Partial Fractions

Matthew M. Peet
Illinois Institute of Technology

The Partial Fraction Expansion

Introduction

In this Lecture, you will learn: The Inverse Laplace Transform

• Simple Forms

The Partial Fraction Expansion

• How poles relate to dominant modes
• Expansion using single poles
• Repeated Poles
• Complex Pairs of Poles
I Inverse Laplace

M. Peet Lecture 7: Control Systems 2 / 27

Recall: The Inverse Laplace Transform of a Signal

To go from a frequency domain signal, û(s), to the time-domain signal, u(t), we

use the Inverse Laplace Transform.

Definition 1.
The Inverse Laplace Transform of a signal û(s) is denoted u(t) = Λ−1 û.
Z ∞
u(t) = Λ−1 û = eıωt û(ıω)dω
0

• Like Λ, the inverse Laplace Transform Λ−1 is also a Linear system.

• Identity: Λ−1 Λu = u.
• Calculating the Inverse Laplace Transform can be tricky. e.g.

s3 + s2 + 2s − 1
ŷ =
s4 + 3s3 − 2s2 + s + 1

M. Peet Lecture 7: Control Systems 3 / 27

Poles and Rational Functions
Definition 2.
A Rational Function is the ratio of two polynomials:

n(s)
û(s) =
d(s)

Most transfer functions are rational.

Definition 3.
n(s)
The point sp is a Pole of the rational function û(s) = d(s) if d(sp ) = 0.

• It is convenient to write a rational function using its poles

n(s) n(s)
=
d(s) (s − p1 )(s − p2 ) · · · (s − pn )
• The Inverse Laplace Transform of an isolated pole is easy:
1
û(s) = means u(t) = e−pt
s+p
M. Peet Lecture 7: Control Systems 4 / 27
Partial Fraction Expansion
Definition 4.
The Degree of a polynomial n(s), is the highest power of s with a nonzero
coefficient.
Example: The degree of n(s) is 4

n(s) = s4 + .5s2 + 1

Definition 5.
n(s)
A rational function û(s) = d(s) is Strictly Proper if the degree of n(s) is less
than the degree of d(s).

• We assume that n(s) has lower degree than d(s)

• Otherwise, perform long division until we have a strictly proper remainder

s3 + 2s2 + 6s + 7 2
=s+1+ 2
s2 + s + 5 s +s+5
M. Peet Lecture 7: Control Systems 5 / 27
Poles and Inverse Laplace Transforms

A Strictly Proper Rational Function is The Sum of Poles: We can usually

find coefficients ri such that

n(s) r1 rn
= + ... +
(s − p1 )(s − p2 ) · · · (s − pn ) s − p1 s − pn

• Except in the case of repeated poles.

Poles Dominate the Motion: Because a signal is the sum of poles, The
inverse Laplace has the form

u(t) = r1 ep1 t + . . . + rn epn t

• pi may be complex.
I If p are complex, r may be complex.
i i

• Doesn’t hold for repeated poles.

M. Peet Lecture 7: Control Systems 6 / 27

Examples
Simple State-Space: Step Response
s−1 2 1
ŷ(s) = = −
(s + 1)s s+1 s
1
y(t) = e−t − 1(t)
2
Suspension System: Impulse Response  
s
1 1 1 2J  1 1
ŷ(s) = = − 
J s2 − M2Jgl
q q
J M gl s − M gl s + M gl
2J 2J

2J  √ M gl t √ M gl 
s
1
y(t) = e 2J − e− 2J t
J M gl

Simple State-Space: Sinusoid Response  

s−1 s 1
ŷ(s) = = −
(s + 1)(s2 + 1) s2 + 1 s + 1
1 1
y(t) = cos t − e−t
2 2
M. Peet Lecture 7: Control Systems 7 / 27
Partial Fraction Expansion

Conclusion: If we can find coefficients ri such that

n(s) n(s)
û(s) = =
d(s) (s − p1 )(s − p2 ) · · · (s − pn )
r1 rn
= + ··· + ,
s − p1 s − pn

then this is the PARTIAL FRACTION EXPANSION of û(s).

We will address several cases of increasing complexity:

1. d(s) has all real, non-repeating roots.
2. d(s) has all real roots, some repeating.
3. d(s) has complex, repeated roots.

M. Peet Lecture 7: Control Systems 8 / 27

Case 1: Real, Non-repeated Roots
This case is the easiest:
n(s)
ŷ(s) =
(s − p1 ) · · · (s − pn )

where pn are all real and distinct.

Theorem 6.
For case 1, ŷ(s) can always be written as
r1 rn
ŷ(s) = + ··· +
s − p1 s − pn
where the ri are all real constants
The trick is to solve for the ri
• There are n unknowns - the ri .
• To solve for the ri , we evaluate the equation for values of s.
I Potentially unlimited equations.
I We only need n equations.
I We will evaluate at the points s = pi .
M. Peet Lecture 7: Control Systems 9 / 27
Case 1: Real, Non-repeated Roots
Solving

We want to solve
r1 rn
ŷ(s) = + ··· + for r1 , r2 , . . .
s − p1 s − pn
For each ri , we Multiply by (s − pi ).
s − pi s − pi s − pi
ŷ(s)(s − pi ) = r1 + · · · + ri + · · · + rn
s − p1 s − pi s − pn
s − pi s − pi
= r1 + · · · + ri + · · · + rn
s − p1 s − pn
Evaluate the Right Hand Side at s = pi : We get
pi − pi pi − pi
r1 + · · · + ri + · · · + rn
pi − p1 pi − pn
0 0
= r1 + · · · + ri + · · · + rn = ri
pi − p1 pi − pn
Setting LHS = RHS, we get a simple formula for ri :
ri = ŷ(s)(s − pi )|s=pi
M. Peet Lecture 7: Control Systems 10 / 27
Case 1: Real, Non-repeated Roots
Calculating ri

So we can find all the coefficients:

ri = ŷ(s)(s − pi )|s=pi

n(s)
ŷ(s)(s − pi ) = (s − pi )
(s − p1 ) · · · (s − pi−1 )(s − pi )(s − pi+1 ) · · · (s − pn )
n(s)
=
(s − p1 ) · · · (s − pi−1 )(s − pi+1 ) · · · (s − pn )

• This is ŷ with the pole s − pi removed.

Definition 7.
The Residue of ŷ at s = pi is the value of ŷ(pi ) with the pole s − pi removed.

Thus we have
n(pi )
ri = ŷ(s)(s − pi )|s=pi =
(pi − p1 ) · · · (pi − pi−1 )(pi − pi+1 ) · · · (pi − pn )
M. Peet Lecture 7: Control Systems 11 / 27
Example: Case 1
2s + 6
ŷ(s) =
(s + 1)(s + 2)
We first separate ŷ into parts:
2s + 6 r1 r2
ŷ(s) = = +
(s + 1)(s + 2) s+1 s+2
Residue 1: we calculate:
2s + 6 2s + 6 −2 + 6 4
r1 = (s + 1)|s=−1 = |s=−1 = = =4
(s + 1)(s + 2) s+2 −1 + 2 1
Residue 2: we calculate:
2s + 6 2s + 6 −4 + 6 2
r2 = (s + 2)|s=−2 = |s=−2 = = = −2
(s + 1)(s + 2) s+1 −2 + 1 −1
Thus 2

1.8

4 −2 1.6

ŷ(s) = + 1.4

s+1 s+2 y(t) 1.2

0.8
Concluding, 0.6

y(t) = 4e−t − 2e−2t

0.4

0.2

0
0 1 2 3 4 5 6
t

M. Peet Lecture 7: Control Systems 12 / 27

Case 2: Real, Repeated Roots
Sometimes ŷ has a repeated pole:
n(s)
ŷ(s) =
(s − p1 )q (s − p2 ) · · · (s − pn )
In this case, we CAN NOT use simple expansion (As in case 1). Instead we
have
n(s)
ŷ(s) =
(s − p1 )q (s
− p2 ) · · · (s − pn )
 
r11 r12 r1q r2 rn
= + + ··· + + + ··· +
(s − p1 ) (s − p1 )2 (s − p1 )q s − p2 s − pn

• The rij are still real-valued.

Example: Find the rij

(s + 3)2 r11 r12 r13 r21 r22

= + + + +
(s + 2)3 (s + 1)2 s + 2 (s + 2)2 (s + 2)3 s + 1 (s + 1)2

M. Peet Lecture 7: Control Systems 13 / 27

Case 2: Real, Repeated Roots

Problem: We have more coefficients to find.

• q coefficients for each repeated root.

First Step: Solve for r2 , · · · , rn as before:

If pi is not a repeated root, then

n(pi )
ri =
(pi − p1 )q · · · (pi − pi−1 )(pi − pi+1 ) · · · (pi − pn )

M. Peet Lecture 7: Control Systems 14 / 27

Case 2: Real, Repeated Roots

New Step: Multiply by (s − p1 )q to get the coefficient r1q .

(s − p1 )q (s − p1 )q (s − p1 )q
 
ŷ(s)(s − p1 )q = r11 + r12 2
+ · · · + r1q
(s − p1 ) (s − p1 ) (s − p1 )q
(s − p1 )q (s − p1 )q
+ r2 + · · · + rn
s − p2 s − pn
q−1
+ r12 (s − p1 )q−2 + · · · + r1q


= r11 (s − p1 )
(s − p1 )q (s − p1 )q
+ r2 + · · · + rn
s − p2 s − pn
and evaluate at the point s = pi to get .

r1q = ŷ(s)(s − p1 )q |s=p1

• r1q is ŷ(p1 ) with the repeated pole removed.

M. Peet Lecture 7: Control Systems 15 / 27

Case 2: Real, Repeated Roots
To find the remaining coefficients, we Differentiate:

ŷ(s)(s − p1 )q = r11 (s − p1 )q−1 + r12 (s − p1 )q−2 + · · · + r1q

(s − p1 )q (s − p1 )q
+ r2 + · · · + rn
s − p2 s − pn
to get
d
(ŷ(s)(s − p1 )q )
ds
= (q − 1)r11 (s − p1 )q−2 + (q − 2)r12 (s − p1 )q−3 + · · · + r1,(q−1)


 
d r2 rn
+ (s − p1 )q + ··· +
ds s − p2 s − pn

Evaluating at the point s = p1 .

d
(ŷ(s)(s − p1 )q ) |s=p1 = r1,(q−1)
ds

M. Peet Lecture 7: Control Systems 16 / 27

Case 2: Real, Repeated Roots
If we differentiate again, we get

d2
(ŷ(s)(s − p1 )q )
ds2
= (q − 1)(q − 2)r11 (s − p1 )q−3 + · · · + 2r1,(q−2)

d2
 
q r2 rn
+ 2 (s − p1 ) + ··· +
ds s − p2 s − pn

Evaluating at s − p1 , we get

1 d2
r1,(q−2) = (ŷ(s)(s − p1 )q ) |s=p1
2 ds2
Extending this indefinitely

1 dq−j−1
r1,j = (ŷ(s)(s − p1 )q ) |s=p1
(q − j − 1)! dsq−j−1

Of course, calculating this derivative is often difficult.

M. Peet Lecture 7: Control Systems 17 / 27
Example: Real, Repeated Roots
Expand a simple example
s+3 r11 r12 r2
ŷ(s) = 2
= + 2
+
(s + 2) (s + 1) s + 2 (s + 2) s+1
First Step: Calculate r2
s+3 −1 + 3 2
r2 = |s=−1 = = =2
(s + 2)2 (−1 + 2)2 1
Second Step: Calculate r12
s+3 −2 + 3 1
r12 = |s=−2 = = = −1
s+1 −2 + 1 −1
The Difficult Step: Calculate r11 :
 
d s+3
r11 = |s=−2
ds s + 1
 
1 s+3 1 −2 + 3 1 1
= − |s=−2 = − = − = −2
s + 1 (s + 1)2 −2 + 1 (−2 + 1)2 −1 1

M. Peet Lecture 7: Control Systems 18 / 27

Example: Real, Repeated Roots

So now we have
s+3 −2 −1 2
ŷ(s) = = + +
(s + 2)2 (s + 1) s + 2 (s + 2)2 s+1
1
Question: what is the Inverse Fourier Transform of (s+2)2 ?
Recall the Power Exponential:
0.35

1 tm−1 e−at 0.3

→
(s + a)m (m − 1)! 0.25

0.2
y(t)

0.15
Finally, we have: 0.1

−2t −2t −t
y(t) = −2e − te + 2e 0.05

0
0 1 2 3 4 5 6 7 8 9 10
t

M. Peet Lecture 7: Control Systems 19 / 27

Case 3: Complex Roots
Most signals have complex roots (More common than repeated roots).
3
ŷ(s) =
s(s2 + 2s + 5)

has roots at s = 0, s = −1 − 2ı, and s = −1 + 2ı.

Note that:
• Complex roots come in pairs.
• Simple partial fractions will work, but NOT RECOMMENDED
I Coefficients will be complex.

I Solutions will be complex exponentials.

I Require conversion to real functions.

Best to Separate out the Complex pairs as:

n(s)
ŷ(s) =
(s2 + as + b)(s − p1 ) · · · (s − pn )

M. Peet Lecture 7: Control Systems 20 / 27

Case 3: Complex Roots

Complex pairs have the expansion

n(s) k1 s + k2 r1 rn
= 2 + + ··· +
(s2 + as + b)(s − p1 ) · · · (s − pn ) s + as + b s − p1 s − pn

Note that there are two coefficients for each pair: k1 and k2 .
NOTE: There are several different methods for finding k1 and k2 .
First Step: Solve for r2 , · · · , rn as normal.
Second Step: Clear the denominator. Multiply equation by all poles.
 
k1 s + k2 r1 rn
n(s) = (s2 + as + b)(s − p1 ) · · · (s − pn ) 2 + + ··· +
s + as + b s − p1 s − pn

Third Step: Solve for k1 and k2 by examining the coefficients of powers of s.

Warning: May get complicated or impossible for multiple complex pairs.

M. Peet Lecture 7: Control Systems 21 / 27

Case 3 Example
Take the example

2(s + 2) k1 s + k2 r1
ŷ(s) = = 2 +
(s + 1)(s2 + 4) s +4 s+1

First Step: Find the simple coefficient r1 .

2(s + 2) −1 + 2 1 2
r1 = 2
|s=−1 = 2 2 =2 =
s +4 −1 + 4 5 5

Next Step: Multiply through by (s2 + 4)(s + 1).

2(s + 2) = (s + 1)(k1 s + k2 ) + r1 (s2 + 4)

Expanding and using r1 = 2/5 gives:

2s + 4 = (k1 + 2/5)s2 + (k2 + k1 )s + k2 + 8/5

M. Peet Lecture 7: Control Systems 22 / 27

Case 3 Example

Recall the equation

2s + 4 = (k1 + 2/5)s2 + (k2 + k1 )s + k2 + 8/5

Equating coefficients gives 3 equations:

• s2 term - 0 = k1 + 2/5
• s term - 2 = k2 + k1
• s0 term - 4 = k2 + 8/5

An over-determined system of equations (but consistent)

• First term gives k1 = −2/5
• Second term gives k2 = 2 − k1 = 10/5 + 2/5 = 12/5
• Double-Check with last term: 4 = 12/5 + 8/5 = 20/5

M. Peet Lecture 7: Control Systems 23 / 27

Case 3 Example
So we have
 
2(s + 2) 2 1 s−6
ŷ(s) = = −
(s + 1)(s2 + 4) 5 s + 1 s2 + 4
 
2 1 s 6
= − +
5 s + 1 s2 + 4 s2 + 4
2 −t

y(t) = e − cos(2t) + 3 sin(2t)
5

1.5

0.5
y(t)

−0.5

−1

−1.5
0 5 10 15 20 25 30
t

M. Peet Lecture 7: Control Systems 24 / 27

Case 3 Example

Now consider the solution to A Different Numerical Example: (Nise)

3 3/5 3 s+2
= −
s(s2 + 2s + 5) s 5 s2 + 2s + 5

What to do with term:

s+2
?
s2 + 2s + 5

We can rewrite as the combination of a frequency shift and a sinusoid:

s+2 s+2 (s + 1) + 1
= 2 =
s2 + 2s + 5 s + 2s + 1 + 4 (s + 1)2 + 4
(s + 1) 1 2
= +
(s + 1)2 + 4 2 (s + 1)2 + 4

M. Peet Lecture 7: Control Systems 25 / 27

Case 3 Example
s+2 (s + 1) 1 2
= +
s2 + 2s + 5 (s + 1)2 + 4 2 (s + 1)2 + 4
(s+1)
• is the sinusoid s2s+4 shifted by s → s + 1.
(s+1)2 +4
• s → s + 1 means multiplication by e−t in the time-domain.
•    
−1 (s + 1) −t −1 s
Λ =e Λ = e−t cos 2t
(s + 1)2 + 4 s2 + 4
 
2
• Likewise Λ−1 = e−t sin 2t
(s + 1)2 + 4
1

0.8

0.6

0.4
y(t)

0.2

−0.2

−0.4
0 1 2 3 4 5 6 7 8
  t  
−1 s+2 −t 1
Λ =e cos 2t + sin 2t
s2 + 2s + 5 2
M. Peet Lecture 7: Control Systems 26 / 27
Summary

What have we learned today?

In this Lecture, you will learn: The Inverse Laplace Transform

• Simple Forms

The Partial Fraction Expansion

• How poles relate to dominant modes
• Expansion using single poles
• Repeated Poles
• Complex Pairs of Poles
I Inverse Laplace

Next Lecture: Important Properties of the Response

M. Peet Lecture 7: Control Systems 27 / 27