Chapter 3

Time-Domain Analysis of Control

Systems
 Contents

3.1 Time response of continuous-data system

3.2 The unit-step response and time-domain specifications
3.3 Steady-state error
3.4 Time response of a first-order system
3.5 Time response of a second-order system
3.6 Effects of adding poles and zeros to the transfer functions
3.7 Dominant poles and zeros of transfer functions

Nguyen Le Hoa, Ph.D., DUT-UD 2/27

 3.1 Time response of continuous-data system
Transient response vs. steady-state response
Definition: Time response of a control system is the output response which is a
function of time, to the applied input.

Solution: y (t ) = yt (t ) + yss (t ), t  0 yt(t): transient response

yss(t): steady-state response

Transient response, yt(t)

▪ Present in a short period of time immediately after the system is excited
▪ Reflect dynamic behavior of the system
▪ Dependent upon the system poles only and not on the type of input
▪ For the asymptotically stable system, lim yt (t ) = 0
t →

Steady-state response, yss(t):

▪ Remain after the transient response has died out
▪ Reflect final accuracy of the system

Nguyen Le Hoa, Ph.D., DUT-UD 3/27

 3.1 Time response of continuous-data system
Typical test input signals
Unit-step, us(t) Unit-impulse, δ(t)
1 t  0 dus  t = 0
 Lus (t ) =
1
▪ us (t ) =  ▪  (t ) = =  L (t ) = 1
0 t  0 s dt  0 t  0
▪ Useful in evaluating the system’s ▪ Impulse response = L-1 of system’s
quickness in responding the inputs transfer function
with abrupt changes Rt 2
Prabolic-function input r (t ) = u s (t )
2
Step-function input r (t ) = Ru s (t ) u
(t )
Rt

s
=
Ramp-function input r(t ) = Rus (t ) t)
R r(
▪ r (t ) Rtus (t ) δ(t)
▪ Has the ability to test how the us (t)
system would response to a signal 1
Rt 2
r (t ) = u s (t )
that changes linearly with time 2

0 t 0 t

Nguyen Le Hoa, Ph.D., DUT-UD 4/27

 3.2 The unit-step response and time-domain
specifications
3.2.1 Unit-step response
▪ Definition: The response of a system (with all zero initial conditions) to the
unit-step input
▪ Solution:
−1  G ( s ) 
G( s) r(t) = us(t) y(t) = h(t)
H ( s) = G( s)U s ( s) =  h(t ) = L   G(s)
s  s 
Ex. 3.1. Find the unit-step response of the following system
▪ Solution r y
e
Go(s)
Go ( s) 1
G( s) = = 2 -
1 + KGo ( s) s + 5s + 6
K
 G ( s) 
−1 −1  1 
h(t ) = L   = L  2 
 s   s ( s + 5 s + 6)  1
G (s) = ,
1 / 6 3 / 6 2 / 6  1
( ) s + 5s + 3
2
−1 −2t − 3t
=L  − +  = 1 − 3e + 2e
 s s + 2 s + 3 6 K =3

Nguyen Le Hoa, Ph.D., DUT-UD 5/27

 3.2 The unit-step response and time-domain
specifications
3.2.1 Unit-step response
Ex. 3.1. (cont’d)
Matlab statements for plotting the unit-step response of the above system are as
follows: Step Response

clear all 0.16

%Step-unit response 0.14
K=3; 0.12
Go=tf([1],[1 5 3]);

Amplitude
0.1
G=feedback(Go,K); 0.08
step(G); 0.06
xlabel('Time'); 0.04
ylabel('Amplitude'); 0.02
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (sec)

Nguyen Le Hoa, Ph.D., DUT-UD 6/27

 3.2 The unit-step response and time-domain
specifications
3.2.2 Time-domain specifications y(t)
▪ Overshoot Δm
- Maximum overshoot: Δm = ymax – yss 1.0
0.9 ± ε%
- Percent maximum overshoot (PO)
y − yss 
PO = max 100% = m 100% 0.5
yss yss td
- Use to measure relative stability
0.1
▪ Time delay, td : The time required for 0 t
the step response to reach 50% of its tr ts

final value
▪ Rise time, tr : The time required for the step response to rise from 10 to 90% of
its final value
▪ Settling time, ts : The time required for the step response to decrease and stay
with ± ε% (ε = 2 or 5) of its final value.
▪ Steady-state error, ess : ess = limr (t ) − y (t ) = lim e(t )
t → t →

Nguyen Le Hoa, Ph.D., DUT-UD 7/27

 3.3 Steady-state error
3.3.1 Introduction
Definition: The steady-state error is the difference between the output and the
reference (desired output) in the steady state.
Assume that the objective is to have the output y(t) to track the input r(t)
(reference), then
ess = lim r (t ) − y (t ) = lim e(t ) e(t): system error
t → t →

▪ If the closed-loop system is unstable, ess is meaningless:

▪ If the closed-loop system is stable, r u y
G(s)
ess = lim e(t ) = lim sE ( s ) R(s) - U(s) Y(s)
t → s →0

H(s)
▪ If H(s) = 1 (unity feedback):
E ( s ) = R( s ) − Y ( s ) = R( s )1 − M ( s ) M(s): closed-loop transfer function
G(s) R( s)
= R( s) − R( s) =
1 + G ( s) 1 + G(s)

Nguyen Le Hoa, Ph.D., DUT-UD 8/27

 3.3 Steady-state error
3.3.1 Introduction
▪ If H(s) # 1 (non unity feedback):
R(s) U(s) Y(s) R(s) U(s) Y(s) R(s) E(s) G (s) Y(s)
G(s) G(s)
- - - - 1 G ( s) H ( s) G ( s)

H(s)-1
H(s)

G( s) 1 + G( s)H ( s) − 1
E ( s ) = R( s ) − Y ( s ) = R( s ) − R( s ) = R( s )
1 + G( s) H ( s) 1 + G( s) H ( s)

Important notes!
▪ In practice, the input signal r(t) is usually a typical test signal (i.e., step-, ramp-,
parabolic-unit function), but may not the actual reference signal. In this case, the
system error E(s) obtained above cannot be applied.
R(s)
▪ If lim H ( s) H (0) K H , the reference signal can be defined as
s 0 KH
R( s)
▪ If H(s) has Nth-order zero at s = 0, the reference signal can be defined as
KH sN
Nguyen Le Hoa, Ph.D., DUT-UD 9/27
 3.3 Steady-state error
3.3.2 Steady-state error of a system with unity feedback
sR( s)
ess = lim sE ( s) = lim R(s) E(s) Y(s)
s →0 s →0 1 + G ( s ) G(s)
-
ess depends on the characteristics of G(s) and the type R(s)
▪ System type
In general, the system’s transfer function can be expressed as
M
K  (s − si )
G(s) = i =1
e −Td s , M  N + P; si , s j  0; K , Td  R
s N  (s − s j )
P

j =1

N – number of poles of G(s) at s = 0 (system type)

For examples
K (1 + 0.5s)
G( s) = Type 1
s(1 + s)(1 + 2s)(1 + s + s 2 )

K (1 + 2s)
G( s) = Type 3
s 3 (1 + s)
Nguyen Le Hoa, Ph.D., DUT-UD 10/27
 3.3 Steady-state error
3.3.2 Steady-state error of a system with unity feedback
R R R
▪ r(t) = Rus(t): R ( s ) =  ess = = Kp - step-error constant
s 1 + lim G ( s ) 1 + K p
s →0
r(t),
- N = 0 (type 0): Kp = const. ess = const. y(t) Ref. input,
Rus(t)
- N ≥ 1: Kp = ∞ ess = 0. R
R
e ss =
1+ K p
R
▪ r(t) = Rtus(t)  R( s) = output, y(t)
s2
R R R
ess = lim = = 0 t
s →0 s + sG ( s ) lim sG ( s) K v
s →0
r(t),
Kv - ramp-error constant y(t)
R
- N = 0 (type 0): Kv = 0, ess = ∞ ess =
Ref. input, Kv
Rtus(t)
- N = 1 (type 1): Kv = const. ess = const.
output, y(t)
- N ≥ 2: Kv = ∞ ess = 0.
0 t

Nguyen Le Hoa, Ph.D., DUT-UD 11/27

 3.3 Steady-state error
3.3.2 Steady-state error of a system with unity feedback
r(t),
2
t R y(t)
▪ r (t ) = R u s (t )  R( s) = 3
2 s
R R R R
ess = lim 2 2 = = 2
t e ss =
s →0 s + s G ( s ) lim s 2G ( s ) K a ref. input, R u s Ka
2
s →0
output, y(t)
Ka - parabolic-error constant
- N = 0 (type 0): Ka = 0. ess = ∞ 0 t

- N = 1 (type 1): Ka = 0. ess = ∞

- N = 2 (type 2): Ka = const. ess = const.
- N ≥ 3: Ka = ∞. ess = 0
K ( s 3.15)
Ex. 3.2. Consider a system with G ( s ) , and H ( s ) 1
s ( s 1.5)( s 0.5)
N = 1, therefore the system is type 1
R
- Step input: K p = lim G ( s) =   ess = =0
s →0 1+ K p
- Ramp input: R R
K v = lim sG ( s) = 4.2 K  ess = =
s →0 K 4.2 K
R v
- Parabolic input: K a = lim s G ( s) = 0  ess =
2
=
s →0 Ka
Nguyen Le Hoa, Ph.D., DUT-UD 12/27
 3.3 Steady-state error
3.3.2 Steady-state error of a system with unity feedback
Step response error
Ex. 3.2. (cont’d): Let K = 0.5 and R = 2 2

K=0.5;R=2; % Parabolic response 1.5

t=0:0.1:100; [y,t]=lsim(G,parabol,t); 1

e(t)=r(t)-y(t)
step=R*ones(size(t)); figure(3) 0.5
ramp=R*t; plot(t,parabol'-y); 0
parabol=R*t.^2/2; xlabel('Time'); -0.5
Go=tf(K*[1 3.15],[1 2 0.75 0]); ylabel('e(t)=r(t)-y(t)');
G=feedback(Go,1); -1
grid on
% Step response -1.5
0 20 40 60 80 100
[y,t]=lsim(G,step,t); Time
figure(1) Ramp response error Parabolic response error
plot(t,step'-y); 3 100
xlabel('Time'); 2.5
80
ylabel('e(t)=r(t)-y(t)'); 2
grid on
e(t)=r(t)-y(t)

e(t)=r(t)-y(t)
1.5 60
% Ramp response
1
[y,t]=lsim(G,ramp,t); 40

figure(2) 0.5
20
plot(t,ramp'-y); 0
xlabel('Time'); -0.5 0
0 20 40 60 80 100 0 20 40 60 80 100
ylabel('e(t)=r(t)-y(t)') Time Time
grid on
Nguyen Le Hoa, Ph.D., DUT-UD 13/27
 3.4 Time response of a first-order system
dy(t ) 1 K Y (s) K
▪ Model description + y (t ) = r (t )  G ( s) = =
dt   R( s ) s + 1 
K – steady-state gain, τ – time constant.
▪ Steady-state gain, K
1 K K K
If r(t) = us(t) (unit-step), then R( s) =  Y ( s) = = −
s s(s + 1  ) s s + 1 
 y (t ) = K ( )
− Ke −t / = K 1 − e −t / , t  0
yss yt
y(t)
▪ Time constant, τ
K
- If K = 1, then initial response rate
of the system is 0.63K
dy(t )
dt
|t =0 =
d
dt
(−e −t / 
)|t =0 =
1

- At t = τ, y(τ) = 0,63K = 63% of yss 0 τ t

- If τ > 0, G(s) has a only positive pose at s = - 1/τ => the system is always
stable
Nguyen Le Hoa, Ph.D., DUT-UD 14/27
 3.5 Time response of a second-order system
d 2 y (t ) dy(t )
▪ Model description T 2
2
+ 2 T + y (t ) = r (t )
dt dt R(s) E(s)  n2 Y(s)
Y ( s) n 2
s (s + 2 n )
 G(s) = = 2 r(t) - e(t) y(t)
R( s) s + 2n s + n 2

1.8
ζ – damping ratio ζ=0.1

ωn = 1/T – natural (undamped) frequency 1.6

0.2

▪ Unit-step response 1.4 0.3

0.4
If r(t) = us(t), then R(s) = 1/s 1.2 0.5

  
0.6
 
2
−1 G ( s ) −1
 y (t ) = L  =L  2

y (t )
0.7

2 
( )
n 1.0

 s   s s + 2 n s +  n 
0.8
1.0

( )
0.8

e − nt
1.2

= 1− sin n 1 −  2 t + cos−1  0.6

ζ=1.4

1−  2
0.4

▪ ζ ≥ 1, there is no overshoot
▪ ωn has a effect on tr, ts, and td but does not 0.2

effect on the overshoot

0 1 2 3 4 5 6 7 8 9 10 11 12 13
ω nt
Nguyen Le Hoa, Ph.D., DUT-UD 15/27
 3.5 Time response of a second-order system
▪ Root locus of characteristic equation Root
jω

s 2 + 2n s + n2 = 0  s1, 2 = −n  jn 1 −  2 ωn n 1 −  2

▪ Natural frequency q
ζωn σ
When ζ = 0 (undamped), s1,2 = ± jωn => y(t) = 1-cos(ωn t):
Step response is purely sinusoidal with frequency ωn
Root
▪ Damping factor
α = ζ ωn controls the rate of rise or decay jω ζ=0
(damping) of y(t), α is called the damping
factor. If ζ = 1(critical damping), then α = 0<ζ<1 0 > ζ > -1
ωn
ζ = -∞ ζ = -1
▪ Damping ratio ζ>1 ζ>1
ζ→∞ σ
 ζ < -1 ζ = -∞
 = damping ratio = ζ=1 ζ= ∞
 | =1
0> ζ > -1
0<ζ<1
ωn = const.
ζ=0

Nguyen Le Hoa, Ph.D., DUT-UD 16/27

 3.5 Time response of a second-order system
▪ Damping ratio jw y (t )
• ζ > 1 (overdamped)
1
ζ>1
s1, 2 = −n  jn 1 −  2  0
 y (t ) = 1 + k1e −t / 1 + k 2 e −t / 2 0 σ

1 0
 1, 2 = 0 t
n  jn 1 −  2 jw y (t )
• ζ = 1 (critically damped) ζ=1 1
s1 = s2 = −n
σ
 y (t ) = 1 + (k1 + k2t )e −tn
0

0
• 0 < ζ <1 (underdamped) t
jw
s1, 2 = −n  jn 1 −  2 y (t )

e − nt
sin (n  t +  )
1
 y (t ) = 1 − 0<ζ<1
 0 σ
 = 1 −  2 ;  = cos−1 
0 t

Nguyen Le Hoa, Ph.D., DUT-UD 17/27

 3.5 Time response of a second-order system
▪ Damping ratio jw y (t )
• ζ = 0 (undamped)
ζ=0
s1, 2 =  jn
1
 y (t ) = 1 − cos nt 0 σ

• -1 < ζ < 0 (negatively damped) 0 t

jw y (t )
s1, 2 = −n  jn 1 −  , − n  0
2

e − nt -1 < ζ < 0

 y (t ) = 1 − sin (n  t +  ) 1
 0 σ
0
t
• ζ < -1 (negatively damped)
jw y (t )
s1, 2 = −n  jn 1 −   0 2

ζ < -1
 y (t ) = 1 + k1e t /1 + k 2 et / 2
1
1 0 σ
 1, 2 = 0
− n  jn 1 −  2 0 t

Nguyen Le Hoa, Ph.D., DUT-UD 18/27

 3.5 Time response of a second-order system
▪ Maximum overshoot
Consider a typical response of a 2nd oder system (i.e., 0 < ζ <1)
e − nt
y (t ) = 1 − sin (n  t +  ),  = 1 −  2 ;  = cos−1 

dy n e −nt
 =  sin (n t +  ) −  cos(n t +  )
dt 
Use  = cos    = 1 −  = sin  , we obtain
2

dy n e −nt  n e −  n t
= sin (n t +  )cos − cos(n t +  )sin   = sin n t
dt  y(t) 
dy n
=0t = , n = 0, 1, 2,  m
dt n 
The maximum overshoot appears at 1
 
the first peak of y(t), i.e., at t = =
n  n 1 −  2
  m = y (t ) |  −1 = e − 1− 2
t=
n 
Percent maximum overshoot
 2 3 4
PO = e − 1− 2
100 0
n 1−  2
n 1 −  2
n 1 −  2
n 1 −  2
t

Nguyen Le Hoa, Ph.D., DUT-UD 19/27

 3.5 Time response of a second-order system
▪ Maximum overshoot PO, %
100
The dependence of PO on ζ is plotted 90
in the figure below. It shows that, very 80

roughly, we can approximate PO as 70

60   
follows: PO = 1 −  100
 0.6 
50
  
PO  1 −  100 40

 0.6  30
PO = e− 1− 2
100
20
10
ζ
0.0 0.2 0.4 0.6 0.8 1.0

Nguyen Le Hoa, Ph.D., DUT-UD 20/27

 3.5 Time response of a second-order system
▪ Settling time, ts
The settling time can be approximated as
 3 .2
   for  = 5%
1−
1
1− 
e − nt s = 1 −   ts =
1
 n
(
ln  1 −  2 )
−1
 n
4

2
for  = 2%
n

▪ Delay time, td and rise time, tr y(t)

1
1+ e −  n t
The delay time and the rise time 1− 2

can be approximated by the

following expressions ±ε
1+ε
1 + 0.7 0.8 + 2.5 1-ε
td  ; tr 
n n 1
1− e −nt
1− 2
 nt
0
ts

Note: The above approximations of ts are applied for systems with 0 < ζ < 0.69
Nguyen Le Hoa, Ph.D., DUT-UD 21/27
 3.6 Effects of adding poles and zeros to the TFs
3.6.1 Adding a pole to the forward-path transfer function
Consider a unity-feedback system with the transfer function as follows
n2
Go ( s ) = where the pole at s = -1/Tp is considered to be
s (s + 2n )(1 + T p s ) added to the second-order transfer function
Go ( s ) n2
The closed-loop TF: Gc ( s ) = =
1 + Go ( s ) T p s 3 + (1 + 2nTp )s 2 + 2n s + n2
Effects of the adding pole:
y(t) Step response
• Increase the maximum
2.0
overshoot of the closed-loop ζ = 1, ωn = 1
system’s response Tp = 5
• Increase the rise time of Tp = 2
Tp = 1
the closed-loop system’s 1.0
Tp = 0
response
Note: If Tp is too large, the
t
closed-loop system becomes 0 5 10 15 20
unstable
Nguyen Le Hoa, Ph.D., DUT-UD 22/27
 3.6 Effects of adding poles and zeros to the TFs
3.6.2 Adding a pole to the closed-loop transfer function
Consider a closed-loop transfer function as follows
n2
Gc ( s ) = 2 where the pole at s = -1/Tp is considered to be
(s + 2n s + n )(1 + Tp s ) added to the second-order transfer function
2

Effects of the adding pole:

y(t) Step response
• Decrease the maximum 1.8
overshoot of the system’s ζ = 0.5
Tp = 0 ωn = 1
response 1.2
Tp = 0.5
• Increase the rise time of the 1.0
Tp = 1
system’s response 0.8
Tp = 2 Tp = 4

Note: When the maximum 0.4

overshoot is concerned, adding a
pole to the closed-loop transfer t
0 3 6 9 12
function has the opposite effect
to that of adding a pole to the
forward-path transfer function

Nguyen Le Hoa, Ph.D., DUT-UD 23/27

 3.6 Effects of adding poles and zeros to the TFs
3.6.3 Adding a zero to the closed-loop transfer function
Consider a closed-loop system with the transfer function
n2 (1 + Tz s )
Gc ( s) = 2 where the zero at s = -1/Tz is considered to be
(s + 2n s + n2 ) added to the second-order transfer function

Effects of the adding zero: y(t) Step response

• Increase the maximum 5

Tz = 10

overshoot of the system’s ζ = 0.5

ωn = 1
response 4
• Decrease the rise time of the Tz = 6

system’s response 3

Tz = 3
2
Tz = 1
1
Tz = 0
t
0 3 6 9 12

Nguyen Le Hoa, Ph.D., DUT-UD 24/27

 3.6 Effects of adding poles and zeros to the TFs
3.6.4 Adding a zero to the forward-path transfer function
Consider a system with the forward-path transfer function
6(1 + Tz s )
Go ( s) = where the zero at s = -1/Tz is considered to be
s(s + 1)(s + 2) added to the third-order transfer function
The closed-loop transfer function can be obtained as
y(t) ζ = 0.5, ωn = 1
6(1 + Tz s ) 2.0
Gc ( s) = Tz = 5 Tz = 3 Tz = 0
s 3 + 3s 2 + (2 + 6Tz ) s + 6 Tz = 0.2
1.6
Effects of the adding zero: Tz = 0.5

• Tz appears in the numerator of 1.2

Gc(s) increases the maximum

overshoot. Tz appears in the 0.8

denominator of Gc(s) decreases

0.4
the maximum overshoot.
t
0 5 10 15

Nguyen Le Hoa, Ph.D., DUT-UD 25/27

 3.7 Dominant poles and zeros of TFs
The location of poles and zeros of a transfer function greatly effects the transient
response of the system
Dominant poles:
•Give rise relatively slow decay terms in the jω

transient response
•Locate close to the imaginary axis in the
Region of Region of
left-half s-plane insignificant dominant
unstable
Insignificant poles: region
poles poles
•Contribute fast decay terms to the transient
0 σ
response
D
•Locate far away from the imaginary axis in
the left-half s-plane
s-plane
Zeros that are close to imaginary axis in the left-half s-plane affect significantly
the transient response (dominant zeros)

Question: How large a pole is considered to be insignificant?

Answer: If the real part of a pole is at least 5 to 10 times that of a dominant pole

Nguyen Le Hoa, Ph.D., DUT-UD 26/27

 3.7 Dominant poles and zeros of TFs
For example: Consider the following third-order system
6000 Step Response
G( s) =
( )
100

(s + p ) s 2 + 12s + 100 80

Poles of the system: third-order system

60

Amplitude
s1 = − p; s2,3 = −6  j8
approximate first-order system

• If p = 0.6  Res2,3   10 s1
40

20

s1 is considered to be a dominant pole

0
6000 100 0 1 2 3 4 5 6 7 8 9 10

G( s)  Time (seconds)

s+ p Step Response
1.4
The system exhibits dominant first- 1.2
order behavior
• If p = 65  s1  10 Res2,3 
1
Amplitude

0.8
third-order system
0.6
s2,3 are considered to be dominant approximate second-order system
0.4
poles
6000 p 0.2

G(s)  2
s + 12s + 100
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time (seconds)

The system exhibits dominant second-order behavior

Nguyen Le Hoa, Ph.D., DUT-UD 27/27