Digital Signal Processing Reference
In-Depth Information
0.5
V
− 0.5
V
x
x
x
−3
A
−
A
A
3
A
Figure 5.25
. Examples showing the application of the mod
V
operator. The points
±A
and
±
3
A
correspond to the words in a 2-bit PAM constellation.
What makes the pre-equalization system successful is an important property
of the modulo operator. Given two real numbers
u
1
and
u
2
, it is readily verified
that
(
u
1
+
u
2
)mod
V
=
u
1
mod
V
+
u
2
mod
V
mod
V.
(5
.
92)
Using this we will show that for the transceiver in Fig. 5.23, as long as
−
0
.
5
V
≤
s
(
n
)
<
0
.
5
V
for all
n,
the signal
s
(
n
) at the receiver is exactly equal to
s
(
n
)in
absence of channel noise:
s
(
n
)=
s
(
n
)
.
(5
.
93)
Thus, even though
x
(
n
) may have any value in the range
x
(
n
)
<
0
.
5
V,
after it passes through the channel
H
(
z
) and the modulo operator again, the
result is precisely the transmitted symbol
s
(
n
) from the constellation. When
there is channel noise, this is modified to
−
0
.
5
V
≤
s
(
n
)=[
s
(
n
)+
q
(
n
)] mod
V
=
s
(
n
)+[
q
(
n
)mod
V
]
mod
V.
(5
.
94)
Thus, even though the modulo operator is nonlinear, the transceiver reproduces
s
(
n
) perfectly at the receiver, except for channel noise. If
q
(
n
)issmallcompared
to the room between the largest PAM symbol and the boundary 0
.
5
V
, Eq. (5.94)
canbewrittenas
s
(
n
)=
s
(
n
)+
q
(
n
)
.
In this case it is as though there is no
modulo operation at all, and the channel has been perfectly pre-equalized!
Proof of Eq. (5.93).
Since
B
(
z
) is as in Eq. (5.89), we see from Fig. 5.23
that
x
(
n
)=
L
)+
s
(
n
)
mod
V,
−
h
(1)
x
(
n
−
1)
−
...
−
h
(
L
)
x
(
n
−
which implies
x
(
n
)+
h
(1)
x
(
n
L
)
mod
V
=
s
(
n
)
.
−
1) +
...
+
h
(
L
)
x
(
n
−
(5
.
95)
But the output of the channel
H
(
z
) in Fig. 5.23 is
y
(
n
)=
x
(
n
)+
h
(1)
x
(
n
−
1) +
...
+
h
(
L
)
x
(
n
−
L
)
.
Search WWH ::
Custom Search