Digital Signal Processing Reference
In-Depth Information
x[n]
y[n]
+
+
+
X
a
+
X
y[n-m]
X
a 2
a M
+
X
a M-1
Feed forward part
Figure 7.29 Look-ahead transformation to add M registers in the feedback path of a first-order IIR filter
The generalized IIR filter implementing this difference equation with M registers in the feedback
loop is given in Figure 7.29. This reduces the IPB of the orginal DFG by a factor of M. The difference
equation implements the same transfer function H(z):
X M 1
0 a i z i
1 az 1
1
HðÞ¼
az 1 ¼
ð 7 : 9 Þ
Þ M
1
ð
The expression in (7.8) adds an extra M
1 coefficients in the numerator of the transfer function of
the systemgiven in (7.9), requiringM 1multiplications andM 1 additions in its implementation.
These extra computations can be reduced by constraining M to be a power of 2, so that M ¼ 2 m .
This makes the transfer function of (7.9) as:
Y m
2 i
þ az 1
1
ð
Þ
1
1
HðÞ¼
az 1 ¼
ð
7
10
Þ
:
1
Þ M
1 az 1
ð
The numerator of the transfer function now can be implemented as a cascade of m FIR filters each
requiring only one multiplier, where m ¼ log 2 M.
Example: This example increases the IPB of a first-order IIR filter by M
8. Applying the
look-ahead transformation of (7.9), the transfer function of a first-order system with a pole at 0.9 is
given as:
¼
1 þ 0 : 9 z 1
þ 0 : 81 z 2
þ 0 : 729 z 3
þ 0 : 6561 z 4
þ 0 : 5905 z 5
þ 0 : 5314 z 6
þ
z 7
1 0 : 4305 z 8
1
1 0 : 9 z 1 ¼
0
4783
:
HðÞ¼
Using (7.9), this transfer function can be equivalently written as:
ð
1 þ 0 : 9 z 1
Þ 1 þ 0 : 81 z 2
ð
Þð 1 þ 0 : 6561 z 4
Þ
1
1 0 : 9 z 1 ¼
HðÞ¼
1 0 : 4305 z 8
Search WWH ::




Custom Search