ME2142 Feedback Control System

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

August 11, 2023

1 / 20
Preliminaries
Lecturer: Asst. Prof. Velusamy
Lecturer: A/Prof. C.J. Ong
Subramanian
Rm: EA-05-27
Rm: EA-05-24
Tel: 6516-2217
Tel: 6516-2272
Email: [email protected]
Email: [email protected]
References:
1 Kuo, BC, “Automatic Control Systems” 7th Edition, Prentice Hall
2 Dorf, RC, and Bishop, RH, “Modern Control Systems,” Prentice Hall Inc.,
2008
3 Shinners, MS, “Modern Control System Theory and Application”,
Addison-Wesley
4 Ogata, K, “Modern Control Engineering”, Prentice Hall
5 Raven, E, “Automatic Control Engineering”, McGraw Hill.
6 Norman S Nise, “Control Systems Engineering” 5th Edition, John Wiley &
Sons

2 / 20
Preliminaries
Outline of course:
1 General Introduction to Automatic Control:
2 Review of Laplace Transformation:
- Laplace Transformation and its inverse. Solution of Di!erential Equations.
3 Mathematical Model of Physical Systems:
- Transfer functions. Block diagrams. Modeling of mechanical systems,
electrical systems, motors.
4 Transient Response Analysis:
- Standard time response test functions. Time responses of first-order,
second-order and higher-order systems.
5 System Stability and Steady State Characteristics:
- Routh’s stability criterion. System Types. Steady state error analysis
6 Controller Design:
-Using time responses for the design of P, PI and PID controller
7 Frequency Response Analysis:
- Forced sinusoidal response. Graphical frequency response methods Bode and
Nyquist plots. Nyquist stability criterion. Gain and phase margins.
Closed-loop frequency response.
3 / 20
Preliminaries

Assessment:
1 Final Examination (70%)
2 Lab Report (30%)
↭ Lab 1: DC Speed and Position Control
↭ Lab 2: Frequency Response

4 / 20
Module Learning Outcome

The student is expected to be able to

obtain dynamic models of simple physical systems in the form of transfer
functions.
understand and characterize the transient response of systems.
determine the stability of a system using Routh’s stability criterion and Root
locus method.
determine the steady state errors of systems.
understand the frequency response of physical systems.
obtain the frequency response in the form of Bode diagram and Nyquist plots.
determine the stability of physical systems using Nyquist stability criterion.
determine the relative stability measures such as gain and phase margins.

5 / 20
Introduction

What is a Control System?

A Control System is a system in which some physical quantities are
controlled by regulating certain energy inputs.
A system is a group of physical components assembled to perform a specific
function. The components may be electrical, mechanical, thermal, biomedical,
chemical, pneumatic or any combination of the above.
A system need not be limited to an ensemble of physical components. It can
refer to abstract, dynamic phenomena such as economics (COE) or ecology
systems.
A physical quantity may be temperature, pressure, electrical voltage,
mechanical position, velocity, pH value, liquid level etc.
There are many applications of Control System including simple devices like
Aircon, microwave, tra”c light control to sophisticated systems like Space
vehicle, missile guidance, autopilot, robotic systems, Power system, assembly
line, Chemical processes, Quality Control etc.

6 / 20
Introduction
A Simple Control System
One specific feature of Control system is the concept of feedback.
The output quantity to be controlled is measured and compared with the
desired for corrective action.

Each block represents a functional element of the control system but it may
not have a corresponding physical element.
7 / 20
Introduction

Antenna System
A more practical example is the position control of an Antenna system.
To increase the number of channels, the dish is designed to aim at di!erent
satellites - requires ability to control the motors to any desired position.

A large number of such systems can be similarly modeled and analyzed.

8 / 20
Introduction
Position Control System
There are other types of position control systems. Consider the simplified
nuclear power plant.
Intensity of the reaction inside the reactor is controlled by the vertical position
of the control rods.
The deeper the control rods are submerged, the more heat the reactor
generates.

Other control systems exist in the nuclear power plant. Maintenance of the
boiler’s water level and/or regulation of the generated voltage at a fixed
voltage are all control systems.
9 / 20
Introduction
Besides position control system, there are also velocity control systems.
Earliest example is the flyball-governor speed control of steam engine.
Driving tapes in video or audio recorders at a constant speed is important in
ensuring the quality of the output. The problem is complicated because of the
load of the driver varies from a full to an empty reel.
The speed of conveyor in a production line must be precisely controlled;
The speed of a CD-ROM drive or a harddisk drive has to be precisely
controlled for the proper retrieval and writing of data;
An error in the roller speed of just 0.1% in a paper production drive system
may cause the paper to tear of pile up on the roller;
The velocity of the rotor of power plant generator is to be kept at 50 or 60 Hz

10 / 20
Introduction
Clothes Dryer Control System
There are two types of clothes dryers – manual and automatic.
In the manual dryer, the dryer is set on for some pre-specified duration, say 40
minutes.
At the end of the 40 minutes, this dryer will automatically turn o! even if the
clothes are still damp or wet.
In the automatic dryer, a desired degree of dryness is set and the dryer will
automatically turn o! when the clothes reach the desired level of dryness.
The amount of time needed is automatically determined by this type of dryer.

11 / 20
Some examples of Control System

Figure: Robotics
Figure: Harddisk drive

Figure: Automobile

12 / 20
Simple Classification of Control Systems
Open-loop and Closed-loop control system

Output has no e!ect on control action. Examples: Tra”c light, microwave

oven, dryers, toasters. Requires careful calibration for good performance.

Actuating signal and hence output, depends on the input and the current
output. Examples: Servo Controlled motor, missile, robots etc.
13 / 20
Control Systems

What are the e!ects of Feedback ?

Advantages are somewhat intuitive.

How to quantify them?

14 / 20
Control System

For simplicity, consider a static system (ultimate objective is for dynamical systems)

Assume that G, H are static values,

then

e(t) = r(t) → b(t) = r(t) → Hc(t)

c(t) = Ge(t) = Gr(t) → GHc(t)
c(t) G
↑ =
r(t) 1 + GH

Compare this to the openloop

system where
c(t)
=G
r(t)
c(t)
Let M := r(t) , the relation between
c(t) and r(t).

15 / 20
Control Systems

E!ect on Stability
c(t)
If GH = →1 then r(t)
= ↓.
Closed loop system can become unstable if choices are G,H are not chosen
carefully.
For the simple case above, unstable condition of GH = →1 can be avoided
using an appropriate value of H.
However, in the systems that we are going to analyze, elements of G and H are
functions of s (frequency) and a proper handling is needed for such systems.
Bottom line is that feedback can destabilize a system if it is not carefully
designed.
Open loop system seldom has such a problem! Why?

16 / 20
Control Systems
E!ect on Parameter Sensitivity
How does system response when parameters of the system change?
M
Define a measure of sensitivity: SG is the percentage change in M with
respect to a percentage of change in G.
!M
M M #M G ↔ ωM G
SG = !G
= =
G
#G M ωG M

For a closed-loop system:

M 1
SG =
1 + GH
For openloop system:
M
SG =1

Implications:
↭ Openloop system “feels” the changes in G directly.
↭ Closedloop system can be made to be less sensitive to changes in G
using an appropriately large value of H.

17 / 20
Control Systems
Another significant advantage of feedback system is its superior performance in the
presence of disturbance or noise.

E!ect of input noise

Suppose H = 0 and r = 0 (Openloop, no feedback and zero reference) then
c(t)
c(t) = G2 n(t), or, = G2 C G GzHc
n(t) = -

Suppose H ↗= 0 but with r = 0, we have + Gra

c(t) = G2 (f (t) + n(t)) = G2 (→G1 Hc(t) + n(t)),

c(t) =
G2
n(t), or,
c(t)
=
G2 (1 + G GrH) c .

1 + G1 G2 H n(t) 1 + G1 G2 H -Gra
If H or G1 is large, then 1 + G1 G2 H is large and c(t) due to n(t) is small.
18 / 20
Control Systems
However, feedback is not e!ective against sensor noise.

E!ect of sensor noise

Suppose H ↗= 0 but with r = 0, we have
c(t) = G2 G1 e(t) = →G2 G1 H(n(t) + c(t)), or
→G2 G1 H
c(t) = n(t)
1 + G2 G1 H
If H is large, both numerator and denominator are large, n(t) is not
attenuated by large H!
19 / 20
Control Systems
In summary:
An openloop system
is simple to construct and maintain
is less expensive
has no problem with respect to stability
is very convenient when output is hard to measure or economically infeasible
is very sensitive to disturbances, system parameter variations, etc
requires re-calibration from time to time.
A Closed-loop system
can achieve good performance or “accurate control”
less sensitive to external disturbances, system parameter variations, etc
does not require recalibration
is more expensive to maintain since it uses more components and sensors
stability may be a problem and needs careful design of the system

20 / 20
 ME2142
Chapter 2 Laplace Transform

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

August 22, 2023

1 / 27
Signals and Systems
Systems
In the examples of physical systems, we realize that the behavior of the
variables of interest are represented by a set of di!erential equations.
While their respective physical domain can be di!erent (thermal, hydraulic,
electrical etc), they can be seen generically as a system.
When seen in this light, the system then takes some input function and
produces some output function.
Signals
In the most general definition, these input and output functions can be seen as
signals.
Hence, system takes a signal and produces another signal or signals.
More formally, a signal is a function that represents some variables that
describe behavior of the system (natural or artificial).
Signals are part of the whole and are meaningless without the systems to
interpret them. Similarly, systems are useless without signals to process.

2 / 27
Signals and Systems

There are two types of signals: continuous time and discrete-time signal.

Discrete-time signal is when information is available at discrete interval of time.

Our focus is on continuous time signal.
In order to analyse responses of system when it is subjected to various signals,
one approach is to use some typical representative signals (consider fourier
series).
We now consider various typical signals and focus on those that are
representable by Laplace transform.
We begin with a review of Laplace transform.

3 / 27
Complex numbers, variables and functions
Complex Numbers and Complex Variables
A complex number consists of two parts: the real and
the imaginary and is typically represented as

s = 3 + 4i, s = 3 + 4j

It can also be represented in terms of magnitude and

phase; for example imaginary
! 4
3 + 4j = |3 + 4j|eiω = 32 + 42 eiω = 5eiω , ω = tan→1
3
real
A complex variable is one where the real and
imaginary parts are both variables

s = ε + ϑi where ε, ϑ are both variables

Similarly, it can be represented as s = r↭(ω)

The complex conjugate of s is s̄ = ε → ϑi = r↭(→ω)
where
s ↑ s̄ = ε 2 + ϑ 2 = |s|2 = r2
4 / 27
Complex numbers, variables and functions

Basic operations of complex numbers

Given z1 = a1 + b1 i and z2 = a2 + b2 i,
Addition & Subtraction :

z1 ± z2 = a1 ± a2 + (b1 ± b2 )i

Multiplication :

z1 ↑ z2 = (a1 + b1 i) ↑ (a2 + b2 i) = (a1 a2 → b1 b2 ) + i(a1 b2 + a2 b1 )

Division :
z1 z1 z̄2 z1 ↑ z̄2
= ↑ =
z2 z2 z̄2 |z2 |2
(a1 a2 + b1 b2 ) + i(a2 b1 → a1 b2 )
=
a22 + b22

5 / 27
Complex numbers, variables and functions
Complex Function
A complex function is a function of the complex variable s, given by

G(s) = Gx + Gy i (Gx and Gy are real numbers)

where the values of Gx , Gy depends on the value of s.

G(s) is also
! a complex number with G
|G(s)| = (Gx )2 + (Gy )2 , ω = ↭G(s) = tan→1 Gxy or G(s) = |G(s)|↭ω

↭ Example: G(s) = s2 + 2s + 2. Then

G(s)|s=2+j = (2+j)2 +2(2+j)+2 = 9+6j

G(s)|s=→1+j = (→1+j)2 +2(→1+j)+2 = 0

↭ When s takes di!erent values, so does
G(s)

6 / 27
Laplace Transform
Laplace Transformation
Given a real time function f (t) where t ↓ 0.
Let s = ε + jϑ be a complex variable where ε, ϑ are both variables.
The Laplace transform of f (t) denoted by F (s) is
" ↑
F (s) = L[f (t)] = f (t)e→st dt
0

Existence: F (s) exists if the integral converges.

The integral converges if f (t) is piecewise continuous in every finite interval in
the range t > 0 and it is of exponential order as t ↔ ↗.
A function f (t) is said to be exponential order if a real, positive constant ε
exists such that
e→εt |f (t)| ↔ 0 as t ↔ ↗
or equivalently |f (t)| ↘ M eεt for some positive constant M . This is so since
" ↑ " ↑ " ↑
M
|F (s)| = | f (t)e→st dt| ↘ |f (t)|e→st dt ↘ M eεt e→st dt =
0 0 0 s→ε

7 / 27
Laplace Transform

Properties and examples of Laplace Transform

1 L[Af (t)] = AL[f (t)] where A is a constant.
2 L[A1 f1 (t) + A2 f2 (t)] = A1 L[f1 (t)] + A2 L[f2 (t)]
3 Exponential function : f (t) = Ae→ϑt , t ↓ 0, then
" ↑ " ↑
F (s) = L[f (t)] = Ae→ϑt · e→st dt = Ae→(ϑ+s)t dt
0 0

A →(ϑ+s)t ↑ A
=→ e |0 =
s+ϖ s+ϖ

8 / 27
Laplace Transform
Properties and examples of Laplace Transform
4 Step function: f (t) = A, t ↓ 0, then
" ↑
A
F (s) = L[f (t)] = Ae→st dt =
0 s
A unity step function is when A = 1 for t ↓ 0 or f (t) = (t) with
corresponding F (s) = 1s .
Since Laplace transform of any f (t) is only defined from t ↓ 0, the use of (t)
will be common.
5 Ramp function: f (t) = At, t ↓ 0 (or f (t) = At (t)), then
" ↑
A
F (s) = Lf (t) = Ate→st dt = 2
0 s
1
A unity ramp function is when A = 1 or f (t) = t (t) with F (s) = s2
.

9 / 27
Laplace Transform

Properties and examples of Laplace Transform

6 Sinusoidal function: f (t) = Asinϑt, t ↓ 0 with A, ϑ
being constants. Note that
1 iϖt
sinϑt = (e → e→iϖt )
2i
" ↑
A Aϑ
F (s) = (eiϖt → e→iϖt )e→st dt =
2i 0 s2 + ϑ 2
As
Similarly, L[Acosϑt] = s2 +ϖ 2

7 Power of t: f (t) = tn , t ↓ 0 where n is a positive integer, then

" ↑
n!
F (s) = L[tn ] = tn e→st dt = n+1
0 s
tn→1 1
Similarly, if f (t) = (n→1)!
, then F (s) = sn
.

10 / 27
Laplace Transform

8 Translated function:
Suppose a function f (t) is given with
L[f (t)] = F (s),
f (t → ϖ) (for some ϖ > 0) is a time-shifted
function of f (t) by ϖ.
In addition, f (t) = 0 for t < 0. This means that f (t → ϖ) = 0 for t < ϖ.
For this purpose, let (t → ϖ) = 1 for t ↓ ϖ and (t → ϖ) = 0 for t < ϖ. Then,
" ↑ " ↑
L[f (t → ϖ) (t → ϖ)] = f (t → ϖ) (t → ϖ)e→st dt = f (t → ϖ)e→st dt
0 ϑ

Let ϱ := t → ϖ and consider a change of variable, the above becomes

" ↑ " ↑
f (ϱ )e→s(ϱ +ϑ) dϱ = e→ϑs f (ϱ )e→sϱ dϱ = e→ϑs F (s)
0 0

Hence, the Laplace transform of a translated time function by ϖ(↓ 0)

corresponds to the multiplication of e→ϑs to F (s).

11 / 27
Laplace Transform

9 Pulse function: Consider the pulse function

# A
t0
, 0 < t < t0 ;
f (t) =
0, t ↘ 0, or t > t0 .

where A and t0 are constants.

A A
Hence, f (t) = t0
(t) → t0
(t → t0 ), or
#
A A A A →st0 A
L[f (t)] = L[ (t)] → L[ (t → t0 )] = → e = (1 → e→st0 )
t0 t0 st0 st0 st0

12 / 27
Laplace Transform
10 Impulse function: the special case of pulse function where t0 ↔ 0.
A
Recall that height of the pulse is t0
while the duration is t0 , hence area is A.
When t0 approaches 0, this area remains unchanged. In addition,

A A(1 → e→st0 )
L[ lim f (t)] = lim (1 → e→st0 ) = lim
t0 ↓0 t0 ↓0 st0 t0 ↓0 st0
d
dt0
A(1 → e→st0 ) As
= lim d
= =A
t0 ↓0
dt0
st0 s

The special case where A = 1 is known as the unit-impulse function, ς(t).

When the impulse occurs at t = t̄ is denoted by ς(t → t̄). Note that ς(t → t̄)
satisfies

ς(t → t̄) = 0 for t ≃= t̄

ς(t → t̄) = ↗ for t = t̄, and
" ↑
ς(t → t̄)dt = 1
→↑

13 / 27
Laplace Transform
More Properties of Laplace Transform
Given L[f (t)] = F (s). It can be shown that
11 L[f (t)e→at ] = F (s + a) where a is a constant. For example:
ϑ
L[e→at sinϑt] =
(s + a)2 + ϑ 2

shittyoperty
L[e→at cosϑt] =
s+a
(s + a)2 + ϑ 2
12 Change of time scale: L[f ( ϑt )] = ϖF (ϖs) where ϖ is a constant. E.g.,
1 2
L[e→t ] = ; L[e→t/2 ] =
s+1 2s + 1
13 Di!erentiation:
df (t)
L[ ] = sF (s) → f (0)
dt
d2 f (t)
L[ ] = s2 F (s) → sf (0) → f ↔ (0)
dt2
dn f (t)
L[ ] = sn F (s) → sn→1 f (0) → sn→2 f ↔ (0) → · · ·
dt
14 / 27
Laplace Transform
The above di!erentiation rule can be seen from, using the integration by parts:
" ↑ " ↑
e→st ↑ df (t) e→st
F (s) = f (t)e→st dt = f (t) |0 → [ ] dt
0 →s 0 dt →s
f (0) 1 df (t)
= + L[ ]
s s dt
Hence, L[ dfdt(t) ] = sF (s) → f (0)
$t
14 Integration: L[ 0 f (t)dt] = F (s) s
. This results follows from
" t " ↑" t
L[ f (t)dt] = [ f (t)]e→st dt
0 0 0
" t " ↑
e→st ↑ e→st
=[ f (t)] |0 → f (t) dt
0 →s 0 →s
" t " ↑
1 1 f →1 (0) F (s)
= f (t)dt|t=0 + f (t)e→st dt = +
s 0 s 0 s s
$t
where f →1 (0) = 0 f (t)dt|t=0 and f →1 (0) = 0 for most f except when
f (0) = ↗ (impulse function).

15 / 27
Laplace Transform
15 Final value theorem: What is f (↗) (if exists) given F (s)?
To see the result, take the limit of the above
df (t)
lim L[ ] = lim (sF (s) → f (0))
s↓0 dt s↓0
$↑ df (t) →st $↑
In addition, since lims↓0 0 dt
e dt = 0 dfdt(t) dt = f (↗) → f (0)
Using this property in the above shows that
f (↗) → f (0) = lims↓0 (sF (s) → f (0)) or f (↗) = lims↓0 sF (s)
16 Initial value theorem: What is f (0) given F (s)?
The above is answered using similar analysis:
" ↑
df (t) df (t) →st
L[ ]= e dt = sF (s) → f (0)
dt 0 dt
Taking the limit as s ↔ ↗ on both sides, we have
" ↑
df (t) →st
lim e dt = 0
s↓↑ 0 dt
Hence, 0 = lims↓↑ sF (s) → f (0) or
lim sF (s) = f (0)
s↓↑

16 / 27
Laplace Transform

Convolution Integral
One important property of Laplace transform is the convolution integral.
A convolution integral of two time functions f1 (t) and f2 (t) is
" t
f1 (t) ⇐ f2 (t) := f1 (t → ϱ )f2 (ϱ )dϱ
0

Convolution integral is used in many applications besides control system -

computer science, signal processing and others.
One special application is the convolution neural networks (CNN), an artificial
intelligent tool for image processing.
What is the convolution operator of two time functions? Let’s look at the
convolutional integral of a simple function.
Consider the case where f1 (t) = f2 (t) = (t)

17 / 27
Laplace Transform
Convolution Integral
For simplicity, let the step function be written as u(t).
The figures below show u(ϱ ) and u(→ϱ ) and u(t → ϱ ).

Note that u(t → ϱ ) is the translated function of u(ϱ ) by t and the figure
changes for values of t.
Depending on the value of t, the convolution integral can be obtained.
When t ↘ 0,
" t
f1 (t) ⇐ f2 (t) = u(ϱ ) · u(t → ϱ )dϱ = 0( ↫ u(t → ϱ ) = 0 for t ↘ 0)
0

When t > 0,
" t " t
f1 (t) ⇐ f2 (t) = u(ϱ ) · u(t → ϱ )dϱ = 1dϱ = t
0 0

18 / 27
Laplace Transform
In Laplace transform, the above complicated computation is greatly simplified.
" t
L[f1 (t) ⇐ f2 (t)] = L[ f1 (t → ϱ )f2 (ϱ )dϱ ] = F1 (s) · F2 (s)
0

Check (earlier example): f1 (t) ⇐ f2 (t) = L→1 (F1 (s) · F2 (s)) = L→1 ( s12 ) = t (t)
To show the above (for those interested), note that
" t " ↑" t
L[f1 (t) ⇐ f2 (t)] = L[ f1 (t → ϱ )f2 (ϱ )dϱ ] = [ f1 (t → ϱ )f2 (ϱ )dϱ ]e→st dt
0 0 0
" ↑" ↑
= f1 (t → ϱ ) (t → ϱ )f2 (ϱ )dϱ e→st dt
0 0
" ↑ " ↑
= f1 (t → ϱ ) (t → ϱ )e→st dt f2 (ϱ )dϱ
0 0
ς:=t→ϱ " ↑ " ↑
%&'(
= f1 (φ) (φ)e→s(ς+ϱ ) dφ f2 (ϱ )dϱ
" ↑0 " ↑ 0

=[ f1 (φ)e→sς dφ][ f2 (ϱ )e→sϱ dϱ ] = F1 (s) · F2 (s)

0 0

19 / 27
Inverse Laplace Transform
" c+j↑
L→1 [F (s)] = F (s)est ds; t ↓ 0
c→j↑

where c is a chosen constant, larger than the real parts of all singular points of F (s).
However, the above is seldom used to compute L→1 F (s). Instead, partial
fraction and inverse Laplace table are preferred.

Laplace Transform Table

f (t) F (s)
1 ς(t) 1
1
2 (t) s
n n!
3 t , n = 1, 2, · · · sn+1
4 e→at 1
s+a
n →at n!
5 t e (s+a)n+1
ϖ
6 sinϑt s2 +ϖ 2
s
7 cosϑt s2 +ϖ 2
→at ϖ
8 e sinϑt (s+a)2 +ϖ 2
9 e→at cosϑt s+a
(s+a)2 +ϖ 2
All f (t) are assumed to be 0 for t < 0.

20 / 27
Example of finding the inverse Laplace Transform
Example 1
1 5 8
Find the inverse Laplace of F (s) = s2
+ (s+3)2
+ s2 +16
.
From inverse Laplace Table:
1
L→1 [ ]=t
s2
5 1
L→1 [ ] = 5L→1 [ ] = 5te→3t (L[e→at f (t)] = F (s + a))
(s + 3)2 (s + 3)2
8 4
L→1 [ 2 ] = 2L→1 [ 2 ] = 2sin4t
s + 16 s + 42
Hence, f (t) = t + 5te→3t + 2sin4t for t ↓ 0
Example 2
4
Find the inverse Laplace of F (s) = s2 +2s+5
.

4 2
L→1 [ ] = L→1 [2 ] = 2e→t sin2t for t ↓ 0
s2 + 2s + 5 (s + 1)2 + 22

21 / 27
Example of finding the inverse Laplace Transform
Example 3
4e→8s
Find the inverse Laplace of F (s) = s2 +2s+5
.

4e→8s 2
L→1 [ ] = 2L→1 [e→8s 2 ]
s2 + 2s + 5 s + 2s + 5
2
= 2L→1 [e→8s ]
(s + 1)2 + 22
= 2e→(t→8) sin(2(t → 8)) (t → 8)

where the following two properties are used in the last step:

L[e→at f (t)] = F (s + a) and L[f (t → a) (t → a)] = e→as F (s)

Hence, the solution is

#
2e→(t→8) sin(2(t → 8)), t ↓ 8;
f (t) =
0, t < 8.

What about the general case?

22 / 27
The inverse Laplace Transform
Case 1: All distinct poles

B(s) (5s + 3)
F (s) = =
A(s) (s + 1)(s + 2)(s + 3)

Since all poles are real and distinct, we know

a1 a2 a3
F (s) = + +
s+1 s+2 s+3
(5s + 3)
where a1 = (s + 1)F (s)|s=→1 = |s=→1 = →1
(s + 2)(s + 3)
(5s + 3)
where a2 = (s + 2)F (s)|s=→2 = |s=→2 = 7
(s + 1)(s + 3)
(5s + 3)
where a3 = (s + 3)F (s)|s=→3 = |s=→3 = →6
(s + 1)(s + 2)

Hence, f (t) = L→1 [ s+1

→1
+ 7
s+2
+ →6
s+3
] = →1e→t + 7e→2t → 6e→3t for t ↓ 0.

23 / 27
The inverse Laplace Transform

Case 2: Repeated poles

B(s) s2 + 2s + 3
F (s) = =
A(s) (s + 1)3

Since poles are repeated, we have

a1 a2 a3
F (s) = + +
(s + 1) (s + 1)2 (s + 1)3
where a3 = (s + 1)3 F (s)|s=→1 = s2 + 2s + 3|s=→1 = 2
d
where a2 = [(s + 1)3 F (s)]|s=→1 = 2s + 2|s=→1 = 0
ds
1 d
where a1 = (s + 3)3 F (s)|s=→1 = 1
2! ds

1
Hence, f (t) = L1 [ s+1 + 2
(s+1)3
] = (1 + t2 )e→t for t ↓ 0

24 / 27
The inverse Laplace Transform
Case 1:
B(s) K(s+z1 )(s+z2 )···(s+zm )
Given F (s) = A(s)
= (s+p1 )(s+p2 )···(s+pn )
with m ↘ n and pk are all distinct.
Obtain a partial fraction expansion of F (s) in the form of
a1 a2 an
F (s) = + + ··· + an+1
s + p1 s + p2 s + pn

Then,

f (t) = a1 e→p1 t + a2 e→p2 t + · · · + an e→pn t + an+1 ς(t); t↓0

The values of ak can be obtained from

ak = (s + pk )F (s) evaluated at s = →pk for any k = 1, · · · , n

an+1 = F (s) evaluated at s = ↗

Note that

an+1 = 0 if m < n
an+1 = constant if m = n

25 / 27
The inverse Laplace Transform
Case 2:
B(s) K(s + z1 )(s + z2 ) · · · (s + zm )
F (s) = =
A(s) (s + p1 )r (s + pr+1 ) · · · (s + pn )
with m ↘ n and pk , k = 1, r + 1, · · · , n are all distinct, p1 is repeated r times.
Obtain a partial fraction expansion of F (s) in the form of
ar ar→1 a1
F (s) = + + ··· +
(s + p1 )r (s + p1 )r→1 s + p1
ar+1 an
+ + ··· + an+1
s + pr+1 s + pn

Then,
ar ar→1 r→2
f (t) = ( tr→1 + t + · · · + a2 t + a1 )e→pi t
(r → 1)! (r → 2)!
+ar+1 e→pr+1 t + ar+2 e→pr+2 t + · · · + ar+n e→pr+n t + an+1 ς(t); t↓0

Note that

an+1 = 0 if m < n
an+1 = constant if m = n

26 / 27
The inverse Laplace Transform

The values of ak can be obtained from

ar = (s + p1 )r F (s) evaluated at s = →p1

d
ar→1 = [(s + p1 )r F (s)] evaluated at s = →p1
ds
..
.
1 dj
ar→j = [(s + p1 )r F (s)] evaluated at s = →p1
j! dsj
..
.
1 dr→1
a1 = [(s + p1 )r F (s)] evaluated at s = →p1
(r → 1)! dsr
aj = (s + pj )F (s) evaluated at s = →pj for any j = r + 1, · · · , n
an+1 = F (s) evaluated at s = ↗

27 / 27
 ME2142
Chapter 3 Mathematical Modeling of Physical
Systems

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

August 18, 2022

1 / 23
Transfer Function Representation

Systems
In this chapter, we show how physical systems can be characterized by their
input-output relationship.
Such input-output relationships are known as transfer functions.
This characterization is applicable to many physical systems in di!erent
domains so long as they are described by linear di!erential equations.
They are also represented graphically as

with C(s) or c(t) being the output and R(s) or r(t) being the input.

2 / 23
Transfer Function

Suppose you are asked to find out the behaviour of a certain physical quantity
when the input (energy source) is suddenly increased by 10%.
How would you answer such a question?
If you know the relationship between the input (y(t)) and the output (x(t)),
then you might be able to do that.
Lets say they are related by a di!erential equation -

d2 x(t) dx(t) dy(t)

+2 + 5x(t) = 2y(t) +
dt2 dt dt
How to proceed? - search your bag of tricks!
We now show how this can be done.

3 / 23
Transfer Function
Taking L on both sides and assume that all initial conditions are zeroes,

d2 x(t) dx(t) dy(t)

L[ ] + L[2 ] + L[5x(t)] = L[2y(t)] + L[ ]
dt2 dt dt
s2 L[x(t)] → sx(0) → ẋ(0) + 2sL[x(t)] → x(0) + 5L[x(t)] = 2L[y(t)] + sL[y(t)] → y(0)
(s2 + 2s + 5)L[x(t)] = (2 + s)L[y(t)]

2+s
or X(s) = ·Y (s)
s2 + 2s + 5
! "# $
G(s)

To get the output expression,

x(t) = L→1 (G(s) · Y (s)) = convolution integral of g(t) and y(t)

The term G(s) appears commonly - known as the Transfer Function

X(s) L[x(t)] 2+s

G(s) = = = 2
Y (s) L[y(t)] s + 2s + 5

4 / 23
Transfer Function

What about the general case?

Consider the di!erential equation:

dn dn→1 dm dm→1
a0 n
c(t) + a1 n→1 c(t) + · · · + an c(t) = b0 m r(t) + b1 m→1 r(t) + · · · + bm r(t)
dt dt dt dt
with n ↑ m.
Doing the same procedure as before, the transfer function is

TraestarG(s) = L[output]
=
C(s)
=
b0 sm + b1 sm→1 + · · · bm→1 s + bm
L[input] R(s) a0 sn + a1 sn→1 + · · · an
function
Note that in deriving the above, all initial conditions are assumed zero.

5 / 23
TF satisfies superposition

We now show that TF satisfies superposition theorem:

Specifically, this means that if c1 (t), c2 (t) are the outputs of a TF having
inputs r1 (t), r2 (t) respectively of the following system:

ω1 c1 (t) + ω2 c2 (t) is the output of the same system with input ω1 r1 (t) + ω2 r2 (t)
Note that this superposition property is for Linear system only, nonlinear
systems do not have such property.

6 / 23
TF satisfies superposition

We now show this superposition property for the general case:

Suppose (R1 (s), C1 (s)) and (R2 (s), C2 (s)) are two input-output pairs, or

C1 (s) = G(s)R1 (s), C2 (s) = G(s)R2 (s)

Clearly, it also implies that

ω1 C1 (s) = G(s)(ω1 R1 (s)), ω2 C2 (s) = G(s)(ω2 R2 (s))

for any scalars ω1 and ω2 .

Consider now an input function R(s) = ω1 R1 (s) + ω2 R2 (s) on the system
G(s), we have

C(s) = G(s)(ω1 R1 (s) + ω2 R2 (s)) = G(s)(ω1 R1 (s)) + G(s)(ω2 R2 (s))

= ω1 C1 (s) + ω2 C2 (s)

7 / 23
TF satisfies superposition
For example, consider the following two systems (I and II)

(I) ċ(t) = r(t) (II) ċ(t) = r2 (t)

with two separate inputs r1 (t) = (t) and r2 (t) = t (t).

t2
For system (I), the outputs are c1 (t) = t and c2 (t) = 2
respectively.
If we use as input r(t) = ω1 r1 (t) + ω2 r2 (t), then
% t
t2
c(t) = (ω1 1 + ω2 t)dt = ω1 t + ω2 = ω1 c1 (t) + ω2 c2 (t)
0 2

Hence, system I satisfy superposition.

t3
For system II, the outputs are c1 (t) = t and c2 (t) = 3
respectively.
If we use as input r(t) = ω1 r1 (t) + ω2 r2 (t), then
% t % t
c(t) = (ω1 1 + ω2 t)2 dt = (ω12 + 2ω1 ω2 t + ω22 t2 )dt
0 0
3
t
= ω12 t 2
+ ω1 ω2 t + ω22 ↓= ω1 c1 (t) + ω2 c2 (t)
3

8 / 23
Transfer Function

Some Additional Definitions:

Order of a system - The highest power of “s↑↑ in the denominator of G(s).
Poles of a system - roots of the denominator polynomial of G(s).
Zeros of a system - roots of the numerator polynomial of G(s).
Proper Transfer Function: The transfer function G(s) is called proper if
n ↑ m (i.e., G(s)|s=↓ = constant).
Strictly proper transfer function: The transfer function G(s) is called strictly
proper if n > m(i.e., G(s)|s=↓ = 0).
Example: Suppose
s2 + 2s + 2
G(s) =
s(s2 + 3s + 2)
By definition, n = 3, m = 2 ↔ G(s) is strictly proper.
Order of system is 3, Poles of system are s = 0, →2, →1 and Zeros of system are
s = →1 ± j1.

9 / 23
Transfer Function
E!ect of Poles and Zeroes of a system:
Consider a system with transfer function given by
C(s) 3s → 1
G(s) = =
R(s) (s + 1)(s + 2)

The output function of this system with input being a unit step function is
3s → 1 1 4 3.5 0.5
C(s) = G(s)R(s) = · = → →
(s + 1)(s + 2) s s+1 s+2 s

Taking the inverse Laplace of each of the terms

c(t) = 4e→t → 3.5e→2t → 0.5

The response consists of

↭ (i) exponential corresponding to each pole of the system,
↭ (ii) contribution due to the input
What about the e!ect of the zeros?

10 / 23
E!ect of Poles and Zeroes

To answer the above, consider the following systems:

2 0.2(s + 10)
G1 (s) = , G2 (s) =
(s + 1)((s + 1)2 + 1) (s + 1)((s + 1)2 + 1)
→0.2(s → 10) 10(s2 + 0.1s + 0.2)
G3 (s) = , G4 (s) =
(s + 1)((s + 1)2 + 1) (s + 1)((s + 1)2 + 1)

with input R(s) = 1s .

Note that all four systems above have the same denominator.
The output can be shown to be of the form

c(t) = k1 e→t + k2 e→(1+j)t + k̄2 e→(1→j)t + k4

11 / 23
E!ect of Poles and Zeroes
The exact responses are given here:

2.5
G1(s)
G2(s)
G3(s)
2
G4(s)

1.5
y(t) of four systems

0.5

-0.5
0 1 2 3 4 5 6 7 8
time (s)

Even though the poles of G(s) determine the basic form of the system
responses, exact responses are determined by the poles, zeros and the input
applied to the system.
Therefore, the zeros of a transfer function cannot be completely ignored in the
analysis and design of control systems.

12 / 23
Modeling of Physical Systems

A mathematical model of a physical system is defined as a set of equations

that represent the dynamics of the system accurately or, at least, fairly well
within the operating region of interest.
A dynamic system is a system whose response at any time depends on the
input of the system at the present as well as in the past times.
The mathematical model is not unique - a system may be represented in a
number of di!erent ways.
The mathematical model of many systems may be described in terms of
di!erential equations. Such equations are obtained using laws of physics such
as Newton’s laws, Kircho!’s laws etc.
When possible, the first step in the analysis of a physical system is to derive its
mathematical model.
In more complicated system, system model can be hard to obtain - use system
identification techniques.

13 / 23
Example 1: Mechanical Translational System
Consider a mechanical translation system given here:

x(t)

k
Mass m f(t)

viscous friction b

Using Newton’s&law (mass ↗

acceleration = forces applied):
we are interested in the TF between
f (t) (input) and x(t) (output). d2 x(t) dx(t)
m = →b → kx(t) + f (t)
f (t) : force applied dt2 dt

x(t) : displacement of the mass Taking Laplace Transform (with

k : spring constant zero initial states):
b : viscous friction coe”cient of wheel ms2 X(s) + bsX(s) + kX(s) = F (s)
m : mass of the cart
X(s) 1
↔ =
F (s) ms2 + bs + k
14 / 23
Example 2: Mechanical Translational System II
we are interested in the TF between
y(t) x(t) (input) and y(t) (output).
x(t)
x(t) : displacement of cart
k
y(t) : displacement of the mass
Mass m
b k : spring constant
b : viscous friction coe”cient of wheel
m : mass of the cart
Using Newton’s law again:

d2 y(t) d
m = →b (y(t) → x(t)) → k(y(t) → x(t))
dt2 dt
Taking Laplace Transform (with zero initial states):

ms2 Y (s) + bsY (s) + kY (s) = bsX(s) + kX(s)

Y (s) bs + k
↔ =
X(s) ms2 + bs + k

15 / 23
Example 3: Mechanical Rotational System
Interested in the TF between ε (t)
(input) and ϑ(t) (output).
τ (t) ω(t)
Inertia J ϑ(t) : angular vel. of shaft
ε (t) : torque applied to shaft
Damping b
J : moment of inertia
b : viscous friction coe!. of bearings
Using Euler’s equation:
dϑ(t)
J = →bϑ(t) + ε (t)
dt
Taking Laplace Transform (with zero initial states):

Js#(s) + b#(s) = T (s)

#(s) 1
↔ =
T (s) Js + b

16 / 23
Example 4: Mechanical Rotational System
Interested in the TF between ε (t)
(input) and ϖ(t) (output).

ϑ(t) : angular vel. of shaft

ϖ(t) : angular displacement of shaft
ε (t) : torque applied to shaft
J : moment of inertia
b : viscous friction coe!. of bearings
k : spring constant of shaft
Using Euler’s equation:

d2 ϖ(t) dϖ(t)
J = →b → kϖ(t) + ε (t)
dt2 dt
Taking Laplace Transform (with zero initial states):

(Js2 + bs + k)$(s) = T (s)

$(s) 1
↔ =
T (s) Jss + bs + k

17 / 23
Example 5: Thermal System
Interested in the TF between u(t)
(input) and ϖ(t) (output).

Qi1 (t), Qi2 (t) : heat gain through windows

Qo (t) : heat loss thru AC
ϖ(t), ϖe (t) : temperature inside and out
R1 , R2 : heat resistances
C : heat capacity of room
The heat balance equation is
dϖ(t)
Qi1 (t) + Qi2 (t) → Qo (t) = C
dt
Heat into the room from outside is proportional temperature di!erence - ϖe → ϖ.
Heat removed by air conditioning - u(t).
ϖe (t) → ϖ(t) ϖe (t) → ϖ(t) dϖ(t)
+ → u(t) = C
R1 R2 dt
Taking the Laplace transform of all the above and rearranging
1 1
( + )($e (s) → $(s)) → u(s) = Cs$(s)
R1 R2
(R1 + R2 )$e (s) → R1 R2 u(s)
$(s) =
R1 + R2 + CR1 R2 s 18 / 23
Example 5: Electrical RLC Circuit
Consider a RLC circuit:

Using the KVL around the circuit, it follows that

% %
di(t) 1 t 1 t
ei (t) = i(t)R + L + i(t)dt, eo (t) = i(t)dt
dt C 0 C 0

Taking the Laplace transform and assuming zero initial conditions:

1 Eo (s) 1
Ei (s) = (R + Ls + )I(s), =
Cs I(s) Cs

I(s) Cs Eo (s) 1
↔ = , =
Ei (s) LCs2 + RCs + 1 Ei (s) LCs2 + RCs + 1

19 / 23
Example 6: Operational Amplifier system

Assume if = 0 because input For this system,

impedance of op-amp is high.
1
If so, i1 (t) + if (t) = 0 or, zf = + Rf , zi = Ri
vi (t) Cs
zi
+ vozf(t) = 0.
Using the expression,
This implies that
Vo (s) 1 + Rf Cs
zf Vo (s) zf =→
vo (t) = → vi (t), =→ Vi (s) Ri Cs
zi Vi (s) zi

20 / 23
DC motor
Example 7: Armature-controlled D.C. motor

The DC motor is an extension of the RLC circuit in combination with

mechanical rotational system.
Note that the back e.m.f is proportional to the angular speed of the rotor.
The torque generated is proportional to the product of the magnetic fields -
proportional to armature current for permanent magnet motor.
From the electrical circuit side:
dia
R a i a + La + vb (t) = va (t) = u(t)
dt
dia
or, Ra ia + La + kb ϑ(t) = u(t)
dt
21 / 23
DC motor
As for the mechanical side, we have
dϑ(t)
J + f ϑ(t) = T (t)
dt
For a permanent magnet DC motor, we have
T (t) = kt ia (t)
Taking the Laplace transform of the above two, we arrive at
(Js + f )#(s) = T (s) = kt Ia (s)
U (s) → kb #(s)
(Ra + La s)Ia (s) + kb #(s) = U (s), Ia (s) =
(Ra + La s)
Noting that #(s) = s$(s), the following diagram is obtained

22 / 23
DC motor

Eliminating the Ia (s) from the two equations:

$(s) kt
G(s) = =
U (s) s[(Js + f )(Ra + La s) + kt kb ]

In most cases, the inductance, La is small ignored. The above becomes

$(s) kt km
G(s) = = = (↘)
U (s) s[JRa s + f Ra + kt kb ] s(ε s + 1)
where
kt
km = = motor gain constant
kt kb + f R a
JRa
ε = = motor time constant
kt kb + f R a

Expression (*) is known as the motor transfer function.

Note that ε depends on load J and parameters Ra and kt .

23 / 23
 ME2142
Chapter 4 Block Diagrams Algebra

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

August 25, 2022

1 / 13
Block Diagram Representation of System
A block diagram is a graphical tool to help visualize the various parts of a
system and their mathematical relationships.
The easiest block is

where C(s) = G(s)R(s)

Recall a more complicated block diagram from last Chatper.

This chapter deals with the simplification of block diagram in order to arrive
at the transfer function between an input and its output.
A complicated block diagram involving many feedback loops can be simplified
by a step-by-step rearrangement using the rules of block diagram algebra.
2 / 13
Basic rules

The best way to explain the algebra of block diagram is through examples.
However, useful to remember a few basic principles to the manipulation of
block diagrams:
(i) the product of the transfer functions in the feedforward direction must be
the same before and after the manipulation.
(ii) the product of the transfer functions around the loop must remain the
same before and after the manipulation.
These rules will be highlighted in the next few examples.
We begin with a few simple known results.

3 / 13
 Blocks in parallel
Blocks in series

Z(s) = G1 (s)G2 (s)X(s)

Y (s) = (G1 (s) + G2 (s))X(s)

4 / 13
Shifting across a block
Shifting a block to the left over a summing junction

Shifting a block to the left over a tap

Shifting a block to the right over a summing junction

5 / 13
Closed-loop Feedback System

C(s)
Interested in TF of R(s)
. Note that

R(s), C(s) input and output of C = GE = G(R → B) = G(R → HC)

system. ↑ (1 + GH)C = GR
B(s) feedback signal. C(s) G
↑ =
E(s) error signal. R(s) 1 + GH
G(s) feedforward TF E(s) 1
In addition, C(s)
= G(s)
. Hence,
G(s)H(s) openloop TF
E(s) E(s) C(s) 1
= · =
R(s) C(s) R(s) 1 + GH

6 / 13
Closed-loop Feedback System

Gc : the transfer function for the controller

Gp : the transfer function for the plant
M : is the output of the controller
D : is the external disturbance
Gc Gp : is the feedforward transfer function
Gc Gp H : is the openloop transfer function

7 / 13
Closed-loop Feedback System

Assume D = 0, we have
C(s) G Gc Gp
= =
R(s) 1 + GH 1 + Gc Gp H

Assume R = 0, we have
C(s) G Gp
= =
D(s) 1 + GH 1 + Gc Gp H

8 / 13
Example 1:

G1 G2
Y (s) 1→G2 G3 G1 G2
= G1 G2
=
R(s) 1 + 1→G 1 → G 2 G3 + G1 G2
2 G3

9 / 13
Example 1: Alternative approach

R(s) Y(s)
+ +
- GI - G2 -

¼i - G3

R(s) + + Y(s)
G 1 G2
'
+ ·•

G G
ia1

R(s) +

Y (s) G1 G2 G1 G2
= G3
=
R(s) 1 + G1 G2 (1 → G1
) 1 → G2 G3 + G1 G2

10 / 13
Example 2:

11 / 13
Example 2:

C(s) G1 G2 G3
=
R(s) 1 → G 1 G 2 H 1 + G 2 G 3 H2 + G 1 G 2 G 3

12 / 13
Example 3:

Y (s) G1 G1
= =
R(s) 1 + G1 ( G
G1
2
+ G 3 ) 1 + G 2 + G1 G3

13 / 13
 ME2142
Chapter 5 Time Domain Analysis

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

September 14, 2023

1 / 31
System Responses

As systems are dynamical in nature, their performances are hard to measure or

quantify.
One accepted approach is to use a reference input to the system, and measure
its performances in time - time domain analysis.
Let c(t) be the output of the system with input r(t). The time response of a
system consists of two parts: the transient and the steady state response

2 / 31
System Responses

The responses can be written as

c(t) = ct (t) + css (t)

where css (t) denotes the steady state response and ct (t) denotes the transient
response.
The transient response is defined as the part of the time response that goes to
zero as time becomes very large, or,

lim ct (t) = 0
t→↑

while the steady state response is the part of the total response that remains
after the transient had died out.
All real control systems exhibit transient phenomena (response) to some extent
before the steady state is reached. Since inertia, mass, inductance, capacitance
etc. are unavoidable in physical systems, the responses of a typical control
system cannot follow sudden changes in the input instantaneously and
transients are usually observed.

3 / 31
System Responses

Hence, having a good transient response of a system is necessarily important,

as it is a significant part of the dynamic behavior of the system;
It determines the deviation between the output response and the input (or the
desired response) before the steady state is reached.
The steady state response is also very important - it determines where the
system output response ends up at when time becomes large.
If the steady state response of the output does not agree with the steady state
of the input or desired response exactly, a steady state error occurs (see figure).
The design of a controller should include considerations of the transient and
steady state performance specifications.
This chapter deals with the transient response of systems. Chapter 6 discusses
the steady state response.

4 / 31
System Responses

C(s)
Given a transfer function G(s) = R(s)
. How do we
specify the characteristics of the response required?
compare it with another system?
know whether it’s response will adequately meet our needs?
Our approach is to
use a family of standard test signals.
responses of systems over this family of signals are used for characterization
and comparison purposes.
characterize system responses to these test signals by the parameters of the
system.

5 / 31
Test Signals
Step Input,
!
A, t → 0;
r(t) = Step Input
0, t < 0.

↭ R(s) = As .
↭ When A = 1, known as unit step input.
↭ useful in studying changes in input signals.
Ramp Input
! Ramp input
At, t → 0;
r(t) =
0, t < 0.

↭ R(s) = sA2 .
↭ When A = 1, known as unit ramp input.
↭ useful in studying system ability to following
increasing input value.

6 / 31
Test Signals

Impulse Input,

r(t) = Aω
Impulse Input

↭ R(s) = A.
↭ When A = 1, known as unit impulse input.
↭ useful in studying sudden shock or impact
and system identification
Sinusoidal Input
Sinusoidal input
!
Asinεt, t → 0;
r(t) =
0, t < 0.

↭ ω
R(s) = s2 +ω 2.
↭ When A = 1, known as unit sinusoid.
↭ useful in frequency response of system.

7 / 31
Responses of First-order system
We now look into the responses of systems under the test signals discussed. Focus is
on First-Order system :
1
G(s) =
Ts + 1

Recall the TF of DC motor with armature voltage as input and angular

velocity as output:
!(s) km
=
U (s) ϑm s + 1
Similarly, we have the TF of mechanical rotational system with torque as input
and angular velocity as output:
!(s) 1
=
U (s) Js + b

As well as the TF for Thermal System (with u(s) = 0):

”s (s) R1 + R2
=
”e (s) R1 + R2 + CR1 R2 s

8 / 31
Responses of First-order system
In terms of Block Diagram, First-order system can be represented in the following
two forms:

where T, k > 0.

Consider the input as a unit impulse, then

1 1
C(s) = R(s) = ·1
Ts + 1 Ts + 1

In time domain, it is
1 ↓ Tt
c(t) = e ,t → 0
T

9 / 31
Responses of First-order system

In the case where the input is a unit

step, R(s) = 1s .

1 1 T
C(s) = = ↑
s(T s + 1) s Ts + 1

In time domain, it is
t
c(t) = 1 ↑ e↓ T , t → 0

Note that

c(0) = 1 ↑ 1 = 0
The smaller the T , the faster the
c(T ) = 1 ↑ e↓1 = 0.632 response.
c(↓) = 1
For t > 4T , response within 2% of
dc(t) 1 t 1 final value
|t=0 = e↓ T |t=0 =
dt T T

10 / 31
Responses of First-order system

In the case where the input is a unit

ramp, R(s) = s12 .

1 1 T T2
C(s) = = ↑ +
s2 (T s + 1) s 2 s Ts + 1
1
-

2 In time domain, it is
t
# c(t) = t ↑ T + T e↓ T , t → 0

Note that

c(0) = 0 ↑ T + T = 0
c(↓) = ↓ error at steady state is

The error e(t) = r(t) ↑ c(t) = e(↓) = T

t t
T ↑ T e↓ T = T (1 ↑ e↓ T )

11 / 31
A property of Laplace Transform
Note : An important property of linear time-invariant systems is the following:
Suppose c1 (t) is the output response corresponding to an input r1 (t). Then if
the input is changed to
dr1 (t)
r2 (t) =
dt
then the corresponding output c2 (t) is given by

dc1 (t)
c2 (t) =
dt
This property can be verified easily when expressed in the Laplace domain. In
this case, we have

C1 (s) = G(s)R1 (s) ↔ sC1 (s) = G(s)sR1 (s) = G(s)R2 (s)

This above implies that the output of the system with R2 (s) as input is
sC1 (s). In time domain, this means output is dcdt
1 (t)
when input is drdt
1 (t)
.
This fact manifests in the responses of first order system with ramp, step and
impulse inputs - step function is the derivative of the ramp function, and the
impulse function is the derivative of the step input.

12 / 31
Responses of second-order system
We now look into the responses of a more general class of systems under the same
test signals. This class of system is Second-Order system represented as

εn2
G(s) =
s2 + 2ϖεn s + εn2
They are represented by
mechanical translation system with force as input and position as output:
X(s) 1
=
F (s) ms2 + bs + k

mechanical rotational system with toque as input and angular displacement as

output:
”(s) 1
=
T (s) Js2 + bs + k
RLC circuit with voltage as input and capacitor voltage as output:
Eo (s) 1
=
Ei (s) LCs2 + RCs + 1

13 / 31
Responses of second-order system
In terms of Block Diagram, second-order system can be represented in the following
two forms:

They are two parameters that characterize second-order system

εn = natural frequency
ϖ = damping ratio

In both block diagrams above, the TF between input and output is

εn2
G(s) =
s2 + 2ϖεn s + εn2

14 / 31
Responses of second-order system
Consider the case where ϖ, ε → 0 (other cases?) and input is a unit step. In this
case, R(s) = 1s and the output is

εn2
C(s) = G(s)R(s) =
s(s2 + 2ϖεn s + εn2 )

To obtain the steady state value, the final value theorem is used (assuming
steady state exists),

sεn2
lim c(t) = lim sC(s) = lim =1
t→↑ s=0 s=0 s(s2 + 2ϖεn s + εn2 )

To obtain expression of the transient response, expression of the poles of G(s)

is needed.
The poles of G(s) are the roots of denominator polynomial or roots of

s2 + 2ϖεn s + εn2 = 0

Consider the roots of the above under di#erent ranges of ϖ.

15 / 31
Responses of second-order system

1 0 < ϖ < 1 (underdamped case): roots are

"
s1,2 = ↑ϖεn ± jεn 1 ↑ ϖ2

2 ϖ = 0 (undamped case): roots are

s1,2 = ±jεn

3 ϖ = 1 (critically damped case): roots are

s1,2 = ↑εn

4 ϖ > 1 (overdamped case): roots are

" "
s1,2 = ↑ϖεn + εn ϖ 2 ↑ 1, ↑ϖεn ↑ εn ϖ2 ↑ 1

16 / 31
Responses of second-order system
We begin with the underdamped case (0 < ϖ < 1).
Expression of the output is

εn2 εn2 1
C(s) = = ·
s(s2 + 2ϖεn s + εn2 ) (s + ϖεn )2 + εd2 s
"
where εd = εn 1 ↑ ϖ 2 , known as the damped natural frequency. Simplifying,
1 s + 2ϖεn 1 s + ϖεn ϖεn
C(s) = ↑ 2 = ↑ ↑
s s + 2ϖεn s + εn2 s (s + ϖεn )2 + εd2 (s + ϖεn )2 + εd2

Taking the inverse Laplace, we get

ϖ
c(t) = 1 ↑ e↓ωεn t cosεd t ↑ e↓ωεn t " sinεd t
1 ↑ ϖ2
e↓ωεn t "
=1↑ " (ϖsinεd t + 1 ↑ ϖ 2 cosεd t)
1 ↑ ϖ2
↓ωεn t
"
e ↓1 1 ↑ ϖ2
=1↑ " sin(εd t + tan ), t → 0
1↑ϖ 2 ϖ

17 / 31
Responses of second-order system
Consider now the other cases.
Expression of the output for ϖ = 0 is

εn2 1 ε2 1 1 s
C(s) = · = 2 n 2 · = ↑ 2
(s + jεn )(s ↑ jεn ) s (s + εn ) s s s + εn2
↔ c(t) = 1 ↑ cosεn t, t → 0

Expression of the output for ϖ = 1 is

εn2 1 1 1 εn
C(s) = · = ↑ ↑
(s + εn )2 s s s + εn (s + εn )2
↔ c(t) = 1 ↑ (1 + εn t)e↓εn t , t → 0

Expression of the output for ϖ > 1 is

εn2 1 1 1 b a
C(s) = · = + ( ↑ )
(s + a)(s + b) s s a↑b s+a s+b
1
↔ c(t) = 1 + (be↓at ↑ ae↓bt ), t → 0
a↑b
" "
where a = ϖεn + εn ϖ 2 ↑ 1, b = ϖεn ↑ εn ϖ 2 ↑ 1
18 / 31
Responses of second-order system
How does c(t) look like for these cases?

Note the x-axis is in units of εn t.

19 / 31
Responses of second-order system

Assuming that εn is fixed, a few notable points are:

underdamped system with ϖ between 0.5 and 0.8 gets close to the final value
more rapidly than a critically damped or overdamped system.
Among the systems responding without oscillation, a critically damped system
exhibits the fastest response.
An overdamped system is always sluggish in responding to any inputs.

20 / 31
Definition of Transient Response

Systems with energy storage cannot respond instantaneously and will exhibit
transient responses whenever they are subjected to inputs or disturbances.
In many applications, the desired performance characteristics of control
systems are specified in terms of time-domain quantities (both transient and
steady state).
As systems with energy storage cannot respond instantaneously to input
demand, transient responses results.
Frequently, the performance characteristics of a control system are specified in
terms of transient response to a unit-step input since it is easy to generate and
is su$ciently drastic.
The transient response of a system to a unit-step input depends on the initial
conditions. For purposes of comparison, a standard set of initial conditions are
used.

21 / 31
Definition of Transient Response
Several quantities are often used for specifying transient-response characteristics of a
control system to a unit-step input.

1 Delay time, td : The delay time is the time required for the response to reach
half the final value the very first time.
2 Rise time, tr : The rise time is the time required for the response to rise from
10 % to 90 %, 5% to 95%, or 0% to 100% of its final value. For underdamped
second-order systems, the 0% to 100% rise time is normally used. For
overdamped systems, the 10% to 90% rise time is commonly used.

22 / 31
Definition of Transient Response

3 Peak time, tp : The peak time is the time required for the response to reach the
first peak of the overshoot.
4 Maximum overshoot, Mp : The maximum overshoot is the maximum peak
value of the response curve measured from unity. If the final steady-state value
of the response di#ers from unity, then it is common to use the maximum
percent overshoot. It is defined by
c(tp ) ↑ c(↓)
Mp = ↗ 100%
c(↓)

5 Settling time, ts : The settling time is the time required for the response curve
to reach and stay within a range about the final value of size specified by
absolute percentage of the final value (usually 2% or 5%). The settling time is
related to the largest time constant of the control system. Which percentage
error criterion to use is usually determined from the objectives of the system
design in question.

23 / 31
Comments on Transient Response

The time-domain specifications given are quite important since the responses
in time are what are observed.
Hence, the transient response should be satisfactory. Note that if values of
td , tr , tp , Mp and ts are specified, the shape of the response curve is virtually
determined.
Clearly, not all quantities are applicable for any given case. For example,
tp , Mp are not defined for an overdamped system; ts may not be reached if
systems have steady-state errors to step input, etc. Discussions on steady-state
errors are given in Chapter 6.
Except for special applications where oscillations cannot be tolerated, the
transient response should be su$ciently fast and lightly damped (minimal
overshoot). In the case of a second-order system, this means ϖ is between 0.5
and 0.8. Small values of damping ratio (less that 0.4) yield excessive overshoot
and large values of damping ratio (greater than 1) yields sluggish response.

24 / 31
Transient Quantities for Second-order systems
In the special case of standard second-order system with 0 ↘ ϖ ↘ 1, the
afore-mentioned quantities can be connected to the system parameters. Recall that
"
↓ωεn t ϖ e↓ωεn t ↓1 1 ↑ ϖ2
c(t) = 1 ↑ e (cosεd t + " sinεd t) = 1 ↑ " sin(εd t + tan )
1↑ϖ 2 1↑ϖ 2 ϖ

Note that tr is defined by the smallest solution to c(tr ) = 1. Hence,

"
e↓ωεn t ↓1 1 ↑ ϖ2
1=1↑ " sin(εd tr + tan )
1↑ϖ 2 ϖ

which implies that

sin(εd tr + ϱ) = 0
→ωε t
since ≃ n
e
⇐= 0. Simplifying, εd tr + ϱ = 0, ς, 2ς, · · · For the first tr > 0,
1↓ω2
≃
1↓ω2
ς↑ϱ ς ↑ tan↓1 ω
tr = =
εd εd
.

25 / 31
Transient Quantities for Second-order systems

dc(t)
Peak time tp is defined by dt tp
| = 0 (why?). Hence,

dc(t) εn e↓ωεn tp "

|t=tp = " (↑ϖsin(εd tp + ϱ) + 1 ↑ ϖ 2 cos(εd tp + ϱ)) = 0
dt 1↑ϖ 2

which implies that

"
ϖsin(εd tp + ϱ) = 1 ↑ ϖ 2 cos(εd tp + ϱ)
"
1 ↑ ϖ2
tan(εd tp + ϱ) =
ϖ
εd tp + ϱ = ϱ + nς
ς 2ς
or, tp = , ,···
εd εd

26 / 31
Transient Quantities for Second-order systems

The maximum overshoot occurs at

the peak time (assuming there is
overshoot). Hence,
ϖ
Mp = c(tp )↑1 = ↑e↓ωεn tp (cosεd tp + " sinεd tp )
1 ↑ ϖ2

↓ωεn ( εϑ ) ϖ
= ↑e d (cosς + " sinς)
1 ↑ ϖ2
which implies that
↓ ≃ ωϑ
1→ω2
Mp = e

or in terms of percentage is
↓ ≃ ωϑ
1→ω2
e ↗ 100%

27 / 31
Transient Quantities for Second-order systems

The settling time ts can be obtained by considering the envelope of the

response, c(t).
→ωεn t
↭ Specifically, note that 1 ± e≃
1↓ω2
contains the response of c(t) at all
time t.
↭ The speed of decay of the envelope
curves depends on the value of the
time constant T = ωε1n
↭ Using results of First-order system,
ts for ±2% and ±5% tolerant band
are approximated by
4 3
ts = 4T = , and ts = 3T =
ϖεn ϖεn

28 / 31
Transient Quantities for Second-order systems

Note that the settling time is inversely proportional to the product of the
damping ratio and the undamped natural frequency of the system.
Since the value of ϖ is usually determined from the requirement of permissible
maximum overshoot, ts can be determined by εn .
This means that the duration of the transient period may be varied, without
changing the maximum overshoot.
From the preceding analysis, it is evident that εn must be large for rapid
response.
To limit the maximum overshoot and to make the settling time small, the
damping ratio ϖ should not be too small.
Note that if ϖ is between 0.4 and 0.8, then the maximum percent overshoot for
the step response is between 25% and 2.5%, see figure shown two slides before.

29 / 31
Summary of transient response

Consider the unit-step response of the second order system defined by

εn2
G(s) =
s2 + 2ϖεn s + εn2
" εd
define εd = εn 1 ↑ ϖ 2 and ϱ = tan↓1 ωε n
ϑ↓ϖ
Rise time, tr = εd
Peak time, tp = εϑd
↓ ≃ ωϑ
1→ω2
Maximum Overshoot, Mp = e ↗ 100%
3 4
Settling time, ts = 3T = ωεn
, for 5% criterion and ts = ωεn
for 2% criterion

30 / 31
Responses of Higher-order systems
We have discussed responses of First and Second order systems. What about higher
order systems?
The location of the poles of the transfer function in the s-plane has great
e#ects on the transient response of the system. So it is important to sort out
the poles that have dominant e#ect on the transient response, known as the
dominant poles.
The poles that are close to the imaginary axis in the left half of the s-plane
give rise to transient responses that will decay relatively slowly, whereas the
poles that are far away from the imaginary axis (relative to the dominant
poles) correspond to fast decaying time responses.
10
To see this e#ect, take G(s) = (s+1)(s+10)
, the pole locations and system
responses are given below:

31 / 31
 ME2142
Chapter 6 Steady State Analysis

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

September 5, 2022

1/7
Steady State Analysis of Systems

Consider the feedback control system shown above where the loop transfer
function G(s)H(s) is written as

K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)

G(s)H(s) =
sN (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)

The system is called type 0, type 1, type 2, type 3, etc. if

N = 0, N = 1, N = 2, N = 3, etc. respectively.
This chapter shows the relationship between the error signal E(s) and system
type (N ) under di!erent R(s).
Note that the error is for a closed-loop system with system type defined by the
openloop transfer function (or the loop transfer function)

2/7
Error Analysis

Note that
C(s) G(s) E(s) 1
= and =
R(s) 1 + G(s)H(s) C(s) G(s)
Hence, the error signal is given by
R(s)
E(s) =
1 + G(s)H(s)

The steady state error can be found using the final-value theorem :
sR(s)
ess = lim e(t) = lim sE(s) = lim
t→↑ s=0 s=0 1 + G(s)H(s)

The final value theorem can be applied only if all the roots of the denominator
sR(s)
polynomial of 1+G(s)H(s) have negative real parts.

3/7
Error Analysis
Consider the case where R(s) = 1s , or r(t) = 1 for t → 0. Then

s 1s 1 1
ess = lim sE(s) = lim = =
s=0 s=0 1 + G(s)H(s) 1 + G(0)H(0) 1 + Kp
where Kp := lims=0 G(s)H(s) = G(0)H(0) is the step-error constant.
For Type 0 system,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Kp = lim =K
s=0 s0 (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)

For Type 1 system or higher,

K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Kp = lim =↑
s=0 sN (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
In summary, for step input
1
ess = , for Type 0 system
1 + Kp
ess = 0 for Type 1 system or higher

4/7
Error Analysis
1
Consider the case where R(s) = s2
, or r(t) = t for t → 0. Then
s s12 1 1
ess = lim sE(s) = lim = lim =
s=0 s=0 1 + G(s)H(s) s=0 s + sG(0)H(0) Kv
where Kv := lims=0 sG(s)H(s) is the ramp-error constant
For Type 0 system,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Kv = lim s =0
s=0 s0 (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
For Type 1 system,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Kv = lim s =K
s=0 s1 (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
For Type 2 system or higher,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Kv = lim s =↑
s=0 sN (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
In summary, for ramp input
ess = ↑, for Type 0 system
1
ess = for Type 1 system
K
ess = 0 for Type 2 system or higher
5/7
Error Analysis
1 t2
Consider the case where R(s) = s3
, or r(t) = 2
for t → 0. Then
s s13 1 1
ess = lim sE(s) = lim = lim =
s=0 s=0 1 + G(s)H(s) s=0 s2 + s2 G(0)H(0) Ka
where Ka := lims=0 s2 G(s)H(s) is the parabolic-error constant
For Type 0 and Type 1 system,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Ka = lim s2 N =0
s=0 s (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
For Type 2 system,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Ka = lim s2 =K
s=0 s2 (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
For Type 3 system or higher,
K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
Ka = lim s2 N =↑
s=0 s (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
In summary, for parabolic input
ess = ↑, for Type 0 and Type 1 system
1
ess = for Type 2 system
K
ess = 0 for Type 3 system or higher
6/7
Summary
Given a closed-loop Feedback System

with the openloop or loop TF given by

K(ωa s + 1)(ωb s + 1) · · · (ωm s + 1)
G(s)H(s) =
sN (ω1 s + 1)(ω2 s + 1) · · · (ωp s + 1)
The error table for di!erent system type and input function are
Type of system to step input to ramp input to parabolic input
1
Type 0 1+K
↑ ↑
1
Type 1 0 K
↑
1
Type 2 0 0 K
Type 3 0 0 0

7/7
 ME2142
Chapter 7 Stability Analysis

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

February 8, 2023

1 / 19
System Stability

This chapter is concern with the stability of a system in the form of

Note that the above block diagram representation can be that of an open-loop or
closed-loop system.
A few basic definitions are needed before the concept of stability can be
introduced.
Definition: A signal, y(t), is said to be bounded if there exists a finite constant
µ (µ < →) such that
|y(t)| < µ for all t ↑ 0
Examples of bounded signals include (t), sinωt, e→t .
Examples of unbounded signals are t, tsinωt, et .

2 / 19
Stability Analysis

Intuitively, a stable system is one whose response does not “blow up” during
its operation.
Another intuitive understanding is the output of a system have “finite energy”
if the input to the system has only “finite energy”.
Since “energy” is related to the size (magnitude) of a signal, this leads to the
following stability concept.

Definition
A system is Bounded-Input Bounded-Output (BIBO) stable if whenever the input
signal is bounded, so is the output signal.

In other words, if a stable system is driven by an input that doesn’t blow up,
then the output doesn’t blow up either.
On the other hand, if a system is unstable, then some bounded input produces
an unbounded output.

3 / 19
Stability Analysis

What are the conditions for a system to be BIBO stable?

Again, a few definitions are needed before the conditions can be stated.
Open left-half of s-plane: The region to the left of the imaginary axis in the
s-plane excluding the imaginary axis.
Closed left-half of s-plane: The region to the left of the imaginary axis in the
s-plane including the imaginary axis.
Open right-half of s-place: The region to the right of the imaginary axis in the
s-plane excluding the imaginary axis.
Closed right-half of s-place: The region to the right of the imaginary axis in
the s-plane including the imaginary axis.

Theorem
The system G(s) is BIBO stable if and only if all its poles are in the open left-half
s-plane, or equivalently, the poles have negative real parts.

4 / 19
Stability Analysis

To better appreciate the definition, let’s consider a simple case where the poles of
the system are all distinct and that the system is driven by a unit step function (i.e.,
r(t) = (t)).
In that case, the output C(s) is
N (s) 1
C(s) =
D(s) s
or, by partial fraction, we get
n
1 ! ai
C(s) = +
s i=1 s + pi

↓ c(t) = 1 + ai e→pi t
What about other input functions?

5 / 19
Stability Analysis

Problem: Given a system with transfer function

s2 + 2s + 1
G(s) =
s3 + 3s2 + 5s + 7

We wish to know if this system is BIBO stable.

To check stability one must factor the denominator polynomial and find its
roots, which are the poles of G(s). If all its poles are in the open left half of
the s-plane (i.e. have negative real parts), then the system is stable. Otherwise
the system is unstable.
Hence, the determination of stability is converted into a problem of finding the
roots of the denominator polynomial
Possible for low order polynomials, need a computer for higher-order
polynomials.

6 / 19
Routh Table

Question: Can we determine the stability of G(s) without computing the

poles of G(s)?
Answer: Yes, the approach is known as Routh’s stability criterion.
This approach is general and is applicable to any general polynomial.
When the polynomial is the denominator polynomial of G(s), the approach
also determines the stability of G(s).
To see this, consider the following system
N (s)
G(s) =
D(s)

Note that G(s) can be the TF of an openloop or closed-loop system.

The approach relies on the interpretation of the Routh Table.

7 / 19
Routh Table
Forming the Routh Table
(1) Suppose D(s), the denominator polynomial of G(s) is
a0 sn + a1 sn→1 + · · · + an = 0
(2) If any of the coe!cients of this polynomial is zero, or if all the coe!cients are
not of the same sign, then there is a root or roots of polynomial D(s) that is
either imaginary or have positive real part (i.e. in the closed right half of the
s-plane). Therefore, in such a case, the system G(s) is not stable. (Thus for
BIBO stability of G(s) we must have no ai equal to zero and all ai must be of
the same sign).
Construct the Routh Table as follow:

sn a0 a2 a4 a6 ··· a1 a2 ↔ a0 a3
n→1 b1 = ,
s a1 a3 a5 a7 ··· a1
sn→2 b1 b2 b3 b4 ··· a1 a4 ↔ a0 a5
b2 = ,
sn→3 c1 c2 c3 c4 ··· a1
.. .. a1 a6 ↔ a0 a7
. . b3 = ,
a1
s2 e1 e2
s1 f1 ..
.
s0 g1
8 / 19
Routh Table

sn a0 a2 a4 a6 ··· b1 a3 ↔ a1 b2
c1 = ,
sn→1 a1 a3 a5 a7 ··· b1
sn→2 b1 b2 b3 b4 ··· b1 a5 ↔ a1 b3
c2 = ,
sn→3 c1 c2 c3 c4 ··· b1
.. .. b1 a7 ↔ a1 b4
. . c3 = ,
b1
s2 e1 e2
s1 f1 ..
.
s0 g1
The process is continued until the n-th row has been completed.
Routh’s stability criterion states that the number of roots of D(s) (or
equivalently, number of poles of G(s)) with positive real part (i.e. in the open
right half s-plane) equals the number of sign changes in the first column of the
Routh table.
Hence, the system G(s) is BIBO stable if all the entries in the first column of
the Routh table have no sign change (i.e. all the entries in the first column are
all positive or all negative).

9 / 19
Routh Table
s2 +2s+1
As an example, consider the case where G(s) = s3 →4s2 +s+6
The denominator polynomial,D(s), is

s3 ↔ 4s2 + s + 6

Since all the coe!cients of D(s) are not of the same sign, G(s) is not BIBO
stable.
We can verify this using the Routh Table.
s3 1 1
s2 -4 6
(→4)(1)→(6)(1)
s1 →4
= 2.5 0
s0 6 0
2 sign changes in the first column ↓ two roots in the open right half s-plane.
Equivalently, G(s) has 2 poles with positive real parts. (The poles are
s = 2, 3, ↔1)

10 / 19
Routh Table

s+1
G(s) =
s4 + 2s3 + 5s2 + 4s + 9

The denominator polynomial,D(s), is s4 + 2s3 + 5s2 + 4s + 9.

Since all the coe!cients of D(s) have the same sign, no conclusion can be
drawn. Proceed to Routh Table.
s4 1 5 9
s3 2 4 0
s2 3 9 0
s1 -2 0
s0 9
2 sign changes in the first column ↓ two roots in the open right half s-plane.
Equivalently, G(s) has 2 poles with positive real parts.
Poles are ↔1.2381 ± 1.5148 and 0.2381 ± 1.5148

11 / 19
Routh Table
Special Cases: Case 1: The first element in any one row in the Routh table is zero,
but not all elements of that row are zero. This is handled by the ε - method.
s2 +1
Consider G(s) = s4 +s3 +2s2 +2s+3
. Char eqn: s4 + s3 + 2s2 + 2s + 3 = 0

s4 1 2 3
s3 1 2 0
s2 0 3

Since the first element in the s2 row is zero, the elements in the next row
would all be infinite.
Remedy: Replace the zero by a small number ε, and proceed as usual;
s4 1 2 3
s3 1 2 0
s2 ε 3
→3
s1 2ω→3
ω
= ω
0
s0 3
There are 2 sign changes for ε > 0 or ε < 0.
Conclusion: D(s) has two roots in the open right half s-plane.

12 / 19
Routh Table

Special Cases: Case 2: All elements in one row of the Routh table are zero. This
indicates that one or more of the following conditions are true:
(i) G(s) has at least one pair of poles which are purely real with equal magnitude
but opposite signs.
(ii) G(s) has one or more pairs of purely imaginary poles.
(iii) G(s) has pairs of complex conjugate poles forming symmetry about the origin
of the s-plane; e.g. s = ↔1 ± j1, s = 1 ± j1.
Thus G(s) is an unstable system in these cases.
Form the auxiliary equation A(s) = 0 from the coe!cients of the row just
above the row of zeros in the Routh table. This auxiliary equation is always an
even polynomial, i.e., only even powers of s appear.
The roots of the auxiliary equation are poles of G(s) which correspond to the
conditions (i), (ii) or (iii) listed earlier.

13 / 19
Routh Table
The nature of poles of G(s) in this case is handled by the auxiliary equation
method.
dA(s)
(1) Take the derivative of the auxiliary equation with respect to s, i.e. ds
= 0.
(2) Replace the row of zeros with the coe!cients of this equation and proceed with
the Routh table as usual.
(3) Interpret the resulting table as before.
s+1
For example, consider G(s) = s5 +4s4 +8s3 +8s2 +7s+4
.

s5 1 8 7
s4 4 8 4
s3 6 6 0
s2 4 4 0
s1 0 0 0

Since all the elements of the s1 row are zero, G(s) is an unstable system.
Forming the auxillary equation, A(s) = 4s2 + 4 = 0.
Solving the auxiliary equation we get two roots s = ±j which are the two poles
of G(s) on the imaginary axis.

14 / 19
Routh Table
dA(s)
To proceed further, take the derivative of A(s) with respect to s; ds
= 8s
1
Replace the row of zeros (i.e. the s row) with the coe!cients of this equation
and proceed with the Routh table as usual.
s5 1 8 7
s4 4 8 4
s3 6 6 0
s2 4 4 0
s1 8 0
s0 4
There are no sign changes in the entries of the first column of the Routh table.
Thus among the remaining poles of G(s) (i.e. excluding the poles obtained as
solutions to the auxiliary equation) there is none in the open right half of the
s-plane.
Therefore G(s) is an unstable system with
(i) 2 poles on the imaginary axis (obtained as solutions of A(s) = 0),
(ii) no poles with positive real part,
(iii) 3 poles with negative real part.
Indeed, the poles of G(s) are ±1j, ↔1, ↔1.5 ± 1.3229j.
15 / 19
Routh Table

Example 1: Application of Routh Table for Stability of Closed-loop Control Systems

Find the range of K for which the system is stable.

The closed loop TF is
K
s4 + 3s3 + 3s2 + 2s + K
Hence, char. eqn. is
s4 + 3s3 + 3s2 + 2s + K = 0

16 / 19
Routh Table

Forming the Routh Table, we have

s4 1 3 K
s3 3 2 0
s2 7
3
K 0
s1 2 ↔ 9K
7
0
s0 K
For closed loop stability, all entries of the first column of the Routh table must
be positive. Therefore we must have
9K
2↔ > 0 and K > 0
7
Combining, we have
14
0<K<
9

17 / 19
Routh Table

Example 2: Application of Routh Table for Stability of Closed-loop Control Systems

Find the range of ϑ for which the system is stable.

The closed loop TF is
s+ϑ
s3 + 3s2 + 3s + ϑ
The poles of the CLTF are given by the roots of

s3 + 3s2 + 3s + ϑ

18 / 19
Routh Table

Forming the Routh Table, we have

s3 1 3
s2 3 ϑ
s1 9→ε
3
0
s0 ϑ 0
For closed loop stability, all entries of the first column of the Routh table must
be positive. Therefore we must have
9↔ϑ
> 0 and ϑ > 0
3
Hence, for stability, we need
0<ϑ<9

19 / 19
 ME2142
Chapter 8 Time Domain Design of Control System

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

February 17, 2022

1 / 26
Time Domain Design
Consider a feedback system shown below

Controller System

Sensor

Design a controller such that the following performance specifications are satisfied:
(i) closed loop system must be BIBO stable.
(ii) the steady-state error ess < some fixed value.
(iii) the maximum percentage overshoot Mp < some fixed value.
(iv) the rise time tr < some fixed value.
(v) the settling time ts < some fixed value.
To satisfy the steady-state error requirement, the controller structure must be
chosen such that the resulting loop transfer function has the required “type”.
The other performance criteria may be satisfied by choosing the correct values
for the parameters of the controller so that the damping ratio and natural
frequency of the closed loop transfer function have the required values.
2 / 26
Time Domain Design
The process of designing a controller generally involves many steps and tradeo↵s. A
typical scenario is as follows:
1 Study the system to be controlled and decide what type of sensors and
actuators will be used and where they will be placed.
2 Model the system to be controlled.
3 Simplify the model if necessary so that it is tractable.
4 Analyze the resulting model and determine its properties.
5 Decide on performance specifications.
6 Decide on the type of the controller to be used.
7 Design a controller to meet the performance specifications, if possible; if not,
modify the specifications or go back to step 6 to choose a di↵erent type of
controller.
8 Simulate the resulting closed loop controlled system, either on a computer or
in a pilot plant.
9 Repeat from step 1 if necessary.
10 Choose hardware and software and implement the controller.
11 Tune the controller on-line if necessary to get good performance.
3 / 26
Time Domain Design
We will illustrate the design process using the following examples.
Example 1: Position Control of a DC Motor Using Proportional Controller
dc motor
Reference position controller
E(s) U(s)
in volts r(t) + km
KP C(s)
R(s) - s(τ m s + 1)

k1
Motor position
in volts potentiometer

Suppose we are given the plant and wish to design a controller that satisfies the
following criteria:
(i) closed loop system must be BIBO stable.
(ii) the steady-state error ess (step) = 0.
We begin with the simplest controller, the P -controller with the T.F. of
U (s)
= Kp
E(s)

4 / 26
P controller

With P controller, the loop T.F. (or openloop T.F.)is

K p km k1
s(⌧m s + 1)

The openloop T.F. is type 1, hence, steady-state error for the closed-loop
system is ess (step) = 0.
The closed-loop transfer function is
Kp km
C(s) K p km ⌧m
= = K km k1
R(s) ⌧ m s 2 + s + K p km k1 s2 + s
+ p⌧m
⌧m

If we choose the controller gain Kp > 0 (and since ⌧m , k1 , km > 0), the poles of
the closed loop system are in the open left half of the s-plane and the closed
loop system is BIBO stable.
Both design conditions are satisfied.

5 / 26
P controller
The above specifications, however, do not include the performances on
transient characteristics.
Suppose we wish to include an additional specification of having
(iii) the maximum overshoot Mp (step) = 10%.
Then, from the closed-loop characteristic equation of
s K p km k1
s2 + + =0
⌧m ⌧m
we have r
K p km k1 K p km k1
!n2= ) !n =
⌧m ⌧m
1 1
2⇠!n = )⇠= p
⌧m 2 K p km k1 ⌧ m
We can, therefore, choose a value for Kp to achieve the required damping ratio
for the closed-loop system (from Mp and ⇠ graph, Mp  10% for ⇠ > 0.6.).
Note that ⇠ is inversely related to Kp : ⇠ # as Kp ", and system is sluggish if
Kp is small!

6 / 26
P controller

Now, suppose the system is to follow a moving target - typically represented by

a ramp input.
Need system to have low or no errors to ramp input. What shall we do?
If the input is a unit-ramp then the ramp error constant is
K p km k1
Kv = lims!0 s = K p km k1
s(⌧m s + 1)

The steady state error is equal to

1 1
ess (ramp) = =
Kv K p km k1

Since ess (ramp) is inversely proportional to the value of Kp . To reduce

ess (ramp), Kp must be large. However if Kp is large, ⇠ is small, which in turn
implies that Mp is large. This is one limitation of the P -controller.

7 / 26
An example

we show an example of the above conflicting requirements

Let ⌧m = 0.0206, k1 = 1 and km = 8.247 (realistic numbers for a DC motor).
The closed-loop transfer function is
C(s) 400Kp
= 2
R(s) s + 48.5s + 400Kp
p
Hence, !n = 400Kp and ⇠ = p48.5
2 400Kp
1
Also, ess (step)=0, ess (ramp)= 8.247Kp

If Kp = 2.94, then ⇠ = 0.707 which is good damping. But ess (ramp)=0.0412,

larger that 4%.
If Kp = 100, then ess (ramp)=0.0012125 which is very good. But ⇠ = 0.12125
will lead to Mp > 90%

8 / 26
An example

The responses to step and ramp inputs are given below (Y - c(t), X - time):

9 / 26
P D controller
We now use a di↵erent controller, known as the P D controller.

The P D-controller has a T.F. of

U (s)
= Kp + KD s
E(s)
and is a sum of an amplifier with gain Kp and a di↵erentiator with gain KD .
With this controller, the loop transfer function is of type 1 given by
(Kp + KD s)km k1
s(⌧m s + 1)
and hence, the steady-state error ess (step) = 0.
10 / 26
P D controller
The closed-loop transfer function is
(Kp +KD s)km
C(s) (Kp + KD s)km ⌧m
= = K km k1
R(s) ⌧m s2 + (1 + KD km k1 )s + Kp km k1 s2 + 1+KD km k1
s + p⌧m
⌧m

If Kp , KD > 0 (and since ⌧m , k1 , km > 0), the poles of the closed loop system
are in the open left half of the s-plane and the closed loop system is BIBO
stable.
From the closed-loop characteristic equation
1 + K D km k1 K p km k1
s2 + s+ =0
⌧m ⌧m
we have r
K p km k1 K p km k1
!n2 = ) !n =
⌧m ⌧m
1 + K D km k1 1 + K D km k1
2⇠!n = )⇠= p
⌧m 2 K p km k1 ⌧ m

11 / 26
P D controller

The ramp error constant,

(Kp + KD s)km k1
Kv = lim s = K p km k1
s)0 s(⌧m s + 1)

The steady state error

1 1
ess (ramp) = =
Kv K p km k1
which is inversely proportional to Kp . Hence, ess (ramp) can be kept small by
making Kp large.
What about ⇠?
⇠ can be chosen as the required damping ratio by the choice of KD .
Thus, both the Mp and ess can be achieved using P D controller.

12 / 26
PD example
Using the previous example. We have

DC
motor

potentiometer

The closed-loop transfer function is

C(s) 400(Kp + KD s)
= 2
R(s) s + (48.5 + 400KD )s + 400Kp
p 48.5+400KD
Hence, !n = 400Kp and ⇠ = p
2 400Kp
1
Also, ess (step)=0, ess (ramp)= 8.247Kp

If Kp = 100, then ess (ramp)=0.0012125 which is very good for tracking ramp.
Choose KD = 0.8788, then ⇠ = 1 which is critically damped with no overshoot.
13 / 26
PD example
Based on the example, the step and ramp responses are

Comments: A PD-controller is used to improve the transient response of a

closed-loop system as follows:
1 overshoot is reduced by changing the damping ratio.
2 rise time is made better (faster).
3 has no additional e↵ect on steady state error as compared to P-controller.
4 adds a zero to the closed-loop transfer function.

14 / 26
Speed Feedback

The previous approach was done using a P D controller. The same e↵ect can
be achieved using a P controller with speed feedback.
The speed feedback design has the added advantage of not introducing a zero
to the CLTF.

DC
motor
tachometer

potentiometer

With this controller, the loop transfer function is of type 1 given by

Kp km (k1 +kt s)
s(⌧m s+1)
. Hence, ess (step) = 0.

15 / 26
Speed Feedback

The closed-loop transfer function is

Kp km
C(s) K p km ⌧m
= = 1+Kp km k1 Kp k m k1
R(s) ⌧m s2 + (1 + Kp km kt )s + Kp km k1 s2 + s +
⌧m ⌧m

q
Kp k m k1 1+K k k
Hence, !n = ⌧m
and ⇠ = p p m t
2 K p k m k 1 ⌧m

K k (k1 +Kt s) 1 1
Also, Kv = lims)0 s p m
s⌧m +1
= Kp km k1 , and ess (ramp)= Kv
= Kp km k 1
.
Hence, Kp " leads to ess (ramp) #.
Choose Kt to adjust ⇠.

16 / 26
Speed Feedback

Using the same example before with k1 = 1, ⌧m = 0.0206, km = 8.247, we get

p 48.5 + 400Kp kt 1
!n = 400Kp , ⇠= p , ess (ramp) =
2 400Kp 8.247Kp

Choose Kp = 100, then ess (ramp)=0.0012125 (same as PD controller).

Choose kt = 0.008788, we have ⇠ = 1, critically damped.
In summary, a rate feedback controller is used to improve the transient response of
the DC motor system:
overshoot is reduced by changing the damping ratio.
rise time is made faster (by choice of Kp .
has no additional e↵ect on steady state error as compared to proportional
controller.

17 / 26
P I controller
The P and P D can make ess (ramp) small but not zero. If we want ess (ramp)=0, a
P I controller can be used.

DC
motor

potentiometer

The P I-controller has a T.F. of

U (s) KI
= Kp +
E(s) s
and is a sum of an amplifier with gain Kp and an integrator with gain KI .
With this controller, the loop transfer function is of type 2 given by
(Kp s + KI )km k1
s2 (⌧m s + 1)
and hence, the steady-state error ess (step) = ess (ramp)=0.
18 / 26
P I controller
The closed-loop transfer function is
KI
C(s) (Kp s + KI )km (s + Kp
)km
= = KI k m k1
R(s) ⌧ m s 3 + s 2 + K p km k1 s + K I km k1 ⌧m 3
Kp
s + 1 2
Kp
s + km k1 s + Kp

The CLTF function is a third-order system and the design becomes

complicated.
KI
One method is to use an approximation: choose Kp , KI such that Kp
' 0. In
that case, the CLTF
C(s) km
' ⌧m 2 1
R(s) Kp
s + Kp
s + km k1

Choose Kp > 0 (since km , k1 > 0 so that the approximation above is BIBO

stable. Need to check that the original system is stable too!)
Use the same design steps as before, we have the char. equation being
⌧m 2 1
s + s + km k1 = 0
Kp Kp
q
Kp k m k1 1 1
Hence, !n = ⌧m
, 2⇠!n = ⌧m
or ⇠ = p
2 K p k m k 1 ⌧m

19 / 26
PI example
Using the previous example. We have

DC
motor

potentiometer

The closed-loop transfer function is

KI
C(s) 400(s + Kp
)
= 1 3 48.5 2 KI
R(s) Kp
s + Kp
s + 400s + 400 K p

KI
choose KP
= 0.1, then the CLTF approximation is

C(s) 400
' 1 2 48.5
R(s) Kp
s + Kp
s + 400

20 / 26
PI example

we have
p 48.5
!n = 400Kp , and 2⇠!n = 48.5 ! ⇠ = p
2 400Kp
ess (step)=ess (ramp)=0.
If Kp = 100, KI = 10 and ⇠ = 0.12125 (low damping).
If Kp = 2, KI = 0.2 and ⇠ = 0.857 (good damping).

Note the initial error to ramp!

21 / 26
PI example
Note that there is a large initial error to ramp for the case of Kp = 2, KI = 0.2.
Note that the error get close to 0 only when t > 50 seconds.

A PI-controller is used to improve the response of a closed-loop system as follows:

type of the system is increased to 2. So ess (step) = 0 and also ess (ramp) = 0.
overshoot is reduced by changing the damping ratio.
the response is generally sluggish.
characteristic equation is 3rd order which may cause stability problems.

22 / 26
P ID controller
We now introduce the PID controller for the DC motor system.

DC
motor

potentiometer

The P ID-controller has a T.F. of

U (s) KI
= Kp + KD s +
E(s) s
and is a sum of an amplifier with gain Kp , a derivative term with gain KD and
an integrator with gain KI .
With this controller, the loop transfer function is of type 2 given by
(KD s2 + Kp s + KI )km k1
s2 (⌧m s + 1)
and hence, the steady-state error ess (step) = ess (ramp)=0.
23 / 26
P ID controller

The CLTF is
C(s) (KD s2 + Kp s + KI )km
=
R(s) ⌧m s + (1 + KD km k1 )s2 + Kp km k1 s + KI km k1
3

(K
Kp
D 2
s +s+ KI
Kp
)km
= ⌧m 3 (1+KD km k1 ) 2 KI
Kp
s + Kp
s + km k1 s + km k1 K p

KI
If we choose the controller gain KI , Kp such that the ratio K p
' 0 , then the
closed-loop transfer function may be approximated by a second order T.F.

(K
Kp
D
s + 1)km
= ⌧m 2 (1+KD km k1 )
Kp
s + Kp
s + km k1

If we choose the controller gain Kp , KD > 0 (and since ⌧m , km , k1 > 0) then

the poles of the CLTF are in the open left half of the s-plane and so the closed
loop system is BIBO stable.

24 / 26
P ID controller
The characteristic equation of the CLTF is
⌧m 2 (1 + KD km k1 )
s + s + km k1 = 0
Kp Kp

Hence, r
K p km k1
!n = ,
⌧m
1 + K D km k1 1 + K D km k1
2⇠!n = )⇠= p
⌧m 2 K p km k1 ⌧ m
The presence of the derivative term helps in faster rise time and reduced
overshoot compared to a PI-controller.
While PID controller can combine the advantages of PD and PI controllers,
however, the design is difficult and relies heavily on approximation,
assumptions and experience .
The approximation adopted here: First design a PD-controller by choosing
appropriate Kp , KD > 0 to achieve the required damping ratio for the
closed-loop system. Thus the overshoot and oscillations in the system response
can be controlled. Next choose a positive number ✏ close to zero such that
KI = ✏Kp . This completes the design.
25 / 26
PID example

A PID-controller is now discussed for the same example.

DC
motor

potentiometer

First choose Kp = 100, KD = 0.8788 as we did for a PD-controller and this

gives critically damped response.
KI
If we choose the controller gain such that Kp
= 0.1, then KI = 10 as we did
for the PI-controller.
Also, ess (step)=0 and ess (ramp)=0.

26 / 26
 ME2142
Chapter 9 Root Locus Analysis

Chong Jin Ong

Department of Mechanical Engineering,

National University of Singapore

September 12, 2022

1 / 20
8.1 Introduction

This method of plotting is developed in 1948 by W.R. Evans.

Depends on the relationship between locations of poles and time responses.
Used as a design tool for the selection of a particular parameter value.
As a design tool, we assume a controller configuration and search for one
parameter (usually the gain) hoping that a good response can be obtained.
In recent decades, the plotting of the loci is replaced by software packages.
This fact brings into question the need for this method.
My inclusion of this topic is based mainly on the insight/understanding that
the analysis brings. Techniques learnt can be used to address, for example
design of systems with one unknown parameter. For example, in PID control
where values of Kp and KD are determined by transient specifications, value of
KI can be determined from root-locus method.

2 / 20
Example 1

Consider the control system as represented below.

From the diagram above, the closed-loop transfer function of the system is
Y (s) K
= 2 (8.1.1)
R(s) s + 4s + K
Therefore, the closed-loop characteristic equation is

s2 + 4s + K = 0 (8.1.2)

3 / 20
Example 1
The roots of the characteristic equation are
↑
s = →2 ± 4 → K for 0<K<4
s = →2 for K = 4
↑
and s = →2 ± j K → 4 for K>4

4 / 20
8.2 Basis of Root Loci
Consider a general feedback system as shown below.

The closed-loop transfer function is

KG(s)
(8.2.1)
1 + KG(s)H(s)
with its characteristic equation being

1 + KG(s)H(s) = 0 (8.2.2)

Equation (8.2.2) can be expressed as

KG(s)H(s) = →1 (8.2.3)

5 / 20
8.2 Basis of Root Loci

Hence, values of s for which (8.2.3) is satisfied are the roots of the characteristic
equation, or the closed-loop poles. The above equation can also be seen as two
conditions since s is a complex quantity. These two conditions are
the Magnitude condition
|KG(s)H(s)| = 1 (8.2.4)
and the Angle condition

↭(KG(s)H(s)) = ↭(G(s)H(s)) = (2n + 1)ω (8.2.5)

Based on these two conditions, some properties of root locus can be derived.

6 / 20
Magnitude and Angle Criteria

In general, G(s) and H(s) are usually rational functions of s. That is

K(s + z1 )(s + z2 ) · · · (s + zm )
KG(s)H(s) = (8.2.6)
(s + p1 )(s + p2 ) · · · (s + pn )
where →zi and →pi are the zeros and poles of the open-loop transfer function.
In this form, the magnitude and angle conditions (8.2.4) and (8.2.5) can be
rewritten as
Magnitude Condition
|s + z1 ||s + z2 | · · · |s + zm | 1
= (8.2.7)
|s + p1 ||s + p2 | · · · |s + pn | |K|

Angle Condition
m
! n
!
↭(G(s)H(s)) = ↭(s + zi ) → ↭(s + pi ) = (2n + 1)180→ (8.2.8)
i=1 i=1

7 / 20
Magnitude and Angle Criteria
To see this, for example, we have
↑ ↑
(1 + j)(2 + j) 2↭(45→ ) 5↭(26.56→ )
= ↑
(3 + j)(4 + 3j) 10↭(18.43→ )5↭(36.86→ )
1
= ↭(45→ + 26.56→ → 18.43→ → 36.86→ )
5
1
= ↭(16.27→ )
5
There is also a clear geometrical interpretation to the two conditions. Consider the
example of
K(s + z1 )
KG(s)H(s) =
(s + p1 )(s + p2 )(s + p3 )

8 / 20
Magnitude and Angle Criteria

The Angle condition is

↭(G(s)H(s)) = ε1 → ϑ1 → ϑ2 → ϑ3 = (2n + 1)180→

and the Magnitude condition is

|s + z1 | 1
|G(s)H(s)| = =
|s + p1 ||s + p2 ||s + p3 | |K|

9 / 20
Useful facts for designing using Root Loci

Root locus has n branches -

N (s)
1 + KG(s)H(s) = 0 ↓ 1 + K = 0 ↓ D(s) + KN (s) = 0
D(s)

(s + p1 )(· · · )(s + pn ) + K(s + z1 )(· · · )(s + zm ) = 0 (1)

Since, n ↔ m, D(s) + KN (s) = 0 is an nth -order equation with n roots.

The starting point of the root loci are the open-loop poles. This follows from
(1) when K = 0:
(s + p1 )(· · · )(s + pn ) = 0
The ending point of the root loci are the open-loop zeros. Similarly from (1)
when K = ↗:
(s + z1 )(· · · )(s + zm ) = 0.
What happens when n ↘= m? n → m of the branches go to ↗ along some
asymptotes.

10 / 20
Useful facts for designing using Root Loci

The above fact is useful for designing controller.

If we wish to have a faster settling time or response time, how do we

accomplish this?
One possibility is to use a di!erent controller and put one zero to the left.

How will the root locus look like?

11 / 20
8.4 The root-locus design method
In the above sections, the underlying principles of the various properties of plotting
root-locus diagram have been discussed. We now show how the root-locus diagram
can be used in the design process.

Root-locus method is particularly useful when the specifications are given in time
domain quantities like damping ratio, natural frequency, maximum overshoot, rise
time and settling time.

Recall that for a standard second-order system, the maximum overshoot condition
depends on ϖ, and the region for a max. value of ϖ corresponds to sector on the
s-plane as shown.

12 / 20
8.4 The root-locus design method
Settling time conditions depends on the quantity ϖϱn . For a complex pole, this
quantity corresponds to the real part of the pole. Hence, the region looks like

The above provides good guidelines as to where the closed-loop poles should be
located. They are exact for the special 2nd order systems but can be used as
approximations for higher order system. In general, a proper design will require
several iterations of computer simulations and based on the computer response,
make changes based on the response. The following is a typical design procedure.

13 / 20
8.4 The root-locus design method
Step 1: Choose a configuration and a compensator with one open parameter k.

Step 2: Compute the overall transfer function and then find the range of k for the
system to be stable and to meet steady-state specifications. If no such k exists, go
back to step 1.

Step 3: Plot the root loci that yield the poles of the overall system as a function of
the parameter.

Step 4: Find the desired pole region from the specification on overshoot and
settling time requirements.

Step 5: Find the range of k in which the root loci lie inside the desired pole region.
If no such k exists, go to Step 1 and choose a more complicated compensator or a
di!erent configuration.

Step 6: Find the range of k that meets 2 and 5. If no such k exists, go to Step 1.

Step 7: From the range of k in Step 6, find a k to meet the remaining

specifications, such as the rise time or the constraint on the actuating signal. This
step may require computer simulation of the system.
14 / 20
Examples of root locus design

Example 1. Suppose we are given a plant with a transfer function

s+4
G(s) =
(s + 2)(s → 1)
The design task is to design an overall system that meet the following specifications:
Position error due to step ≃ 10%
Overshoot ≃ 5%
Settling time ≃ 4.5 seconds
Rise time as small as possible
We try the unity feedback configuration shown below

15 / 20
Examples of root locus design

The overall transfer function is

k(s + 4)
G0 (s) =
s2 + (k + 1)s + 4k → 2

The conditions for G0 (s) to be stable are (can be verified using Routh table)

4k → 2 > 0
k+1>0

which implies that k > 0.5. For range of k to have position error less than 10%, we
note that
1
ess (t) = lim sess (s) =
s↑0 1 + G(0)H(0)
4k(0.25s+1)
Since lims=0 G(s)H(s) = lims=0 2(0.5s+1)(s↓1)
= →2k Hence,
1 1
ess (t) = | 1↓2k | = | 2k↓1 | ≃ 0.1 which implies that k ↔ 5.5

16 / 20
Examples of root locus design
Plot the root locus using the properties of root locus or through some kind of
computer program like MATLAB.

Find the ranges of k that meet the specifications on overshoot and settling time.
To summarize
k > 0.5: stable
k > 5.5: meet specification (1).
k > 5 or 1 > k > 0.5 meet specification (2)
k > 1: meet specification (3).
Clearly, in order to meet (1), (2) and (3), k must be larger than 5.5.

17 / 20
Examples of root locus design

The last step of the design is to find a k in k > 5.5 such that the system has
the smallest rise time. To achieve this, we choose k such that the closed loop
poles are farthest away from the origin.
For this reason, we choose k = 13.3, the breakin location.
It is important to stress that the desired pole region is developed for the
standard second order system. The G0 (s) system above is not such a transfer
function. Therefore, it is advisable to simulate the resulting system to verify
that the constraints are satisfied.

18 / 20
Examples of root locus design

Example 2. In this example, we have chosen a particular controller (PI controller)

and the overall system is of the following form

Question: Determine the value of ς such that the damping ration ϖ = 0.5. closed
loop characteristic equation of the system is

1 + GH = 0 = s(s + 1)(s + 3) + 2(s + ς) (*)

or
2ς
1+ =0
s3 + 4s2 + 5s
in standard root locus form.

19 / 20
Examples of root locus design
For stability requirement, we can obtain from the Routh Table with char. eqn
s3 + 4s2 + 5s + 2ς = 0
s3 1 5
s2 4 K
s1 20↓K
4
s0 k
⇐ K > 0 and K < 20.

Generate the root locus graph using a package like MATLAB, it shows that ϖ = 0.5
corresponds to s ⇒ →0.63 ± 1.09j. The value of k = 4.32 = 2ς or ς = 2.16

20 / 20