You are on page 1of 3

268 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS AND CONTROL INSTRUMENTATION, VOL. IECI-27, NO. 4, NOV.

1980

Equivalent Time Constants for Nonoscillatory


Second-Order Processes
CHARLES P. NEUMAN, SENIOR MEMBER, IEEE

Abstract-Equivalent time constants are obtained for nonoscillatory i) Solve the initial-value problem (2) for the unit-step
second-order processes. In analogy with first-order processes, the response x(t).
equivalent time constant is defined to be the time at which the
monotonic step-response of the second-order process reaches ap- I
ii) Calculate Te by setting x(t) t=Te = [ - e- 1 1 .
proximately 63.3 percent of its steady-state value. These time The critically damped (t = 1) process is treated in Section-III
constants can be applied to the analysis and design of feedback control and overdamped processes (D > 1) are investigated in Section
systems in which oscillations cannot be tolerated.
IV.

I. INTRODUCTION III. THE CRITICALLY DAMPED PROCESS


THE TIME CONSTANT r of a first-order linear time- When ¢ = 1, the unit-step response
invariant process completely characterizes its transient
response [1I. The time constant is the time required for its X(t) X(cJn t) I 1 (1 + Wn OC In ti
= = (3)
output to reach [1 - e-1 _ 63.3 percent of its steady-
state value in response to a step input. approaches the steady-state value of unity asymptotically;
To build upon this universal concept, the objective of there is neither overshoot nor oscillation. The slope x(t) It=O
this paper is to determine an equivalent time constant Te of (3) at t = 0 is zero. This is in contrast to a first-order proc-
of the nonoscillatory second-order process characterized by ess whose unit-step response has its maximum slope r- 1 at
the transfer function t=0.
At t = re, the unit-step response (3) becomes
I]
G(s) X(s) - X(Te) = X(W)nre) = [ 1-(1 + WnTe)eCJnTe] = [ 1-e
U(s) S )2 (S)( (4)
- +2t - + I
Con / n/ from which
In (1), on is the undamped natural frequency (radians/ (nTe = 1 ±ln (1 + C)nTe). (5)
second) and r is the dimensionless damping ratio. The equiv-
alent time constant Te is defined to be the time at which the Solving (5) numerically yields wnTeo 2.146. The super-
monotonic unit step response x(t) of the second-order process script zero is introduced to indicate the equivalent time-
(1) reaches [1 - e- 1 ] of its steady-state value xs5 = 1. The constant for the critically damped process. It is recalled that
unit step-response of (1) is the solution of the initial-value the speed of response of nonoscillatory second-order proc-
problem esses (1) decreases as the damping ratio r increases. Since the
critically damped process reaches steady-state without over-
d2x dx shoot faster than all overdamped processes, w.nTe0 is the
t)22~- + x(t) = f(t), for t>0 minimum value of the [undamped natural frequency] - [equiv-
d(wnt d(Unt) alent time-constant] product for all nonoscillatory second-
(2)
order processes.
x(0) = 0 and - =0
dt t-=O IV. OVERDAMPED PROCESSES
where J(t) is the unit-step function. When D > 1, the eigenvalues of (2a) are si = (-D +
\flTTTi)¾n and s2 = (-D /-E)own The unit-step
II. PROCEDURE response
The procedure for determining the equivalent time constant
Te is as follows: e-twnt f/ \

Manuscript received November 11, 1978; revised January 16, 1980.


x(t) = X 1t
=
-nt 2

This work was supported in part by the National Science Foundation


under Grant ENG 76-81346. .2-.j Wnt
The author is with the Department of Electrical Engineering,
+ 1--
N ..2-.je (6)
Carnegie-Mellon University, Pittsburgh, PA 15213.
0018-9421/80/1100-0268$00.75 i 1980 IEEE
NEUMAN: SECOND-ORDER PROCESS TIME CONSTANTS 269

approaches the steady-state value of unity asymptotically;


there is neither overshoot nor oscillation. As for all second-
order processes, the slope x(t) It=o of (6) at t = 0 is zero.
For large values of ( 1), the unit-step response (6) be- 3
comes1
_ Z

x(t) = x(w.ot) 1-e-e7nI2 )t) Zz


l 0<
< Z D
> 0p-
Hence, as ¢ increases, the behavior of (6) approaches that 0i
wn ITS
2r
of a first-order process with the equivalent time constant f orirf 2 1.5
re = 2¢/w, C.
At t = re, the unit-step response (6) becomes z 2.6 F

e7-rwn,re
X(Te) = X(wnTe) = 1 a04

<
2.4I
0

Z cr

j(1 +epL
)e @J Twnre z
L I

I 2.2
+ (1-- - he e
2.1 46
l.l 1.2 1.3 1.4 1.5 2 3 4 5
= [1-e-l] (8) DAMPING RATIO C O

Fig. 1. wnTe product versus damping ratio for nonoscillatory second-


from which order processes.

1(1 L)e~f ticj'nre TABLE I


EQUIVALENT TIME CONSTANTS OF OVERDAMPED SECOND-
ORDER PROCESSES
+ (9)2
Undamped Equi valent
Damping Ratio Natural Frequency * Time-Constant
. _1 .2
Equation (9) is solved numerically for the product CunTe as a
Wn TO - 2.146193221
function of the damping ratio t. The numerical solutions are
summarized in Table I and illustrated in Fig. 1. 2. 1477
1.001
The product &.nre is a monotonically increasing function 1.01 2.1616
of t. This reflects the fact that the speed of response of over- 1.025 2.1850
damped second-order processes decreases as ¢ increases. For 1.05 2.2245
¢ > 1, (9) simplifies to &nTe- 2. The curve of Fig. 1 rises 1.075 2.2648
from the value n'reo - 2.146 for the critically damped 1.1 2.3057
1.2 2.4761
process (D = 1) and asymptotically approaches nTe I r> 1 + 2.6551
1.3
2¢. The numerical solutions compiled in Table I indicate 1.4 2.8405
that3 Cwnr,e- 2 to within- one percent for ¢ > 1.5. For 1.5 3.0305
= 1.5, the eigenvalue ratio is 6.85. The eigenvalue ratio 1.6 3.2235
1.7 3.4185
1.8 3.6149
AS2 _ +\/-2-I = 1 +2v
\ +\ iI
>1I
1.9
2
3.8122
4.0101
3 6.0026
(10) 4 8.0010
5 10.001
6 12.000
From the binomial series for r 1 10 20

2r2 2r >> I
Wn re 24
When r -+ 1, (9) reduces to (5).
2
3 It is noted that the negative of the sum of the reciprocals of the
process eigenvalues equals 2f/wn; that is, -[(1Is1) + (l/s2)J = 2r/wn.
270 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS AND CONTROL INSTRUMENTATION, VOL. IECI-27, NO. 4, NOV. 1980

of the overdamped second-order process (1) is a monotonically (3) and (6) are calculated at integral multiples of the equiva-
increasing function of ¢ and asymptotically approaches a* = lent time constant Te. The outputs x[N(&nre)I indicate
1 + 4r2. Solving (10) for the damping ratio yields that the equivalent tirne constant re indeed does charac-
I terize the transient response of the nonoscillatory second-
I order process (1). For ¢ 1, the second-order process rises
t-2 _ 4 >1. (11)
-

slightly faster than does the first-order process. In the case


of the critically damped process (D = 1), which shows the
V. APPROXIMATIONS FOR THE EQUIVALENT greatest disparity, x [N(wnre0)] is seven, four, and two
TIME CONSTANT percent larger than the corresponding output of a first-order
The product w,nTe of the undamped natural frequency process for N = 2, 3, and 4, respectively. For ¢ > 1.5, the
Con and the equivalent time constant Te can be approximated difference between the responses of first- and second-order
by processes is negligible. The essential difference is that the
I 2.146+175 [-1], for 1I <1.5 slope of the first-order process response is a maximum at
t = 0 while the slope of the second-order process response
&JnTre 2~, for ~> 1.5. (12) is zero.
A first-order process settles to within two percent of its
For 1 < ¢ < 1.6, the straight line steady-state value in four time constants. The corresponding
(nTe-2.146 + 1.75 [l-1I settling times t5 for the nonoscillatory second-order process
(1) are
fits the numerical solutions compiled in Table I with a maxi-
mum error of less than one percent. 3Te for <l
I. 1.1 and 4,re for > 1.1.
VI. PERFORMANCE EVALUATION OF THE
EQUIVALENT TIME CONSTANT CHARACTERIZATION REFERENCES
To evaluate the equivalent time-constant characterization [l] J. C. Gille, M. J. Pelegrin, and P. Decaulne, Feedback Control
Systems: Analysis, Synthesis, and Design. New York: McGraw-
of the nonoscillatory second-order process (1), the outputs Hill, 1959, p. 63.

You might also like