Digital Signal Processing Reference
In-Depth Information
Example 8.3: Channels with zeros on the unit circle
Consider the channel C ( z )= n =0
z −n , which has all seven zeros on the
unit circle at the points
z k = e −j 2 πk/ 7 ,
0
k
6 .
We have L = 7, and choosing M =8wehave P = M + L =15 . Calculations
show:
A K 2 /M
K
8
1 . 87
9
1 . 75
10
1 . 59
11
1 . 45
12
1 . 31
13
1 . 18
14
1 . 05
15
0 . 89
A K 2 /M gets smaller, though not as
dramatically as the case where the zeros of C ( z ) are outside the unit circle.
For unit circle zeros with higher multiplicity, the choice of large K becomes
very crucial. For example let
Thus as K increases the quantity
C ( z )=1+3 z 1 +3 z 2 + z 3 ,
which has three zeros at z =
1 . With M =8sothat P = M + L = 11,
calculations show:
A K 2 /M
K
8
734 . 25
9
21 . 03
10
5 . 03
11
3 . 81
In this example there is a major improvement as K increases from M to
M + 1. Again, the noise gain is least when K = M + L =11 .
In practice of course, an arbitrary channel can have zeros inside, on or outside
the unit circle. The above examples show that, as long as we choose K = P (i.e.,
use all the P samples in the output block at the receiver), the quantity
A K 2
is insensitive to these details, and the equalization can be done without undue
noise amplification.
 
Search WWH ::




Custom Search