EMS507 Control System

Analysis and Design

Week 3 – System Stability

Yunpeng Zhu
System Stability
© Copyright 2023 Yunpeng Zhu. All Rights Reserved
Edition: v1.1
Table of Contents
1 Stability ............................................................................................... - 1 -
1.1 Transient state and steady state responses ........................................... - 1 -
1.2 System stability ...................................................................................... - 2 -
2 Poles of the transfer function............................................................... - 3 -
3 Routh-Hurwitz method ........................................................................ - 7 -
3.1 Characteristic equation .......................................................................... - 7 -
3.2 Routh and Hurwitz ................................................................................. - 7 -
3.3 Routh-Hurwitz criteria ......................................................................... - 10 -
4 Root Locus ......................................................................................... - 16 -

i
1 Stability
1.1 Transient state and steady state responses

We can analyze the response of the control systems in both the time domain
and the frequency domain. We will discuss frequency response analysis of
control systems in later chapters. Let us now discuss about the time response
analysis of control systems.
Time Response:
If the output of control system for an input varies with respect to time, then it
is called the time response of the control system. The time response consists
of two parts:
 Transient response
 Steady state response
The response of control system in time domain is shown in the following figure.

Here, both the transient and the steady states are indicated in the figure. The
responses corresponding to these states are known as transient and steady
state responses.
Mathematically, we can write the time response y (t ) as

y (t )  ytr (t )  yss (t )

where ytr (t ) is the transient response, yss (t ) is the steady state response.

After applying input to the control system, output takes certain time to reach
steady state. Transient response will reduce to zero:
lim ytr (t )  0
t 

i.e. y (t )  10  5e t .
-1-
Quiz 1.1: Find the transient and steady state responses.

y (t )  5e t  sin(3t ) ; y (t )  (5e 2t  1)cos(3t ) ; y (t )  10  cos(t )

Solution:

y (t )  5et  sin(3t ) ;

y (t )  5e2t cos(3t )  cos(3t ) ;

y (t )  10  cos(t )

1.2 System stability

Discussion: Why we discuss stability?

Stability is the cornerstone of a control system – performance cannot be
achieved without stability.
Without wanted to get tied up in the multiple possible definitions of “stability”,
for an introductory course it is sufficient to understand that stability means
that: for a bounded input signal u(t), the output signal y(t) is also bounded.
EXAMPLES:

my (t )  ky(t )  u (t )

Consider the input is impulse. The mass is positive.

It is obvious that if the stiffness k  0 , If the stiffness k  0 , spring force will
the system is stable (Oscillation). increase the oscillation

-2-
 my(t )  cy (t )  ky (t )  u (t )

Consider the input is impulse. The mass and stiffness are positive.
If the damping c  0 , the system is If the damping c  0 , damping force
stable. will increase the oscillation (Flutter).

(Watch a video)

2 Poles of the transfer function

Stability analysis is the analysis of a system that can passing through the
transient state to reach steady state.
Consider the above examples:
1
Transfer function: G ( s) 
ms 2  k

k
Pole: s   
my (t )  ky(t )  u (t ) m

Plot the pole-zero diagram:

-3-
 1
Transfer function: G ( s) 
ms  cs  k
2

c  c 2  4mk
my(t )  cy (t )  ky (t )  u (t ) Pole: s 
2m

Plot the pole-zero diagram:

Let’s consider some other examples. First, complete the following Quizzes.
Quiz 2.1: Evaluate the following inverse Laplace Transform and plot all poles in
one pole-zero diagram.
Laplace 1 1 2 2 s 1
Transfrom s4 s  0.2 ( s  5) 2  22 ( s  0.1) 2 ( s  1)2  1.52
Signal e4t  0 e0.2t  0 e5t sin 2t  0 2te0.1t  0 et cos1.5t  0
Poles s  4 s  0.2 s  5  2j s  0.1 s  1  1.5j

What do these signals have in common?

The real part of the poles is less than zero.

-4-
Quiz 2.2: Evaluate the following inverse Laplace Transform and plot all poles in
one pole-zero diagram.
Laplace 1 1 2 1 3 1 2 1
  
Transfrom s s2 s s5 s ( s  2) 2 s 2
s 1
Signal 11 t  2  e5t  2 3  te2t  3 2t  et  
Poles s0 s  0,0 s  0, 5 s  0, 2, 2 s  1,0,0

As long as the other poles have negative real parts, a single pole at the origin
results in a non-zero steady-state.
More than one pole at the origin results in divergence.
Quiz 2.3: Evaluate the following inverse Laplace Transform and plot all poles in
one pole-zero diagram.
Laplace s 1 3s 3s 2

Transfrom s 2  12 s 2  22 s 2  1.52 s 2  32 s 2  22
Signal cost 0.5sin 2t 3cos1.5t 3cos3t  sin 2t
Poles s  j s  2 j s  1.5j s  3j, s  2j

If all poles are on image axis, the signal oscillate.

-5-
Quiz 2.4: Evaluate the following inverse Laplace Transform and plot all poles in
one pole-zero diagram.
Laplace 1 2 2 s 1
Transfrom s4 ( s  5)2  22 ( s  0.1) 2 ( s  1)2  1.52

Signal e 4t   e5t sin 2t   2te0.1t   et cos1.5t  

Poles s4 s  5  2j s  0.1 s  1  1.5j

If the real part of any pole is greater than zero, the signal is divergent.
Summary: If the output signal of a system is divergent, assuming a convergent
input, we talk about the system being unstable. This statement is equivalent to
simply checking the position of the poles:
• All poles must have real parts less than zero for the system to be stable.

• If any poles have positive real parts, the system is unstable.

Transfer functions Left Half Plane (LHP); Right Half Plane (RHP)

s 1 Both poles are in the LHP

G( s) 
( s  6)( s  4)  System is stable

s2  4 All poles are in the LHP

G ( s)  2
( s  s  2)( s  1)  System is stable

s  0.4 One pole is in the RHP

G(s) 
( s  0.2)( s  0.6)  System is unstable

-6-
 ( s  2)( s  2) One pole is in the Right Half Plane (RHP)
G(s) 
( s  5)( s  1)( s  4)  System is unstable
One pole is on the origin and the other pole is in
s 1 the LHP. Would be considered borderline stable,
G ( s) 
s ( s  10) although some industries would classify this as
unstable

3 Routh-Hurwitz method
Solving the roots of complex transfer functions is painful!
In this section, let us discuss the stability analysis in the ‘s’ domain using the
Routh-Hurwitz stability criterion. In this criterion, we require the characteristic
equation to find the stability of the systems.

3.1 Characteristic equation

In a general form of transfer function:

N ( s ) b0 s m  b1s m1   bm1s  bm

G(s)  
D( s ) a0 s n  a1s n1   an1s  an

where a0 , , an and b0 , , bm are real numbers. The denominator

D( s )  a0 s n  a1s n1   an1s  an

is known as the characteristic equation.

It can be found that the stability of the underlying system can be assessed by
solving the roots of the characteristic equation (Poles).

3.2 Routh and Hurwitz

Before introducing these two guys, let’s go back to the stability analysis using
poles. We have learned that the system is stable (we don’t consider marginally
stable cases) if and only if all poles of the transfer function are located in the
left half plane in the pole-zero diagram. BUT, do you know who invented this
method?

-7-
 Monument in Edinburgh

Story:
Stage 1: Let’s Go back to the very beginning of the story:

There is a popular story that James Watt was inspired to invent the steam
engine by seeing a kettle boiling, the steam forcing the lid to rise and thus
showing Watt the power of steam.
This story is told in many forms; in some Watt is a young lad, in others he is
older, sometimes it’s his mother’s kettle, sometimes his aunt’s.
Watt did not actually invent the steam engine, as the story implies, but
dramatically improved the efficiency of the existing Newcomen engine by
adding a separate condenser.
This is difficult to explain to someone not familiar with concepts of heat
and thermal efficiency. It appears that the story was created, possibly by
Watt’s son James Watt Jr., and persists because it is easy for children to
understand and remember.
In this light, it can be seen as akin to the story of Isaac Newton and the
falling apple and his discovery of gravity.

-8-
In 1788, Watt invented the Fly-ball Governor to control his steam engine. But

when he increased the speed of the steam engine, the governor became
unstable. It cyclically speeds up and slows down or even destroys. Engineers
noticed this phenomenon, but didn’t know how to resolve it.

Stage 2: During the time completing his electromagnetism theory, James C.

Maxwell also tried to solve the unstable issue of the Fly-ball Governor. In 1868,
Maxwell published a paper, “On Governors,” in Proceedings of the Royal
Society of London. This paper was overlooked for a long time because it was
deemed by many to be difficult to comprehend. What we have learned about
Poles and Stability was discussed in this paper.
However, since Norbert Wiener drew attention to this paper in 1948, it has
been recognized as the first significant paper on control theory; as a result,
Maxwell has been regarded as the “father of control theory”
Stage 3: During 1850–1856, Maxwell studied at Trinity Colleague, Cambridge.
Yes, Newton graduated here too. Because of Newton’s impact, Trinity
Colleague held a math competition called “Senior Wrangler”. In 1854, Maxwell

-9-
graduated from Trinity with a degree in mathematics. He scored second
highest in the Wrangler. The first award was Edward Routh.

In 1877, Based on Maxwell’s work, Routh proposed the Routh criteria to

analyse the stability of linear systems, while in 1895, Adolf Hurwitz derived the
Hurwitz criteria, independently, to do the same thing by using a different
method. These two criteria are equivalent to each other, and we named them
the Routh–Hurwitz criteria.

3.3 Routh-Hurwitz criteria

What did Routh say?

All roots lie in LHP if and only if a certain set of algebraic combinations of the
a a  a3a0
coefficients have same sign. (i.e. 1 2 )
a1

- 10 -
What did Hurwitz say?
All roots lie in LHP if and only if a certain set of determinants of the
a1 a3
coefficients have same sign. (i.e.  a1a2  a3a0 )
a0 a2

Basically, they are equivalent to each other.

The Routh-Hurwitz criteria are sufficient and necessary conditions for the
stability of a linear time invariant system. We often use Routh’s idea because
calculating determinants is complex. This includes 2 Rules below for the
characteristic equation
Rule 1:
All the coefficients ai must be present and positive

(Equivalently all of them must be negative, with no sign change)

Rule 2:
All the elements of the first column of the Routh array defined by the
coefficients ai should have the same sign.

(Any sign changes, therefore, mean the system is unstable).

Rule 1 is easy, for example:

In Rule 2, forming the Routh array from coefficients ai is introduced as follows.

Routh array:
Without losing generality, we assume a0  0 in the characteristic equation:

D( s )  a0 s n  a1s n1   an1s  an

- 11 -
Routh array like the table above. We need to find all elements in the Routh
array following some algebraic calculations, so that we can implement Rule 2
to assess the stability.
The process of determining a Routh array is as follows.
Step 1: Determine the top two rows of the Routh array:

The first row contains a0 , a4 , a6 , a8 , of the characteristic equation. The

second row contains a1 , a3 , a5 , a7 , of the characteristic equation.

Take zeros if there is no value in the Routh array. For example, for the
characteristic equation

s 4  3s 3  3s 2  2s  1  0
s4 1 3 1
s3 3 2 0

Step 2: Determine the remaining rows of the Routh array by the rules:

- 12 -
In equations:

where

1 a0 a2 a0 a3  a1a2 a1a2  a0 a3
b1     (Hurwitz determinant)
a1 a1 a3 a1 a1

1 a0 a4 a0 a5  a1a4 a1a4  a0 a5
b2     ,…
a1 a1 a5 a1 a1

Repeat this process to the row s 0 .

For example, for the characteristic equation

s 4  3s 3  3s 2  2s  1  0
the Routh array can be written as

s4 1 3 1
s3 3 2 0
1 2  3  3 7 1 0  1 3
s2    1 0
3 3 3
3 7 5
s1  (3  1  2  )  0 0
7 3 7
7 7 5
s0  (  0  1 )  1 0 0
5 3 7

All coefficients of the characteristic function are positive; All elements in the
first column are positive. Thus the underlying system is stable.

- 13 -
Quiz 3.1: Assess the stability of the following systems

s 1 s 2  2s  3
1) G ( s)  5 ; 2) G ( s )  5 4
s  3s 4  2s 3  7.5s 2  s  1 s  s  5s 3  s 2  3s  1
1
3) G ( s) 
s 3  s 2  3s  1
1) There is a negative coefficient in the characteristic equation. Thus the
system is unstable.
2) All coefficients in the characteristic equation are positive. Evaluate the
Routh array as follows:

s5 1 5 3
s4 1 1 1
1 1  1 5 1 1  1 3
s3  4  2 0
1 1
1 1 1
s2  (1  2  4  1)   (1  0  4  1)  1 0
4 2 4
1
s1 2(4  1   2)  6 0 0
2
1 1
s0 (  0  6  1)  1 0 0
6 2

The system is unstable

3) All coefficients in the characteristic equation are positive. Evaluate the
Routh array as follows:

s3 1 3

s2 1 1
s1 (1 1  1 3)  2 0
1
s0  (1  0  2  1)  1 0
2

The system is stable.

- 14 -
Special cases: (Optional)

If any row of the Routh array contains only the first element as zero and at
least one of the remaining elements have non-zero value, then replace the first
element with a small positive integer,  . And then continue the process of
completing the Routh table.
Now, find the number of sign changes in the first column of the Routh table by
substituting  tends to zero.
If all the elements of any row of the Routh array are zero, follow these two
steps:

For example,

s5  3s 4  s3  3s 2  s  3  0

If there are common factors in one row, you can cancel them to simplify the
row.

- 15 -
4 Root Locus
The Root locus is the locus of the roots of the characteristic equation by
varying system gain K from zero to infinity.
Root locus diagram is used to observe the path of the closed loop poles to
identify the nature of the control system. In this technique, we will use an open
loop transfer function to know the stability of the closed loop control system.
Quiz 4.1: Find the transfer function of the close loop control system

G( s)
Y ( s)  G ( s)[U ( s)  H ( s)Y ( s)]  Y ( s)  U ( s)
1  G ( s) H ( s)
where G( s) H ( s) is known as the open loop transfer function in a close loop
control system.

It can be seen that the characteristic equation of the closed loop control
system is
1  G( s) H ( s)  0

- 16 -
Rewrite the open loop transfer function into the Head-1 standard form:
N (s) ( s  z1 )( s  z2 ) ( s  zm )
G(s) H (s)  K K
D( s ) ( s  p1 )( s  p2 ) ( s  pn )

where K is the gain of the open loop transfer function.

The characterization equation becomes:
N (s)
1 K 0
D( s )
The roots of this equation change with respected to K from 0 to  . The
changing path is the root locus.
MATLAB command:
num = [1, a1, a2, …];
den = [1, b1, b2, …];
G = tf(num, den);
rlocus(G);
For example, consider
s 1
G(s) H (s) 
s 4  s 2  2s  5
Root locus:

Please ignore the unit seconds-1.

- 17 -
 Walter was a noted American control theorist and the
inventor of the root locus method in 1948.
The world’s first “personal computer” invented in 1973
How to plot the root locus by hand?
Walter R. Evans
There are about 10 rules to plot the root locus by hand.

Rule 1): As K increases from 0 to  , the roots move from the poles of
G( s) H ( s) to the zeros of G( s) H ( s)

N ( s) 1

D( s) K
When K  0 , D( s)  0 , that’s the poles of open loop transfer function

When K   , N ( s)  0 , that’s the zeros of open loop transfer function

Rule 2): There are n lines where n is the degree of D( s) .

Rule 3-10): I know you cannot remember more than 2 rules. Check the web
when you have time.
https://www.mit.edu/people/klund/weblatex/node8.html
What you need to know is that:
 The root-locus plot clearly shows the contributions of each open-loop pole
and zero to the locations of the closed-loop poles.
 The root-locus plot also shows the manner in which the open-loop poles
and zeros should be modified so that the response meets system
performance specifications.

- 18 -