1
0.1
0.1 p (15)
p 10 1
0.1
10 s 1
log 0.1
N 10 p 1 (16)
log s
p
Since this N values is normally not an integer value, therefore, roundoff N to the neat integer value,
0.1
10 s 1
log 0.1
N 10 p 1 (17)
log s
p
log
N (18)
log s
p
where
0.1 p
10 1
(19)
0.1 s
10 1
For simplicity,
0.1
10 s 1
A 0.1
10 p 1
p
k
s
Therefore the eqn. (18) can be represented as
log A
N (20)
1
log
k
The cut-off frequency can be obtained using
p s
c 0.1 p
0.1 s
(21)
(10 1) 1/2 N
(10 1)1/2 N
The following Table 1 gives the Butterworth polynomials for various values of N for c=1 rad/sec.
3 (s 1)( s 2 s 1)
4 (s 2 0.76537 s 1)( s 2 1.8477 s 1)
5 (s 1)( s 2 0.61803 s 1)( s 2 1.61803 s 1)
Example 1
For the following specification p 1dB; s 30dB ; p 200rad / sec ; s 600rad / sec . Determine
the order of the filter.
Solution
We know that
0.1
10 s 1
log 0.1
N 10 p 1
log s
p
log(62.115)
log(3)
3.758
Example 2
Design an analog Butterworth filter for the following specifications.
1
1/ 2
H (e j ) 1 0 0.2
2
H (e j ) 0.1 0.4
Solution
0.1
10 s 1
log 0.1
N 10 p 1
log s
p
log(0.1005)
log(0.5)
3.314
The order of the filter is N = 4.
Since N = 4 (even) the T.F of the fourth order analog filter is
Bk c
N /2 2
H a ( s)
s 2 bk c s C k c
2
k 1
Bk c
2 2
s 2 bk c s C k c
2
k 1
B1 c B2 c
2 2
.
s 2 b1 c s C1 c s 2 b2 c s C 2 c
2 2
We Know That,
(2k 1)
bk 2 sin , C1 C 2 1.
2N
b1 2 sin 0.7649 0.765
2
3
b2 2 sin 1.8473
8
B1 c B2 c
2 2
H a ( s) .
s 2 0.765 c s C1 c s 2 1.847 c s C 2 c
2 2
The cut off frequency c can be estimated from the magnitude response of the butterworth filter as
p
c 1/2 N
1
A 2 1
1
Note:
j 1
H (e ) 1/2
p 2 N
1
c
1
p 1/2
p 2 N
1
c
1
p2 2N
1 p
c
1
1/2 N
p 2 1
c p
The analog filter pass band cut off frequency can be obtained from the digital and analog filter frequency relation,
p pT
Assume B1 B2 1
0.15553
H a ( s)
( s 0.4804 s 0.3943)( s 2 1.1599 s 0.3943)
2
Example 3
Design an analog Butterworth filter that has a – 2 dB passband attenuation at a frequency of 20 rad/sec and an atleast –
10 dB stopband attenuation at 30 rad/sec.
Solution
Given p 2dB; s 30 dB ;
p 20rad / sec ; s 30rad / sec .
We know that
0.5
10 0.1 s 1
log 0.1
10 p 1
N
log s
p
0.5
10 1
log 0.2
10 1
log s
p
3.37
Rounding off N to the next higher integer we get N=4.
The normalized lowpass Butterworth filter for N=4 can be found from Table 1
1
H a ( s)
(s 0.76537s 1)(s 2 1.8477s 1)
2