230 Control Lnginering

CHAPTER
4
Stability of Linear
Control Systems

4.1 INTRODUCTION

(Most control systems that require the attention of control

engineers are large, complex,
A
and involve many variables. The factors are related to one another in complicated math- bounded-output.
ematical modeis, quantitative answers to probtems are Obtained by numerical solutions of
Fig. 4.1 A bounded-input and a
model equations.) to G2, but the response is
Besides this approach to the
analysis of control systems, engineers need a m
8 4.2, the same bounded pulse function is applied
qualitative investigation which will yield a yes or no type of answer and serve assimplified seen to be unbounded; therefore, the system is unstable.
to their
approach in the analysis as well as the design.
a
guide .

Will the ship float or sink? Will the machine run or

run away? Will the
or fall? Wll the kettle boil or blow bridge stand
up? All these systems can be analyzed in the light
of a generalized and unified
theory known as stability analysis. All these questions then
becomes one: Is the system stable ?
As a prelude to the definition of
stability, we define two types of responses for LTI
systems:
1. Zero-state response: The zero-state is due to
response input only; all the initial
conditions of the system are at zero.

2. Zero-input response: The zero-input response is due to initial conditions only; all
the inputs are zero.
Fig, 4.2 Abounded-inputand an unbounded-output.
Then, according to principle of superposition, when a system is
inputs and initial conditions, the total response is written as subjected to both
4.3 RELATIONSHIP BETWEEN CHARACTERISTIC EQUATION
ROOTS AND BIBO STABILITY
Total response Zero-state response Zero-input response
=
+
In general, a transfer function such as
4.2 BOUNDED-INPUT, BOUNDED-OUTPUT (BIBO) STABILITY
s+b sm-l+ b2 sm-=2+ + bn
A linear system is said to be
stable, if and only if,
Cs)= s+ag s"-+a1 s"-z +..+ an n>T
(4.1)
output.
any bounded input produces a bounded
As an illustration, consider, may be decomposed by partial fractions into the following form:
Fig.4.1 which shows a system to which a pulse function
has been applied and a response has been
produced. Both are bounded, the system is
said to be stable.
so
Gs)= +2..K,
s-Sn (4.2)
where s; are the roots of the characteristic
equation or the denominator of G(s)
 Stability of Linear (ontrol Systems 231 232 ControlLngineering
InverseLaplace transformation gives
9(t) K^e$1+K2c** ++K,c*
=

fct)lot)|M,dr
It is obvious that
the BIBO stability of tlhe (4.3) -C
impulse response g(t), which in turn system depends on the boundness of
In
depends upon the signs of si. If S Io()| < oo, then, Jolt)| < o
general, si are complex numbers. If all their 0
impulse response will die away in real parts are less than
time, thereby exhibiting a bounded zero, the which means a bounded output. Hence, the systern under investigation is stable, if its
9t) and a impulse
corresponding stable system. For example, response impulse response g(t) is absolutely integrable. That is, the area under lg(t)! v/s t curve
should be a finite quantity.
91(t) Kiet +K2el-3+j0)f + K3el--j6 +
=

is a bounded Ke-6
impulse response; therefore G(s) is a BIBO stable 4.4 ZERO-INPUT STABILITY
But if one
(or more) s; has a positive real part or roots on thesystem.
below, ju-axis, as in equation Zero-input stability refers to the stability of the system when the input is zero, and the
system response is solely due to initial conditions. The zero-input stability also depends
92(t)= Kaet+ Kiels+57)t + Kee37
then the
+Ke3t + Ke"i3t on the roots of the characteristic equation.
impulse response, ast goes to infinity, will have no bound A system is said to be zero-input stable, if and only if, the zero-input response c(t)is
that is unstable. and represents a
system bounded and approaches zero ast approaches infinity; otherwise, the system is unstable.
Thus, for BIBO stability, the roots of the characteristic Mathematically, an LTI system is zero-input stable if for any set of finite initial con-
cannot be located in the equation, or the poles of G(s), ditions, c(to), there exists a positive number M, which depends on cl)(to), such that
all be in the left-half
right-half s-plane or on the jw-axis. In other words, they must
s-plane. 1. e(t)| M < o for all t2 to and
Let us now
prove mathematically that a system is BIBO stable, if and only if, its 2. lim |ot)] =0
impulse response is bounded or
absolutely integrable.
The output of a linear time-invariant system Since, the output e(t) approaches zero as t - oo, the zero-input stability is also known
is given by
et) = g(t) * r(t) as the asymptotic stability.
(4.4) For a system to be zero-input stable, the roots of the characteristic equation must all
where g(t) =
L{G(s)} is defined the
impulse response of the system.
as be in the left-half s-plane.
The above equation (4.4) is nothing but the convolution of Hence, we conclude that for a system to be BIBO stable and zero-input stableor
written as g(t) and r(t) and may be
asymptotically stable, all the roots of the characteristic equation must lie to the left-
half of the s-plane. Thus, if a system is BIBO stable, it must also be zero-input or
dt)= 9trr(t-rdr (4.5) asymptotically stable. For this reason, we shall simply refer to the stability condition of
a linear system as stable or unstable.
If input r(t) is bounded, then
When the roots of the characteristic equation has simple roots on the ju-axis and none
rt-7)S M, <0 in the right-half of the s-plane, the system is said to be marginally stable or marginally
unstable. For example, when the system is an oscillator, the characteristic equation would
Taking magnitudes on both sides of equation (4.5), we get
have simple roots on the jw-axis, and the system is considered to be stable.

lct)=| 9(r) r(t- r) dr| 4.5 THE CHARACTERISTIC EgUATiON OFA NEGATIVE FEED
BACK SYSTEM AND CRITERION FOR STABILITY
A typical cosed-loop control system may be represented by the simplified block diagram
-co
of Fig. 4.3.
 Stability of Linear Control Systems 3 34 Control Enginering
The Routh's array is
R$) C()

a2 44

a3 45
Fig. 4.3 Simplified system block diagram. b3
The closed-loop transfer function of the system is C1 C3

Cs) G(s) d
R( 1+G(s) H(s)
e1
The characteristic equation of the system is 1+G(s)H(s) = 0. The roots of
the The first two rows are nothing but the coefficients of the characteristic equation. The
characteristic equation are called the zeros of
1+G(s)H(s). remaining entries are entered by computation with the following pattern.
The procedure for investigating the
stability of a system is to search the right-half
plane for zeros of 1+G(s)H(s). If there are no zeros of 1+G(s)H(s) in the aoa2
then there are no roots of the right-half plane,
characteristic equation in the right-half of the s-plane, and b1 =1a3
= a1a2aoa3
the system is stable. Obviously, it is
impractical, if not impossible, to investigate everyy a1
point on the right-half of the s-plane and so it is necessary to devise some shorter
methods.
Methods like Routh-Hururitz criterion, Nyquist criterion and Bode
criterion are well known
for determining the stability of linear
systems without involving b11= _5 a1d4a0@5
root solving.
4.6 ROUTH-HURWITZ CRITERION Similarly,
In 1887, Routh developed a method for determining whether or not an equation has C
C1 1G3-a1 b3
roots with positive real parts without actually solving the equation for the roots. When
used with the characteristic equation of a feedback 5 1x 0
control system, Rouths criterion offers Ca
a
simple means for detecting system instability, because roots with 1
indicate transients which grow, rather than positive real parts
decay, with time. Entries di and ej are obtained in a similar
manner.
The usual characteristic
equation can be written in general form as Routh's criterion states that the number of
column equals the number of roots with sign changes in the elements of the first
ags +a1st+.+a18 +n =0 positive
real parts.
Limitations
in which the coeficients typically are
positive numbers. An unstable system or a system 1. The Routh's criterion does not
with marginal stability (simple roots on give the location of the roots that lie
ju-axis), is immediately indicated if any cocfficient the s-plane and also does not tell
is either zero ornegative. whether these roots are real or to
right-half of
The first step in applying Routh's criterion is to form the Routh 2. The Routh's criterion cannot be complex.
array. For the sake applied to any other
of illustration, a characteristic
equation with n =
5 is taken. plane, such as the unit circle in the z-plane which is stability boundaries in a complex
the
control systems. stability of boundary digital
ag+a18* +a2s° +a35 +a4s +ag = 0
4.6.1 Relative Stability Analysis
The question of absolute stability is answered by the
practical cases, information about the relative Routh's criterion.
stability is However, in many
required. A useful way for
 Stability of Linear Control Systems235 236 Control Lngineering

checking relative stability of a system is to shift the s-plane axis and then apply Routh's Solution
criterion. That is to put
The Routh's array is
(=a constant)
into the characteristic equation of the system. Now
and apply Routh's criterion to the new write the polynomial in terms ofs1
the first column of the polynomial in s1. The number of sign changes in
Routh's array using the new 3 2 8
of roots that are polynomial is equal to the number
situated to the right of the vertical line s = o. Thus, the number of
-

roots that lie to the

right of the vertical line is revealed this test.
by -12
The
ample 4Investigate the stability of closed-loop system whose characteristic
a
12
equation is given by

s +3s+7s +20s2+ 6s + 15 =0 2
There are two sign changes (+ to - and - to +) in the first column of the Houth's

WSolution array. This implies that there are two roots with positive, real parts. Hence, the system
in question is unstable.
The Routh's array is
Important note:

or divided by a
It can be
shown that the entries in any row can be multiplied positive
constant without affecting sign changes in the first column. This procedure can be used
3 3 20 15
to reduce the mathematical labour involved in forming the array.

4.6.2 Difficulties Encountered in Routh's

Criterion
1115 Two difficulties arise that are
worthy to note:

1. The first element in any one row is

st zero, but the others are not.
15 2. All the elements in one row of Routh's
array are zero.

There are no sign changes in the first column of the Routh's array, and so there are
no roots with positive real parts. Hence, the system under investigation is stable. EXGLMple (Dificulty-1) Investigate the stability of a closed-loop system whose
characteristic equation is given by
Kampe2 Investigate the stability of a closed-loop system whose characteristic
equation is given by
s$+s+2»3 +2s2+3s+5 = 0

s+2s+ 3s+8s+2 =0
 Stability of Lincar Contnol Systems237 288 Control Lngineering
In the present context, we have
Solution =0
The Routh's array is (s+ 1)(s+s+2s+2+3s +5)
=0
s+2854+3s+4s+5+8s +5

2 3
The Routh's array is

2 5
0-2 5
2

2 5
Dificulty-1 is being addressed as the first element in a row being zero. To overcome
this difficulty, put s = in the given characteristic equation. Accordingly, we get

5
that there
Again, we see that there are two sign changes in the first column showing
Multiplying throughout by z" and rearranging, we get are two roots with positive real parts. Therefore, the system is unstable.

55+3A+223+222+2+1 =0 +8s*+
Example (Dificulty-2) Considérthercharacteristic equation s +2s
The Routh's array is 12s3+20s2 + 165 +16 of the
= 0. Using Routh's criterion, investigate the stability
closed-loop system.
5 2 Solution

A32 The Routh's array is

8 20 16
22 12 16
2 12 16

There are two sign changes in the first column showing that there are two roots with
positive real parts. Therefore, the system is unstable.
Sometimes the above described method of overcoming difficulty-1 may not work. In
that case,multiply the given characteristic equation by (s+1) and then repeat the Routh's
criterion.
 Stability of Linear Control Systems230
240 Control Enginering
Since, a row of zeros
appears prematurely
mial using the coefficients of s* row. (dificulty-2) we form an
auxiliary polyno- Solution
The Routh's array is
A(s) =2s4+ 12s2 + 16
dAls) =8s+24s
ds 1 15 K
From the above
equation, the coefficients 8 and 24 replace the 25 20
original tabulation. Thus, the Routh's zeros of s3 row of the
array becomes:
14.2 K

8 20 1.76 K
20 16
22 12 16 K
For stability:
2 1 2 16 i. 20 1.76K >0 » K <11.36
Coefficients of dA(s)
8 24 ds i. K>0
6 16 Thus, the system is stable for 0 < K < 11.36. When K = 11.36, the system is
marginally stable. The auxiliary equation is A(s) = 14.2s2+11.36 =0. Thesolution of
this equation gives s = tj0.894. Hence, the frequency of sustained ascillations is 0.894
rad/sec.
16

Fxanple16 Refer the characteristic equation given by

Since, there are no sign changes in the first column of the entire Rouths array, the
given characteristic equation has no roots with positive real parts. Solving the auxiliary s+Ks+2s+(K+1)s + 10 = 0
equation, A(s) 0, we get s
=
+j2 and tjv2. Thus, the characteristic equation has Determine:
four nonrepeated roots on the jw-axis, and no roots with
positive real parts and hence a. the range of K so that the system is stable and
the system is marginally stable. These roots on the jw-axis are
responsible for the row b. the value of K so that the system is marginally stable and the frequencies of sustained
of zeros in the s* row. Also, these roots on the jw-axis indicate that the system has two
oscillations, if any.
frequencies of sustained oscillations of V2 rad/sec and 2 rad/sec.
Solution
Xempe Refer the characteristic equation given by The Routh's array is
+25s3+ 15s +20s +K = 0 10
2
K K+1|
Determine:
10
a. the range of value ofK so that the system is asymptotically stable and
-9K-
b. the value of K that the
so system is marginally stable and the frequencies of sustained
oscillations, if applicable.
10
 Stability of Linear Control Systems 241 BZe Control Engineering
For stability:

i. K>0 For stability:

K> 1 3x 10- K>0 K <3 x 10

600
i >0
-9K2- 1> 0 K2 tjV 2.1408 10"K - K2 x
K <2.1408 x 10
. >0
For linear system, K is
a 3x 107 K
always real and hence, the last
any real value of K.
Thus, the system is always unstable. condition cannot be met by
i.-5 x 10 +2.1607x 107K -
K2
0 K?-2.1607 x 10'K+5 10 x <0
0.04166(2.1408 x 107 - K)
EXample Refer the negative feedback control system shown in The roots of the
Fig. Ex.4.7. equation
Using Routh's criterion determine the K2-2.1607 x 107K +5 x 10 = 0
of K. stability of the
closed-loop system as a function
are
K 2.34 x 105 and K =2.1373 x 10
R(s) Ks+4) (s+ 20) Hence, (K 2.34 x 105)(K - 2.1373 x 10")<0
sls +100)(s+ 500) C() Thus, 2.34 x 103 <K <2.1373 x 10
iv. 80K >0 K>0
Thus, for stability:
Fig. Ex.4.7
Find the value of K that
will cause sustained oscillations 2.34 x 10 <K <2.1373 x
10
frequency of oscillations. in the
system. Also, find the When K 2.34
= x
103, the s' row becomes zero.

Solution
The characteristic
equation of the system given by
is
Then,
At)=20 + 8OK K=2.34x 105
=

s tj10.6 rad/sec
1+Gls)H(s) = 0 w =
10.6rad/sec
14K(st+4(6+20) when K 2.1373x 10', the
3(s +100) (s +500) = 0
=
s row
again becomes zero.
s+600s+ 50000s3+Ks*+ 24Ks +
The Routh's array 80K = 0
is Then, A)- S[2.1408 x 107 K-
gives
x10 +80K SK-2.1373x 107 = 0

50000 24K
S=+j140.42
w =
140.42 rad/sec
600 K 80K
3x10-K
600, 14320K
600 Reinforcement Problems
2.1408x
3x10-
10 K-K2 80K
-5XI042.1607
0.04166 2.1408 xx10 K-K
10-K
RP 4 Apositional servomechanism is
characterized by an
80K
open-loop transfer function
G()H()=Ko +2)
s(s-1)
 Stability of Linear Control Systems243 244 Control Erngineering
Determine the Rouths criterion, determine how many roots of the characteristic
a. the value of the
gain K when
RP-42 Using
of the closed-loop roots is cqual at 0.707 and to the right of the line s -1 in the s-plane.
equation
=
are
b. the value of the gain K when the closed-loop system has two roots on the ju-axis.
Fls)=s +5s +3 = 0
Solution
a . The characteristic equation of the closed-loop system Solution
is given by
S+a. Then, when s -a,$1 =0. This transforms s - a axis in the s-plane
=
Let s1 =

1+G(s)H(s) = 0 onto the imaginary axis of the 81-plane.

1, then s 1. Hence, letting s $1 -1 in the
1+t2),
-
=
In the present context, a =
81 =

s(s-1) characteristic equation, we get

s+ (K-1)s +2K =0
(s1-1+5(s1-1) +3=0
Comparing the above equation with the reference characteristic equation, we get s+3s11 =0
2wn = (K-1) and =2K The Routh's array is

Given = 0.707. Hence, 2 x0.6

x V2K =
(K -1)

K-6K+1 =0 3

Solving for K yields, K = 5.83 and K = 0.17.

-
The Routh's array is
Since, there is one sign change in the first column of the Routh's array, there is one
root to the right of s =-l in the s-plane. In fact, the roots ofthe characteristic equation
2K 2K are s=-3.303 and s = 0.303.

sK-1
RP43 A negative feedback system is represented by the equation
s 2K

The condition for stability: F(s)=s+10s+29s+K

i.K-1>0 K>1
Shift the vertical axis to the right by 2 by using s = s1 -2, and determine the value
ii. 2K > 0 K > 0
ofgain K so that the complex roots are s=-2+1j.
Hence, for stability, K >1. Thus, the value of Kkeeping in mind that the criteria for
stability and = 0.707, we require K = 5.83. Solution
b . When K =1, the system is marginaly stable. Then the auiliary equation is The transformed characteristic equation is given by

A(s)= +2K\K=1 =0 (61-2+10(1 -2 +29(61 -2) +K =0

solving, we get $=tjv2 s+4s+s1 +K -26 =0
 Stabilitu of Linear Contrel Systems 246 Control Tngineering
we ge
polynomial must
he of e v e n order,
The Routh's array is Since, the auxiliary
=0
A(3) = as + (K+ 1)

4 K -K 26
Hence,
30-K Given, w =2 rad/sec. (4.6)
s K- 26
If K = 30. then A(s) = 4s+ K -26|K=30= 0 gives for A(s) to be of second order.
must be zero
Also, the s row
(4.7)
4s+4 =0
S1 =tj Hence,
a(2+K)-(K +1) =0
Hence, s= S1 -2 implies s =-2+j
Solving equations (4.6) and (4.7),
we get

RP A A system oscillates with a frequency w if it has poles at s = tjw and no poles

a =0.75 and K =2
in the right half of the splane. Determine the values of K and a so that the system shown characterized by
in Fig. RP.44 oscillates at a frequency of 2 rad/sec. RP25 A negative feedback control system is

Ks+1)
s3 tas2+ 2s+1 Co) Glo)H) s(s+5s +9)
Determine the maximum value of K for stability.

Solution
write e 1- Hence,
Fig. RP.4.4
For small values of s, we can s.

Solution Gs)H() K(1-s)

s(s +5s+9)
The systern characteristic equation is given by The system characteristic equation is given by
1+G(s)H(s) = 0

1+ Ks+l =0 14
*" S=0
s +as +2s + 1 s(s2 +5s +9)
s+as+ (2+ K)s + (K+1) =0 s+5s+(9- K)s +k= 0
The Routh's array is The Routh's array is

1 2+K
3 9-K
K+1
K
af2+K)-(K+1)
s 45-6K
K+1
K
 Stability of Lincar Control Systems 247 Control Lngineerin
Conditions for stability:
Conditions for stability:
i 4 0, K< 75 0.205/*>0 K< 17.6
i. 3.6
ii. K>0
ii. 2+K>0 K> -2
Hence, for closed-loop stability K >0. Hence, for stability,
Since, K is always positive, the second condition means
0<K <7.5 0<K<17.6.
The maximum value of K for trial and error)
stability is 7.5. b . For K =
9, the roots of the characteristic equation are (found by
s1200, s2,3 -0.33 t j2.23 and s4 =-4.35
RR Consider the closed-loop control system shown in Fig. RP. 4.6.
=

a. Calculate the maximum value of K for the C . Assuming the complex roots are dorninant,
system that will result in a stable system.
b. For 0.51 of the maximum value
of K found in part (a), determine the roots of
the
S= cos p23
characteristic equation. 0.33
C. Estimate the overshoot of the system of part V(0.33)2+ (2.23)*
(b) when it is subjected to a step input. = 0.15 0.33

Controller Plant Mp =eVi-

R()- Hence, 2.23
0.S8+ 1)6+1 ) = 0.62
Fig. RP4.6a)
= 62%

0.005s+1 RP.4.7. The input to the

Controller 7 Refer the positional servomechanism shown in Fig.
Fig. RP 4.6 system is the reference shaft position, 0r and the system output is load shaft position 6.
Draw a block diagram of the system and then 6R{5) Find the range of Ka for the
fnd
closed-loop stability using Routh's criterion.
Solution
The system characteristic The parameters of the system are as given below:
a.
equation is given by
1+G(s)H(s) =0 1. Sensitivity of the input and output potentiometer: Kp=10 V/rad.
K 2. Amplifer gain: KA V/V.
6+2)0.5s+ 1 ( 6 + 1 ) * (0.0065s + 1 ) = 0
3. Motor field resistance: R = 50 .

0.00254+0.5125s +2.52+4.01s + (2 +K) = 0 4. Motor field inductance: Ls = 10 H.

The Routh's array is given by 5. Motortorque constant: Ky = 10 N-m/A.
6. Moment of inertia ofload: JL = 200 Kg-m?.
0.0025 2.52 2+K
7. Coefficient of viscous friction of the load: B = 2000 N-m/rad/sec.
S 0.5125 4.01 8. Motor to load gear ratio: n= =

25 2.5 2+K 9. Load to potentiometer gear ratio:=1.

10. Motor inertia and friction ar negligible.
s3.6-0.205K
2+K_
 Stability ofLinearControl Systems02 Control Engineering
Ti for the following
K> 0 and
RP 12 Refer the block diagram shown in Fig. RP.4.12. Find the
range of K >0 as
Determine the stability limits on the control parameters
a function of K1 > 0 such that the
system is stable. Comment on the effect of increasing CAses
K1 on the limit on gain K for stability. 0); and
Proportional-only control (Ti
=

(a)
100 (b) Proportional-derivative control (Ti> 0).
s(s+5)s+10) C()
Solution
characteristic equation is
The system
1+G(3)H(s) = 0
Fig. RP.4.12
1K(1+Tis)=0
Solution *2 (Ts+1)
The system characteristic equation is
T8+s+KT1s +K =0
The Routh's array tabulation is as shown below:
1+ G(s)H(s) = 0

s+15s +(50+ 100K,)s + 100K =0 T KT1

The Routh's array is as shown below:
1

500+100K1 s KT-TK
15 100K K

15(500+100K1)-100OK For closed-loop stability, the following conditions must be true:

1. 7>0
100K
2. K>0
For closed-loop stability, the following conditions must be satisfied: 3. KT-TK >0
1. 100K> 0 K > 0
() Tf Ti= 0, then condition (3) gives
15(500+ 100K1) -100K ,a
15 -TK >0

K< K <0
(7.5 +15K1)
Since the above condition on K contradicts the condition K > 0, it implies tha
Comment: Limit on K for stability increases closed-loop system is always unstable.
as K1 increases.

RPE A negative feedback control has (b) When Ti>0, condition (3) gives
KT1>TK
G()=1+71s)
TiT
rs +1) for closed-loop stability.