Lecture notes 8

Stability of Systems
Objectives

The main objectives of this chapter are:

1. To investigate the relation between system poles and stability

2. To determine the stability of the system

8.1 Introduction
The stability of a system is determined by its response to inputs or
disturbances. Intuitively, a stable system is one that will remain at rest
unless excited by an external source and will return to rest if all
excitations are removed. Stability can be precisely defined in terms of the
impulse response of a system as follows:
Definition 1: A system is stable if its impulse response approaches zero as
time approaches infinity.
Definition 2: A system is stable if every bounded input produces a bounded
output.
We can explain the stability by using the Cone example. If a cone is placed
on its bottom side on a flat surface, and then a small force is applied, the
cone swings on its bottom but goes back to its initial position thereafter.
This is an example to a stable dynamics. If the cone is placed on its side, it
is called it is in neutral equilibrium state. However, if we try to keep the
cone standing on its top, we will not be able to achieve this easily. Even a
small disturbance or unbalance will cause the cone fall off. This behavior is
called unstable equilibrium state.

Kararlı - Stable Yansız (Nötr) - Neutral Kararsız - Unstable

Figure 8.1. Types of stability (a cone example)

Another good example to explain the stability of systems is the
dynamics of a ball on convex and concave shaped surfaces.

Kararsız - Unstable

Yansız (Nötr) - Neutral

Kararlı - Stable

Figure 8.2. Types of stability (a ball example)

 Stability of linear control systems can be inferred from the poles of the
closed loop transfer functions, i.e. the roots of the characteristic
equation. Whether a system is stable or not is independent of its
inputs. The poles of the input function have effect only on the transient
response of a system not on its stability.
Block diagram of a closed loop system and its transfer function are
given as follows:

R(s) + C(s)
G(s)
- 𝐶 𝑠
𝑅 𝑠
=
𝐺 𝑠
1+𝐺 𝑠 𝐻 𝑠

H(s)
If we apply a unit impulse input 𝑟(𝑡) = 𝛿(𝑡) , 𝑅(𝑠) L 𝑟(𝑡) = 1, the system
response becomes.

𝐺(𝑠) 𝐵(𝑠) 𝑏0 𝑠 𝑚 +𝑏1 𝑠 𝑚−1 + ... +𝑏𝑚

𝐶(𝑠) = = =
1+𝐺(𝑠)𝐻(𝑠) 𝐴(𝑠) 𝑎0 𝑠 𝑛 +𝑎1 𝑠 𝑛−1 + ... +𝑎𝑛

The characteristic equation can be obtained by setting A(s)=0.

𝐴(𝑠) = 𝑎0 𝑠 𝑛 + 𝑎1 𝑠 𝑛−1 + 𝑎2 𝑠 𝑛−2 + . . . +𝑎𝑛 = 0
𝑠 is a complex variable in general form it is represented as s=r+jω
Since the order of A(s) is n, it has n roots which are 𝑠1 , s2 , 𝑠3 , . . . , s𝑛 . A(s) can also
be represented as
𝐴(𝑠) = (𝑠 − 𝑠1 )(𝑠 − 𝑠2 ) . . . (𝑠 − 𝑠𝑛 )
We will investigate the 3 situations arising from the solution of A(s)=0.
a) 𝒔𝒊 are all real and positive.

Let 𝑠1 = 𝑟1 , 𝑠2 = 𝑟2 , . . . , 𝑠𝑛 = 𝑟𝑛 . Then,
𝐴(𝑠) = (𝑠 − 𝑠1 )(𝑠 − 𝑠2 ) . . . (𝑠 − 𝑠𝑛 )

𝐴(𝑠) = (𝑠 − 𝑟1 )(𝑠 − 𝑟2 ) . . . (𝑠 − 𝑟𝑛 )
The output of the closed loop system is

𝐴1 𝐴2 𝐴𝑛 𝐿−1
𝐶(𝑠) = + + ... + 𝑐(𝑡) = 𝐴1 𝑒 +𝑟1 𝑡 +
(𝑠−𝑟1 ) (𝑠−𝑟2 ) (𝑠−𝑟𝑛 )

𝐴2 𝑒 +𝑟2𝑡 +. . . . . . . . . . +𝐴𝑛 𝑒 +𝑟𝑛𝑡

If the roots of the closed loop
transfer function are all real and
positive, then the response of the
closed loop system c(t) to the unit
impulse function 𝑟(𝑡) = 𝛿(𝑡) is an
exponentially increasing curve. The
response of the system to a
bounded input is unbounded; hence
the system is said UNSTABLE.
b) 𝒔𝒊 are all real and negative.

Let 𝑠1 = −𝑟1 , 𝑠2 = −𝑟2 , . . . . . . . . . . , 𝑠𝑛 = −𝑟𝑛 , Then

𝑨 𝒔 = 𝒔 − 𝒔𝟏 𝒔 − 𝒔𝟐 . . . . . . . . . . 𝒔 − 𝒔𝒏

𝐴(𝑠) = (𝑠 + 𝑟1 )(𝑠 + 𝑟2 ). . . . . . . . . . (𝑠 + 𝑟𝑛 ).

The output of the closed loop system is

𝐴1 𝐴2 𝐴𝑛 𝐿−1
𝐶(𝑠) = + + ...+ 𝑐(𝑡)
𝑠 + 𝑟1 𝑠 + 𝑟2 𝑠 + 𝑟𝑛
= 𝐴1 𝑒 −𝑟1 𝑡 + 𝐴2 𝑒 −𝑟2 𝑡 + . . . +𝐴𝑛 𝑒 −𝑟𝑛𝑡

Since the response of the system is a decaying exponential,

the system is STABLE

c) Only one root is positive 𝑠1 = +𝑟1

and the rest is all negative real 𝑠2 = −𝑟2 , . . . . . . . . . , 𝑠𝑛 = −𝑟𝑛

𝐴(𝑠) = (𝑠 − 𝑠1 )(𝑠 − 𝑠2 ). . . . . . . . . . (𝑠 − 𝑠𝑛 )
𝐴(𝑠) = (𝑠 − 𝑟1 )(𝑠 + 𝑟2 ). . . . . . . . . . (𝑠 + 𝑟𝑛 )
𝐴1 𝐴2 𝐴𝑛 𝐿−1
𝐶 𝑠 = + +. . . . . . . . . . + 𝑐 𝑡
𝑠 − 𝑟1 𝑠 + 𝑟2 𝑠 + 𝑟𝑛
= 𝐴1 𝑒 +𝑟1𝑡 + 𝐴2 𝑒 −𝑟2𝑡 +. . . . . . . . . . +𝐴𝑛 𝑒 −𝑟𝑛 𝑡
𝑧𝑎𝑚𝑎𝑛𝑙𝑎 𝑧𝑎𝑚𝑎𝑛𝑙𝑎𝑦𝑜𝑘𝑜𝑙𝑢𝑟𝑙𝑎𝑟
𝑔𝑖𝑡𝑔𝑖𝑑𝑒𝑏ü𝑦ü𝑟

The system is UNSTABLE.

 8.2. Absolute and Relative Stability and
Stability axis
For a closed loop control system, if all of its poles must lie in m ( j)
the left half of the complex plane, then it is absolutely stable.
If a system has a pair of complex conjugate roots s1,2=±𝑗𝜔 , Stable region Unstable region
the system response will be oscillatory with constant
amplitude. Then the system is said to be Marginally stable,
which is not desired in control systems design. − R e ( )

The stability increases relatively when the system poles that

lie on the left half of the s-plane, move away to the left from
the Im axis. It is desired that the system has a strong relative
stability. In order to achieve a strong relative stability, we Stability
axis
want the roots of the characteristic equation are on the left of
a line crossing the Real axis at −𝜎 . Having the roots on the
left of this line guarantees the relative stability.
8.3. Routh’s Stability Criterion
It is also known Routh-Hurwitz criterion. Routh’s stability criterion tells us whether or
not there are unstable roots in a polynomial equation without actually solving it. The
information about the absolute stability of a control system can be obtained from the
coefficients of the characteristic polynomial of the closed loop transfer function.

In order to have an absolutely stable system, the necessary condition is the characteristic
polynomial has nonzero and positive coefficients. If this condition is satisfied then we can
implement the Routh’s stability test. The sufficient condition for the characteristic
polynomial has stable roots is explained on the following example:
 Example:
𝑎0 𝑠 5 + 𝑎1 𝑠 4 + 𝑎2 𝑠 3 + 𝑎3 𝑠 2 + 𝑎4 𝑠 + 𝑎5 = 0
a1a2 − a0 a3 a1a4 − a0 a5
s5 : a0 a2 a4 b1 = b2 =
a1 a1
s4 : a1 a3 a5 b1a5 − a1 0
ba −ab
c1 = 1 3 1 2 c2 = = a5
s3 : b1 b2 0 b1 b1
s2 : c1 c2 c1b2 − b1c2
d1 =
c1
s1 : d1 0
d1c2 − c1 0
0
s : e1 0 e1 =
d1
The first column of the coefficients table must have all positive coefficients for the system to be
absolutely stable. It is the sufficient condition for system to be stable. If there is a sign change in the first
column of the coefficients table, we say that the characteristic polynomial has positive roots and hence
the system is unstable.
 Problem 1: Find the sufficient condition for all the roots of the 3rd
order polynomial given below to have negative real parts.
𝑎0 𝑠 3 + 𝑎1 𝑠 2 + 𝑎2 𝑠 + 𝑎3 = 0

Solution:
s3 : a0 a2
s2 : a1 a3 a1a2 − a0 a3
Then, the polynomial must satisfy 0 a1a2  a0 a3
a1
1 a1a2 − a0 a3
s: 0
a1
s0 : a3
Problem 2: 𝑠1 = +1 + 𝑗 7, 𝑠2 = +1 − 𝑗 7, 𝑠3 = −3 For these
poles the system given below is unstable,
(𝑠 − 𝑠1 )(𝑠 − 𝑠2 )(𝑠 − 𝑠3 ) = 𝑠 − 1 − 𝑗 7 𝑠 − 1 + 𝑗 7 𝑠 + 3 =
𝑠 3 + 𝑠 2 + 2𝑠 + 24 = 0
We see that the necessary condition is satisfied since all of the
coefficients of the polynomial are positive and nonzero.
s3 : 1 2
s2 : 1 24 2 − 24
= −22
s1 : − 22 1
s0 : 24

However, by applying the Routh’s test, we find out that the sufficient
condition is not satisfied since there are two sign changes in the first
column of the coefficient table, which is the indication of the presence of
two poles on the right half part of the Im-Re plane.
Problem 3:
100𝑠 3 + 80𝑠 2 + 17𝑠 + 6 = 0
Solution:
The necessary condition is satisfied, sufficient condition must
be investigated.

s3 : 100 17
2
There is no sign change in the Routh’s
s : 80 6 coefficient table, therefore the system is
s1 : 760 0 stable.
80
0
s : 6
Special Cases 1:

A zero term can appear in the first column of the Routh’s coefficient
table in some cases. If the rest of the coefficients of the corresponding
row are nonzero or there is no remaining term, then the zero term can
be replaced by a small number  and the rest of the array is evaluated
by using the small term instead of zero. At the end, the small term is
then replaced with zero.
Problem 4: Investigate for the given poles
𝑠1,2 = ±𝑗, 𝑠3 = −2
Solution:
(𝑠 − 𝑠1 )(𝑠 − 𝑠2 )(𝑠 − 𝑠3 ) = (𝑠 − 𝑗)(𝑠 + 𝑗)(𝑠 + 2) = (𝑠 2 + 1)(𝑠 + 2)
= 𝑠 3 + 2𝑠 2 + 𝑠 + 2 = 0

If the signs of the coefficients above and

below zero () are the same then it
indicates that the characteristic
polynomial has a pair of imaginary roots.
The example above confirms this
assertion since it has two roots at 𝑠1,2 =
± 𝑗.
 Problem 5:
Investigate the stability of the transfer function given below.
10
𝑇(𝑠) = 5
𝑠 + 2𝑠 4 + 3𝑠 3 + 6𝑠 2 + 5𝑠 + 3

s 5 : (+) 1 3 5
s 4 : (+) 2 6 3
s 3 : (+) 0 7/2 0 There are two sign changes in
6 − 7 the first column; therefore the
s 2 : ( −) 3 0 system has two poles on the

42 − 49 − 6 2 right half Im-Re plane. The
s : (+)
1
0 system is unstable.
12 − 14
s 0 : (+) 3
Problem 6: 𝑠4 + 𝑠3 + 2𝑠2 + 2𝑠 + 3 = 0
The roots of this polynomial are: 𝑠1,2 = −0.09057 ±
𝑗0.902, 𝑠3,4 = 0.4057 ± 𝑗1.2928
s4 : 1 2 3
s3 : 1 2 0
s2 : 0 3
2 − 3 3
s1 : − 0
 
s0 : 3 There are two sign changes in the first
column; therefore the system has two
poles on the right half Im-Re (s-plane)
plane. The system is unstable.
Special Cases 2:
If all the coefficients in any derived row are zero, it indicates that there
are roots of equal magnitude lying radially opposite in the 𝐼𝑚 − 𝑅𝑒 (s-
plane)- that is, two real roots with equal magnitudes and opposite sign
(𝑠1,2 = ±𝑎), and/or two conjugate imaginary roots (𝑠1,2 = ±𝑏𝑗 or
𝑠1,2 = −𝑎 ± 𝑏𝑗). In such a case, the evaluation of the rest of the array
can be continued by forming an auxiliary polynomial with the
coefficients of the last row and by using the coefficients of the
derivative of this polynomial in the next row. Such roots with equal
magnitudes and lying radially opposite in the s plane can be found by
solving the auxiliary polynomial.
 Problem 7:
𝑠 5 + 3𝑠 4 + 10𝑠 3 + 16𝑠 2 + 24𝑠 + 16 = 0
s5 : 1 10 24
s4 : 3 16 16
30 − 16 72 − 16 30 − 16 14 72 − 16 56  14 
s3 : 0  = , = = 4 
3 3 3 3 3 3  3

Dividing the coefficients in the same row by a constant does not affect the Routh’s criterion. For
the sake of simplicity of the calculations, dividing a row by a constant is always preferable if the
coefficients of the corresponding row are large numbers. In this example, we divide the row
belonging to s3 by 14/3.
s5 : 1 10 24
s4 : 3 16 16
s3 : 1 4 0
16 − 12 16 − 0 16 − 12 16 − 0
s2 : 0  =4 , = 4 ( 4)
1 1 1 1

The row belonging to s2 is divided by 4.
s5 : 1 10 24
s4 : 3 16 16
s3 : 1 4 0
s2 : 1 4 0 It is found that all the elements of the row belonging
s1 : 0 0 to s 1 are found to be zero. Let us set an auxiliary
polynomial.
s0 :
dP(s)
P s = s 2 + 4 = 0, = 2s
ds

s5 : 1 10 24 There is no sign change in the

s4 : 3 16 16 first column. However from the
solution of the auxiliary
s3 : 1 4 0
polynomial, we get
s2 : 1 4 0 𝑃(𝑠) = 𝑠 2 + 4 = 0, 𝑠 = ±2𝑗 ,
dP( s ) the corresponding roots show
s1 : 2 0  = 2 s ' in katsayıları yazılır.
ds that the system oscillates with
s0 : 4 constant amplitude. Hence the
system is marginally stable.
Problem 8:
10
𝑇(𝑠) = 5
𝑠 + 7𝑠 4 + 6𝑠 3 + 42𝑠 2 + 8𝑠 + 56

s5 : 1 6 8 s5 : 1 6 8
s4 : 71 42 6 56 8 s4 : 1 6 8
s3 : 0 0 0 s3 : 41 12 3 0
s2 :  s2 : 3 8
s1 : 1
s1 : 0
3
s0 :
s0 : 8

dP ( s )
s 4 + 6s 2 + 8 = 0 = 4 s 3 + 12 s
ds
s2 = x
−6 36 − 32 −4  2 j The system is Marginally stable.
x2 + 6 x + 8 = 0 x1,2 =   s 2
= x 
2  −2  2j
• Problem 9: Find the range of K so that the system is stable.
𝐹(𝑠) = 𝑠 4 + 𝑠 3 + 𝑠 2 + 𝑠 + 𝐾

s4 : 1 1 K
s3 : 1 1 0
s2 : 0 K
 −K
s1 : 0

s0 : K
 0 Parameter K cannot be less than 𝜀
 −K 0 K  𝐾 > 0 𝜀 > 0 ⇒ in this case
K 0
1
−𝐾
𝑠 : <0
𝜀

The system is unstable for all K values.

Problem 10: For what range of K is the closed loop system stable?

R(s) + K 1 C(s)
___ ________________
3 2
_ s s + 6s + 11s + 6
Solution
𝐾 1
𝐶(𝑠) 𝑠 𝑠 3 + 6𝑠 2 + 11𝑠 + 6 𝐶(𝑠) 𝐾
= ⇒ =
𝑅(𝑠) 1 + 𝐾 1 𝑅(𝑠) 𝑠 𝑠 3 + 6𝑠 2 + 11𝑠 + 6 + 𝐾
𝑠 𝑠 3 + 6𝑠 2 + 11𝑠 + 6
Note: For a closed loop system to be stable, negative feedback is required. It is obvious that in the
case of positive feedback, the necessary condition for Routh-Hurwitz Stability criterion is not
satisfied.
Characteristic Equation: 𝑠 4 + 6𝑠 3 + 11𝑠 2 + 6𝑠 + 𝐾 = 0
For the necessary condition 𝐾 > 0.
Sufficient condition:
s4 : 1 11 K
60−6𝐾
s3 : 6 6 0 10
⇒ 𝐾 < 10
s2 : 10 K For the system with integral control 0 < 𝐾 < 10 must be
satisfied.
s1 : ( 60 − 6 K ) 10 0
s0 : K