You are on page 1of 2

# Yet more notes on partial fraction expansion

Revised Edition
The most general case of partial fraction expansion assumes you have a proper
fraction with all positive powers of z (if its not proper, carry out long division
until it is; if it has negative powers multiply by z
n
/z
n
, where n is the highest
negative power).
Lets suppose weve satised the above requirements, then we have an X(z)
that can be written as
X(z) =
B(z)
A(z)
,
where the numerator is of order m and the denominator of order n, m n. The
rst step is to factor the denominator:
X(z) =
B(z)
(z p
1
)(z p
2
) . . . (z p
n
)
,
where the p
i
are poles of X(z). Then we need to expand this into a sum of
terms
X(z) = c
0
+
c
1
z
z p
1
+
c
2
z
z p
2
+ . . . +
c
n
z
z p
n
,
the inverse z-transform of each of which we can look up in a table.
The individual c
i
factors are solved for as follows:
c
0
= X(z) |
z=0
,
which is actually just the ratio of the constant terms in the numerator and
denominator, while for 0 < i n
c
i
=
_
z p
i
z
X(z)
_
z=p
i
.
Example: Find x[n] when
X(z) =
3
5
2
z
1
1
3
2
z
1
+
1
2
z
2
.
First, we need to multiply through by z
2
/z
2
to make all the powers positive,
leaving us with
X(z) =
3z
2

5
2
z
z
2

3
2
z +
1
2
.
Then we factor the denominator:
X(z) =
3z
2

5
2
z
(z
1
2
)(z 1)
.
1
Now solve for the c
i
:
c
0
=
0
1
2
= 0
c
1
=
_
z
1
2
z
3z
2

5
2
z
(z
1
2
)(z 1)
_
z=
1
2
=
3
_
1
4
_

5
2
_
1
2
_
1
2
_
1
2
1
_ = 2
c
2
=
_
z 1
z
3z
2

5
2
z
(z
1
2
)(z 1)
_
z=1
=
3
5
2
1
_
1
1
2
_ = 1.
Thus
X(z) =
2z
z
1
2
+
z
z 1
,
which we can look up in the table to nd
x[n] = u[n]
_
2
_
1
2
_
n
+ 1
n
_
= u[n]
_
_
1
2
_
n1
+ 1
_
.
Repeated poles: The method explained above needs an additional step
when a pole appears more than once in the denominator. Suppose some pole
p
i
shows up r times, or in other words the denominator contains the expression
(z p
i
)
r
. Then the expansion requires all the terms
c
i
1
z
z p
i
+
c
i
2
z
2
(z p
i
)
2
+ . . . +
c
i
r
z
r
(z p
i
)
r
,
which can be iteratively (tediously) solved for as
c
i
r
=
_
(z p
i
)
r
z
r
X(z)
_
z=p
i
,
c
i
(r1)
=
_
(z p
i
)
r1
z
r1
_
X(z)
c
i
r
z
r
(z p
i
)
r
__
z=pi
,
c
i
(r2)
=
_
(z p
i
)
r2
z
r2
_
X(z)
c
i
r
z
r
(z p
i
)
r

c
i
(r1)
z
r1
(z p
i
)
r1
__
z=p
i
,
and so on. Note that you must simplify the expressions inside the braces be-
fore doing the z = p
i
substitution; if you dont, and if there are any (z p
i
)
denominators, you will nd it dicult to evaluate the result correctly.
2