Systems Analysis and Control

Matthew M. Peet
Arizona State University

Lecture 5: Calculating the Laplace Transform of a Signal

Introduction

In this Lecture, you will learn:

Laplace Transform of Simple Signals.
Step, Exponentials, Ramps.
Impulse

Properties of the Laplace Transform.

delay, scaling
convolution
etc.

Transfer Functions
State Space to Transfer Function

M. Peet Lecture 5: Control Systems 2 / 30

Previously:
The Laplace Transform

Definition 1.
The Laplace Transform of the signal u(t) is
Z
u(s) = est u(t)dt
0

The Laplace Transform is not the same as the Fourier transform.

Assumes that signals begin at time t = 0 and u(t) = 0 for t < 0.
Tends to flatten out peaks.
Signal:
u(t) = sin at
10

7
magnitude of u(iw)
6

Laplace Transform: 5

a 1

u(s) = 2 0
0 1 2 3 4 5
Frequency(w)
6 7 8 9 10

s a2
M. Peet Lecture 5: Control Systems 3 / 30
The Laplace Transform
Example: Step Function
Consider the input signal:
(
0 t<0
u(t) =
1 t0

The Laplace transform is

Z
u(s) = est dt
0
1  st  1
= e 0
= [0 1]
s s
1
=
s
Note that limt est = 0.
Because we assume the real part of
s is negative.
M. Peet Lecture 5: Control Systems 4 / 30
Review: Integration by parts

A very useful formula for finding the Laplace Transform is

integration by parts.
Z b Z b
0 s=b
u(s)v (s)ds = [u(s)v(s)]s=a u0 (s)v(s)ds
a a

M. Peet Lecture 5: Control Systems 5 / 30

The Laplace Transform
Example: Ramp Function
Consider the input signal: 10

( 9

0 t<0 8

u(t) = 7

t t0

Ramp Function
6

= t 1(t) 4

Recall 1(t) is the step function. 2

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

The Laplace transform is t

Z
u(s) = test dt
0

1 st
 Z
1 st
= te + e dt
s 0 s 0
1
= [0 0] 2 [est ]
0 dt
s
1 1
= 2 [0 1] = 2
s s
M. Peet Lecture 5: Control Systems 6 / 30
The Laplace Transform
Example: Quadratic Function

The Quadratic is similar to the Ramp. 25

(
20
0 t<0
u(t) = 2

Quadratic Function
t t0 15

= t2 1(t) 10

5
Recall 1(t) is the step function.
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t
The Laplace transform is
Z
u(s) = t2 est dt
0

1 st
 Z
1 2 st
= t e + te dt
s 0 s 0
1
= 3
s

M. Peet Lecture 5: Control Systems 7 / 30

The Laplace Transform
Example: exponential

The Sinusoid is created from the complex exponentials.

Lets do exponentials first. 1

0.9
Consider the input signal: 0.8
( 0.7

0 t<0

Exponential Function
0.6
u(t) = at
e t0 0.5

0.4

= eat 1(t)
0.3

0.2

0.1

Recall 1(t) is the step function. 0

0 1 2
t
3 4 5

The Laplace transform is

Z
u(s) = e(as)t dt
0 
1 (as)t
= e
as 0
1
=
sa

M. Peet Lecture 5: Control Systems 8 / 30

The Laplace Transform
Linearity

Input Output
u u^=u

The Laplace Transform is itself a LINEAR SYSTEM.

The Input is u
The Output is u
Lets call the system L
u = Lu
As for any linear system,

L(u1 + u2 ) = Lu1 + Lu2

= u1 + u2
We can compute the Laplace transform by using simple parts.
Example: Sinusoids
et + et
cos t =
2
M. Peet Lecture 5: Control Systems 9 / 30
The Laplace Transform
Example: Sinusoids
Recall that
et + et
cos t =
2
But the Laplace transform of eat is
1
L(eat ) =
sa

So the Laplace transforms of et and et are

1 1
u1 (s) = L(et ) = and u2 (s) = L(et ) =
s s +
Which shows us how to find L(cos(t))
e + et
 t   t   t 
e e u1 u2
L(cos(t)) = L =L +L = +
2 2 2 2 2
   
1 1 1 1 2s s
= + = = 2
2 s s + 2 (s )(s + ) s + 2
M. Peet Lecture 5: Control Systems 10 / 30
The Laplace Transform
Example: Sinusoids

Likewise, for the sin function,

et et
sin t =
2i

Which shows us how to find L(cos(t))

u1 u2
L(sin(t)) =
 2 2i 
1 1 1
=
2i s s +
 
1 2
= = 2
2i (s )(s + ) s + 2
M. Peet Lecture 5: Control Systems 11 / 30
The Laplace Transform
Example: Impulse functions

Definition 2.
An Impulse is a significant strike which occurs in an infinitesimally short time.
Mathematically, we model the impulse as a Dirac delta function, (t).
Question: What is the Laplace transform of an
Impulse?
Answer: By definition of the Dirac Delta:

magnitude
Z
(s) = est (t)dt
0
= est |t=0
=1

0 time
The Impulse has a Uniform Frequency Response!!!!
The opposite of the step function (1(s) = L(t)).

M. Peet Lecture 5: Control Systems 12 / 30

Laplace Transform Table
Simple Signals

Summary: We can find the Laplace Transform for many simple signals
u(t) u(s)
1
Step 1(t) s
m!
Power tm sm+1
1
Exponential eat s+a
m1 at
t e 1
Power Exponential (m1)! (s+a)m
a
Sine sin(at) s2 +a2
s
Cosine cos(at) s2 +a2
Impulse (t) 1
Remember that all functions are only for t 0.
We always have u(t) = 0 for t < 0

Question: How to use these functions to solve for more complex functions?

M. Peet Lecture 5: Control Systems 13 / 30

The Laplace Transform
Other Properties

Properties: What properties of the Laplace transform can we exploit?

We have already discussed Linearity.
Other Examples:
Delay: The signal is delayed

u(t) u(t )

Time-Scale: The signal is slowed down or speeded up

u(t) u(at)

Differentiation: The signal is a differential in time:

u(t) u0 (t)

How do these changes affect the the Laplace transform?

u(s) ?????

M. Peet Lecture 5: Control Systems 14 / 30

The Laplace Transform
Delay

What is the effect of a delay on the Laplace Transform?

Use a change of variables: t0 = t , dt0 = dt

Z
udelayed (s) = L(u(t )) = est u(t )dt
Z0
0
= es(t + ) u(t0 )dt0
0
Z
0
= es est u(t0 )dt0
0
= es u(s)
M. Peet Lecture 5: Control Systems 15 / 30
The Laplace Transform
Time-Scaling

What is the effect of scaling time?

Use a change of variables: t0 = at, dt0 = adt

uscaled (s) = L(u(at))

Z
= est u(at)dt
0
1 st0
Z
= e a u(t0 )dt0
a 0
1 s
= u
a a

M. Peet Lecture 5: Control Systems 16 / 30

Example: Watch The Order of Operations
What is the Laplace Transform of

u(t) = sin(a(t ))
1
We know that usin (s) = L(sin(t)) = s2 +1 .
The change is sin t sin(at) sin(a(t ))
1 s


Time-Scaling: us (s) = a u a
Time-Delay: ud (s) = e s
u (s)
Thus we have

uds (s) = e s us (s)

!
1
s a
=e
s 2


a + 1
a
= e s 2
s + a2

M. Peet Lecture 5: Control Systems 17 / 30

The Laplace Transform Properties
Differentiation

A very common case is when a signal has been differentiated. e.g.:

x(t) = Ax(t) + Bu(t)

In this case we use integration by parts:

Z
L (x(t)) = est x(t)dt
0
Z
 st 
= e x(t) 0 + s est x(t)dt
0
Z
= x(0) + s est x(t)dt
0
= x(0) + sx(s)

M. Peet Lecture 5: Control Systems 18 / 30

The Laplace Transform Properties
Multiple Differentiation

The differentiation property can be applied recursively:

...
L ( x (t)) = x(0) + sL(x(t))
= x(0) sx(0) + s2 L(x(t))
= x(0) sx(0) s2 x(0) + s3 x(s)

General Formula:

L(x(n) (t)) = sn x(s) sn1 x(0) xn1 (0)

M. Peet Lecture 5: Control Systems 19 / 30

The Laplace Transform Properties
Integration

Importantly, the Laplace transform can also be applied to integration:

An immediate consequence of differentiation
 
L f(t) = f (0) + sL(f (t))

Rt
If f (t) = 0
x( )d , then f(t) = x(t), and f (0) = 0, so
 
x(s) = L(x(t)) = L f(t)
= f (0) + sL(f (t))
Z t 
= sL x( )d
0

Hence Z t 
x(s)
L x( )d =
0 s

M. Peet Lecture 5: Control Systems 20 / 30

The Laplace Transform Properties
Convolution

Recall the solution of a state-space system:

Z t
y(t) = CeA(ts) Bu(s)ds + Du(t)
0

This is a special kind of integral called the Convolution Integral.

Definition 3.
The Convolution Integral of u(t) with v(t) is defined as
Z t
(u v)(t) := u( )v(t )d
0

Thus roughly speaking

y(t) = u(t) CeAt B

M. Peet Lecture 5: Control Systems 21 / 30

The Laplace Transform Properties
Convolution

Z t
(f g)(t) := f ( )g(t )d
0
M. Peet Lecture 5: Control Systems 22 / 30
The Laplace Transform Properties
Convolution

The convolution of two signals has a remarkable property:

L(u v) = u(s)v(s)

Proof.
Since u(t) = v(t) = 0 for t < 0, we have that u( )v(t ) = 0 for < 0 or
> t. Thus: Z t Z
u( )v(t )d = u( )v(t )d
0 0
Now using the delay property:
Z Z t Z Z
L(u v) = est u( )v(t )d dt = est u( )v(t )d dt
Z0 Z0 0 0

= u( ) est v(t )dtd

0 0
Z
= u( )es v(s)d = u(s)v(s)
0

M. Peet Lecture 5: Control Systems 23 / 30

Properties of the Laplace Transform
Summary

Summary: Simple functions and properties can be combined to calculate the

Laplace Transform.
u(t) u(s)
s
Delay u(t ) e u(s)
1 s


Time-Scaling u(at) a u a
Differentiation u0 (t) su(s) u(0)
Rt 1
Integration 0
u( )d s u(s)
d
Frequency Differentiation tu(t) ds u(s)
Frequency Shift eat u(t) u(s a)
Rt
Convolution 0
u( )v(t )d u(s)v(s)

M. Peet Lecture 5: Control Systems 24 / 30

The Inverse Laplace Transform
An Important Case

Consider the simple case of a cosine with an exponential:

u(t) = et sin t
Approach:

L(sin t) = s2 + 2

Exponential means frequency shift: et u(t) 7 u(s ).

Hence the Laplace transform is

u(s) = = 2
(s )2 + 2 s 2s + ( 2 + 2 )

Problem: Can we go the other direction? Calculate u(t) if

1
u(s) =
s2 + bs + c
Solution: Use a combination sinusoid and frequency shift First recall that

L(sin t) = 2
s + 2
M. Peet Lecture 5: Control Systems 25 / 30
An Important Case, continued
1
If u(s) = s2 +bs+c , the roots of the denominator are
1p 2
s1,2 = b/2 b 4c
2
Hence
1
u(s) = 1

(s + b/2 + 2 4c)(s + b/2 12 b2 4c)
b2
1
=
((s + b/2) + 2 b 4c)((s + b/2) 12 b2 4c)
1 2

1
=
((s + b/2)2 14 (b2 4c))

Now assume 4c > b2 . Then u(s) = h(s + b/2), where

1
h(s) = 1 .
s2 + 4 (4c b2 )

M. Peet Lecture 5: Control Systems 26 / 30

Numerical Example
1
h(s) = .
s2 + 14 (4c b2 )
This is a simple sine function:
1
h(t) = sin (t)
q
b2
where = c 4
b
Recall the frequency-shift property: L1 u(s) = L1 h(s + 2b ) = e 2 t h(t).
We conclude that
r !
1 2b t b2
u(t) = q e sin 4c t
4c b2 4
4

1
Numerical Example: Let u(s) = s2 +2s+2 . Taking the roots, we find
1
u(s) = .
(s + 1)2 + 1
This is a sine function with a frequency shift of 1. Therefore
u(t) = et sin t
M. Peet Lecture 5: Control Systems 27 / 30
The Laplace Transform of a State-Space System
Transfer Functions

Recall the State-Space System

x = Ax(t) + Bu(t)
y(t) = Cx(t) + Du(t) x(0) = 0

Applying the Laplace Transform to the first signal:

sx(s) x(0) = Ax(s) + B u(s)

where we used
Linearity
Differentiation Property
Let x(0) = 0 and we have

(sI A)x(s) = B u(s) or x(s) = (sI A)1 B u(s)

M. Peet Lecture 5: Control Systems 28 / 30

The Laplace Transform of a State-Space System
Transfer Functions

Applying the Laplace Transform to the output equation,

y(t) = Cx(t) + Du(t)

we get
y(s) = C x(s) + Du(s)
Solving for y(s), we get

y(s) = C(sI A)1 B + D u(s)

Thus we have a Transfer Function for the system which can be used to
construct a solution for any input using the frequency-domain representation.

Theorem 4.
The Transfer Function for a state-space system is

G(s) = C(sI A)1 B + D

M. Peet Lecture 5: Control Systems 29 / 30

Summary

What have we learned today?

Laplace Transform of Simple Signals.

Step, Exponentials, Ramps.
Impulse

Properties of the Laplace Transform.

delay, scaling
convolution
etc.

Transfer Functions
State Space to Transfer Function

Next Lecture: More on Transfer Functions

M. Peet Lecture 5: Control Systems 30 / 30