Time Domain Analysis

Consider the engineering system

X ( s) B( s ) bm s m + bm −1s m −1 + m + b0
= G( s) = = n .
U (s) A( s ) s + an −1s n −1 + m + a0
We can note the following points.

1. System descriptions

The description shown above is known as transfer function form. It involves a denominator
polynomial of order n , and a numerator or order m . The coefficient of the s n term in the
denominator is always one. All of the other coefficients are arbitrary and some of them may be zero.

It is noted that n ≥ m always, and for real physical systems, n > m . This implies that inputs cause
changes in the outputs of systems and not the other way round. We call systems that obey this rule
causal systems and the mathematical descriptions of such systems, that is G ( s ) , are strictly proper.
The physical reasoning for the need for causal systems is also seen in examining the frequency
response of such systems.

Alternative descriptions than the transfer function form shown are also sometimes useful.

• Pole-zero form

m
K ∏ (s + z ) i
K ( s + z1 )( s + z2 )m( s + zm )
G (s) = i =1
=
n
( s + p1 )( s + p2 )m( s + pn )
∏ (s + p )
i =1
i

This form has the advantage that all of the roots of both the denominator and numerator are explicitly
identified. The names pole and zero will be explained in the next section.

• Time constant form

m−q
2
q 2  s  s
K ∏ (τ i s + 1) ⋅ ∏ (  + 2ξ k + 1)
i =1 k =1  ω nk  ω nk
G (s) = n−r
2
r 2  s  s
∏
i =1
(τ i s + 1) ⋅ ∏ (
k =1  ω nk 
 + 2ξ k
ω nk
+ 1)

m−q
where there are q purely real roots of the numerator and pairs of complex conjugate roots,
2
n−r
and there are r purely real roots of the denominator and pairs of complex conjugate roots in
2
the denominator. This form can used to describe the various first order time constants and natural
frequencies and damping ratios of the system.

Note that all forms:

• Transfer function form
 • Pole-zero form, and
• Time constant form
are entirely equivalent in that they can be used to describe exactly the same underlying physical
system. Therefore they will all share exactly the same mathematical properties. The only difference is
in the way the system is described which allows for some insight into the characteristics of the
system.

Example

Given
  s 2 s 
15 (3s + 1)   + 2(0.7) + 1
 2  2 
G (s) = 
  s 2 s   s  2 
(10s + 1)(65s + 1)   + 2( 0.1) + 1   + 2(0.9) s + 1
  0.5  0.5   6  6 

in time constant form, the same systems can also be represented in pole-zero form as

0.156 ( s + 0.33)( s + 1.4 + 1.43 j )( s + 1.4 − 1.43 j )

G (s) =
( s + 0.1)( s + 0.015)(s + 5.4 + 2.62 j )( s + 5.4 − 2.62 j )( s + 0.05 + 0.5 j )( s + 0.05 − 0.5 j )

and also in transfer function form as

0.156s 3 + 0.488s 2 + 0.769 s + 0.208

G (s) = 6 .
s + 11.015s 5 + 38.589s 4 + 10.624s 3 + 9.784s 2 + 1.048s + 0.014

All forms of G (s ) are entirely equivalent.

2. Poles and zeros

Given a system described in pole-zero form

K ( s + z1 )( s + z2 )m( s + z m )
G (s) =
( s + p1 )( s + p2 )m ( s + pn )
if we choose to evaluate this transfer function, G (s ) , for particular (possibly complex) values of
s = s* , then G ( s* ) will simply be a complex number.

Example

1
Given G ( s ) = , let s = 2 + j .Then
s +1
1 1 1 3− j 1 3 1
G (s) s = 2 + j = = = ⋅ = 2 2 (3 − j ) = − j.
2 + j +1 3 + j 3 + j 3 − j 3 +1 10 10
If, for example, we evaluate G (s ) at s = − zi , i = 1,l, m , then
K (− zi + z1 )m(− zi + zi )m(− zi + zm ) K (− zi + z1 )m (0)m (− zi + zm )
G (s) s = − z = = = 0.
i
(− zi + p1 )m (− zi + pn ) (− zi + p1 )m (− zi + pn )
Therefore when G (s ) is evaluated at each s = − zi , i = 1,l, m , the function is zeroed. Thus each
s = − zi , i = 1,l, m is called a zero of G (s ) .

Similarly, evaluating G (s ) at each s = − pi , i = 1,l, n gives

K (− pi + z1 )m(− pi + zm ) K (− pi + z1 )m (− pi + zm )
G (s) s = − p = = = ±∞ .
i
(− pi + p1 )m (− pi + pi )m(− pi + pn ) (− pi + p1 )m(0)m (− pi + pn )
Therefore the function G (s ) gets large in magnitude when evaluated at s = − pi , i = 1,l, n . If we
consider a graph of G (s ) versus s , the graph will go to zero at the zeros of the transfer function and
will tend to infinity at the values s = − pi , i = 1,l, n . In a colloquial way, one can view this as if
someone stuck a tent pole under the graph and pushed it up to infinity at these points. Thus we call
the points s = − pi , i = 1,l, n poles of the transfer function.

Example

( s + )( s − 20)
Let G ( s ) = and evaluate G (s ) for various particular inputs of s . The results are shown
( s + 8)( s − 6)
in the figure. Note that the function goes to zero at the zeros of G (s ) and the plot goes to infinity at
the poles of G (s ) .

plot of G(s) for varying s

10

) 0
s(
G
-2

-4

-6

-8

-10
-20 -15 -10 -5 0 5 10 15 20
s

It is noted that there are exactly m zeros and n poles of the transfer function G (s ) , which had a
numerator of order m and a denominator of order n .
The relative order of the system is RD = n − m . This is sometimes called the pole excess of the
system. Causal systems are those with more poles than zeros, and so consequently another way of
describing the collection of all causal systems is those systems with RD > 0 .

3. System stability

Stability is, for obvious reasons, a key issue in engineering systems. Thus we should have a clear and
precise definition of stability and a methodology for determining if a given system is stable.

For our purposes, it suffices to consider a form of stability analysis called bounded-input-bounded-
output stability. This implies that there is a guarantee that the output of a system will be bounded
(that is will not blow up to infinity) at all times if there are bounds on the size of the input at all times.
A test for bounded-input-bounded-output (BIBO) stability is as follows.

Consider a system described by

X ( s)
= G( s)
U (s)
where X (s ) is the output, U (s ) is the input, and G (s ) is the system transfer function. The transfer
function contains poles and zeros as in
K ( s + z1 )( s + z2 )( s + z m )
G (s) = .
( s + p1 )( s + p2 ) ( s + pn )
Further, U (s ) is an input signal that has the Laplace domain description
K ( s + zu1 )( s + zu 2 )( s + zum )
U ( s) = u
( s + pu1 )( s + pu 2 )( s + pun )
1
For example, if u (t ) is a unit step input then U ( s ) = and if u (t ) = sin ωt then
s
ω ω
U (s) = 2 = .
s +ω 2
( s + jω )( s − jω )

Given this system and input, the output can be described by

K ( s + z1 )(s + z 2 )( s + zm ) K u ( s + zu1 )( s + zu 2 )( s + zum )

X (s) = ⋅ .
( s + p1 )(s + p2 )( s + pn ) ( s + pu1 )(s + pu 2 ) ( s + pun )
Now, taking the inverse Laplace transform, the time domain solution for the output is
x(t ) = c1e − p1t + c2e − p 2 t +  + cne − p n t + cu1e − pu 1t + cu 2e − pu 2 t +  + cun e − pun t .
Each of the poles of both the system and input are, in general, complex valued. If they are real valued
then the complex part is simply zero. Further, if any poles are complex valued, they will appear as
complex conjugate pairs, and imply an oscillating or trigonometric time function solution. This
trigonometric function will have an exponential term in front of it whose exponent is given by the
real part of the pole.

In summary, a portion of the terms in the sum solution for the output will look like
xri (t ) = ci e − ai t
where pi = ai ∈ ℜ or is a purely real number. The remaining terms will appear in pairs and will look
like
xck (t ) = ck e − a k t (e − jbk t + ρ k e jbk t )
where pk = ak + jbk and pk +1 = ak − jbk .

The important point to recognize here is that all of the terms e jbk t for all values of bk are purely
oscillatory with constant unity magnitude. That is they are unit vectors in the complex plane that
rotate about the origin.

−a t
Note that if any of the exponential terms with real exponents (that is, all of the e j , j = 1,, n ) are
growing with time then the output will grow in magnitude with time. Therefore, for stability we will
require that none of the terms in the sum grow with time. This is sufficiently satisfied if
a j ≥ 0 ∀ j = 1,, n .
That is all of the real parts of the exponents are positive for all of the system and input poles.
However, by definition of BIBO stability, the input poles will necessarily meet this condition since
u (t ) . Is a bounded function of time by construction and so all poles of U (s ) will have negative real
parts.

Thus we only need to check the poles of the system.

Thus a precise definition of BIBO stability is that for all poles, pi = − ai + jbi of G (s ) , if
real ( pi ) ≤ 0
then the system is BIBO stable.

A slightly stronger sense of stability is that of asymptotic stability, which implies that the output of
the system tends to zero for energy bounded input signals. The bottom line is that for asymptotic
stability the condition on the poles changes to
real ( pi ) < 0 for all i = 1,, n .

4. Characteristic equations

From the above discussion, we have seen that the system poles play a major role in determining
system stability. In fact, all of the dynamic modes of the system response are entirely determined by
the system poles.

The zeros of the system influence the relative weighting of each of the modes. That is, the system
poles determine the ai in the exponents, and the zeros (along with the poles) determine the ci in the
partial fraction expansion description of the output and also in the sum of the modes found in the
output.

B( s)
It is noted that the system poles are the roots of the denominator of the transfer function G ( s ) = .
A( s )
That is the roots of the equation A( s ) = 0 . We will label this special polynomial used in determining
the dynamic response of systems, A(s ) , as the characteristic polynomial. Further we will label the
equation
A( s ) = 0
as the characteristic equation of the system transfer function G (s ) .