Time Domain Performance

In order to analyze the time domain performance of systems, we will examine the characteristics of
the output of the system when a particular input is applied. The input we will choose is a unit step
input, that is
1 t ≥ 0
u (t ) =  .
0 t < 0
The Laplace transform of this input is
1
U (s) = .
s
This input is selected because it
• is simple
• is relatively easy to experimentally implement to test systems and to implement on system
simulations and
• contains a rich frequency content in that the Fourier transform of the step function contains all
frequency signals. Thus the step input is a signal rich input potentially capable of providing a
wealth of information about the system.

This unit step input will be applied to first, second and higher order systems in order to determine the
general behavior of these types of processes.

First Order Systems

Consider the system

X ( s) k
= G( s) =
U (s) τs + 1
1
and using U ( s ) = gives
s
k
X ( s ) = G ( s )U ( s ) = .
s (τs + 1)
Constructing a partial fraction expansion of the output,
c c2
X ( s) = 1 +
s s+ 1
τ
where
k
c1 = sX ( s ) s = 0 = =k
τs + 1 s = 0
k
c2 = ( s + 1 ) X ( s ) = = −k .
τ s=− 1
τ τs s = − 1
τ
Thus
x(t ) = k 1 − e τ  .
−t

 
It is quickly verified that the dynamic response of this system is entirely governed by τ or
equivalently the pole s = − 1 .
τ
We will label τ the time constant of this first order system.

−t
τ
By evaluating the dynamic part of the response, the term e , we can compute that

t −t
e τ
0 0
τ 0.37
3τ 0.05
4τ 0.02

This says that the initial value of the response is zero and when t = τ that 63% of the dynamics have
dissipated. Also by the time t = 3τ , 95% of the dynamics have occurred. Finally, when t = 4τ has
elapsed, 98% of the dynamic part of the response has disappeared.

We can also note that

k −t
xD (t ) = e τ .
τ
Therefore the initial slope of the response is k
τ and the slope monotonically decreases at the same
rate as the dynamics fade. Eventually the slope becomes very small and asymptotically approaches
zero. This implies that the system eventually reaches at steady state.

It is important to recognize that as the time constant gets smaller, the pole location, s = − 1 moves
τ
more to the left on the real number line. That is, the pole becomes more negative. Further, as τ gets
smaller, the system response effectively gets faster.

Thus there is a fundamental connection between fast system response and system poles that are more
to the left in the complex plane.
Step Response
Step response of two first order systems

0.8

e
d
ut 0.6
i
pl
m
A
0.4

0.2

0
0 5 10 15 20
Time (sec.)

The figure shows the response to a step input of two first order systems with different time constants.
The first and faster responding system has a time constant of 1 second. The slower system has a time
constant of 0.33 seconds. It is noted that this system responds roughly three times slower than the
first system.
Second Order Systems

Consider the engineering system

X ( s)
= G( s)
U (s)
where
kω n
2
G (s) = 2 .
s + 2ξω n s + ω n
2

where the variable ω n is usually called the system natural frequency and the variable ξ is often
referred to as the damping ratio. This standard form case can be used to cover essentially all possible
alternatives or variations on particular second order system cases. These variations include the
following.

1. Underdamped, overdamped and critically damped systems

The poles of the system in standard for are the roots of the characteristic equation
s 2 + 2ξω n s + ω n = 0
2

which are
− 2ξω n ± (2ξω n )2 − 4ω n 2
s1, 2 =
2
2
( )
= −ξω n ± ω n ξ − 12 = −ξω n ± jω n 1 − ξ 2 .
From this it is seen that there are three basic possibilities:
• ξ <1 Under damped case. Here there will be two complex conjugate poles with
equal real parts and complex parts that are negative inverse of each other.
• ξ =1 Critically damped case. Here there will be two poles that are identical since the
discriminant in the quadratic form is precisely zero. The two poles will be s1, 2 = −ξω n .
• ξ >1 Over damped case. Here there will be two real poles. The poles will both be on
the negative real axis (for stable systems) and they will be distinct. This case can be
treated as the cascaded or series combination of two first order systems.

2. Addition of a zero

as + b
If the actual system is G ( s ) = , this can be written as
s + 2ξω n s + ω n2
2

as b
G (s) = 2 + 2 .
s + 2ξω n s + ω n s + 2ξω n s + ω n2
2

The second term is exactly as per the standard form, given that b = kω n , and the first term is simply
2

the Laplace version of the time derivative of the standard form (again with a correction for the gain).
To verify this, recall that taking the derivative in the time domain is equivalent to multiplying by s in
the Laplace domain. Thus this system can be broken into two parts one that is exactly equivalent to
the standard form of the problem and one that is the derivative of the standard form in the time
domain.
Thus if the second term has an inverse transform that is a decaying sine wave, then the first term will
be a decaying cosine wave with a different amplitude but the same decay rate and the same
oscillation frequency.

3. Explicit pole form

c
If the actual system is written G ( s ) = , then this can be expressed as
( s + p1 )( s + p2 )
c
G (s) = .
s + ( p1 + p2 ) s + p1 p 2
2

Thus, using the relations

p + p2 c
ω n = p1 p2 , ξ = 1 , k=
2 p1 p2 p1 p2
the two forms are seen to be equivalent. Furthermore the poles of the system can be evaluated as
s = − p1 = −ξω n + jω n 1 − ξ 2 , and s = − p2 = −ξω n − jω n 1 − ξ 2 .
It is quickly verified that the poles will always appear as either a set of real poles or as a pair of
complex conjugates.

4. Real-imaginary pole form

c
If the actual system is written G ( s) = , then this can be expressed as
(s + a)2 + b 2
c
G (s) = .
s + 2as + a 2 + b 2
2

Thus, using the relations

a c
ω n = a 2 + b2 , ξ = 2 , k= 2
a + b2 a + b2
the standard and this form are found to be equivalent. In this form, the poles of the system are
computed to be
s = − a + jb, and s = − a − jb . Therefore we can see that
a = ξω n , b = ω n 1 − ξ 2 .
In other words, the real part of the system poles is − a = −ξω n and the imaginary part is
± b = ±ω n 1 − ξ 2 .

5. Pole-zero plots

The poles and zeros of a dynamic system can be plotted on a complex plane. From the locations of
these poles and zeros, one can assess aspects of the dynamic performance of the underlying physical
system.

For example, from the discussion of stability analysis, where it was shown that for asymptotic
stability of a system,
Re( pi ) < 0 ∀ i = 1,l, n .
This says that the real parts of all of the system poles must be negative, or lie in the closed left half
plane of the complex plane.

Similarly for BIBO stability, the poles of the system must lie in the open left half plane of the
complex plane. The difference being that poles with zero real parts are allowed. Poles with real parts
equal to zero are often called marginally stable poles since they imply dynamic system modes that are
on the verge of instability.

6. Performance regions

Natural frequency

For a second order system written in standard form with poles

s = − p1 = −ξω n + jω n 1 − ξ 2 , and s = − p2 = −ξω n − jω n 1 − ξ 2 ,
the magnitude of these poles is
M = Re( p1 ) 2 + Im( p1 ) 2
The magnitude is the same for both poles. Substituting in this data,

M= (− ξω n )2 + (ω n 1−ξ 2 )=
2
(ξω n )2 + ω n2 − (ξω n )2 = ωn .

Thus systems with equal values for ω n (often called the natural frequency of the system) have equal
magnitudes for the pairs of complex conjugate poles. In other words, the poles lie on circles centered
at the origin and of magnitude M = ω n . We usually restrict discussion to stable systems, or those
with poles only in the left half plane. Thus systems of equal natural frequency are those with poles on
the semi-circle centered at the origin and of magnitude M = ω n . Systems with natural frequencies
higher than this value appear outside of the semicircle.

Damping ratio

The angle of the poles is given by

 Re( p1 ) 
φ1 = cos −1   .
 M ( p1 ) 
The angle of the other pole can be found to be φ2 = −φ1 since the real parts are precisely the same and
the imaginary parts are negative inverses. On substitution,
 ξω 
φ1 = cos −1  n  = cos −1 (ξ ) .
 ωn 
The parameter ξ is often called the damping ratio of a system. Thus systems with equal damping
ratios have poles that lie along two mirror (in the real axis) vectors emanating from the origin at
angles given by ± φ = ± cos −1 (ξ ) . Systems with higher damping ratios will appear inside the cone
shaped region bounded by the limiting vectors.
Figure of real and imaginary parts of poles, lines of constant natural frequency and damping ratio.

We continue the time domain performance analysis of second order systems by examining the
application of a unit step input to the second order system in standard form. That is, let
1
U (s) =
s
in
ωn2 1
X ( s ) = G ( s )U ( s) = 2 ⋅ .
s + 2ξω n s + ω n s
2

Using inverse Laplace transforms, we can find that

 ξ 
x(t ) = 1 − e −σt  cos(ω d t ) + sin( ω t ) 
 1−ξ 2 d

 
where ω d = ω n 1 − ξ 2 which is often called the damped natural frequency, and σ = ξω n , which is
equal to the magnitude of the real part of the system poles.

The rate of change of this function is given by

 ξ   
xD (t ) = −σe −σt  cos(ω d t ) + sin( ω t )  + ω e −σt  − sin(ω t ) + ξ cos(ω t ) 
 1−ξ 2 d
 d
 d
1−ξ 2 d

   
Collecting terms,
  − ξ 2ω − ω 
Dx(t ) = e −σt  (− σ + σ )cos(ω d t ) +  n d 
sin(ω d t )

 = −
(
 n )
 ω ξ 2 + 1 − ξ 2  −σt
e sin(ω t ) .
  − ξ 2    − ξ 2  d
  1    1 

It is noted that the initial value of the system is

 ξ 
x(0) = 1 − e0  cos(0) − sin( 0)  = 1 − 1(1 − 0) = 1 − 1 = 0 .
 1−ξ 2 
 
The initial slope of the output is
xD (0) = 0 .
Also, the final value of the output is
lim x(t ) = x f = 1 − 0 = 1 .
t →∞

Step Response
step response of a second order system

0.8

e
d
ut 0.6
i
pl
m
A
0.4

0.2 note: initial zero slope

0
0 2 4 6 8 10 12
Time (sec.)

Beyond these basic calculated artifacts of the output, the shape of the response of the system to a unit
step input is essentially defined by four parameters:
• The rise time, tr , which is defined here as the first time the system crosses its eventual final
value after the application of the step input. This is a measure of the speed of response of the
system.
• The peak time, t p , which is the time at which the system reaches its highest or peak value
after a step input before returning back down and eventually settling at a final value. This is
also a measure of the speed of response of a system, and is usually only used in conjunction
with finding the overshoot in a system.
• The maximum percentage overshoot, M p , is defined as how far the system goes past its
eventual final value at the peak after a step input. This is a measure of how oscillatory the
system is.
• The settling time, t s , which is defined as the time beyond which the system stays within a pre-
specified percentage bound of the final value after a step input. This is a combined measure of
the speed of response of the system and its oscillatory nature. It is a good practical measure of
how fast the system can reach a new steady operating condition after a disturbing input.

• The rise time, tr .

This parameter can be defined in several ways, but for our purposes will be defined as the time after a
step input is applied until the system first crosses its eventual final value. Therefore in order to
compute the rise time, we need to find the first time at which x(t ) = x f . By observing the system
output solution, this will occur when the trigonometric parts of the output are zero, leaving only the
constant out front, which is precisely the final value. That is at the rise time,
 ξ
cos(ω d t ) + sin(ω d t ) = 0 .
1−ξ 2
By rearranging this we get
ξ
1+ tan(ω d t ) = 0
1−ξ 2
1−ξ 2
tan(ω d t ) = −
ξ
1  2 
−1  1 − ξ 
t=− tan
ωd  ξ 
 
Thus the rise time can be expressed
1 1  2 
−1  1 − ξ 
tr = − tan .
ωn 1 − ξ 2  ξ 
 

This function can be very roughly approximated by

1 .8
tr =
ωn
where the approximation takes into account the possibility of extra poles and zeros affecting the
response of the system. Thus the approximation is only a conservative measure. However, the
approximation is useful in that it makes clear that the rise time is essentially not a function of
damping ratio at all, but rather only a function of the system natural frequency.

Rise time variation with damping

2

1.8

1.6

1.4

r 1.2
ot
c
af 1
e
m
ti
e 0.8
si
r
0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
damping ratio

The figure indicates the value β in the function

β
tr = .
ωn
where
1  2 
−1  1 − ξ 
β =− tan .
1−ξ 2  ξ 
 
It shows that β does not vary significantly over a wide range of damping ratios. Thus considering
this factor as a constant is a reasonable approximation. It is also noted that β appears to vary from
1.4 to 1.1. This is in contrast to the chosen value of 1.8 in the approximation, a difference which can
be attributed to other factors such as extra poles and zeros in real systems.

• The peak time, t p .

The peak time is another measure of the speed of response of a system, much like to the rise time. It
is most often used in conjunction with a measure of the maximum overshoot of a system. The peak
time is defined as the moment when the system reaches its maximum value. This is equivalent to
saying when the slope of the system reaches zero for the first time after the initial step is applied.
That is the first time after t = 0 when
( )
 ω ξ 2 + 1 − ξ 2  −σt
x (t ) = − n e sin(ω t ) = 0 .
 1−ξ 2  d
 
This condition will occur when
sin(ω d t ) = 0 .
for the second time. The first time occurs when t = 0 . Thus the peak time, t p , satisfies
ωdt p = π
or
π π
tp = = .
ωd ωn 1 − ξ 2
It can be shown that this function has stronger variation with damping ratio than the rise time, but
also it is seen that the variation with natural frequency is precisely the same as for rise time.

• The maximum percentage overshoot, M p .

The maximum percentage overshoot is the largest excess, expressed as a percentage, by which the
output of the system exceeds its final value. This will occur at precisely the peak time. That is, the
maximum overshoot will occur when the output is
 ξ 
x(t p ) = 1 − e p  cos(ω d t p ) + 
−σt
sin(ω t )
 1−ξ 2 d p

 
and given that
π π
tp = = ,
ωd ωn 1 − ξ 2
 π 
x(t p ) = 1 − e  cos(ω π ) + ξ
−σt p
sin(ω
−σt −σ t
) = 1 − e p (−1 + 0) = 1 + e p .
 d
ωd 1−ξ 2 d

ωd 

Also, noting that the final value of the output is x f = 1 , the excess above the final value, divided by
the final value, and expressed as a percentage is
 π −πξ
−σt p −ξω n ⋅
e ω n 1−ξ 2 1−ξ 2
Mp = × 100% = 100e = 100e .
1
It is noted that the maximum percentage overshoot is a function of damping ratio alone, and not one
of natural frequency.

The inverse function for damping ratio from overshoot can be found as
2
  M p 
 ln 
 100  
   
ξ= 2
.
  M  
π 2 +  ln p  
  100  

• The settling time, t s .

The settling time is a measure of the time that it takes before the output of a system enters a region
near to its final value and no longer leaves that region. From examination of the output equaition
 ξ 
x(t ) = 1 − e −σt  cos(ω d t ) + sin(ω t ) 
 − ξ 2 d

 1 
it is seen that the trigonometric terms in the bracket will continuously oscillate with a constant
amplitude. Therefore if the system is to settle to some steady state condition, the exponential term
must decay to zero. To find the settling time it is sufficient to find the time at which the exponential
term has decayed to smaller than the bound required for the settling region. Let us define the settling
region bound as b about the final value, x f . Thus the settling time occurs at
t = t s such that e −σt ≤ b .
At the limit of the condition,
e −σt s = b
− σt s = ln(b)
− ln(b) − ln(b)
ts = = .
σ ξω n
Typical values for the bound are 1 %, 2 %, and 5 % depending on designer preferences. Using these
conditions, the settling time formula can be explicitly stated as

Bound, b Settling time formula, t s

1% 3
ξω n
2% 4
ξω n
5% 4.6
ξω n
Example

A system of general form is constructed with

1
X ( s ) = G ( s )U ( s ), G ( s ) = 2
 s  s
  + 2ξ +1
ω
 n ω n

and a unit step input or U ( s ) = 1 . The time response of four versions of this system are shown in
s
the figure. The table summarizes the time domain performance properties of the systems as predicted
by the design equations.

System Natural Damping Rise time, tr Peak time, Maximum Settling

frequency, ratio, ξ tp percentage time, t s
ωn overshoot,
Mp
1 1 2/5 1.8 2.4 25 10
2 2 2/5 0.9 1.7 25 5
3 1 1/5 1.8 3.2 53 20
4 5/3 2/3 1.1 2.5 6 3.6

Step Response
Second order systems step responses
1.6

1.4

1.2

e
d
ut 0.8
i
pl
m
A

0.6

0.4

0.2

0
0 2 4 6 8 10 12 14 16 18 20

Time (sec.)