Transient Response of a Second-Order System

ECEN 2830
Spring 2012

1. Introduction
In connection with this experiment, you are selecting the gains in your feedback loop to
obtain a well-behaved closed-loop response (from the reference voltage to the shaft
speed). The transfer function of this response contains two poles, which can be real or
complex. This document derives the step response of the general second-order step
response in detail, using partial fraction expansion as necessary.

lnLegral urlver& g
compensaLor moLor
Speed
reference
Wheel
speed
Speedsensor
Sensed Sensed
speed

2. Transient response of the general second-order system
Consider a circuit having the following second-order transfer function H(s):

v
out
(s)
v
in
(s)
= H(s) =
H
0
1 + 2
s

0
+
s

0
2
(1)
where H
0
, , and
0
are constants that depend on the circuit element values K, R, C, etc.
(For our experiment, v
in
is the speed reference voltage v
ref
, and v
out
is the wheel speed )
In the case of a passive circuit containing real positive inductor, capacitor, and resistor
values, the parameters and
0
are positive real numbers. The constants H
0
, , and
0

are found by comparing Eq. (1) with the actual transfer function of the circuit. It is
common practice to measure the transient response of the circuit using a unit step
function u(t) as an input test signal:
v
in
(t) = (1 V) u(t)
(2)
ECEN 2830
2
The initial conditions in the circuit are set to zero, and the output voltage waveform is
measured.
This test approximates the conditions of transients often encountered in actual
operation. It is usually desired that the output voltage waveform be an accurate
reproduction of the input (i.e., also a step function). However, the observed output
voltage waveform of the second order system deviates from a step function because it
exhibits ringing, overshoot, and nonzero rise time. Hence, we might try to select the
component values such that the ringing, overshoot, and rise time are minimized.
The output voltage waveform v
out
(t) can be found using the Laplace transform.
The transform of the input voltage is

v
in
(s) =
1
s
(3)
The Laplace transform of the output voltage is equal to the input v
in
(s) multiplied by the
transfer function H(s):

v
out
(s) = H(s) v
in
(s) =
1
s
H
0
1 + 2
s

0
+
s

0
2
(4)
The inverse transform is found via partial fraction expansion.
The roots of the denominator of v
out
(s) occur at s = 0 and (by use of the quadratic
formula) at

s
1
, s
2
=
0

0

2
1
(5)
Three cases occur:
> 1. The roots s
1
and s
2
are real. This is called the overdamped case.
= 1. The roots s
1
and s
2
are real and repeated: s
1
= s
2
=
0
. This case
is called critically damped.
< 1. The roots s
1
and s
2
are complex, and can be written

s
1
, s
2
=
0
j
0
1
2
(6)
This is called the underdamped case.
ECEN 2830
3
Figure 2 illustrates how the positions of
the roots, or poles, vary with . For =
, there are real poles at s = 0 and at s =
. As decreases from to 1, these
real poles move towards each other
until, at = 1, they both occur at s =

0
. Further decreasing causes the
poles to become complex conjugates as
given by Eq. (6). Figure 3 illustrates
how the poles then move around a circle
of radius
0
until, at = 0, the poles
have zero real parts and lie on the
imaginary axis. Figure 2 is called a root
locus diagram, because it illustrates how
the roots of the denominator polynomial of H(s) move in the complex plane as the
parameter is varied between 0 and .
Several other cases can be defined that are normally not useful in practical
engineering systems. When = 0, the roots have zero real parts. This is called the
undamped case, and the output voltage waveform is sinusoidal. The transient excited by
the step input does not decay for large t. When < 0, the roots have positive real parts
and lie in the right half of the complex plane. The output voltage response in this case is
unstable, because the expression for v
out
(t) contains exponentially growing terms that
increase without bound for large t.
Partial fraction expansion
is used below to derive the output
voltage waveforms for the cases
that are have useful engineering
applications, e.g. the overdamped,
critically damped, and
underdamped cases.
2.1. Overdamped case, > 1
Partial fraction expansion of Eq.
(4) leads to

Re (s)
Im (s)
LHP
RHP

0
=
=
=
=
=
j
0
j
0

Fig. 2. Location of the two poles of H(s) vs. , as described
by Eqs. (5) and (6).
Re (s)
Im (s)

0
j
0

0

0
j
0
1
2
j
0
j
0
1
2

Fig. 3. For 0 < 1, the complex conjugate poles
lie on a circle of radius
0
.
ECEN 2830
4

v
out
(s) =
K
1
s
+
K
2
s s
1
+
K
3
s s
2
(7)
Here, s
1
and s
2
are given by Eq. (5), and the residues K
1
, K
2
and K
3
are given by

K
1
= s
1
s
H
0
1 + 2
s

0
+
s

0
2
s = 0
K
2
= s s
1
1
s
H
0
1 + 2
s

0
+
s

0
2
s = s
1
K
3
= s s
2
1
s
H
0
1 + 2
s

0
+
s

0
2
s = s
2 (8)
Evaluation of these expressions leads to
K
1
= H
0
K
2
= s s
1
1
s
H
0
1
s
s
1
1
s
s
2
s = s
1
=
s
2
s
2
s
1
H
0
K
3
= s s
2
1
s
H
0
1
s
s
1
1
s
s
2
s = s
2
=
s
1
s
1
s
2
H
0
(9)
The inverse transform is therefore

v
out
(t) = H
0
u(t) 1
s
2
s
2
s
1
e
s
1
t

s
1
s
1
s
2
e
s
2
t
(10)
In the overdamped case, the output voltage response contains decaying exponential
terms, and the rise time depends on the magnitudes of the roots s
1
and s
2
. The root
having the smallest magnitude dominates Eq. (10): for | s
1
| << | s
2
|, Eq. (10) is
approximately equal to

v
out
(t) H
0
u(t) 1 e
s
1
t
(11)
This is indeed what happens when >>1. Equation (11) can be expressed in terms of
0

and as

v
out
(t) H
0
u(t) 1 e

0
t
2
(12)
ECEN 2830
5
When >> 1, the time constant 2/
0
is large and the response becomes quite slow.
2.2. Critically damped case, = 1
In this case, Eq. (4) reduces to

v
out
(s) =
H
0
s 1 +
s

0
2
(13)
The partial fraction expansion of this equation is of the form

v
out
(s) =
K
1
s
+
K
2
s +
0
2
+
K
3
s +
0
(14)
with the residues given by
K
1
= H
0
K
2
= s +
0
2 H
0
s 1 +
s

0
2
s =
0
=
0
H
0
K
3
=
d
ds
s +
0
2 H
0
s 1 +
s

0
2
s =
0
= H
0
(15)
The inverse transform is therefore

v
out
(t) = H
0
u(t) 1 1 +
0
t e

0
t
(16)
In the critically damped case, the time constant 1/
0
is smaller than the slower time
constant 2/
0
of the overdamped case. In consequence, the response is faster. This is
the fastest response that contains no overshoot and ringing.
2.3. Underdamped case, < 1
The roots in this case are complex, as given by Eq. (6). The partial fraction expansion of
Eq. (4) is of the form

v
out
(s) =
K
1
s
+
K
2
s +
0
j
0
1
2
+
K
2
*
s +
0
+ j
0
1
2
(17)
The residues are computed as follows:
ECEN 2830
6
K
1
= H
0
K
2
= s +
0
j
0
1
2
H
0

0
2
s s +
0
j
0
1
2
s +
0
+ j
0
1
2
s =
0
+
j
0
1
2
(18)
The expression for K
2
can be simplified as follows:

K
2
=
H
0

0
2

0
+ j
0
1
2
2j
0
1
2
=
H
0
+ j 1
2
2j 1
2
=
H
0
2 1
2
+ j2 1
2
(19)
The magnitude of K
2
is

K
2
=
H
0
2 1
2
(20)
and the phase of K
2
is

K
2
= tan
1
2 1
2
2 1
2
= tan
1

1
2
(21)
The inverse transform is therefore

v
out
(t) = H
0
u(t) 1
1
1
2
e

0
t
cos 1
2

0
t + tan
1

1
2
(22)
In the underdamped case, the output voltage rises from zero to H
0
faster than in the
critically-damped and overdamped cases. Unfortunately, the output voltage then
overshoots this value, and may ring for many cycles before settling down to the final
steady-state value.
ECEN 2830
7
In some applications, a moderate amount of ringing and overshoot may be
acceptable. In other applications, overshoot and ringing is completely unacceptable, and
may result in destruction of some elements in the system. The engineer must use his or
her judgment in deciding on the best value of .
3. Step response waveforms
Equations (10), (16), and (19) were employed to plot the step response waveforms of
Fig. 4. Underdamped, critically damped, and overdamped responses are shown. It can be
deduced from Fig. 4 that the parameter
0
scales the horizontal (time) axis, while H
0

scales the vertical (output voltage) axis. The damping factor determines the shape of
the waveform.

0
0.5
1
1.5
2
0 5 10 15

0
t, radians
= 0.05
= 0.01
= 0.125
= 0.25
= 0.5
= 0.67
= 1
= 1.67
= 2.5
= 5
= 10
= 50
v
out
(t)
H
0

Fig. 4. Second-order system step response, for various values of damping factor .

Three figures-of-merit for judging the step response are the rise time, the percent
overshoot, and the settling time. Percent overshoot is zero for the overdamped and
critically damped cases. For the underdamped case, percent overshoot is defined as

percent overshoot =
peak v
out
v
out
()
v
out
()
100%
(20)
One can set the derivative of Eq. (19) to zero, to find the maximum value of v
out
(t). One
can then plug the result into Eq. (20), to evaluate the percent overshoot. Note that the
ECEN 2830
8
final (steady-
state) value of the
output v
out
() is
H
0
. The following
equation for the
percent overshoot
results:

percent overshoot = e
/ 1
2
100%
(21)
Again, this
equation is valid
only in the underdamped case, i.e., for 0 < < 1. It can be seen from Fig. 4 that
decreasing the damping factor results in increased overshoot. The overshoot is 0% for
= 1. In the limit of = 0 (the undamped case), the overshoot approaches 100%.
As illustrated in Fig. 5, the rise time is defined as the time required for the output
voltage to rise from 10% to 90% of its final steady-state value. When the system is
underdamped, the output waveform may pass through 90% of its final value several
times; the first pass is used in computation of the rise time. It can be seen from Fig. 4
that the rise time increases monotonically with increasing .
The settling time is the time required by an underdamped system for its output
voltage response to approach steady state and stay within some specified percentage (for
example, 5%) of the final steady-state value. As can be seen from Fig. 4, systems having
very small values of have short rise times but long settling times.
4. Experimental measurement of step response.
The difficulty in measuring a transient
response is that it happens only once
if you blink, you will miss it! This
problem can be alleviated by causing
the step input to be periodic: apply a
square wave (Fig. 6) to the circuit
input. The duration T/2 of the positive
portion of the square wave is chosen
to be much longer than the settling time of the output response, so that the circuit is in
steady-state just before each step of the input waveform occurs. In consequence, the
v
out
(t)
t
settling time
rise
time
overshoot
final
value
v()
10%
of
v()
90%
95%
105%

Fig. 5. Salient features of step response, second order system.
t T T/2 0
0 V
1 V
v
in
(t)

Fig. 6. Use of a square wave input, with sufficiently
long period T, allows the output transient to be
observed on any oscilloscope.
ECEN 2830
9
output voltage waveform is identical to the waveform observed when a single step input
is applied, except that the output transient occurs repetitively. The output transient
waveform can now be easily observed on an oscilloscope, and can be studied in detail.

BIBLIOGRAPHY
[1] BENJAMIN C. KUO, Automatic Control Systems, New Jersey: Prentice-Hall.
[2] J. DAZZO and C. HOUPIS, Linear Control System Analysis and Design: Conventional and
Modern, New York: McGraw-Hill, 1995.