Laplace Transforms and use

in Automatic Control

P.S. Gandhi
Mechanical Engineering
IIT Bombay

Acknowledgements: P.Santosh Krishna, SYSCON

 Recap

Fourier series
Fourier transform: aperiodic
Convolution integral
Frequency response
 Convolution Integral
Linear, time-invariant (LTI) systems can be
completely characterized by their impulse response

Consider an LTI system whose impulse response

is h(t), the response of the system to arbitrary
input u(t) is given by:

y(t)=h(t)*u(t)

y (t ) = h(t )u( )d

 Convolution Integral
Fundamentals

Convolution as sum
of inverted and shifted
impulse responses
at various input points
 Frequency Response

e jt LTI ? H(j) e jt

Frequency response of the system = fourier transform of the

impulse response
 Complex Exponentials
and LTI systems

e jt LTI ? H(j) e jt

What will be the response of the system to a sum

of complex exponentials?
 Why Laplace transforms ??
Integration and differentiation can be replaced by
algebraic operations in complex plane.
It converts complex functions such as sinusoidal,
exponential into algebraic functions of complex
variable s
These algebraic equations can be easily solved in
s for dependent variable whose time domain
response can be found with inverse Laplace
transforms MA 203

Q: cant we do these by using Fourier transform?

A: Not always. We will consider some cases later
 Definition of L.T.
If f ( t ) is a function of time , f ( t ) = 0 for t < 0
and s is a complex variable
and F(s) is the laplace transform of function f ( t )

Then the laplace transform of f ( t ) is given by,

F ( s ) = L f ( t ) = f (t ).e dt
st

What if s = jw ????? Fourier transform

formula
 Inverse Laplace Transforms

If F(s) is the laplace transform of a function,

then its time domain response is given by

c + j
1
L F ( s ) = f (t ) =
1

st
F ( s )e ds
2 j c j
for t >0
This is called inverse Laplace transform of F(s)
 Advantages of L.T over F.T
We have already seen Fourier transform which is a
function of the complex variable jw . Laplace trans-
form is a function of the complex variable s denoting
' + jw ' in which if = 0, then Laplace transforms
equals Fourier transforms.
What does this mean physically???

Laplace transforms are introduced to fill the gaps

which Fourier transform does not; We see
these two cases with examples.
 Advantages of L.T over F.T
Case 1: Unstable systems
Stable system is one which gives bounded output
for bounded input. Consider the unstable system
y ( t ) = t .u ( t ) whose output diverge for a unit step
input.
Applying Fourier transforms on both sides, we get

Y ( jw ) = tu (t ).e dt
jwt

 Advantages of L.T over F.T
But Dirichlets first condition states that [t.u(t)] should
be absolutely integrable for the Fourier transform to
exist. That is,

| tu ( t ) | dt <

which is violated in this case. Thus, it is clear that we

cant find Fourier transform for unstable systems. Now,
taking Laplace transforms,
 Advantages of L.T over F.T

Y (s) = tu (t ).e
st
dt


t.e
st
e st

= t.e dt
st
= 2
0 s s 0
1
= 2 for Re{s}> 0
s
So, it is clear that for unstable systems, even though
Fourier transform does not exist, Laplace transform
exits which helps to investigate the stability of a system
 Advantages of L.T over F.T
Case 2:
Let us take the input signal as below instead of unit step
x (t ) = e .u (t )
at

Since we need to find F.T of both the system and the

input signal in order to obtain the output response, we
need to find F.T of this input signal given above

X ( jw ) = e
at
u (t ).e jwt
dt

 Advantages of L.T over F.T

X ( jw ) = e at jwt
dt
0
The right side part is tending to infinite, so we cant
find F.T for this input signal. We can conclude this
without any calculations directly from Dirichlet first
condition as in case-1. Taking Laplace transforms,

X (s) = e at u (t ).e st dt

 Advantages of L.T over F.T

X ( s ) = e t ( s a ) dt
0

1 for Re {s} > a

=
sa
So, for some values of Re{s}, we are able to find L.T
for the given signal. This is called Region of Convergenc
(ROC). Here, it is the area in s-plane whose real part is
greater than a. In case-1, ROC in the right half of the
S-plane.
 How to find laplace
transform ??
Consider a sinusoidal function,
f ( t ) = 0, for t < 0
= A sin wt , for t >= 0

But we know that,

1 jwt jwt
sin wt = (e e )
2j
 contd.
Hence

A
L [ A sin wt ] = ( e jwt
e jwt
)e st
dt
2j 0
A 1 A 1
= . .
2 j s jw 2 j s + jw
Aw
= 2
s +w 2
 Laplace transforms for some
commonly used functions
Function Transform
Unit impulse 1
Unit step 1/ s
n n!
t s n +1
at 1
e
s+a
w
sin wt
s 2 + w2
contd.
Function Transform
s
cos wt s 2 + w2
w
sinh wt s 2 w2
s
cosh wt s w
2 2
 Some other functions
at
Observe how e affects these functions,
Function Transform
n!
at n
e t ( s + a )n +1
at w
e sin wt
( s + a ) 2 + w2

at s+a
e cos wt
( s + a )2 + w2
 Contd.
Function Transform
at w
e sinh wt
( s + a ) 2 w2

e at cosh wt s+a
( s + a ) 2 w2

i.e,

L e at . f ( t ) = e at . f (t ).e st dt = F ( s + a )
0
 L.T of a translated function

let F(s) be the L.T of f(t) and l(t) is a function

such that
f (t ).l (t ) = 0 for t <
 Contd.

then L.T of f (t ).u (t ) is given by,

L [ f ( t ).l ( t )] =
0
f ( t ) l ( t ).e st dt

=

f ( t ) l ( t ) .e s ( t + ) d t

=
0
f ( t ).e s t .e t d t

s
=e .F ( s )
 Change of time scale

If t is changed into t / , where is a

positive constant. Then function f (t ) has
changed into f t , then L.T of f t is,


t t st
L f =


0
f

.e d t


= . f ( t ).e st dt = F ( s )
0
 Laplace transform theorems
Real differentiation theorem
d
L f (t ) = sF ( s ) f (0) IMP for LTI
dt
Real integration theorem
1
F ( s ) f (0)
L f (t )dt = +
s s
Complex differentiation theorem
d
L t . f ( t ) = F (s)
ds
 Initial value theorem
From this theorem, we can find the value of
f(t) at t=0+ directly from the L.T of f(t).
This theorem will not give the value of f(t)
exactly at t=0 but slightly greater than zero

df (t )
Def: If f(t) and are laplce transformable
dt
and if lim
s
sF ( s ) exists then,

f (0 + ) = lim sF ( s )
s
 Final value theorem
This theorem gives the value of f(t) when t
This theorem applies if and only if lim f (t ) exits
t
All the poles of sF ( s ) should lie in the left half of
s-plane for lim f (t ) to exist.
t

df (t )
Def: If f(t) and are laplce transformable
dt
and if lim f (t ) exists then,
t

lim f (t ) = lim sF ( s )
t s 0
 Convolution Integral
t

The integral 0 f1 ( t ) f 2 ( ) d can also be

written as f1 (t ) f 2 (t ) which is known as
convolution integral.
If F1 ( s) & F2 ( s) are L.Ts of f1 (t ) & f 2 (t )
respectively, then L.T of convolution
integral is given by,
t
L f1 (t ) f1 (t ) d = F1 ( s ).F2 ( s )
0
Blocks can be multiplicative in Laplace domain
 L.T of product of two
time functions
If F(s) and G(s) are the laplace transforms
of f(t) and g(t) respectively, then the
laplace transform of f (t ) g (t ) is given by,
c + j
1
L [ f (t ) g (t ) ] = F ( p ).G ( s p ) dp
2 j c j
 Example Problem
We have already seen block diagram representation to
a system with integrators, differentiators and multiplier
blocks. But this block diagram representation becomes
much simpler with this Laplace transforms as shown.
Let us consider the system given in exam,
.. .
y (t ) + 5 y (t ) + 2 y (t ) = 10.x(t )
First let us find the system response applying L.Ts
s Y ( s ) + 5s.Y ( s ) + 2Y ( s ) = 10. X ( s )
2
 Ex. Problem (Contd)
Y (s) 10
= 2
X ( s ) s + 5s + 2
This ratio of output to input is called Transfer Function
After finding the partial fractions and applying Inverse
Laplace transforms to above transfer function with unit
Step as input, we get the following output,
0.439 t 4.562 t
y (t ) = 2.425(e e ) + 5u (t )
Solve and prove in class
 Ex. Problem (Contd)
The block diagram representation for the above system
with integrator and multiplier blocks is,
 Ex. Problem (Contd)
For designing this small system, we needed many
integrator, gain and summation blocks leading to big
block diagram. Instead, we can form it in S-domain
with only a single transfer function block which is
obtained in previous slide.
 Ex. Problem (Contd)
This reduces the complexity of the block diagram very
much and thus representation of much bigger plants
also became simpler with just these transfer function
blocks and multiplier blocks. The output response of
the system will be same in both the block diagram
representations with MATLAB. These output responses
are shown in next slide.
 Ex. Problem (Contd)
Response in time domain block diagram
 Ex. Problem (Contd)
Response in transfer function block diagram