Digital Signal Processing Reference
In-Depth Information
n is even-even:
0
w
d
[n]
=
adw
d
[n]
+ bdw
d
[n1]
+ cdw
d
[n2]
+ ddw
d
[n3]
+ abw
d
[n2]
+ bbw
d
[n3]
+ bcw
d
[n4]
+ bdw
d
[n5]
adw
d
[n]
+ acw
d
[n1]abw
d
[n2]
+ aaw
d
[n3]
cdw
d
[n2]
+ ccw
d
[n3]bcw
d
[n4]
+ acw
d
[n5]:
n is odd-even:
0
=
acw
d
[n1]
w
d
[n]
+ bcw
d
[n2]
+ ccw
d
[n3]
+ cdw
d
[n4]
+ aaw
d
[n3]
+ abw
d
[n4]
+ acw
d
[n5]
+ adw
d
[n6]
+ bdw
d
[n1]bcw
d
[n2]
+ bbw
d
[n3]abw
d
[n4]
+ ddw
d
[n3]cdw
d
[n4]
+ bdw
d
[n5]
adw
d
[n6]:
Canceling out terms gives us:
n is even-even:
0
w
d
[n]
= bdw
d
[n1]
+ ddw
d
[n3]
+ bbw
d
[n3]
+ bdw
d
[n5]
+ acw
d
[n1]
+ aaw
d
[n3]
+ ccw
d
[n3]
+ acw
d
[n5]:
n is odd-even:
0
w
d
[n]
= acw
d
[n1]
+ ccw
d
[n3]
+ aaw
d
[n3]
+ acw
d
[n5]
+ bdw
d
[n1]
+ bbw
d
[n3]
+ ddw
d
[n3]
+ bdw
d
[n5]:
Examining the above equations, we see that we have the same expression for
when n is even-odd as when it is even-even. Therefore, we can rewrite the above
expression, with the note that n must be even. We know from the previous section
that ac =bd, so the terms with w
d
[n1] and w
d
[n5] will cancel out. This leaves
us with w
d
[n]
0
= (aa+bb+cc+dd)w
d
[n3]. As previously noted, (aa+bb+cc+dd) =
1, so we get the nal expression for w
d
[n]
0
0
is
the same as a delayed version of the original w
d
[n]. This is no surprise, since we have
= w
d
[n3]. The reconstructed w
d
[n]