Digital Signal Processing Reference
In-Depth Information
After filter coefficient quantization, we have the quantized digital IIR filter transfer function:
H
q
ðzÞ¼
b
0
þ b
1
z
1
1
þ a
1
z
1
(8.58)
Solving for the pole and zero, we get
z
1
¼
b
1
(8.59)
b
0
p
1
¼a
1
(8.60)
Now considering a second-order IIR filter transfer function as
HðzÞ¼
b
0
þ b
1
z
1
þ b
2
z
2
(8.61)
1
þ a
1
z
1
þ a
2
z
2
and its quantized IIR filter transfer function
H
q
ðzÞ¼
b
0
þ b
1
z
1
þ b
2
z
2
(8.62)
1
þ a
1
z
1
þ a
2
z
2
solving for poles and zeros yields
z
1
;
2
¼
0
:
5
$
b
1
b
2
b
1
b
0
2
1
2
b
0
j
b
0
0
:
25
$
(8.63)
a
2
0
:
25
$
a
1
2
1
2
p
1
;
2
¼
0
:
5
$a
1
j
(8.64)
With Equations
(8.59) and (8.60)
for the first-order IIR filter, and Equations
(8.63) and (8.64)
for the
second-order IIR filter, we can study the effects of location changes of the poles and zeros, and the
frequency responses due to filter coefficient quantization.
EXAMPLE 8.24
Given the first-order IIR filter
HðzÞ¼
1:2341 þ 0:2126
z
1
1 0:5126z
1
and assuming that we use 1 sign bit and 6 bits for encoding the magnitude of the filter coefficients, find the
quantized transfer function and pole-zero locations.
Solution:
Let us find the pole and zero for infinite precision filter coefficients. Solving 1:2341z þ 0:2126 ¼ 0 leads to
a zero location z
1
¼0:17227. Solving z 0:5126 ¼ 0 gives a pole location p
1
¼ 0:5126.
Now let us quantize the filter coefficients. Quantizing 1.2341 can be illustrated as
1:2341 2
5
¼ 39:4912 ¼ 39 ðrounded to integerÞ
Search WWH ::
Custom Search