Lecture 5 Notes

ME 147
System Dynamics and Control

1 General Summary

The derivations of the time response can be a little bit tedious. The characteristics are then
summarised in this section.

1.1 First Order Systems

Time response
c(t) = 1 − e−at (1)

Characteristics
1
Time constant = (2)
a
4
Ts = (3)
a
2.19
Tr = (4)
a

1.2 Second Order Systems

Time response
p ζ p
c(t) = 1 − e−ζωn t cos ωn 1 − ζ 2 t − e−ζωn t p sin ωn 1 − ζ 2 t (5)
1 − ζ2

Characteristics
π
Tp = p (6)
ωn 1 − ζ 2
√ 2
%OS = e(−ζπ/ 1−ζ ) × 100% (7)
4
Ts = (8)
ζωn
ln(%OS/100)
ζ=q (9)
π 2 + ln2 (%OS/100)
2 Response of First Order Systems

A first order system is characterised by the general transfer function:

C(s) a
= (10)
R(s) s+a

If we have a step input R(s) = 1s , we should be able to get the time response of the system using
an inverse Laplace.
a
C(s) = R(s) (11)
s+a
1 a
C(s) = (12)
 s s+a  
1 a A B
c(t) = L −1 = L −1 + (13)
ss+a s s+a

Solving for the partial fractions:

A B A(s + a) + Bs
+ = (14)
s s+a s+a
As + Bs = 0 (15)
Aa = a (16)
A = 1, B = −1 (17)

We then get an inverse Laplace for c(t):

 
−1 1 −1
c(t) = L + (18)
s s+a
c(t) = u(t) − e−at u(t) (19)
−at
c(t) = u(t)(1 − e ) (20)

If we take u(t) as 1 at this point and multiply a necessary gain K to amplify the result at a later
time, we can then simplify the basic form for the response of the system to a step input R(s) = 1s :

c(t) = 1 − e−at (21)

Equation 21 is important as we can use this equation to plot the response of the system in axes
that we are able to observe in the real world: the measured response and time. Notice that this is
a growth function that approaches the value of 1 as t → ∞.

A sample plot of the time response is shown in Figure 1 with several parameters that can be used
to described the response of the system. These are:

1
1. Time constant, a

2
 2. Settling Time, Ts
3. Rise Time, Tr

Figure 1: Plot of time response for a first order system

2.1 Time Constant, 1/a

We can define a parameter called the time constant which is the time it takes for the step response
to rise to 63% of its final value. This may seem arbitrary but it is conveniently defined since this
takes the value of at = 1 such that the response can be plotted in fixed intervals of na for every
n  Z as shown in Figure 1.
1
The time constant, a characterises the response by:
e−at |t=1/a = e−1 = 0.37 (22)
−at
c(t)|t=1/a = 1 − e |t=1/a = 1 − 0.37 = 0.63 (23)

As discussed, the time constant describes the time it takes for the step response to rise to 63% of
the initial value. This applies to a reference point of t = 0 such that the percentage corresponds to:
c(t)f inal − c(t = 1/a)
63% = (24)
c( t)f inal
This would be the rise from t = 0 to t = 1/a but since this is all relative, it should follow that
the rise from a reference of t = 1/a towards t = 2/a should also be 63% of the final value c(t)f inal
referenced from c(t = 1/a).

2.2 Settling Time, Ts

The settling time is the time it takes for a system to stabilise or settle down to a certain level. It
should be noted that there would be a minimum error set to tell whether the system has settled

3
or not. This threshold is defined as 2% of the final level but should be differentiated from the
steady-state error which could be less than 2% of the final value.

Equation 21 is used to determine the expression for Ts . The value of c(t) is set to 0.98 to reflect
the value to be within 2% of the final value. This is more appropriate than using 1.02 since
other systems may have responses that do not go over the final value. Nonetheless, it is another
requirement that the level of c(t) should stay within 2%. Equation 21 achieves this since it does
not have any oscillation at this point. Then,
c(t) = 0.98 = 1 − e−at (25)
−at
e = 0.02 (26)
−at = ln 0.02 = 3.912 (27)
3.912 4
Ts = t = ≈ (28)
a a

2.3 Rise Time, Tr

The risetime is the time it takes for the system response to go from a level of c(t) = 0.1 to
c(t) = 0.9. Other levels are possible and can be used as long as long as the time is referenced
properly. Tr is then derived:
c(t)=0.9
c(t)|c(t)=0.1 = 0.9 − 0.1 (29)
e−at |c(t)=0.9 = 0.1, e−at |c(t)=0.1 = 0.9 (30)
−at|c(t)=0.9 = ln 0.1, −at|c(t)=0.1 = ln 0.9 (31)
2.30 0.11
tc(t)=0.9 = , tc(t)=0.1 = (32)
a a

Finally, Tr is just the difference between these two times1 :

2.30 0.11 2.19
Tr = − = (33)
a a a

2.4 Additional Notes

You may have noticed that the transfer function has a pole located at s = −a and a has also been
used several times in the equations above. This is exactly the case and all time characteristics for
a first order system can be solved using a.

A good example of a first order system is a heat exchanger. Recall the transfer function for a heat
exchanger:
T (s) 1
= (34)
Q(s) RCs + 1
1
where a = RC . Thus, one can solve for the time characteristics using just the thermal resistance
and thermal capacitance of system.
1 2.2
The book Control Systems Engineering by Nise writes this as a
, which is also acceptable.

4
3 Response of Second Order Systems

A second order system is characterised by the transfer function:

C(s) b
= 2 (35)
R(s) s + as + b

From past discussions (Intake Notes), this form can is equal to:

c k
a= = 2ζωn , b= = ωn2 (36)
m m
C(s) ωn2
= 2 (37)
R(s) s + 2ζωn s + ωn2

which then describes the response of the system solely though the combination of the mass, damp-
ing, and spring constant or the derived quantities of damping ratio ζ and natural frequency ωn .
Examples of a second order system include (a) underdamped, (b) undamped, (c) critically damped,
and (d) overdamped mass, damper, and spring systems.

The response of a second order system can be characterised by the following:

1. Peak Time, Tp

2. Percent Overshoot, %OS

3. Settling Time, Ts

4. Rise Time, Tr

Figure 2 shows an underdamped system with legends for the time characterisations. An under-
damped system is used as an example in this case to demonstrate the presence of an overshoot
corresponding to a cmax that may not necessarily happen in other second order systems.

Figure 2: Plot of time response for a second order system

5
 1
Deriving c(t) from the Equation 37 and setting a step input of R(s) = s and using the same method
of partial fractions:
1 ωn2 A Bs + C
C(s) = = + 2 (38)
s s + 2ζωn s + ωn2
2 s s + 2ζωn s + ωn2
As2 + 2ζωn s + Aωn + Bs2 + Cs
C(s) = (39)
s2 + 2ζωn s + ωn2

As2 +Bs2

 =0
2Aζωn s +Cs = 0 (40)
Aωn2 = ωn2


A = 1, B = −1, C = −2ζωn (41)

1 s + 2ζωn
C(s) = − 2 (42)
s s + 2ζωn s + ωn2

The next step is to complete the square of the denominator s2 + 2ζωn s + ωn2 . The usual case is to
look at a from s2 + as + b which is a = 2ζωn . From this we know:
s2 + 2ζωn s + ωn2 = (s + ζωn )2 + D (43)
(s + ζωn )2 = s2 + 2ζωn s + ζωn2 (44)
D= −ωn2 = ζωn2 ωn2 (1
− ζ) (45)
1 s + 2ζωn
C(s) = − (46)
s (s + ζωn )2 + ωn2 (1 − ζ)

The time response requires an inverse Laplace and Equation 46 can further be processed to yield
familiar forms of Laplace: sω2 + ω 2 and s2 +ω
s
2 , which is sin ωtu(t) and cos ωtu(t) respectively.

√ ζ ωn 1 − ζ 2
p
1 s + ζωn 1−ζ 2
C(s) = − − (47)
s (s + ζωn ) + ωn (1 − ζ) (s + ζωn )2 + ωn2 (1 − ζ)
2 2

Obtaining the inverse Laplace of each term:

 
1
L −1 = u(t) (48)
s
 
−1 s + ζωn −ζωn t
p
L − = −e cos ωn 1 − ζ 2 tu(t) (49)
(s + ζωn )2 + ωn2 (1 − ζ)
√ ζ ωn 1 − ζ 2
 p 
 = −e−ζωn t p ζ
1−ζ 2 p
L −1 − sin ωn 1 − ζ 2 tu(t) (50)
(s + ζωn )2 + ωn2 (1 − ζ) 1 − ζ2

Such that the total response is the sum total of Equations 48 to 50. Normalising with respect to
u(t), we have the total response as:
p ζ p
c(t) = 1 − e−ζωn t cos ωn 1 − ζ 2 t − e−ζωn t p sin ωn 1 − ζ 2 t (51)
1 − ζ2

6
 1 − ζ 2 for both sin and cos, and √ ζ
p
Take note that ω = ωn is a constant gain to be multiplied
1+ζ 2
to the amplitude of the sine wave.

3.1 Peak Time, Tp

The peak time is essentially finding a local maxima. It is convenient that the location of the peak,
or cmax , is always the at the time of the first peak in an oscillating response due to the natural
decay in the response.2

Essentially, finding Tp involves solving for the derivative of the time response and setting the
derivative value to 0. Differentiation in the frequency domain is simply done by multiplying the
transfer function by s. Given a step input R(s) = 1s :
1 ωn2
C(s) = R(s)G(s) = (52)
s s2 + 2ζωn s + ωn2
ωn2
L [ċ(t)] = sC(s) = 2 (53)
s + 2ζωn s + ωn2

The process of solving for Tp is similar although multiplying C(s) by s removes the first term 1s
that is present in Equation 42. Thus, there is no need to solve for partial fraction but rather just
getting a familiar inverse Laplace from Equation 37. Completing the square and getting a familiar
inverse Laplace in the form of sin ωtu(t):
p
√ωn ωn 1 − ζ 2
ωn2 1−ζ 2
sC(s) = = (54)
(s + ζωn )2 + ωn2 (1 − ζ) (s + ζωn )2 + ωn2 (1 − ζ)

where √ωn is gain multiplied to the sine waveform, still normalised by u(t):
1−ζ 2
ωn p
ċ(t) = L −1 [sC(s)] = p e−ζωn t sin ωn 1 − ζ 2 t (55)
1 − ζ2

And since we are looking for local maxima, we set the derivative ċ(t) = 0 where the only term that
can equal to zero is the sin θ(t). Since the sin θ(t) equates to zero every π, we then have:
p
ωn 1 − ζ 2 t = nπ (56)
nπ
t= p (57)
ωn 1 − ζ 2

Finally, since we are only interested in the first local maxima3 , we can set n = 1 to get:
π
t = Tp = p (58)
ωn 1 − ζ 2
2
In case of multiple inputs, there is a possibility that the peak of the overall response will not be at the first local
maxima. However, this does not go against the definition of the peak time since Tp is characterised for a response
to a single input and it is possible to decompose a response into different time response components by spectral
decomposition.
3
We will have the same equation for Tp if we are looking a for the first local minima or the lowest level of response.
This happens when we are concerned with an input that lowers the level of response rather than growing it.

7
3.2 Percent Overshoot, %OS

The peak time, Tp corresponds to the maximum level of response that is more than the final level
of the system response. The amount of overshoot is the amount that this maximum level exceeds
the final level of the system response. The Percetage Overshoot, %OS, is then defined as:
cmax − cf inal
%OS = × 100% (59)
cf inal

Since we already know Tp , wep can then evaluate c(t) from Equation 51 at t = Tp , c(Tp ). This
convenient cancels out the ωn 1 − ζ 2 inside the sin and cos functions leaving only:
√
!
2 ζ
cmax = c(Tp ) = 1 − e(−ζπ/ 1−ζ ) cos π + p sin π (60)
1 − ζ2
√ √
2 2
= 1 − e(−ζπ/ 1−ζ ) (−1 + 0) = 1 + e(−ζπ/ 1−ζ ) (61)

With cf inal = 1 and substituting Equation 61 to Equation 59:

√ 2
%OS = e(−ζπ/ 1−ζ ) × 100% (62)

The above discussion lets you solve for the overshoot given and time to peak given the system
characteristics. The reverse is also true such that if you are given a time response data of a system,
you should be able to obtain cmax and Tp through data processing. After which, you will be able
to solve for the overall damping, and subsequently the natural frequency of the system given %OS:

ζ 2π2
ln2 (%OS/100) = − (63)
1 − ζ2
ln (%OS/100)(1 − ζ ) = ζ 2 π 2
2 2
(64)
ln2 (%OS/100) = ζ 2 π 2 + ζ 2 ln2 (%OS/100) (65)
ln(%OS/100)
q =ζ (66)
π 2 + ln2 (%OS/100)

This conveniently illustrates that the %OS tends to zero as the damping ratio ζ increases making
the system more overdamped or even critically damped.

8
 Figure 3: Relation of overshoot and damping

3.3 Settling Time and Rise Time

After some time, t = Ts , the response will stabilise around 2% of the final value. This is the same
as the definition of Ts for a first order system. It so follows that the process of solving for Ts is the
same. Equation 27 is used but a is substituted with = ζωn such that:

−ζωn t = ln 0.02 = 3.912 (67)

3.912 4
Ts = t = ≈ (68)
ζωn ζωn

Unfortunately, Tr cannot be directly calculated for a second order system. The above equations are
all functions of ζ but rise time do not have a simple analytical relationship with ζ. This, however,
can be obtained by processing data.

4 Additional Note

You may have noticed that the numerator is set to either a for a first order system, or b for a second
order system. As you may have realised, not all transfer functions do have a numerator of either
a or b only. However, it still stands the the numerator can be a constant multiplied by a constant
gain K. That said, it might be the case that either of the two can appear:

C(s) Ka
= (69)
R(s) s+a
C(s) Kb
= 2 (70)
R(s) s + as + b

9
The block diagram for these would be:

Figure 4: Block diagram with gain K

5 MATLAB

Data can be generated using a transfer function. The reverse is also true in that a transfer function
can be determined from data. We first look at how we can generate a time response and its
characteristics using a transfer function.

Consider a second-order system:

b 9
G(s) = = 2 (71)
s2 + as + b s + 2s + 9

We setup the system by

Figure 5: Setting up a transfer function

For a step input, we can directly determine all time characteristics with just one function: stepinfo(sys)

10
 Figure 6: Getting time characteristics

Finally, we can plot the time response using another function: step(sys)

Figure 7: Plotting the time response

11