3.

Modeling Physical Systems

Reference: Dorf, Richard C., and Robert H. Bishop. Modern control systems.
Pearson, 12th Edition (Chapter 2)
Outline
• System description
• Linear Time-Invariant (LTI) systems and
linearization
• Transfer functions
• Block diagrams and reductions

2
Outline
• System description
• Linear Time-Invariant (LTI) systems and
linearization
• Transfer functions
• Block diagrams and reductions

3
System: General Description
• A system is physical or artificial, that (takes an input and) produces some
output

Input Output
System

• Examples (e.g., process/plant, controller, the closed-loop system)

Disturbance Noise

Reference + +
+ Process / + Output
Controller
+- Plant

Feedback

4
Example: An Electrical System
Obtain the governing equation of current i(t) for any given input voltage e(t)

𝑢𝑅 = 𝑅 ∗ 𝑖;
𝑢𝑅 𝑢𝐿 𝑢𝐶 𝑑𝑖
𝑢𝐿 = 𝐿 ;
𝑑𝑡
𝑢𝐶 = 𝐶 න 𝑖𝑑𝑡

e = uR + uL + uC
di 1
e = R  i + L  +  idt
dt C
1
e − R  i − L  i − i = 0
C
g (i, i, i; e) = 0 5
Example: A Mechanical System
A propeller with moment of inertia 𝐽 is driven by a
DC motor. Assume the motor has input voltage
𝑢, armature resistance 𝑅𝑎 , and back electromotive
force (i.e., back e.m.f.) 𝐸𝑎 . It is known that 𝐸𝑎 =
1 1 2
𝜔, 𝜏 = 𝐼𝑎 , and 𝜏 𝑑 = 𝑐𝑞 𝜔 , where 𝐾𝑣 > 0, 𝑐𝑞 >
𝐾𝑣 𝐾
𝑣
0 are two constants. Obtain the dynamic equation
of the motor speed.

1 𝜇
𝐽𝜔ሶ + 𝑅 2𝜔 + 𝑐𝑞 𝜔2 = 𝐾 𝑢,
𝑎 𝐾𝑣 𝑣 𝑅𝑎

6
Example: A mass-spring-damper system
Obtain the dynamic equation of the mass position x(t) (relative to its free
length) for any given external force f(t)
f(t)
𝑚𝑥 = 𝑓total = 𝑓 − 𝑘𝑥 − 𝑐𝑥ሶ − 𝑚𝑔

Rearranged to form the dynamic equation:

𝑚𝑥 + 𝑐𝑥ሶ + 𝑘𝑥 − 𝑓 + 𝑚𝑔 = 0

Or represented as a function:

𝑔(𝑥 , 𝑥ሶ , 𝑥; 𝑓) = 0

7
System: Mathematical Description
• A system is physical or artificial, that (takes an input and) produces some
output

Input Output
System

• Mathematical notation
• Input: 𝑢 𝑡 , 𝑡 ≥ 0
• Output: 𝑦 𝑡 , 𝑡 ≥ 0

• Mathematical model

𝑔 𝑦 𝑡 , 𝑦′ 𝑡 , … , 𝑦 𝑛 𝑡 , 𝑢 𝑡 , 𝑢′ 𝑡 , … , 𝑢 (𝑚) 𝑡 = 0; 𝑚 ≤ 𝑛

8
Equilibrium Points
• Mathematically, an equilibrium point is a pair (𝑢𝑒𝑞 , 𝑦𝑒𝑞 ) satisfying

𝑔 𝑦𝑒𝑞 , 0, … , 0, 𝑢𝑒𝑞 , 0, … , 0 = 0

• (𝑢(𝑡) ≡ 𝑢𝑒𝑞 , 𝑦(𝑡) ≡ 𝑦𝑒𝑞 ) is a valid solution of

𝑔 𝑦 𝑡 , 𝑦 ′ 𝑡 , … , 𝑦 𝑛 𝑡 , 𝑢 𝑡 , 𝑢′ 𝑡 , … , 𝑢(𝑚) 𝑡 =0

• Given input 𝑢 0 ≡ 𝑢𝑒𝑞 , system initial condition 𝑦 0 = 𝑦𝑒𝑞 , 𝑦 𝑘 0 =

0, the only system solution is 𝑦(𝑡) ≡ 𝑦𝑒𝑞

• Intuitively, if the system starts from an equilibrium point, it will stay

at the equilibrium point for all future time.

9
Example
Solve the equilibrium points of the below system:
The solution to finding the equilibrium
′′ ′ points for the system:
𝐿𝑖 + 𝑅𝑖 + 𝐶𝑖 − 𝑒′ = 0
𝐿𝑖″ + 𝑅𝑖′ + 𝐶𝑖 − 𝑒′ = 0

Set , , , and similarly, , , at equilibrium.

Substitute these into the equation:

0 + 0 + 𝐶 ⋅ 𝑖𝑒𝑞 − 0 = 0

This implies:

𝑖𝑒𝑞 = 0

Since 𝑒𝑒𝑞 is not specified, we conclude:

(𝑒𝑒𝑞 ∈ ℝ, 𝑖𝑒𝑞 = 0)

10
Example
Solve the equilibrium points of the below system:
The equilibrium points for the system:
1 1
𝐽𝜔ሶ + 𝑘 2 𝑅 𝜔 + 𝑐𝑞 𝜔2 − 𝑘 𝑢 =0
𝑣 𝑎 𝑣 𝑅𝑎

At equilibrium, set , , , , .

Substitute into the equation:

Solve for :

11
Example
Solve the equilibrium points of the below system:

𝑚𝑥 + 𝑐𝑥ሶ + 𝑘𝑥 − 𝑓 + 𝑚𝑔 = 0

At equilibrium, set , , , , and .

Substitute these equilibrium values:

0 + 0 + 𝑘𝑥𝑒𝑞 − 𝑓𝑒𝑞 + 𝑚𝑔 = 0

Thus, the equilibrium equation is:

𝑘𝑥𝑒𝑞 − 𝑓𝑒𝑞 + 𝑚𝑔 = 0

12
Outline
• System description
• Linear Time-Invariant (LTI) systems and
linearization
• Transfer functions
• Block diagrams and reductions

13
LTI system
• General:

𝑔 𝑦 𝑡 , 𝑦′ 𝑡 , … , 𝑦 𝑛 𝑡 , 𝑢 𝑡 , 𝑢′ 𝑡 , … , 𝑢 (𝑚) 𝑡 = 0; 𝑚 ≤ 𝑛

• Linear time-invariant (LTI system): 𝑔(∙) is a linear function

𝑔 𝑦 𝑡 , 𝑦′ 𝑡 , … , 𝑦 𝑛
𝑡 , 𝑢 𝑡 , 𝑢′ 𝑡 , … , 𝑢 (𝑚) 𝑡 =
𝐴0 𝑦 𝑡 + 𝐴1 𝑦 ′ 𝑡 … + 𝐴𝑛 𝑦 𝑛 𝑡 + 𝐵0 𝑢 𝑡 + 𝐵1 𝑢′ 𝑡 + ⋯ + 𝐵𝑚 𝑢 𝑚 𝑡 + 𝐶 = 0;

𝑚 ≤ 𝑛 and 𝐴𝑖 , 𝐵𝑖 , 𝐶 are time-invariant real numbers

14
Example
Is the below system LTI?

𝐿𝑖 ′′ + 𝑅𝑖 ′ + 𝐶𝑖 − 𝑒′ = 0

Yes

15
Example
Is the below system LTI?
1 1
𝐽𝜔ሶ + 𝑘 2 𝑅 𝜔 + 𝑐𝑞 𝜔2 − 𝑘 𝑢 =0
𝑣 𝑎 𝑣 𝑅𝑎

No

16
Example
Is the below system LTI?
𝑚𝑥 + 𝑐𝑥ሶ + 𝑘𝑥 − 𝑓 + 𝑚𝑔 = 0

Yes

17
Linearization
• LTI system
𝐴0 𝑦 𝑡 + 𝐴1 𝑦 ′ 𝑡 … + 𝐴𝑛 𝑦 𝑛
𝑡 + 𝐵0 𝑢 𝑡 + 𝐵1 𝑢′ 𝑡 + ⋯ + 𝐵𝑚 𝑢 𝑚
𝑡 +𝐶 =0

Ordinary differential equation, easy to solve (e.g., by Laplace transform)

• General nonlinear system

𝑔 𝑦 𝑡 , 𝑦′ 𝑡 , … , 𝑦 𝑛 𝑡 , 𝑢 𝑡 , 𝑢′ 𝑡 , … , 𝑢 (𝑚) 𝑡 =0

Difficulty to solve in general!

Possible to transform a nonlinear system into a linear one?

18
Error Analysis – Definition
• Key observation: Practical systems typically operate near to a known
equilibrium point (𝑢𝑒𝑞 , 𝑦𝑒𝑞 )

• Solve 𝑦 𝑡 in terms of input 𝑢 𝑡 is equivalent to solving 𝑦 𝑡 − 𝑦𝑒𝑞 in terms

of input 𝑢 𝑡 − 𝑢𝑒𝑞

• Error output: 𝑦෤ 𝑡 = 𝑦 𝑡 − 𝑦𝑒𝑞 → 𝑦 = 𝑦෤ + 𝑦𝑒𝑞 , 𝑦 𝑘 = 𝑦෤ 𝑘 ,𝑘 ≥ 1

• Error input: 𝑢෤ 𝑡 = 𝑢 𝑡 − 𝑢𝑒𝑞 → 𝑢 = 𝑢෤ + 𝑢𝑒𝑞 , 𝑢 𝑘 = 𝑢෤ 𝑘 ,𝑘 ≥ 1

19
Error Analysis – Approximation
• Key observation: Practical systems typically operate near to a known
equilibrium point (𝑢𝑒𝑞 , 𝑦𝑒𝑞 )

𝑦෤ = 𝑦 − 𝑦𝑒𝑞 ≈ 0 𝑢෤ = 𝑢 − 𝑢𝑒𝑞 ≈ 0

𝑦෤ ′ = 𝑦 ′ ≈ 0 𝑢෤ ′ = 𝑢′ ≈ 0

⋮ ⋮

𝑦෤ (𝑛) = 𝑦 (𝑛) ≈ 0 𝑢෤ (𝑚) = 𝑢(𝑚) ≈ 0

• In sum, 𝑢෤ (𝑘) and 𝑦෤ (𝑘) are small for 𝑘 = 0,1,2, …

20
 Error Analysis – Approximation
• Taylor series expansion
𝑓 𝑥1 , … , 𝑥𝑘 = 𝑓 𝑥ҧ1 , … , 𝑥ҧ 𝑘 The constant term
𝜕𝑓 𝜕𝑓
+ ቤ 𝑥1 − 𝑥ҧ1 + ⋯ + ቤ 𝑥 − 𝑥ҧ 𝑘 First order terms
𝜕𝑥1 0 𝜕𝑥𝑘 0 𝑘

+𝑜 (𝑥1 − 𝑥ҧ1 )2 + ⋯ + 𝑜 (𝑥𝑘 − 𝑥ҧ 𝑘 )2 Second order terms

+⋯ Third order terms and more

Higher order terms (H.O.T.)

𝑔 𝑦, 𝑦 ′ , … , 𝑦 𝑛
, 𝑢, 𝑢′ , … , 𝑢(𝑚) = 𝑔 𝑦𝑒𝑞 , 0, … , 0, 𝑢𝑒𝑞 , 0, … , 0
𝜕𝑔 𝜕𝑔 𝜕𝑔 𝜕𝑔 𝜕𝑔 𝜕𝑔
+ ቤ 𝑦 − 𝑦𝑒𝑞 + ′ ቤ 𝑦 ′ + ⋯ + ቤ 𝑦 𝑛
+ ቤ 𝑢 − 𝑢𝑒𝑞 + ′ ቤ 𝑢′ + ⋯ + ቤ 𝑢 𝑚
𝜕𝑦 0 𝜕𝑦 0 𝜕𝑦 𝑛 0 𝜕𝑢 0 𝜕𝑢 0 𝜕𝑢 𝑚 0
2 𝑛 2 2 𝑚 2
+𝑜 (𝑦 − 𝑦𝑒𝑞 )2 + 𝑜 𝑦 ′ + ⋯+𝑜 𝑦 + 𝑜 (𝑢 − 𝑢𝑒𝑞 )2 + 𝑜 𝑢′ +⋯+ 𝑜 𝑢
+⋯ 21
Error Analysis – Approximation
• Taylor series expansion

𝑔 𝑦, 𝑦 ′ , … , 𝑦 𝑛
, 𝑢, 𝑢′ , … , 𝑢(𝑚) = 𝑔 𝑦𝑒𝑞 , 0, … , 0, 𝑢𝑒𝑞 , 0, … , 0
𝜕𝑔 𝜕𝑔 𝜕𝑔 𝜕𝑔 𝜕𝑔 𝜕𝑔
+ ቤ 𝑦 − 𝑦𝑒𝑞 + ′ ቤ 𝑦 ′ + ⋯ + ቤ 𝑦 𝑛
+ ቤ 𝑢 − 𝑢𝑒𝑞 + ′ ቤ 𝑢′ + ⋯ + ቤ 𝑢 𝑚
𝜕𝑦 0 𝜕𝑦 0 𝜕𝑦 𝑛 0 𝜕𝑢 0 𝜕𝑢 0 𝜕𝑢 𝑚 0
2 𝑛 2 2 𝑚 2
+𝑜 (𝑦 − 𝑦𝑒𝑞 )2 + 𝑜 𝑦 ′ + ⋯+𝑜 𝑦 + 𝑜 (𝑢 − 𝑢𝑒𝑞 )2 + 𝑜 𝑢′ +⋯+ 𝑜 𝑢
+⋯
𝜕𝑔 𝜕𝑔
𝐴𝑖 = ቤ ; 𝐵 = ቤ
𝜕𝑦 (𝑖) 0 𝑖 𝜕𝑢(𝑖) 0

𝑔 𝑦, 𝑦 ′ , … , 𝑦 𝑛
, 𝑢, 𝑢′ , … , 𝑢(𝑚) = 0 + 𝐴0 𝑦෤ + 𝐴1 𝑦෤ ′ + ⋯ + 𝐴𝑛 𝑦෤ 𝑛
+ 𝐵0 𝑢෤ + 𝐵1 𝑢෤ ′ + ⋯ + 𝐵𝑚 𝑢෤ 𝑚

2 𝑛 2 2 𝑚 2
+𝑜 𝑦෤ 2 + 𝑜 𝑦෤ ′ + ⋯ + 𝑜 𝑦෤ + 𝑜 𝑢෤ 2 + 𝑜 𝑢෤ ′ + ⋯ + 𝑜 𝑢෤
+⋯

22
Error Analysis – Approximation
• Truncation:

𝑔 𝑦, 𝑦 ′ , … , 𝑦 𝑛
, 𝑢, 𝑢′ , … , 𝑢(𝑚) = 0 + 𝐴0 𝑦෤ + 𝐴1 𝑦෤ ′ + ⋯ + 𝐴𝑛 𝑦෤ 𝑛
+ 𝐵0 𝑢෤ + 𝐵1 𝑢෤ ′ + ⋯ + 𝐵𝑚 𝑢෤ 𝑚

2 𝑛 2 2 𝑚 2
+𝑜 𝑦෤ 2 + 𝑜 𝑦෤ ′ + ⋯ + 𝑜 𝑦෤ + 𝑜 𝑢෤ 2 + 𝑜 𝑢෤ ′ + ⋯ + 𝑜 𝑢෤
+⋯
𝑢෤ (𝑘) and 𝑦෤ (𝑘) are small
2 2
𝑢෤ (𝑘) and 𝑦෤ (𝑘) are smaller
3 3
𝑢෤ (𝑘) and 𝑦෤ (𝑘) are even smaller
…

𝑔 𝑦, 𝑦 ′ , … , 𝑦 𝑛
, 𝑢, 𝑢′ , … , 𝑢(𝑚) ≈ 𝐴0 𝑦෤ + 𝐴1 𝑦෤ ′ + ⋯ + 𝐴𝑛 𝑦෤ 𝑛
+ 𝐵0 𝑢෤ + 𝐵1 𝑢෤ ′ + ⋯ + 𝐵𝑚 𝑢෤ 𝑚

23
Error Analysis – Approximation
• Hence:

0 = 𝑔 𝑦, 𝑦 ′ , … , 𝑦 𝑛
, 𝑢, 𝑢′ , … , 𝑢(𝑚) ≈ 𝐴0 𝑦෤ + 𝐴1 𝑦෤ ′ + ⋯ + 𝐴𝑛 𝑦෤ 𝑛
+ 𝐵0 𝑢෤ + 𝐵1 𝑢෤ ′ + ⋯ + 𝐵𝑚 𝑢෤ 𝑚

𝑎0 𝑦෤ + ⋯ + 𝑎𝑛−1 𝑦෤ (𝑛−1) + 𝑦෤ 𝑛
= 𝑏0 𝑢෤ + 𝑏1 𝑢෤ ′ + ⋯ + 𝑏𝑚 𝑢෤ 𝑚

𝐴𝑖 𝐵𝑖
𝑎𝑖 = ; 𝑏𝑖 = −
𝐴𝑛 𝐴𝑛

24
Linearization

𝑢(𝑡) 𝑦 𝑡
𝑔 𝑦 𝑡 , 𝑦′ 𝑡 ,…,𝑦 𝑛 𝑡 ,𝑢 𝑡 , 𝑢′ 𝑡 , … , 𝑢 (𝑚) 𝑡 =0

𝜕𝑔 𝜕𝑔
ฬ ฬ
𝜕𝑦 (𝑖) 0 𝜕𝑢 (𝑖) 0 Linearization at (𝑢𝑒𝑞 , 𝑦𝑒𝑞 )
𝑎𝑖 = , 𝑏𝑖 = − ;
𝜕𝑔 𝜕𝑔
ฬ ฬ
𝜕𝑦 (𝑛) 0 𝜕𝑦 (𝑛) 0

𝑢(𝑡) 𝑢(𝑡)
෤ 𝑎0 𝑦(𝑡)
෤ + ⋯ 𝑎𝑛−1 𝑦෤ (𝑛−1) + 𝑦෤ 𝑛 (𝑡) 𝑦(𝑡)
෤ 𝑦 𝑡
+ = 𝑏0 𝑢(𝑡)
෤ + 𝑏1 𝑢෤ ′ (𝑡) + ⋯ + 𝑏𝑚 𝑢෤ 𝑚 (𝑡) +
- +
𝑢𝑒𝑞 𝑦𝑒𝑞
▪ The linearization is valid only when y(t) and u(t) are
near to yeq and ueq, respectively.
▪ ~ will be dropped in the following discussion 25
Example
Linearize the following system around its respective equilibrium point
𝐿𝑖 ′′ + 𝑅𝑖 ′ + 𝐶𝑖 − 𝑒′ = 0

26
Example
Linearize the following system around its respective equilibrium point
1 1
𝐽𝜔ሶ + 𝑘 2 𝑅 𝜔 + 𝑐𝑞 𝜔2 − 𝑘 𝑢 =0
𝑣 𝑎 𝑣 𝑅𝑎

27
Example
Linearize the following system around its respective equilibrium point
𝑚𝑥 + 𝑐𝑥ሶ + 𝑘𝑥 − 𝑓 + 𝑚𝑔 = 0

28
Outline
• System description
• Linear Time-Invariant (LTI) systems and
linearization
• Transfer functions
• Block diagrams and reductions

29
Transfer function (!)
• First, consider a LTI system:
𝑦 (𝑛) + 𝑎𝑛−1 𝑦 (𝑛−1) + ⋯ + 𝑎1 𝑦ሶ + 𝑎0 𝑦
= 𝑏𝑚 𝑢(𝑚) + 𝑏𝑚−1 𝑢(𝑚−1) + ⋯ + 𝑏1 𝑢ሶ + 𝑏0 𝑢 (𝑛 ≥ 𝑚)
• Take Laplace transform to both side, differential equation-> algebraic eq.
• Assuming zero initial conditions:
𝑎0 𝑠 𝑛 + 𝑎1 𝑠 𝑛−1 + ⋯ + 𝑎𝑛−1 𝑠 + 𝑎𝑛 𝑌(𝑠)
= 𝑏0 𝑠 𝑚 + 𝑏1 𝑠 𝑚−1 + ⋯ + 𝑏𝑚−1 𝑠 + 𝑏𝑚 𝑈(𝑠)
• The Transfer Function of the system is then:

ℒ(Output)
𝐺 𝑠 ≜ ቤ
ℒ(Input) zero initial conditions

𝑌(𝑠) 𝑏𝑚 𝑠 𝑚 + 𝑏𝑚−1 𝑠 𝑚−1 + ⋯ + 𝑏1 𝑠 + 𝑏0

= =
𝑈(𝑠) 𝑠 𝑛 + 𝑎𝑛−1 𝑠 𝑛−1 + ⋯ + 𝑎1 𝑠 + 𝑎0

30
Transfer function
• Very important: the foundation of frequency-based control design
• It is a property of the system itself: does not change with the input
• With the system transfer function, its forced response can be calculated for
ANY input: 𝑦 𝑡 = ℒ −1 𝑌 𝑠 = ℒ −1 𝐺 𝑠 𝑈(𝑠)
• Can be obtained by
• modelled from its physical equations
• identified input/output data (system identification)

• Distinguish the three functions of 𝑠:

• 𝑈(𝑠): Laplace transform of the input 𝑢 𝑡 , which is a signal
• 𝑌(𝑠): Laplace transform of the output 𝑦 𝑡 , which is a signal
• 𝐺(𝑠): Representation of the system, which is a system!!

𝑌(𝑠) 𝑏𝑚 𝑠 𝑚 + 𝑏𝑚−1 𝑠 𝑚−1 + ⋯ + 𝑏1 𝑠 + 𝑏0

𝐺(𝑠) ≜ =
𝑈(𝑠) 𝑠 𝑛 + 𝑎𝑛−1 𝑠 𝑛−1 + ⋯ + 𝑎1 𝑠 + 𝑎0

31
Terminologies
• Given a system 𝐺(𝑠)

𝑏𝑚 𝑠 𝑚 + 𝑏𝑚−1 𝑠 𝑚−1 + ⋯ + 𝑏1 𝑠 + 𝑏0 𝑁(𝑠)

𝐺 𝑠 = =
𝑠 𝑛 + 𝑎𝑛−1 𝑠 𝑛−1 + ⋯ + 𝑎1 𝑠 + 𝑎0 𝐷(𝑠)

• Zeros: roots of 𝑁 𝑠 = 0 (i.e., if 𝑁 𝑠 = 𝑏𝑚 𝑠 − 𝑧1 ⋯ (𝑠 − 𝑧𝑚 ), then 𝑧1 , … , 𝑧𝑚

are the zeros)
• Poles: roots of 𝐷 𝑠 = 0 (i.e., if 𝐷 𝑠 = 𝑠 − 𝑝1 ⋯ (𝑠 − 𝑝𝑛 ), then 𝑝1 , … , 𝑝𝑛 are
the poles)
• Characteristic polynomial (CP): 𝐷 𝑠
• Characteristic equation (CE): 𝐷 𝑠 = 0
• Order: the polynomial order of 𝐷 𝑠 (i.e., 𝑛)

32
Impulse response
• For impulse input 𝑢 𝑡 = 𝛿(𝑡)

𝑈 𝑠 =1

𝑌 𝑠 = 𝐺 𝑠 𝑈 𝑠 = 𝐺(𝑠)

𝑦 𝑡 = ℒ −1 𝑌(𝑠) = ℒ −1 𝐺(𝑠) ≜ 𝑔(𝑡)

• 𝑔 𝑡 ≜ ℒ −1 𝐺(𝑠) is called the impulse-response

• Widely used in system identification
• In practice, a pulse input with a very short duration -> impulse

33
Response for Any Input
• For an input 𝑢 𝑡 with 𝑈 𝑠 = ℒ{𝑢 𝑡 },

𝑌 𝑠 = 𝐺 𝑠 𝑈(𝑠)

⇒ 𝑦 𝑡 = ℒ −1 𝑌(𝑠) = ℒ −1 𝐺 𝑠 𝑈(𝑠) = 𝑔 𝑡 ∗ 𝑢(𝑡)

The forced response is the convolution between the impulse

response 𝑔 𝑡 and input 𝑢 𝑡

34
Representation of A LTI System
• Differential Equation

𝑢(𝑡) 𝑦 (𝑛) + 𝑎𝑛−1 𝑦 (𝑛−1) + ⋯ + 𝑎0 𝑦 𝑦 𝑡

= 𝑏𝑚 𝑢(𝑚) + 𝑏𝑚−1 𝑢(𝑚−1) + ⋯ + 𝑏0 𝑢

• Impulse response

𝑢(𝑡) 𝑦 𝑡 = 𝑔 𝑡 ∗ 𝑢(𝑡)
𝑔 𝑡

• Transfer function

𝑈(𝑠) 𝑌 𝑠 = 𝐺 𝑠 𝑈(𝑠)
𝐺 𝑠

35
Example
Obtain the transfer function of the following system
𝐿𝑖 ′′ + 𝑅𝑖 ′ + 𝐶𝑖 − 𝑒′ = 0

36
Example
Obtain the transfer function of the following system
1 1
𝐽𝜔ሶ + 𝑘 2 𝑅 𝜔 + 𝑐𝑞 𝜔2 − 𝑘 𝑢 =0
𝑣 𝑎 𝑣 𝑅𝑎

37
Example
Obtain the transfer function of the following system
𝑚𝑥 + 𝑐𝑥ሶ + 𝑘𝑥 − 𝑓 + 𝑚𝑔 = 0

38
Example (2015-2016 Final Exam, Q5)
Obtain the transfer function of the following system from the input displacement x to
the output displacement y. The reference point of x, y are made such that when x=0
and y=0, the two springs k1 and k2 are at their free length.

39
Use MATLAB
Define a transfer function in MATLAB: “tf”

40
System description
• Examples
• Linear Time-Invariant (LTI) systems and
linearization
• Transfer functions
• Block diagrams and reductions

41
Block Diagrams
• A pictorial representation of a system
• Intuitively illustrate system component relations and signal flows
• Easing the simplification and analysis of larger network consisting of many
systems

𝑈(𝑠) 𝑌 𝑠
𝐺 𝑠

𝑌 𝑠 = 𝐺 𝑠 𝑈(𝑠)

𝑠 can be dropped for notation simplicity

42
Elements (!)
• Block 𝐺 𝑠
Signals are always unidirectional.
• Signal
𝑈(𝑠) Double-arrow lines should ever appear!

𝑈(𝑠)
• Input 𝐺 𝑠

𝑌(𝑠)
• Output 𝐺 𝑠

𝑈1 (𝑠) 𝑈(𝑠) 𝑈1 (𝑠) 𝑈(𝑠)

• Summing point ++/- 𝑈 𝑠 = 𝑈1 𝑠 ± 𝑈2 (𝑠)
+
(Comparator) +/-
𝑈2 (𝑠) 𝑈2 (𝑠)
• Branch (pickoff) point
𝑌(𝑠) 𝑌(𝑠)

𝑌(𝑠) 𝑌(𝑠)
43
Block Diagram Transformations
1. Combining blocks in cascade

𝑋1 𝑋2 𝑋3 𝑋1 𝑋3
𝐺1 𝐺2 𝐺2 𝐺1

44
Block Diagram Transformations
2. Moving a summing point ahead of a block

𝑋1 𝑋3 𝑋1 𝑋3
𝐺 + +
𝐺
± ±
1 𝑋2
𝑋2
𝐺

45
Block Diagram Transformations
3. Moving a summing point behind a block

𝑋1 𝑋3 𝑋1 𝑋3
+
𝐺 𝐺 +
± ±
𝑋2 𝑋2
𝐺

46
Block Diagram Transformations
4. Moving a pickoff point ahead of a block

𝑋1 𝑋2 𝑋1 𝑋2
𝐺 𝐺
𝑋2
𝑋2 𝐺

47
Block Diagram Transformations
5. Moving a pickoff point behind a block

𝑋1 𝑋2
𝑋1 𝑋2 𝐺
𝐺
𝑋1 𝑋1 1
𝐺

48
Block Diagram Transformations
6. Eliminating a negative feedback loop

𝑋1 𝑋2 𝑋1 𝐺 𝑋2
+ 𝐺 1 + 𝐺𝐻
−

49
 Block Diagram Transformations
7. Eliminating a negative feedback loop

𝑋1 𝑋2 𝑋1 𝐺 𝑋2
+ 𝐺 1 + 𝐺𝐻
−

In general, the transfer function of a negative feedback network

Multiplication of all transfer function on the 𝐟𝐨𝐫𝐰𝐚𝐫𝐝 path

1 + Multiplication of all transfer functions on the feedback 𝐥𝐨𝐨𝐩
50
 Block Diagram Transformations
8. Eliminating a positive feedback loop

𝑋1 𝑋2 𝑋1 𝐺 𝑋2
+ 𝐺 1 − 𝐺𝐻
+

In general, the transfer function of a positive feedback network

Multiplication of all transfer function on the 𝐟𝐨𝐫𝐰𝐚𝐫𝐝 path

1 − Multiplication of all transfer functions on the feedback 𝐥𝐨𝐨𝐩
51
 Block Diagram Transformations
9. Eliminating a feedback loop

𝑋1 𝑋2 𝑋1 𝐺 𝑋2
+ 𝐺 1 ∓ 𝐺𝐻
±

In general, the transfer function of a positive/negative feedback network

Multiplication of all transfer function on the 𝐟𝐨𝐫𝐰𝐚𝐫𝐝 path

1 ∓ Multiplication of all transfer functions on the feedback 𝐥𝐨𝐨𝐩
52
Block Diagram Transformations
• Summary

53
Block Diagrams Reduction
• A block diagram may contain multiple loops and blocks
• Reduce one loop at once, using the elementary block diagram
transformations
• Start with the innermost loop

54
Example
Obtain the transfer function of the following network
𝑋1

𝑋2 ++ 𝑌

+
𝐺 𝑃
−

55
 Example
Obtain the transfer function of the following network
𝑋1

𝑋2 ++ 𝑌
𝐺 𝑃
+
−
𝐻

56
Example
Obtain the transfer function of the following network

57
Example
Below shows the structure of a feedback control system with a disturbance observer (DoB). Determine
the transfer function from 𝑅 𝑠 to the tracking error 𝐸 𝑠 . DoB is widely used in high-precision
machines to attenuate the effect of disturbance. The idea of DoB design is to first estimate the total input
𝑉෠ 𝑠 from the system output 𝑌(𝑠) by inverting the system plant model 𝐺𝑛 (𝑠). The estimated total input
𝑉෠ 𝑠 , after removing the known control input 𝑈(𝑠), would give an estimate of the disturbance that is
contaminated by noise in 𝑌(𝑠). To eliminate this noise, a low pass filter denoted by 𝑄 𝑠 is hence used to
obtain the filtered disturbance estimate 𝐷 ෡ 𝑠 . This estimated disturbance 𝐷 ෡ 𝑠 is then added to the
ഥ 𝑠 , hoping that 𝐷
original control input 𝑈 ෡ 𝑠 would cancel 𝐷 𝑠 and that the control input 𝑈 ഥ 𝑠 (i.e.,
controller 𝐶 𝑠 ) can be designed as the system has no disturbance 𝐷 𝑠 at all.

58