3 Time Response Analysis of

Control Systems

3.1 Introduction
The first st~p in the analysis of a control system is, describing the system in terms of a mathematical
model. In chapter 2 we have seen how any given system is modelled by defining its transfer function.
The next step would be, to obtain its response, both transient and steadystate, to a specific input. The
input can be a time varying function which may be described by known mathematical functions or it
may be a random signal. Moreover these input signals may not be known apriori. Thus it is customary
to subject the control system to some standard input test signals which strain the system very severely.
These standard input signals are : an impulse, a step, a ramp and a parabolic input. Analysis and
design of control systems are carried out, defining certain performance measures for the system,
using these standard test signals.
It is also pertinent to mention that any arbitary time function can be expressed in terms of linear
combinations of these test signals and hence, if the system is linear, the output of the system can be
obtained easily by using supersition principle. Further, convolution integral can also be used to determine
the response of a linear system for any given input, if the response is knownfor a step or an impulse
input.
3.2 Standard Test Signals
3.2.1 Impulse Signal
An impulse signal is shown in Fig. 3.1.

f(t)
(A)

--.t
Fig. 3.1 An Impulse signal.
82 Control Systems

The impulse function is zero for all t ;t! 0 and it is infinity at t = O. It rises to infinity at t = 0- and
comes back to zero at t = 0+ enclosing a finite area. If this area is A it is called as an impulse function
of strength A. If A = 1 it is called a unit impulse function. Thus an impulse signal is denoted by
f(t) = A 8 (t).

3.2.2 Step Signal

A step signal is shown in Fig. 3.2.

Af--------

f(t)

o ~t

Fig. 3.2 A Step Signal.

It is zero for t < 0 and suddenly rises to a value A at t = 0 and remains at this value for t > 0: It is
denoted by f(t) = Au (t). If A = 1, it is called a unit step function.

3.2.3 Ramp signal

A ramp signal is shown in Fig. 3.3.

f(t)
1.0 ~t

Fig. 3.3 A Ramp Signal.

It is zero for t < 0 and uniformly increases with a slope equal to A. It is denoted by f (t) = At.
If the slope is unity, then it is called a unit ramp signal.

3.2.4 Parabolic signal

A parabolic signal is shpwn in Fig. 3.4.

A
2
f(t)
1.0 ~t

Fig. 3.4 A unit parabolic signal.

Time Response Analysis of Control Systems 83

At2
A parabolic signal is denoted by f (t) == 2' If A is equal to unity then it is known as a unit
parabolic signal.
It can be easily verified that the step function is obtained by integrating the impulse function from
o to 00; a ramp function is obtained by integrating the step function and finally the prabolic function
is obtained by integrating the ramp function. Similarly ramp function, step function and impulse
function can be to obtained by successive differentiations of the parabolic function.
Such a set of functions which are derived from one another are knowp. as singularity functions. If
the response of a linear system is known for anyone of these input signals, the response to any other
signal, out of these singularity functions, can be obtained by either differentiation or integration of the
known response.

3.3 Representation of Systems

The input output description of the system is mathematically represented either as a differential
equation or a transfer function.
The differential equation representation is known as a time domain representation and the transfer
function is said to be a frequency domain representaiton. We will be considering the transfer function
representation for all our analysis and design of control systems.
The open loop transfer function of a system is represented in the following two forms.
1. Pole-zero form
G(s)==K (s+z,)(S+z2)···(s+zm) ..... (3.1)
I (s+P,)(s+P2) ... (s+Pn)
Zeros occur at s == -zl' -~, - - -, -~

Poles occur at s == -PI' - P2' - - -, -Pm

The poles and zeros may be simple or repeated. Poles and zeros may occur at the origin. In the
case where some of the poles occur at the origin, the transfer function may be written as
K, (s + z,)(s + Z2)"'(S + zm)
G( s) == -r----'--'--'-'--'--"-'----'---"'-'-- ..... (3.2)
S (s+Pr+,)(s+Pr+2) ... (s+Pn)

The poles at the origin are given by the term ~. The term !s indicates an integration in the
sr

1
system and hence - indicates the number of integrations present in the system. Poles at
sr
origin influence the steadystate performance of the system as will be explained later in this
chapter. Hence the systems are classified according to the number of poles at the origion.
If r == 0, the system has no pole at the origin and hence is known as a type - system.
If r == 1, there is one pole at the origin and the system is known as a type - 1 system. Similarly
°
if r == 2, the system is known as type - 2 system. Thus it is clear that the type of a system is
given by the number of poles it has at the origin.
84 Control Systems

2. Time Constant Form

The open loop transfer function of a system may also be written as,
K('tzl s + 1) ('tz2 S + 1) ... ('t
zm S + 1)
G(s) = -----=------=--.,--------=------,--- ..... (3.3)
('tP1s + 1) ('t P2 s + 1) ... ('tPn S + 1)
The poles and zeros are related to the respective time constants by the relation

for i = 1, 2, ..... m

1
p. = - for j = 1, 2, .....n
J 't
PJ
The gain constans KI and K are related by
m
z, 1t
K=K 1 ~
n
1t PJ
J=I
The two forms described above are equivalent and are used whereever convenience demands
the use of a particular form.
In either of the forms, the degree of the denomination polynomial of G(s) is known as the
order of the system. The complexity of the system is indicated by the order of the system. In
general, systems of order greater than 2, are difficult to analyse and hence, it is a practice to
approximate higher order systems by second order systems, for the purpose of analysis.
Let us now find the response of first order and second order systems to the test signals
discussed in the previous section.
The impulse test signal is difficult to produce in a laboratory. But the response of a system to
an impulse has great significance in studying the behavior of the system. The response to a unit
impulse is known as impulse response ofthe system. This is also known as the natural response
of the system.
For a unit impulse function, R(s) = 1
and C(s) = T(s).1
and c(t) = f l [T(s)]
The Laplace inverse ofT(s) is the impulse response of the system and is usually denoted by h(t).
.. f l [T(s)] = h(t)
If we know the impulse response of any system, we can easily calculate the response to any
other arbitrary input vet) by using convolution integral, namely
t
c(t) = f h('t) vet -
.) d't
o
Since the impulse function is difficult to generate in a laboratory at is seldon used as a test signal.
Therefore, we will concentrate on other three inputs, namely, unit step, unit velocity and unit
acceleration inputs and find the response of first order and second order systems to these inputs.
Time Response Analysis of Control Systems 85

3.4 First Order System

3.4.1 Response to a Unit Step Input

Consider a feedback system with G(s) = ~ as show in Fig. 3.5.

1:S
R(s) + C(s)

Fig. 3.5 A first order feedback system.

The closed loop transfer function of the system is given by

T(s) = C(s) = _1_ ..... (3.4)
R(s) 1:S + 1

For a unit step input R (s) = !s and the output is given by

I
C(s)- - - - ..... (3.5)
s(1:S + 1)
Inverse Laplace transformation yields
c(t) = 1 - e-th ..... (3.6)
The plot of c(t) Vs t is shown in Fig. 3.6.

1.0
0.632 -
c(t)

Fig. 3.6 Unit step response of a first order system.

The response is an exponentially increasing function and it approaches a value of unity as t ~ 00.
At t = 1: the response reaches a value,
c(1:) = 1 - e- l = 0.632
which is 63.2 percent of the steady value. This time, 1:, is known as the time constant of the system.
One of the characteristics which we would like to know about the system is its speed of response or
how fast the response is approaching the final value. The time constant 1: is indicative of this measure
and the speed of response is inversely proportional to the time constant of the system.
 86 Control Systems

Another important characteristic of the system is the error between the desired value and the
actual value under steady state conditions. This quantity is known as the steady state error of the
- system and is denoted by ess.
The error E(s) for a unity feedback system is given by
E(s) = R(s) - C(s)

= R (s) _ G(s) R(s)

1 + G(s)

R(s)
..... (3.7)
1 + G(s)

1 1
For the system under consideration G(s) = -, R(s) = - and therefore
'ts s
't
E(s) = --
'ts + 1
e (t) = e- tJ,
As t ~ IX) e (t) ~ O. Thus the output of the first order system approaches the reference input,
which is the desired output, without any error. In other words, we say a first order system tracks the
step input without any steadystate error.

3.4.2 Response to a Unit Ramp Input or Unit Velocity Input

The response of the system in Fig. 3.4 for a unit ramp input, for which,
1
R(s) = 2 '
s
is given by,
1
C(s) = S2 ('ts + 1) ..... (3.9)

The time response is obtained by taking inverse Laplace transform of eqn. (3.9).
c(t) = t - 't (1 - e-t1 ,) ..... (3.10)
If eqn. (3.10) is differentiated we get
dc(t) -tJ,
-- = 1-e ..... (3.11)
dt
Eqn. (3.11) is seen to be identical to eqn. (3.6) which is the response of the system to a step input.
Thus no additional information about the speed of response is obtained by considering a ramp input.
But let us see the effect on the steadystate error. As before,
1 'ts 't
E(s) = ~ . 'ts + 1 = s('ts + 1)

e (t) = 't (1 - e-t1,)

Time Response Analysis of Control Systems 87

Lt
and e ss = e(t) = 1:' ..... (3.12)
t~oo

Thus the steady state error is equal to the time constant of the system. The first order system,
therefore, can not track the ramp input without a finite steady state error. If the time constant is
reduced not only the speed of response increases but also the steady state error for ramp input
decreases. Hence the ramp input is important to the extent that it produces a finite steady state error.
Instead of finding the entire response, it is sufficient to estimate the steady state value by using the
final value theorem. Thus

ess = Lt s E(s)
s~O

Lt 1:'S
S ~O s(1:'s+ 1)
=1:'
which is same as given by eqn. (3.12)
The response ofa first order system for unit ramp input is plotted in Fig. 3.7.

c(t)

ret)

Fig. 3.7 Unit ramp response of a first order system.

3.4.3 Response to a Unit Parabolic or Acceleration Input

The response of a first order system to a unit parabolic input, for which
1
R(s) = - is given by,
S3

1
C( s) = -'s3:-(-1:'s-+-I-)

..... (3.13)
88 Control Systems

Differentiating eqn. (3.13), we get,

1
dc(t) --I
- - =-,+t+, e t
dt

..... (3.14)

Eqn. (3.14) is seen to be same as eqn. (3.10), which is the response of the first order system to
unit velocity input. Thus subjecting the first order system to a unit parabolic input does not give any
additional information regarding transient behaviour of the system. But, the steady state error, for a
prabolic input is given by,
e (t) = r (t) - c (t)

t2 t2 -~t
= - _,2+,t __ +,2 e t
2 2
Lt
ess e (t) = 00
t~oo

Thus a first ordr system has infinite state error for a prabolic input. The steady state error can be
easily obtained by using the final value theorem as :
Lt Lt R(s)
ess s E(s) =
s~O s~O ,+1

Lt s.1
=00
s~O s\,s + 1)
Summarizing the analysis of first order system, we can say that the step input yields the desired
information about the speed of transient response. It is observed that the speed of response is inversely
proportional to the time constant r of the system. The ramp and parabolic inputs do not give any
additional information regarding the speed of response. However, the steady state errors are different
for these three different inputs. For a step input, the steadystate error ess is zero, for a velocity input
there is a finite error equal to the time constant r of the system and for an acceleration input the
steadystate error is infinity.
It is clear from the discussion above, that it is sufficient to study the behaviour of any system to
a unit step input for understanding its transient response and use the velocity input and acceleration
input for understanding the steady state behaviour of the system.

3.5 Second Order System

3.5.1 Response to a Unit Step Input

Consider a Type 1, second order system as shown in Fig. 3.8. Since G(s) has one pole at the origin,
it is a type one system.
Time Response Analysis of Control Systems 89

R(s) + C(s)

Fig. 3.8 Second Order System.

The closed loop transfer function is give by,

T(s) = C(s) = 2 K ..... (3.15)

R(s) 'ts +s+K
The transient response of any system depends on the poles of the transfer function T(s). The
roots of the denominator polynomial in s of T(s) are the poles of the transfer function. Thus the
denominator polynomial ofT(s), given by
D(s) = 't s2 + S +K
is known as the characteristic polynomial of the system and D(s) = 0 is known as the characteristic
equation of the system. Eqn. (3.15) is normally put in standard from, given by,
KI't
T(s)= - - - -
S2 +!s+K/'t
't

00 2
n
..... (3.16)

Where,

OOn = ~ = natural frequency

o= ik = damping factor

The poles of T( s), or, the roots of the characteristic equation

S2 +2 0 00
ns + 00 n2 = 0

_ 2000 n ± ~40200n 2 - 400 n 2

are given by, Sl,2 = ---"'---=--2-"------:::..-

= - 0 OOn ±j OOn ~ (assuming 0 < 1)

= - 0 OOn ±j OOd
90 Control Systems

Where OJd = OJn ~ is known as the damped natural frequency of the system. If 0 > 1, the
two roots sl' s2 are real and we have an over damped system. If 0 = 1, the system is known as a
critically damped system. The more common case of 0 < 1 is known as the under damped system.
If ron is held constant and 8 is changed from 0 to 00, the locus of the roots is shown in Fig. 3.9.
The magnitude of s 1 or s2 is OJn and is independent of 8. Hence the locus is a semicircle with radius OJn
until 0= 1. At 0= 0, the roots are purely imaginary and are given by sl 2 = ± jOJn • For 0= 1, the roots
are purely real, negative and equal to - OJn . As oincreases beyond unity: the roots are real and negative
and one root approached the origin and the other approaches infinity as shown in Fig. 3.9.
1m s
8=0
s-plane

8>1 8=1
----~------~~.-~-----------Res
--(J)n

Fig. 3.9 Locus of the roots of the characteristic equation.

1
For a unit step input R(s) = - and eqn. 3.16 can be written as
s

co 2 1
C(s) = T(s). R(s) = 2 n 2 • - ..... (3.17)
S + 28co ns + COn S
Splitting eqn. (3.17) in to partial fractions, assuming oto be less than 1, we have
- Kl K 2 s + K3
+ ----=------=-------:-
C (S) - -
S S2 + 28co ns + con 2
Evaluating K 1, K2 and K3 by the usual procedure, we have,
1 s + 28co n
C (s) = - - --------7---
S (s + 8CO n)2 + con 2(1- ( 2)

s ~1-82 (S+8CO n)2 + COn 2(1_8 2 )

..... (3.18)
Taking inverse Laplae transform of eqn. (3.18), we have

c(t) = 1- e-'-',' [cosm, JI-6' t+ hsmm,~ t] ..... (3.19)

Time Response Analysis of Control Systems 91

Eqn. (3.19) can be put in a more convenient from as,

Where ffid = ffin ~

and ..... (3.20)

This response is plotted in Fig. 3.10. The response is oscillatory and as t ~ 00, it approaches unity.

Fig. 3.10 Step response of an underdamped second order system.

If t5 = 1, the two roots of the characteristic equations are SI =. s2 = -OJn and the response is
given by

C(s)= (S+ffi )2·;

n

and c (t) = 1 - e--<Ont - t ffin e--<Ont ..... (3.21)

This is plotted in Fig. 3.11.

1.0 - - - - - --

Fig.3.11 Response of a critically damped second order system.

As the damping is increased from a value less than unity, the oscillations decrease and when the
damping factor equals unity the oscillations just disappear. If t5is increaed beyond unity, the roots of
the characteristic equation are real and negative and hence, the response approaches unity in an
exponential way. This response is known as overdamped response and is shown in Fig. 3.12.
92 Control Systems

l0r------------=====~--

c(t)

~t

Fig. 3.12 Step response of an overdamped second order system.

c (t) = Kl e-sl t + ~ e-s2t ..... (3.22)

Where sl and s2 are given by,

SI,2 =-8 ().)n±().)n ~

and Kl and K2 are constants.

3.5.2 Response to a Unit Ramp Input

For a unit ramp input,
1
R (s) = --
S2

and the output is given by,

().) 2
C (S) = n
S2(S2 +28ro ns+ro n2 )
Taking inverse Laplace transform, we get the time response c (t) as,

c (t) =
28
t - -- +
e-IiOO n t (c-;::;))
Sin ron -Jl- 8 2 t + ~ for 8 < 1 ..... (3.23)
ron ron ~l- 8 2

The time response for a unit ramp input is plotted in Fig. 3.13.

215

c(t)

Fig. 3.13 Unit ramp response of a second order system.

Time Response Analysis of Control Systems 93

The response reveals two aspects of the system.

I. The transient response is of the same form as that of a unit step response. No new information
is obtained regarding speed of response or oscillations in the system.

2. It has a steadystate error ess = 28 ,unlike the step response, where the steady state error was
wn
zero. Thus, no new information is gained by obtaining the transient response of the system for a
ramp input. The steadystate error could be easily calculated using final value theorem instead of
laboriously solving for the entire reponse. For the given system, the error E (s) is given by
E (s) = R (s) - C (s)
,., 2
UJ
n _ S
2
+ 21:uWns + Wn 2 - Wn 2
S2 S2(S2 + 28w n s + 00/) - S2(S2 + 28w n s + 00/)

and from the final value theorem,

..... (3.24)

In a similar manner, the unit parabolic input does not yield any fresh information about the transient
response. The steadystate error can be obtained using final value theorem in this case also. For the
given system, for a unit acceleration input,
e55 = ao ..... (3.25)

3.5.3 Time Domain Specifications of a Second Order System

The performance of a system is usually evaluated in terms of the following qualities.
I. How fast it is able to respond to the input,
2. How fast it is reaching the desired output,
3. What is the error between the desired output and the actual output, once the transients
die down and steady state is achieved,
4. Does it oscillate around the desired value,
and 5. Is the output continuously increasing with time or is it bounded.
The last aspect is concerned with the stability of the system and we would require the system to
be stable. This aspect will be considered later. The first four questions will be answered in terms of
time domain specifications of the system based on its response to a unit step input. These are the
specifications to be given for the design of a controller for a given system.
In section 3.5, we have obtained the response of a type I second order system to a unit step input.
The step response of a typical underdamped second order system is plotted in Fig. 3.14.
94 Control Systems

It is observed that, for an underdamped system, there are two complex conjugate poles. Usually,
even if a system is of higher order, the two complex conjugate poles nearest to the j OJ - axis (called
dominant poles) are considered and the system is approximated by a second order system. Thus, in
designing any system, certain design specifications are given based on the typical underdamped step
response shown as Fig. 3.14.
2.0 \

c(t) t \\\ \

"- tolerance band

"
t
--=-----=--=----
1.0
-== 1-=---
0.5

ts t.D
Fig. 3.14 Time domain specifications of a second order system.

The design specifications are:

1. Delay time td: It is the time required for the response to reach 50% of the steady state value
for the first time
2. Rise time tr: It is the time required for the response to reach 100% of the steady state value
for under damped systems. However, for over damped systems, it is taken as the time required
for the response to rise from 10% to 90% of the steadystate value.
3. Peak time tp: It is the time required for the response to reach the maximum or Peak value of
the response.
4. Peak overshoot M : It is defined as the difference between the peak value of the response and
the steady state vafue. It is usually expressed in percent of the steady state value. If the time for
the peak is tp ' percent peak overshoot is given by,
c(tp) - c(oo)
Percent peak overshoot ~ = c(00) x 100. . .... (3 .26)

For systems of type 1 and higher, the steady state value c (00) is equal to unity, the same as the
input.
5. Settling time ts : It is the time required for the response to reach and remain within a specified
tolerance limits (usually ± 2% or ± 5%) around the steady state value.
6. Steady state error ess : It is the error betwen the desired output and the actual output as t ~ 00
or under steadystate conditions. The desired output is given by the reference input r (t) and
Lt
therefore, e ss = [ret) - c(t)]
t~oo
Time Response Analysis of Control Systems 95

From the above specifications it can be easily seen that the time response of a system for a unit
step input is almost fixed once these specifications are given. But it is to be observed that all the above
specifications are not independent of each other and hence they have to be specified in such a way
that they are consistent with others.
Let us now obtain the expressions for some of the above design specifications in terms of the
damping factor 0 and natural frequency (On"
1. Rise time ( tT )
If we consider an underdamped system, from the definition of the rise time, it is the time
required for the response to reach 100% of its steadystate value for the first time. Hence from
eqn. (3.20).

Or

e-&Dntr
Since r:--:::;2 cannot be equal to zero,
"1- 0
Sin (rod tr + 4» = 0
rod tr + 4> = 1t

-I ~1-02
1t - tan
1t-4> o
and t = ----===== ..... (3.27)
r ron~

2. Peak time (tp )

At the peak time, tp, the response attains its maximum value and this can be obtained by
differentiating c (t) and equating it to zero. Thus,
-ocont
dc(t) oro --.'5 t • e
-- = n e (On SIll (rod t + 4» - -~==2= COS (rodt + 4». rod = 0
dt ~1-02 1-0

Simplifying we have,

o Sin (rodt + 4» - ~1- 0 2 cos (rodt + 4» = 0

96 Control Systems

This can be written as,

Cos ~ Sin (rodt + ~) - Sin ~ cos (rodt + ~) =0

~1-o2
where tan~= 0

or for n = 0, 1, 2, ...
Here
n = 0 Corresponds to its minimum value at t = 0
n = 1 Corresponds to its first peak value at t = tp
n = 2 Corresponds to its first undershoot
n = 3 Corresponds to its second overshoot and so on
Hence for n = 1
1t
t = -----;=== ..... (3.28)
p ron~
Thus, we see that the peak time depends on both wn and 15. If we consider the product of wn
and tp ' which may be called as normalised peak time, we can plot the variation of this normalised
peak time with the damping factor t5. This is shown in Fig. 3.15.

4.0

3.8

3.6

0.2 0.4 0.6 0.8 1.0

~o

Fig. 3.15 Normalised peak time ron tp Vs 0 for a second order system.
3. Peak overshoot ( MJT)
The peak overshoot is defmed as
~ = c (tp ) - 1
Time Response Analysis of Control Systems 97

oron lt
e- ron~
Mp = ~ Sin ~ (": Sin (1t +~) = -Sin~)

-ltO

= e
k02
CSin~=~)
Hence, peak overshoot, expressed as a percentage of steady state value, is given by,
-ltO

M p = 100
JI 02
e - % ..... (3.29)
It may be observed that peak overshoot My' is a function of the damping factor 0 only. Its
variation with damping factor is shown in rig. 3.16.

100

80

60
%M p
40
20

0.2 0.4 0.6 0.8 l.0

-'0
Fig. 3.16 Percent overshoot Mp Vs 0 for a second order system.
4. Settling time ( ts )
The time varying term in the step response, c (t), consists of a product of two terms; namely,

an exponentially delaying term, h -0 ron t

1-8
and a sinusoidal term, Sin (OJ; + rP). It is clear that

the response is a decaying sinusoid, the envelop of which is given by b.-oro t

1-8
Thus, the

response reaches and remains within a given band, around the steadystate value, when this
envelop crosses the tolerance band. Once this envelop reaches this value, there is no possibility
of subsequent oscillations to go beyond these tolerane limits. Thus for a 2% tolerance band,
e-0 ronts
r---2 = 0.02
"1-8
For low values of 0, (j « 1 and therefore e-O ronts ::::: 0.02
4
t---=4. ..... (3.30)
s - 8ro
n
where .. is the time constant of the exponential term.
98 Control Systems

Eqn. (3.30) shows that the settling time is a function of both 8and lVn . Since damping factor is an
important design specification, we would like to know the variation of the setting time with 8, with lVn
fixed. Or, in otherwords, we can define a normalised time lVis' and find the variation of this quantity
with respect to 8. The step response of a second order system is plotted in Fig. 3.17 for different
values of 8, taking normalised time lVi, on x-axis. The curves are magnified around the steady state
value for clarity.
0=0.76
1.02
1.0
0.98

c(t)

0.9

Fig. 3.17 C (t) plotted for different value of o.

The settling time monotonically decreases as the damping is decreased from a value greater than
one (over damped) to less than one (under damped). For 2% tolerance band, it decreases until the
first peak of the response reaches the tolerance limit of 1.02 as shown by the curve IV in Fig. 3.17.
Points A, B, C, and D marked on the graph give the values avs' for decreasing values of O. The peak
value of the response reaches 1.02 at a damping factor 8= 0.76. The settling time for this value of 8
is marked as point D on the curve. If 8 is decreased further, since the response crosses the upper
limit 1.02, the point E no longer represents the settling time. The settling time suddenly jumps to a
value given by the point F on the curve. Thus there is a discontinuity at 8 = 0.76. If 8 is decreased
further the setting time increases until the first undershoot touches the lower limit of 0.98. Similarly,
the third discontinuity occurs when the second peak touches the upper limit of 1.02 and so on. The
variation of lVis with 8 for a tolerance band of 2% is plotted in Fig. 3.18.

i
8
~
I~
6
lil n ts
4 I: V
2 I 1 1
0.76 1

0.2 0.4 0.6 0.8 1.0

0.0
-70
Fig. 3.18 Variation of normalised settling time lilnt. Vs o.
Time Response Analysis of Control Systems 99

From Fig. 3.18 it is observed that the least settling time is obtained for a damping factor of
0= 0.76. Since settling time is a measure of how fast the system reaches a steady value, control
systems are usually designed with a damping factor of around 0.7. Sometimes. the systems are
designed to have even lesser damping factor because of the presence of certain nonlinearities which
tend to produce an error under steadystate conditions. To reduce this steadystate error, normally the

system gain K is increased, which in turn.decreases the damping (.: 8 = 2.J~T J. However, for
robotic control, the damping is made close to and slightly higher than unity. This is because the output
of a robotic system should reach the desired value as fast as possible, but it should never overshoot it.
S. Steady state error (e ss )
For a type 1 system, considered for obtaining the design specifications of a second order control
system, the steady state error for a step input is obviously zero. Thus
Lt
ess = 1 - c(t) =0
t~oo

The steady state error for a ramp input was obtained in eqn. (3.24) as ess = 28
ron
As the steadystate error, for various test signals, depends on the type of the system, it is dealt in
the next section in detail.

3.6 Steady State Errors

One of the important design specifications for a control system is the steadystate error. The steady
state output of any system should be as close to desired output as possible. If it deviates from this
desired output, the performance of the system is not satisfactory under steadystate conditions. The
steadystate error reflects the accuracy of the system. Among many reasons for these errors, the
most important ones are the type of input, the type of the system and the nonlinearities present in the
system. Since the actual input in a physical system is often a random signal, the steady state errors
are obtained for the standard test signals, namely, step, ramp and parbaolic signals.

3.6.1 Error Constants

Let us consider a feedback control system shown in Fig. 3.19.

R(s) C(s)

Fig. 3.19 Feedback 'Control System.

100 Control Systems

The error signal E (s) is given by

E (s) = R (s) - H (s) C (s) ..... (3.31)
But C (s) = G (s) E (s) ..... (3.32)
From eqns. (3.31) and (3.32) we have

E (s) - R(s)
- 1 + G(s)H(s)
Applying fmal value theorem, we can get the steady state error ess as,
e Lt s E(s) = Lt
= sR(s) ..... (3.33)
ss
s~0 s ~ 0 I+G(s)H(s)
Eqn. (3.33) shows that the steady state error is a function of the input R(s) and the open loop
transfer function G(s). Let us consider various standard test signals and obtain the steadystate error
for these inputs.
1. Unit step or position input.
1
For a unit step input, R (s) = -. Hence from eqn. (3.33)
s
1
Lt
s.-
s
s ~ 0 1+ G(s)H(s)

1
..... (3.34)
1 + Lt G(s) H(s)
s~

Let us define a useful term, position error constant Kp as,

~ ~ Lt G(s) H(s) ..... (3.35)

s~O
In terms of the position error constant, ess can be written as,
1
e =-- ..... (3.36)
ss I+K p
2. Unit ramp or velocity input.

For unit velocity input, R(s) = ~ and hence,

s
1
Lt s.- Lt 1
e = s
ss s~O I+G(s)H(s> s ~ 0 s + sG(s)H(s)

..... (3.37)
Lt
sG(s)H(s)
s~O
Time Response Analysis of Control Systems 101

Again, defining the velocity error constant K.. as,

Lt
K.. = s G(s) H (s) ..... (3.38)
s~O

e =- ..... (3.39)
55 Kv
3. Unit parabolic or acceleration input.
1
For unit acceleration input R (s) = 3"" and hence
s
Lt s Lt
e ss

..... (3.40)
Lt 2
s G(s)H(s)
s~O

Defining the acceleration error constant Ka as,

Lt
K = s2 G(s) H(s) ..... (3.41)
a s~O

1
e =- ..... (3.42)
55 Ka
For the special case of unity of feedback system, H (s) = 1 and eqns. (3.35) (3.38) or (3.41)
are modified as,
Lt
~ = s~o G(s) ..... (3.43)

Lt
K.. = sG (s) ..... (3.44)
s~O

Lt 2
and K = s G (S) ..... (3.45)
a S~O
In design specifications, instead of specifying the steady state error, it is a common practice to
specify the error constants which have a direct bearing on the steadystate error. As will be
seen later in this section, if the open loop transfer function is specified in time constant form,
as in eqn. (3.3), the error constant is equal to the gain of the open loop system.

3.6.2 Dependence of Steadystate Error on Type of the System

Let the loop transfer function G (s) H (s) or the open loop transfer function G (s) for a unity feedback
system, be giv·en is time constant form.
K(Tz1s + 1)(Tz2s + I) - - --
G(s) = r ..... (3.46)
S (Tp1s + 1)(Tp2 s + I) - - --
102 Control Systems

As s ~ 0, the poles at the origin dominate the expression for G(s). We had defined the type of a
system, as the number of poles present at the origin. Hence the steady state error, which depends on
Lt Lt Lt .
G(s), s G(s) or s2 G(s), IS dependent on the type of the system. Let us therefore
s~O s~O s~o
obtain the steady state error for various standard test signals for type-O, type-I and type-2 systems.
1. Type -0 system
From eqn. (3.46) with r = 0, the error constants are given by

Lt Lt
~ = CJ(s)
s~o s~o

Lt Lt
~ = s CJ(s)
s~o s~o
'"

Similarly K = Lt S2 CJ(s) = 0 ..... (3.47)

a s~O

The steady state errors for unit step, velocity and acceleration inputs are respectively, from
eqns. (3.34), (3.37) and (3.40),

e55 1
= --= --
1 (step mput
. )
I+Kp I+K

e S5 = _1_ = 00 (velocity input)

Kv
1
e = - = 00 (acceleration input)
55 K
a

2. Type 1 system
For type 1 system, r = 1 in eqn. (3.46) and
K
~= Lt (}(s) = Lt - = 00
s~O s
Lt Lt K
~= s CJ(s) = s. - = K
s~O s~O s
Lt Lt K
and K = s CJ(s) = s2. - = 0
a s~O s~O S
The steady state error for unit step, unit velocity and unit acceleration inputs are respectively,
1 1
e = - - = - =0 (position)
S5 l+K p 00
Time Response Analysis of Control Systems 103

1 1
e = - =- (velocity)
55 Ky K

and (acceleration)

3. Type 2-system
For a type - 2 system r = 2 in eqn. (3.46) and
LtK
K = Lt G(s) -=00
-" s ~O s~O S2

Lt Lt sK
~ = s G(s) -=00
s~O s~O S2

Lt Lt s2K
and Ka = s2 G(s) -=K
s~o s~O S2
The steady state errors for the three test inputs are,
1 1
e = - - = - - = 0 (position)
5S I+Kp 1+00

1 1
e=-=-=O (velocity)
ss Ky 00

1 1
and ess = K=
a
K (acceleration)

Thus a type zero system has a fmite steady state error for a unit step input and is equal to
1 1
e =--=-- ..... (3.47)
5S I+K I+K p
Where K is the system gain in the time constant from. It is customary to specify the gain of a
type zero system by ~ rather than K.
Similarly, a type -1 system has a finite steady state error for a velocity input only and is given
by
1 1
eS5 = K = K ..... (3.48)
y

Thus the gain of type -1 system in normally specified as ~.

A type -2 system has a finite steady state error only for acceleration input and is given by
1 1
ess = K =Ka ..... (3.49)

As before, the gain of type -2 system is specified as Ka rather than K.

104 Control Systems

The steady state errors, for various standard inputs for type - 0, type - 1 and type - 2 are
summarized in Table. 3.1.

Table. 3.1 Steady state errors for various inputs and type of systems

Steadystate error e ss

Standard input Type-O Type -1 Type - 2

Lt Lt Lt
~= G(s) ~ = s G(s) K = s2 G(s)
s~O s~O a s~ 0

1
Unit step -- 0 0
I+Kp

1
Unit velocity ao - 0
Kv

1
Unit acceleration ao ao -
Ka

If can be seen from Table. 3.1, as the type of the system and hence the number of integrations
increases, more and more steady state errors become zero. Hence it may appear that it is better to
design a system with more and more poles at the origin. But if the type of the system is higher than
2, the systems tend to be more unstable and the dynamic errors tend to be larger. The stability aspects
are considered in chapter 4.

3.6.3 Generalized Error Coefficients - Error Series

The main disadvantage of defining the.steadystate error in terms of error constants is that, only one
of the constants is finite and non zero for a particular system, where as the other constant's are either
zero or infmity. If any error constant is zero, the steady state error is infinity, but we do not have any
clue as to how the error is approaching infinity.
If the inputs are other than step, velocity or acceleration inputs, we can extend the concept of
error constants to include inputs which can be represented by a polynomial. Many functions which
are analytic can be represented by a polynomial in t. Let the error be given by,
R(s)
E(s)= I+G(s) ..... (3.50)

Eqn. (3.50) may be written as

E (s) = Yes). R(s) ..... (3.51)
. 1
Where Y(s)- - - ..... (3.52)
1+ G(s)