6.

Frequency Response of Control

System
Introduction:
• The design of feedback control systems in
industry is more often accomplished using
frequency-response methods.
– provides good designs in the face of uncertainty in
the plant model,
• for example, for systems with poorly known or changing
high-frequency resonances, we can temper their feedback
compensation to alleviate the effects of those uncertainties.
– experimental information can be used for design
purposes,
• for example, raw measurements of the output amplitude and
phase of a plant undergoing a sinusoidal input excitation are
sufficient to design a suitable feedback control.

Chapter 6 1
 – intermediate processing of the data, such as
finding poles and zeros, is not required to arrive
at a system model.
– for relatively simple systems, it is still often the
most cost-effective design method.
 Note: The method is most effective for
systems that are stable in open-loop.
Advantage of Frequency Response
i. Readly availability of sinusoidal test signals for
various ranges of frequency and amplitudes.
– Thus the experimental determination of the frequency
response of a system is easily accomplished and is the
most reliable and uncomplicated method for the
experimental analysis of a system. Hence:
Chapter 6 2
 Unknown transfer function of a system can be
deduced from the experimentally determined
frequency response of a system.
The design of a system in the frequency domain
provides the designer with control of the bandwidth
of a system and some measure of the response of the
system to undesired noise and disturbance.
ii. The transfer function describing the sinusoidal
steady-state behavior of a system can be obtained
by replacing s with jω in the system TF.
6.1. Frequency Response
• It is the response of a linear system to sinusoidal
inputs.
• It is the variation of the magnitude and phase of
the response with frequency of the input.
Chapter 6 3
• Consider a system described as follows.
Y ( s)
G ( s )
U ( s)
– where the input u(t) is a sine wave with an
amplitude Uo:
u(t ) U o sin t
– which has a Laplace transform:

U ( s ) U o
s2   2
– with zero initial conditions, the Laplace
transform of the output is:

Y ( s ) G ( s )U o 2
s  2
Chapter 6 4
• A partial-fraction expansion of the above
(assuming the poles of G(s) are distinct) is:

1 2 n o o*
Y ( s)    ...   
s  a1 s  a2 s  an s  j s  j
– where a1,a2,…,an are the poles of G(s), αo and αo* are
complex conjugates to be determined.
• The time response that corresponds to Y(s) is:

y (t ) 1ea1t   2ea2t  ...   n ean t  2 o sin(t   )

1 Im(o )
where   tan
Re(o )
Chapter 6 5
• If all the poles are –ve, the natural unforced
response will die out eventually and therefore
the steady-state response of the system will be
due solely to the sinusoidal term, which is
caused by the sinusoidal excitation.
• The remaining sinusoidal term will have the
form:
y (t ) U o A sin(t   )
where:
A  G ( j )  G ( s ) s j   Re  G ( j )  2   Im  G ( j )  2
-1 Im  G (
j ) 
 =tan G ( j )
Re  G ( j ) 

Chapter 6 6
• The equation shows that a stable system with
transfer function G(s) excited by a sinusoid
with unit amplitude and frequency ω will, after
the response has reached steady state, exhibit
a sinusoidal output with a magnitude (A(ω)
) and
a phase at the frequency ω.
Note:
• In frequency response, we not only
understand how a system responds to a
sinusoidal input, but also evaluate G(s) with s
taking on values along the jω axis which is
very useful in determining the stability of
closed loop system.
• jω axis is the boundary between stability and
instability. Example………
Chapter 6
…. 7


Frequency response of 1st Order System

1
G( s) 
1  s
1 1 
 G ( j )    j
1  j 1    2 2
1   2 2
1
 A  G ( j )  ;   tan  1(   )
1   2 2
Frequency response of 2nd Order System
n2
G( s)  2
s  2n s  n2
n2
 G ( j )  2
   j 2n  n2
Chapter 6 8
 n2 1
 G ( j )  2 
 n
   2
  j 2n 
 1    2
 
  

j  2
 

  n   n 

 2
1      j  2  
      
  n   n 
 G ( j )  2
 2  2
    2  
 1        
  n   n 

   
   2 
1 
G ( j )  ; =tan -1   n 
2
2 
 2
      2    
 1    
 1        2     n 
  n    n 
Chapter 6 9
6.2. Frequency Response from a pole-zero
plot
• The magnitude and phase of G(jω) can be
found from the pole-zero plot for a system
k s  z1 s  z2 ....... s  zm 
G ( s) 
s  p1 s  p2 ............ s  pn 

Steps
1. Plot the position of each pole and zero
2. Mark the position s=jω.
3. Draw lines from each pole and each zero to
the point s=jω.
4. Measure the length and angles of each of the
lines
Chapter 6 10
5. The frequency response function is then:
k product of the lengths of the lines from zeros
G ( j ) 
product of the lengths of the lines from poles
G ( j ) sum of angles of lines from zeros -
sum of angles of lines from poles
Example…………

6.3. Frequency Response for series elements

G1(s) G2(s) G3(s)

G ( s ) G1 ( s )G2 ( s )G3 ( s )

G ( j ) G1 ( j )G2 ( j )G3 ( j )
Chapter 6 11
 G ( j )  G1 ( j ) 1. G2 ( j ) 2 . G3 ( j ) 3
 G1 ( j ) G2 ( j ) G3 ( j ) 1  2  3
Thus:
G ( j )  G1 ( j ) G2 ( j ) G3 ( j )
 1  2  3
Example………….

6.4. Bode Plot

• A plot of logarithm of │G(jω)│ and phase angle
of G(jω) both plotted against frequency in
logarithmic scale.

Chapter 6 12
• The standard procedure is to plot 20log│G(jω)│
and (ω) versus log ω.
• The unit of 20log│G(jω)│ is decibel, dB.
• Allows a rapid and reasonably accurate plot to
be made.
• Composite systems will be handled with
relative ease.
• Possible to make the plots as accurate as
required in a given frequency range.
• Plotted for a unit change in frequency
log(ω2/ω1)=1 or ω2=10ω1.
• This range of frequency is called a decade.
Chapter 6 13
• Consider the transfer function

G ( s ) k
 s  1 
 G ( j ) k
 j  1 

 1

k 1  j   1 
.
 s  2   j  2  1  j / 2   2 
1 1  j / 1
G ( j ) k
2 1  j / 2
  2  1/ 2     2  1/ 2

 
G ( j ) db 20 log k
1
2
 20 log 1    
  1  
 20 log 1    
  2  
   

Chapter 6 14
• So, we can add contribution due to the
individual magnitude terms.
• The phase graph when there are a number of
elements is just the sum of the separate
elements.
• The graph covers a greater range of
frequencies and it is approximated as a
straight line.
• Since Bode plots for a system can be built up
from the plots for the individual elements
within the transfer function for that system it
is useful to consider the plots for element
commonly found in the transfer function.
Chapter 6 15
 Let:
k 1  jTa 1  jTb ...
G ( j ) 
    
r  2
 j  1  jT1 1  jT2 ...  1  2 j   j 
n n


• Constant Gain:
– G(s)=k → G(jω)=k → │G(jω)│ =k
– │G(jω)│db = 20 log k; =0
20log│G(jω)│
20 log k

0.1 1 10 100 ω [rad/sec]

(ω)
90

Chapter 6
-90 16
A multiple pole at the origin:
– G(s)=1/sm → G(jω)= 1/(jω)m
– │G(jω)│dB = -20 log ωm= -20m log ω; = - π/2
• That is a straight line of slope -20m
dB/decade.
A zero at the origin:
– G(s)=s → G(jω)= jω
– │G(jω)│db = 20 log ω; =π/2
• That is a straight line of slope 20 dB/decade.

Chapter 6 17
 40
a zero at the origin
20

0.1 1 10 100 ω [rad/sec]

-20
a pole at the origin
-40

90 a zero at the origin

-90 a pole at the origin

A Real Pole:
1 1
G (s)   G ( j ) 
1  s 1   ( j )
 1  2
G ( j ) dB  20 log    10 log  1     
2  
 1     
Chapter 6 18
 tan       tan  1   
2
for   1/  ;     0
 G ( j ) db  10 log1  0db
• This tells us that at low frequency there is a
straight line.
• For ω>>1/τ:
│G(jω)│dB=-20 log ωτ
• We get a straight line of slope -20db/decade
• when ωτ=1 .
│G(jω)│dB ≈ 0
• ω=1/τ is called break point or corner frequency
Chapter 6 19
• The error for 0<ω<1/τ is:
– -10 log (1+ω2τ2)+20log 1;
– at ω=1/ →-10 log (1+1)+10log 1=-3dB
• The error at ω=1/(2 τ) is:
– -10 log (1+1/4) + 20 log 1 = - 1dB
• For 1/ ≤ ω <  ; the error is:
– -10 log (1+ ω22) + 20 log ω ;
at ω = 1/  →-10 log (1+1) + 20 log 1 = - 3dB
• At ω=2/
– -10 log (1+4) + 20 log2 = - 1 dB
• The two straight lines are called asymptotic
approximations
Chapter 6 20
 40 Break Point
a real zero
20

ω [rad/sec]
20 Actual
a real pole
40

a real zero
+90o
ω [rad/sec]
0.1/τ 1/τ 10/τ 100/τ
-90o
a real pole

A real zero:
– G(s)=(1+τs) → G(jω)=(1+ jωτ)
2
– │G(jω)│db=20log 1  ( ) =10log (1+ (ωτ)2)
– tan  = ωτ
Chapter 6 21
 – for ω τ <<1, │G(jω)│dB=10log (1)=0
– for ω τ >>1, │G(jω)│dB=10log (ωτ)2=20log (ωτ)
• i.e a straight line of slope 20 db/decade.
A pair of complex Poles
n2 1
G( s)  2 2
 G ( j )  2
s  2n s  n      j (2 /  )
 1     n
  n  
 2 2
   2
 G ( j ) dB  20log  1      (2 / n ) and
  n  
    
 2    
  tan  1   n 
2
    
 1    
Chapter 6   n  22
 – for (ω/ ωn)<<1, │G(jω)│db=-20log 1=0 dB
– for (ω/ ωn)>>1, │G(jω)│db=-20log (ω/ ωn)2
=-40log (ω/ ωn)
• The above two results shows as that at low
frequencies the magnitude plot is a straight
line at 0 dB, while at high frequencies it is a
straight line of slope -40 dB/decade.
• Break point is at ω = ωn.
• True value depends on the value of ζ.
– for (ω/ ωn)<<1, i.e. (ω/ ωn)=0.1 or less, then
 ≈-tan-10=0o
– for (ω/ ωn)>>1, i.e. (ω/ ωn)=10 or greater, then
 ≈-tan-1∞=-180o
Chapter 6 23
 ω/ ωn

-20

-40 1 10

ω/ωn=0.2

ω/ ωn

-90

-180

Chapter 6 24
ω/ωn=5
A pair of complex zeros

G( s) 
s 2
 2n s  n2
 G ( j ) 
 n2   2   j 2n
n2 n2
 2 2
   2
 G ( j ) db  20log  1      (2 / n ) and
  n  
    
 2    
  tan  1   n 
2
    
 1    
  n 

• The plot of the above case is exactly the

mirror image of the previous case.
Chapter 6 Examples………………. 25
Performance Specifications in the frequency domain
• Qns.
– How does the frequency response of a system relate to
the expected transient response of the system?
– Given a set of time-domain (transient performance)
specifications, how do we specify the frequency response?
• Taking a simple second-order system as an
example, the above questions could be answered
by considering the time–domain performance in
terms of overshoot, settling time and other.

Chapter 6 26
• From the block diagram, the closed-loop TF is:
n2
T (s)  2
s  2n s  n2
• The frequency response of this system was given
as:

Bandwidth

Chapter 6 27
• As shown, the damping ratio of the system is related
to the maximum magnitude or value of the frequency
response, Mpω, which occurs at the resonance (critical)
frequency, ωr .
• Definition:
– Cutoff frequency is the frequency, ωB , at which the
frequency response has declined 3dB from its low-frequency
value.
– Bandwidth is the frequency range 0<ω<ωB in which the
magnitude of the closed loop doesn’t drop -3dB.
• The resonance frequency, ωr ,and the -3dB cutoff
frequency, ωB , can be related to the speed of the
transient response.
– As the cutoff frequency or bandwidth increases, the rise time
of the step response of the system will decrease.
• The overshoot to a step input can be related to Mpω
through the damping ratio ς.
Chapter 6 28
• The following relate the resonance magnitude and
frequency to the damping ratio of the 2nd-order
system.

Resonance frequency:
r n 1  2 2 ,   0.707
Maximum magnitude:

 
1
2
M p  G (r )  2 1   ,   0.707

Fig: The maximum of the frequency response Mpω and

the resonance frequency ωr versus ς for a pair of
complex conjugate poles.

Chapter 6 29
• Step response overshoot could be read from the
following graph or could be utilized using the
equation shown below as function of damping ratio.

  / 1  2
or M p % 100e

Chapter 6 30
• The above shows that as the resonance peak Mpω
increases in magnitude, the overshoot to a step
input increases.
• Also, the bandwidth of a system ωB can be
approximately related to the natural frequency of
the system.

Chapter 6 Fig: Normalized bandwidth ωB /ωn versus ς for a second order system 31
• The greater the magnitude of ωB when ς is
constant, the more rapidly the response
approaches the desired steady-state value.
• Thus the desirable frequency-domain
specifications are:
1. Relatively small resonant magnitudes : Mpω <
1.5
2. Relatively large bandwidths so that the system
time constant τ=1/ ς ωn is sufficiently small.

6.5. Experimental Determination of Transfer

Function
• When the transfer functions of a system are
unknown, Bode Plots are used.
• Frequency response data is first obtained
experimentally in the frequency range of interest. 32
Chapter 6
• The transfer function within a certain degree of
accuracy can then be obtained by fitting an
asymptotic log-magnitude plot to the
experimental data.
• The phase plots are used as a check of the result.
• Asymptotes are drawn to the experimental data.
1. If the slope of the initial asymptote is -20dB/decade
then there is a pole at the origin. If it is +20dB/decade
then there is a zero at the origin.
2. If the slope, in going from one asymptote to the next
changes by + 20dB/decade, then there is a real zero or
pole and the TF is (1+τs) or 1/(1+τs), where the
frequency at which the slope changes is 1/τ.

Chapter 6 33
 3. If the slope, in going from one asymptote to the
next changes by + 40dB/decade, there is a pair
of complex zeros or poles and the transfer
function has a term:
s 2  2n s  n2 n2
or 2
n2 s  2n s  n2
where ωn is the frequency at which slope changes
• Tables are used to give an estimate of ζ.
• All the constituent elements are combined to
give a TF.
• Then the phase graph is plotted and checked
with the experimental data.
Chapter 6 Examples………………. 34
 End of chapter Six

Chapter 6 35