Digital Signal Processing Reference
In-Depth Information
Fig. 1
(
a
) Full adder: truth
table. (
b
) Full adder: symbol
a
a
i
b
i
d
i
s
i
c
i
00000
00110
01010
01101
10010
10101
11001
11111
b
a
i
b
i
d
i
FA
c
i
s
i
1.3.1
Signed-Digit Representation
In a signed-digit (SD) number representation, the digits may have either positive
or negative sign. For a radix-2 representation we have
x
i
∈{−
1
,
0
,
1
}
.Using
W
f
−
1
+
Q
≤
X
≤
1
−
Q
. A number may
now have more than one representation. Consider the number 0
.
25
10
which can be
1 1
SD
,where
¯
1isusedtodenote
written as
1.
digit representation with a minimum number of non-zero positions to simplify
signed-digit (MSD) representation. However, in general it is non-trivial to determine
if a representation is minimum. A specific minimum signed-digit representation is
obtained if the constraint
x
i
x
i
+
1
=
.
01
SD
or
.
−
i
is imposed. The resulting representation is
called canonic signed-digit (CSD) representation and is apart from being minimum
also unique (as the name indicates).
0
,∀
1.3.2
Carry-Save Representation
The carry-save representation stems from the ripple-carry adder, which will be
bits are stored and the data is represented using two vectors. This also leaves an
additional input of the full adder cells unused, so it is possible to add three vectors.
this we can see that
s
i
=
parity
(
a
i
⊕
b
i
⊕
d
i
)=
a
i
⊕
b
i
⊕
d
i
(3)
where
⊕
is the exclusive-OR operation, and
=
(
,
,
)=
+
+
.
c
i
majority
a
i
b
i
d
i
a
i
b
i
a
i
d
i
b
i
d
i
(4)