Second and Higher Order Systems

Chapter 72
72.1 Second Order Dynamics
72.2 Overdamped,  > 1
72.3 Critically Damped,  = 1
72.4 Underdamped, 0 <  < 1
72.5 Undamped,  = 0
72.6 Higher Order Systems

An understanding of the dynamics of second or- the transfer function of a second order system.This
der systems is justified on two counts. First, the is depicted in Figure 72.1.
dynamics of many items of plant and instrumen-
tation are inherently second order.And second, the K.!n2
0 (s) = 1 (s) (72.2)
behaviour of many closed loop systems, whether s2 + 2!n s + !n2
they are second order or not, are often described or
specified in terms of the characteristics of second
order systems. For a more comprehensive treat- θ1 ( s ) Kωn 2 θ 0 (s )
ment the reader is referred to the texts by Dutton 2
s + 2ωn ζs + ωn 2

(1997) and Wilkie (2002).

Fig. 72.1 Transfer function representation of a second order sys-
tem

72.1 Second Order Dynamics Suppose that the system is subjected to a step input
A second order system is one whose dynamic be- of magnitude A. Substituting for 1 (s) = A/s and
haviour may be described by a second order ODE factorising gives:
of the general form:
A.K.!n2
0 (s) = !   " !   "
s. s + !n  +  2 − 1 . s + !n  −  2 − 1
1 d2 0 2 d0
+ + 0 = K.1 (72.1) (72.3)
!n2 dt2 !n dt The solution to Equation 72.3 is found by the usual
procedure of reducing into partial fractions and
where the signals 0 and 1 are both in deviation finding their inverse transforms. The form of the
form, and: solution is determined by the value of the damping
!n is the natural frequency factor . The step response is usually categorised
 damping factor according to the value of , as summarised in Ta-
K steady state gain ble 24.1, and sketched in Figure 72.2.

Transforming both sides of Equation 72.1, assum-

ing zero initial conditions, and rearranging yields
 592 72 Second and Higher Order Systems

θ0 0 < ζ <1 becoming oscillatory. Critical damping is never-

theless used as a basis for design. Typically, the cri-
ζ =1 teria to establish critical damping are determined
and a safety factor is then applied to guarantee
overdamping.

72.4 Underdamped, 0 <  < 1

ζ >1
t Assume that 0 <  < 1 such that the operands
0
of the square roots in the denominator of Equa-
Fig. 72.2 Step response vs damping factor for second order sys- tion 72.3 are negative. The resulting complex con-
tem jugate roots of s yield the step response:
⎛ ⎞
1
72.2 Overdamped,  > 1 0 (t) = AK ⎝1 − ) e−!n t . sin (!d t + )⎠
1− 2
Assume that  > 1 such that the operands of the (72.6)
square roots in the denominator of Equation 72.3 where !d is the damped frequency given by:
are positive. The resulting negative real roots of s 
yield the step response: !d = !n 1 −  2
# $
   and is the phase shift given by:
0 (t) = AK 1 − e −!n t
cosh !n  2 − 1.t  
−1 1 − 2
%& = tan
    
+ sinh !n  2 − 1.t (72.4)
2 − 1 Both of these identities may be represented tri-
gonometrically as shown in Figure 72.3.
A closed loop response of this overdamped form,
in which the output asymptotically approaches its
steady state value, is specified for control systems ωn ωn ζ 1 (1 − ζ )
2

where it is essential that there is no oscillation. Φ

ωd ζ
a b
72.3 Critically Damped,  = 1 Fig. 72.3 a Identity for damped frequency, b Identity for phase
shift
Assume that  = 1 such that the operands of the
square roots in the denominator of Equation 72.3 Inspection of Equation 72.6 reveals an exponen-
are zero. The resulting repeated roots of s = −!n  tially decaying sinusoid which is sketched and la-
yield the step response: belled in Figure 72.4.
' ( Various characteristics of this response are iden-
0 = AK 1 − e−!n t (1 − !n t) (72.5)
tified and defined in Table 72.1,which is largely self
This is the fastest response that can be obtained explanatory.Note that the formulae for all the char-
without any oscillation. However, closed loop sys- acteristics are functions of !n and  only. Thus,
tems should not be designed for critical damping. specifying any two characteristics results in two
Any slight error in the design or measurements equations which may be solved for !n and ,which
could result in a specified exponential response determines the values of the other characteristics.
 72.4 Underdamped, 0 <  < 1 593

θ0

x – p%
+
z
AK

y P
t
0 tr tp ts

Fig. 72.4 Step response for underdamped second order system

Table 72.1 Characteristics of underdamped second order response

Characteristic Symbol Formula

Damped frequency !d !n 1 −  2
2
Period P 
!n 1 −  2
#  &
1 1 − 2
Rise time tr   − tan −1
!n 1 −  2 2


Peak time tp 
!n 1 −  2

−√

x 1− 2
Overshoot e
y
−√
2
z 1− 2
Decay ratio e
x

The response limit is a band of arbitrary width ±p Note that the time constant for the exponential de-
about the equilibrium value. The value of p is ex- cay of the envelope, T = !1n  .
pressed as a percentage of the steady state change The response limit and settling time are related
in output. In practice it is normally taken as either to this time constant by the following useful ap-
±5% or ±2%. The settling time ts is the time taken proximations:
for the output to fall within the response limit and
to remain bounded by it. If p = ±5% then ts ≈ 3T.
The curve which is tangential to the peaks of p = ±2% ts ≈ 4T
the step response is known as its envelope, and is
also sketched in Figure 72.4. The equation of the
envelope is given by:
 
e−!n t
0 = AK. 1 + 
1 − 2
 594 72 Second and Higher Order Systems

72.5 Undamped,  = 0 θ0

Substituting  = 0 into Equation 72.3 results in n=0

imaginary roots of s which reduce into partial frac- 1
tions as follows: 1
  n=
1 1 1

2
0 (s) = AK − −

=
3

n
s 2(s + j!n ) 2(s − j!n )
n=
5
Inverse transforming yields the step response: n=
t
0 (t) = AK 1 − 0.5 e−j!n t + e+j!n t 0
= AK (1 − cos !n t) (72.7) Fig. 72.7 Step response of n first order systems in series

This is a sine wave with oscillations of constant input exactly. If n = 1 the response is that of a
amplitude as depicted in Figure 72.5. For control first order system. With n = 2 the response is that
system design and tuning purposes this clearly of a second order system with repeated roots and
corresponds to the case of marginal stability as critical damping. For n ≥ 3 the response becomes
discussed in Chapter 24. more sluggish and “S shaped”.
θ0
Process plant and control loop elements invari-
ably consist of combinations of non-identical first
order systems, overdamped second order systems
and steady state gains, as depicted in Figure 72.8.
AK
θ1 ( s ) K1 K2 Kn θ 0 (s )
t T1s + 1 T2 s + 1 Tn s + 1
0
Fig. 72.8 A series of n first order systems in series
Fig. 72.5 Sinusoidal response of undamped second order system
Often it is impractical to determine the exact trans-
fer function for each element.A common approach
72.6 Higher Order Systems is to lump them together in an overall transfer
function consisting of a single lag, delay and gain
Consider a series of n identical first order systems, as follows:
as depicted in Figure 72.6. n
0 (s) Kj Ke−Ls
= ≈ (72.8)
1 1 1 1 (s) Tj s + 1 Ts + 1
j=1
Ts + 1 Ts + 1 Ts + 1
The effect of this approximation is to replace the“S
Fig. 72.6 A series of n identical first order systems shaped” response with that of a first order system
The series’ overall transfer function is given by: subject to a time delay, as depicted in Figure 72.9.
The approximation is good enough for most
 n
0 (s) 1 purposes. For example, it provides the basis for the
= reaction curve method of controller tuning, de-
1 (s) Ts + 1
scribed in Chapter 24, in which the control loop
The system’s response to a step input of magnitude is assumed to consist of a 3-term controller in se-
A is sketched in Figure 72.7 as a set of curves with ries with the lumped approximation, as depicted in
n as the parameter. If n=0 the output follows the Figure 72.10.
 72.6 Higher Order Systems 595

θ0

n _> 3

t
0

θ0

0.632 K

t
0 L L+T
Fig. 72.9 Response of first order system subject to time delay

+
KC. 1+
1
+ TDs Ke − Ls +
+ TRs Ts + 1
-

Fig. 72.10 PID controller with lumped approximation of plant