Digital Signal Processing Reference
In-Depth Information
The analog filter s-domain transfer function is H
a
(s) = LT{h
a
(t)}. Now the
poles of H
a
(s) are transformed to the poles of H(z) through the relation z
¼
e
sT
s
:
The method can be summarized as follows (see, for example, A. V. Oppenheim
and R. Schafer, Discrete-Time Signal Processing, Prentice-Hall, 1989):
1. Expand the analog transfer function by partial fractions as
P
m
¼
1
c
m
s
p
m
. Note that
p
m
's can be non-distinct.
2. The required transfer function is
H
ð
z
Þ¼
T
s
X
1
e
p
m
T
s
z
1
¼
T
s
X
M
M
c
m
z
z
z
m
c
m
m
¼
1
m
¼
1
where z
m
¼
e
p
m
T
s
are the poles in the z-domain.
Note 1 If there is no aliasing, H
ð
e
jX
Þ¼
H
a
ð
x
Þ
when
p
X
p
:
Note 2 Since the relation X
¼
xT
s
is applicable, the jx axis is transformed into
the circumference of the unit circle e
jX
¼
e
jxT
s
:
Note 3 Since the poles are transformed according to z
¼
e
sT
s
, stability is
preserved.
Example Using the impulse invariance method, design a digital Butterworth LPF
with the following specifications:
1. T
s
= 0.01 sec (hence, f
s
= 100 Hz),
2. f
c
= 10 Hz (therefore x
c
= 20 p rad/sec),
3. G
m
= 1, gain B 0.1 (i.e., -20 dB) for 20 B f B f
s
/2 = 50 Hz).
Solution:
It is necessary to first find an analog Butterworth LPF with the above specifica-
tions. The gain should be less than -20 dB for a normalized frequency of f
c
C
20/10 = 2 (normalized w.r.t f
c
). From the graph in Tables—Stopband gain for a
B-LPF, the filter order is found to be n = 4. From Tables—Denominator Poly-
nomial Coefficients for Normalized LPF's, the transfer function of this normalized
filter is seen to be:
H
N
ð
s
Þ¼
a
o
1
þ
2
:
6131s
N
þ
3
:
4142s
N
þ
2
:
6131s
N
þ
s
N
Now H(
N
(s)|
s=0
= a
0
= G
dc
= G
m
= 1, and s
N
= s/x
c
. Substituting these val-
ues for a
o
and s
N
into the above equation, yields:
1
:
55e7
1
:
55e7
þ
6
:
48e5s
þ
1
:
34e4s
2
þ
164
:
1s
3
þ
s
4
:
H
a
ð
s
Þ¼
H
a
(s) can then be expanded using partial fractions. This can be easily done in
MATLAB using
B = [1 164.1 1.34e4 6.48e5 1.55e7],A = [1.55e7],
[R,P,K] = residue(A,B)
where
R
gives the coefficients c
m
,
P
gives the poles
Search WWH ::
Custom Search