Automatic Control – Chapter 4

TRANSFER FUNCTION
Tran Anh Tu – Department of Aeronautics
tran-anh.tu@usth.edu.vn
 CONTENTS

01 02 03

Introduction Transfer function Practice with

Matlab / Simulink

2
 CONTENTS

01 02 03

Introduction Transfer function Practice with

Matlab / Simulink

3
 GETTING STARTED

Previously, you have learnt to model a system using differential equations

𝑀𝑥ሷ + 𝑏𝑥ሶ + 𝑘𝑥 − 𝐹 = 0 𝑀1 𝑥ሷ 1 + 𝜇𝑀1 𝑔𝑥ሶ 1 + 𝑘(𝑥1 − 𝑥2 ) − 𝐹 = 0

𝑀2 𝑥ሷ 2 + 𝜇𝑀2 𝑔𝑥ሶ 2 − 𝑘(𝑥1 − 𝑥2 ) = 0

4
 BUT…

How to solve them?

System of second-order DEs Fourth-order DE

𝑀1 𝑥ሷ 1 + 𝜇𝑀1 𝑔𝑥ሶ 1 + 𝑘 𝑥1 − 𝑥2 − 𝐹 = 0 𝑑 4 𝑤(𝑥)
𝐸𝐼 = 𝑞(𝑥)
𝑑𝑥 4
𝑀2 𝑥ሷ 2 + 𝜇𝑀2 𝑔𝑥ሶ 2 − 𝑘(𝑥1 − 𝑥2 ) = 0
(Euler-Bernoulli beam equation)

5
 BUT…

High-ordered differential equations are very difficult

to solve, analyze or manipulate

6
 HOW TO OVERCOME THIS?

• To avoid complex differential equation, we need other "simpler"

ways to represent the mathematical model

• Two most popular other representations:

o Transfer function

o State space representation

7
 MATHEMATICAL MODEL REPRESENTATION

Two most popular representations

Transfer function State-space representation

• Represent model in s-domain • Represent model in time domain
• Classical control theory • Modern control theory

8
 MATHEMATICAL MODEL REPRESENTATION

Two most popular representations

Transfer function State-space representation

• Represent model in s-domain • Represent model in time domain
• Classical control theory • Modern control theory

9
 CONTENTS

01 02 03

Introduction Transfer function Practice with

Matlab / Simulink

10
 ROADMAP TO TRANSFER FUNCTION

FREQUENCY DOMAIN &

FOURIER TRANSFORM TRANSFER FUNCTION

LINEAR TIME-INVARIANT S-DOMAIN & LAPLACE

SYSTEM TRANSFORM

11
 ROADMAP TO TRANSFER FUNCTION

FREQUENCY DOMAIN &

FOURIER TRANSFORM TRANSFER FUNCTION

LINEAR TIME-INVARIANT S-DOMAIN & LAPLACE

SYSTEM TRANSFORM

12
 LTI SYSTEM – INTRODUCTION

You have a juice making machine which produces juice from fruit

1 orange
100 ml orange juice

(Note: For this example, we assume that the machine runs smoothly all the time)

13
 LTI SYSTEM – INTRODUCTION

If you put in 3 oranges, you get 300 ml of orange juice

3 oranges 300 ml orange juice

➢ The machine is said to have homogeneity property

14
 HOMOGENEITY

If you scale the input by a, the output will also be scaled by a

h(ax) = a*h(x)
15
 LTI SYSTEM – INTRODUCTION

If you put in orange and watermelon, you get the mixture of both fruits!

➢ The machine is said to have superposition property

16
 SUPERPOSITION

If you sum multiple separate inputs together, the response will be

the summed outputs of each individual inputs

ℎ 𝑥1 + ℎ 𝑥2 = ℎ(𝑥1 + 𝑥2 )
17
 LINEARITY

A system having two properties homogeneity and superposition

is defined as linear system

18
 LTI SYSTEM – INTRODUCTION

Regardless of time, the machine always behave the same

➢ The machine is said to have time-invariant property

19
 TIME-INVARIANT

If we shift the input by a fixed time T, the output is also

shifted by that fixed time T

y 𝑡 − 𝑇 = ℎ(𝑥(𝑡 − 𝑇))
20
 LTI SYSTEM

A system which has linearity and time-invariant property

21
 LTI SYSTEM

However, no real physical system in the world is LTI system

So why bother?
22
"Linear systems are important,
because we can solve them"
Richard Feynman, legendary physicist
 LTI SYSTEM

• A wide range of real models can be approximated very accurately

using LTI model
• Linearization is popular method in handling non-linear model

24
 CHECKPOINT

Which of the following systems are LTI?

Input Output
1 x mx
2 x x2
3 x mx + b
4 θ mglsinθ
5 θ≈0 mglsinθ
6 f(t) df(t) / dt

25
 ROADMAP TO TRANSFER FUNCTION

FREQUENCY DOMAIN &

FOURIER TRANSFORM TRANSFER FUNCTION

LINEAR TIME-INVARIANT S-DOMAIN & LAPLACE

SYSTEM TRANSFORM

26
 TIME DOMAIN

Analysis of function with respect to time

Function of distance as the man walks

27
 TIME DOMAIN

Analysis of function with respect to time

Function of basketball height during a shoot

28
 TIME DOMAIN

• Time domain is convenient, because we are familiar with it

• But this is not always the case…

Hitting a mass on a spring with a hammer, assume no energy loss

29
 TIME DOMAIN

It's not possible to fully graph this function in time domain

30
 TIME DOMAIN

This signal is very difficult to analyze in time domain

31
 ALTERNATIVE SOLUTION

To deal with complex sinusoidal signal, we should move out of time domain

32
ALTERNATIVE SOLUTION

This system can be fully characterized

by 3 parameters
• Frequency of the bouncing
• Amplitude of the bouncing
• Phase shift corresponding to the
starting position of the mass

33
ALTERNATIVE SOLUTION

𝑥 = 𝐴𝑐𝑜𝑠(𝜔𝑡 + 𝜙)

• 𝜔= 𝑘 Τ𝑀
• 𝜙 = 𝜋/2
• 𝐴 = 1/ 𝑘𝑀

34
 FREQUENCY DOMAIN

If we plot the amplitude and phase with respect to frequency,

the graph looks much simpler

35
 FREQUENCY DOMAIN

𝜋 3𝜋
𝑓 𝑡 = 5𝑐𝑜𝑠 2𝑡 + + 2𝑐𝑜𝑠 4.5𝑡 − + 4𝑐𝑜𝑠(3𝑡 + 𝜋)
3 2

36
 FREQUENCY DOMAIN

In many case, the input signal is not sinusoidal

Sawtooth function

37
 FREQUENCY DOMAIN

• It could still be represented in

frequency domain by converting
the function into infinite number
of sinusoidals, each at different
frequency, amplitude and phase

38
FREQUENCY DOMAIN

(https://youtu.be/r18Gi8lSkfM)
39
 FOURIER TRANSFORM

• The transformation from time

domain to frequency domain can be
done using Fourier Transform
∞
𝐹 𝜔 = න 𝑓(𝑡)𝑒 −𝑗𝜔𝑡 𝑑𝑡
−∞
Visual explanation of formula:
• 3Blue1Brown
Josheph Fourier (1768 – 1830) • Eugene Khutoryansky

40
FOURIER TRANSFORM

• We can go back from frequency

domain to time domain using
Inverse Fourier Transform

1 ∞
𝑓 𝑡 = න 𝐹(𝜔)𝑒 𝑗𝜔𝑡 𝑑𝜔
2𝜋 −∞

41
 FOURIER TRANSFORM

Note: The frequency domain plot may looks more complicated,

but the analyse and manipulate process are simpler

42
 ROADMAP TO TRANSFER FUNCTION

FREQUENCY DOMAIN &

FOURIER TRANSFORM TRANSFER FUNCTION

LINEAR TIME-INVARIANT S-DOMAIN & LAPLACE

SYSTEM TRANSFORM

43
 LAPLACE'S IDEA

• Laplace takes the idea of Fourier further:

Decompose time domain signal into both
sinusoidal and exponential function!
• Define s = σ + jω
• Consider est

Pierre-Simon Laplace (1749 – 1827)

44
 LAPLACE'S IDEA

𝒆𝒔𝒕 = 𝒆𝝈𝒕 𝒆𝒋𝝎𝒕

• 𝑒 𝜎𝑡 is exponential function that change with σ
• 𝑒 𝑗𝜔𝑡 = cos 𝜔𝑡 + 𝑗𝑠𝑖𝑛 𝜔𝑡 → sinusoidal function

45
 WHY IS LAPLACE'S IDEA IMPORTANT?

• A damped harmonic oscillation signal is a combination of sinusoidal

and decay function

SPRING & DAMPER

46
 WHY IS LAPLACE'S IDEA IMPORTANT?

➢ By having one mathematical variable that captures both behaviours,

the analysis process would be simplified a lot

SPRING & DAMPER

47
S-DOMAIN AND S-PLANE

48
 LAPLACE TRANSFORM

• The transformation from time

domain to s-domain
∞
𝐹 𝑠 = න 𝑓(𝑡)𝑒 −𝑠𝑡 𝑑𝑡
0

Visual explanation of formula:

• Zach Star
• Eugene Khutoryansky

49
 LAPLACE TRANSFORM

∞
𝐹 𝑠 = න 𝑓(𝑡)𝑒 −𝑠𝑡 𝑑𝑡
0

• Laplace Transform exists only if the

integral converges
• Region in s-plane in which Laplace
Transform exists is called the Region of
Convergence (ROC)

50
 LAPLACE TRANSFORM

• Example: 𝑓(𝑡) = 𝑒 −𝑎𝑡

∞
𝐹 𝑠 = න 𝑒 −𝑎𝑡 𝑒 −𝑠𝑡 𝑑𝑡
0

𝑒 − 𝑠+𝑎 𝑡 ∞ 1
𝐹 𝑠 = ቤ =
− 𝑠+𝑎 0 𝑠+𝑎

(Region of converge: σ + a > 0)

51
 LAPLACE TRANSFORM TABLE

No. f(t) F(s) No. f(t) F(s)

𝑎
1 δ(t) 1 7 sin 𝑎𝑡
𝑠 2 + 𝑎2
1 𝑠
2 1 8 cos 𝑎𝑡
𝑠 𝑠 2 + 𝑎2
1 2𝑎𝑠
3 𝑡 9 𝑡 sin 𝑎𝑡
𝑠2 𝑠 2 + 𝑎2 2
𝑛! 𝑠 2 − 𝑎2
4 𝑡𝑛 10 𝑡 cos 𝑎𝑡
𝑠 𝑛+1 𝑠 2 + 𝑎2 2
1 𝑏
5 𝑒 𝑎𝑡 11 𝑒 𝑎𝑡 sin 𝑏𝑡
𝑠−𝑎 (𝑠 − 𝑎)2 + 𝑏2
𝑛! 𝑠−𝑎
6 𝑡 𝑛 𝑒 𝑎𝑡 12 𝑒 𝑎𝑡 cos 𝑏𝑡
(𝑠 − 𝑎)𝑛+1 (𝑠 − 𝑎)2 + 𝑏2

52
LAPLACE TRANSFORM THEOREM

53
 LAPLACE TRANSFORM THEOREM

The differentiation and integration theorem are powerful,

since they help to convert differential equation into algebraic equation

𝑑2 𝑦 𝑑𝑦
2
+ 12 + 32𝑦 = 32
𝑑𝑡 𝑑𝑡

2
32
𝑠 𝑌(𝑠) + 12𝑠𝑌(𝑠) + 32𝑌(𝑠) =
𝑠

54
 INVERSE LAPLACE TRANSFORM

• The transformation from s-domain

to time domain

1 𝜎+𝑗∞ 𝑠𝑡
𝑓 𝑡 = න 𝑒 𝐹(𝑠)𝑑𝑠
2𝜋𝑗 𝜎−𝑗∞

• Use table in practice!

55
 EXAMPLE

1. Find the Laplace Transform of 𝑓 𝑡 = 𝑒 −3𝑡 sin(2𝑡)

2. Find the Inverse Laplace Transform of

4𝑠
F 𝑠 = 4
𝑠 + 2𝑠 2 + 1

56
 EXERCISES

1. Find the Laplace transform of 𝑓 𝑡 = 𝑡 2 𝑒 −3𝑡

2. Find the Laplace transform of 𝑓 𝑡 = 4 sin 2𝜋𝑡 − 2cos(𝜋𝑡)

3. Find the inverse Laplace transform of 1/(s+5)2

4. Find the inverse Laplace transform of 3/(s2 + 6s + 12)

57
 ILT OF COMPLICATED SYSTEM

• To find the inverse Laplace 𝑠 3 + 2𝑠 2 + 6𝑠 + 7

𝐹 𝑠 =
𝑠2 + 𝑠 + 5
transform of complicated
function, we convert it to a
2
sum of simpler terms 𝐹 𝑠 =𝑠+1+ 2
𝑠 +𝑠+5
• The result is called Partial-
Fraction Expansion 𝑓1 𝑡 =
𝑑δ(𝑡)
+ δ(𝑡) + ℒ −1
2
𝑑𝑡 𝑠2 + 𝑠 + 5

58
 PARTIAL FRACTION EXPANSION

Case 1: Roots of Denominator F(s) are real and distinct

2
𝐹 𝑠 =
(𝑠 + 1)(𝑠 + 2)

𝐾1 𝐾2
𝐹 𝑠 = +
𝑠+1 𝑠+2

𝑓 𝑡 = 𝐾1 𝑒 −𝑡 + 𝐾2 𝑒 −2𝑡

59
 PARTIAL FRACTION EXPANSION

Case 2: Roots of Denominator F(s) are real and repeated

2
𝐹 𝑠 = 2
(𝑠 + 1) 𝑠 + 2

𝐾1 𝐾2 𝐾3
𝐹 𝑠 = + + 2
𝑠+1 𝑠+2 𝑠+2

𝑓 𝑡 = 𝐾1 𝑒 −𝑡 + 𝐾2 𝑒 −2𝑡 + 𝐾3 𝑡𝑒 −2𝑡

60
 PARTIAL FRACTION EXPANSION

Case 3: Roots of Denominator F(s) are complex and imaginary

3
𝐹 𝑠 =
𝑠(𝑠 2 + 2𝑠 + 5)

𝐾1 𝐾2 𝑠 + 𝐾3
𝐹 𝑠 = + 2
𝑠 𝑠 + 2𝑠 + 5

𝐾3 − 𝐾2
𝑓 𝑡 = 𝐾1 + 𝐾2 𝑒 −𝑡 cos 2𝑡 + sin(2𝑡)
2𝐾2

61
 EXERCISES

Using Partial Fraction Expansion, find the Inverse Laplace Transform of

following functions:
12
1. 𝐹 𝑠 = 3
𝑠 + 5𝑠 2 + 6𝑠
8
2. 𝐹 𝑠 = 3
𝑠 − 4𝑠 2 + 4𝑠
4
3. 𝐹 𝑠 = 3
𝑠 + 2𝑠 2 + 2𝑠

62
 ROADMAP TO TRANSFER FUNCTION

FREQUENCY DOMAIN &

FOURIER TRANSFORM TRANSFER FUNCTION

LINEAR TIME-INVARIANT S-DOMAIN & LAPLACE

SYSTEM TRANSFORM

63
 COMMON DEFINITION

For an LTI system, with single input – single output,

initial condition being zero:

Laplace Transform of output

Transfer function =
Laplace Transform of input

Y(s)
X(s) System Y(s) TF =
X(s)
TF

64
 EXAMPLE

Let's go back to the car suspension system example

Output (response): position (x(t))

Input: Force

𝑀𝑥(𝑡)
ሷ + 𝑏𝑥(𝑡)
ሶ + 𝑘𝑥(𝑡) − 𝐹 = 0
65
 EXAMPLE

Determine the Laplace transform of the system model

f(t) F(s)

F F(s)
𝑀𝑥(𝑡)
ሷ + 𝑏𝑥(𝑡)
ሶ + 𝑘𝑥(𝑡) − 𝐹 = 0
x(t) X(s)

𝑥(𝑡)
ሶ sX(s) – x(0) 𝑀𝑠 2 𝑋(𝑠) + 𝑏𝑠𝑋(𝑠) + 𝑘𝑋(𝑠) − 𝐹(𝑠) = 0

𝑥(𝑡)
ሷ s2X(s) – sx(0) – x(0) (initial condition = 0)

66
 EXAMPLE

Find the transfer function of model

𝑀𝑠 2 𝑋(𝑠) + 𝑏𝑠𝑋(𝑠) + 𝑘𝑋(𝑠) − 𝐹(𝑠) = 0

𝑀𝑠 2 + 𝑏𝑠 + 𝑘 𝑋(𝑠) = 𝐹(𝑠)

𝑋(𝑠) 1
𝐺(𝑠) = =
𝐹(𝑠) 𝑀𝑠 2 + 𝑏𝑠 + 𝑘

This function well captures the

dynamic behaviour of the system

67
 GENERAL FORM OF TRANSFER FUNCTION

• First form (polynomial):

𝑎0 𝑠 𝑚 + 𝑎1 𝑠 𝑚−1 + ⋯ + 𝑎𝑚−1 𝑠 + 𝑎𝑚
𝐺(𝑠) =
𝑏0 𝑠 𝑛 + 𝑏1 𝑠 𝑛−1 + ⋯ + 𝑏𝑛−1 𝑠 + 𝑏𝑛

• Second form (zero – pole – gain):

𝑠 − 𝑧1 𝑠 − 𝑧2 … (𝑠 − 𝑧𝑚 )
𝐺(𝑠) = 𝐾
𝑠 − 𝑝1 𝑠 − 𝑝2 … (𝑠 − 𝑝𝑛 )
where
• m, n: order of polynomials
• z: zeros of the system
• p: poles of the system

68
 POLES AND ZEROS

𝑠(𝑠 + 3)
Give the transfer function 𝐺 𝑠 = 2
𝑠 + 2𝑠 + 5

• Zeros: Values that make numerator equals 0

o s = 0, s = -3
• Poles: Values that make denominator equals 0
o s = -1 + 2j, s = -1 – 2j

3D pole-zero plot

69
 POLES AND ZEROS

𝑠(𝑠 + 3)
Give the transfer function 𝐺 𝑠 = 2
𝑠 + 2𝑠 + 5

• Zeros: Values that make numerator equals 0

o s = 0, s = -3
• Poles: Values that make denominator equals 0
o s = -1 + 2j, s = -1 – 2j

2D pole-zero plot

70
 CHECKPOINT

Determine the zeros and poles of the following transfer functions

𝑠−1
1. 𝐹 𝑠 =
𝑠+3
𝑠2 + 1
2. 𝐹 𝑠 = 2
𝑠 − 4𝑠 + 4
10
3. 𝐹 𝑠 =
𝑠(𝑠 + 2)

71
 CONTENTS

01 02 03

Introduction Transfer function Practice with

Matlab / Simulink

72
 MATLAB PRACTICE

Use Matlab to determine:

1. Laplace Transform

2. Partial Fraction Expansion

3. Inverse Laplace Transform

4. Transfer Function

73
 SIMULINK PRACTICE

Modeling the transfer function of car suspension system in Simulink

1
𝐻(𝑠) =
𝑀𝑠 2 + 𝑏𝑠 + 𝑘

74
 SIMULINK PRACTICE

Follow Tutorial 4.1 to build the model in Simulink

75