Chapter 5

Frequency Domain
Analysis

EE4L005 DIGITAL CONTROL SYSTEMS

N. C. Sahoo
 Introduction
• Historically, the analysis and design of continuous-
time control systems have been well developed.
Practically, all these methods can be extended to
digital control systems.
• Frequency-domain analysis possesses a wealth of
graphical techniques that can be applied to LTI
control systems.

• Such well-known methods as the Nyquist criterion

for stability analysis, the Bode plot, and Nichols
chart can all be extended to analysis and design of
digital control systems.
• The basic feature of the frequency-response method
is that the performance description of a LTI system is
given in terms of steady state response to varying
sinusoidal inputs.

• The characteristics of time-domain performance can

be predicted based on frequency-response. For
example, the bandwidth in frequency-domain is
directly linked to how fast and oscillatory the time-
response will be.

• Therefore, in many design problems, instead of

specifying maximum overshoot, rise time, and settling
time of the time-response, the designer often
chooses to specify the bandwidth of the system.
• Another advantage of frequency-domain analysis is
that the controller can be treated as a filter, and in the
case of a digital controller, it is a digital filter.

• In many situations, the characteristics of a digital

filter are better described in terms of its frequency-
domain properties.
• In Laplace-transform domain, the sinusoidal steady
state analysis is performed by setting s  j .
jT
• Thus, in z-domain, z is replaced by  .

• There are some unique characteristics found in

frequency-domain study of digital control systems.
• For example, the response of a linear continuous-
time system contains only the same frequency
component as the sinusoidal input signal, and only
in systems with nonlinear elements, the harmonics
of input signal are found.

• However, in linear digital control systems, the

sampler (real/fictitious), acts as a harmonic
generator so the system response in general will
contain higher harmonics of the input signal. These
high-frequency components often make the
construction and interpretation of frequency domain
plots more difficult.
• The well-known methods for frequency domain
analysis are as follows.

(a) Nyquist Plot:

• The Nyquist plot of a transfer function, usually the

loop transfer function GH (z ) , is a mapping of the
Nyquist path in z-plane onto the GH (z ) plane in
polar coordinates.

• Thus, the Nyquist plot is also known as polar plots.

Absolute and relative stability of a system can be
determined from Nyquist plot of GH (z ) .
(b) Bode Diagram:

• The Bode diagram is a plot of the amplitude in

decibels (dB) and phase angle of a transfer function,
usually the open-loop transfer function, as a function
of frequency  .

• This can be used to investigate the absolute and

relative stability of system.
(c) Gain-Phase Plot:
• The gain-phase plot of open-loop transfer function is
a plot of amplitude (dB) vs. phase (degree). This plot
also can be used to determine absolute and relative
stability of a system.
 Polar Plot
• Given the loop transfer function GH (z ) of a closed-
loop digital control system, the polar plot of GH (z ) is
obtained by setting
jT
z 
and varying  from zero to infinity. This corresponds
to a mapping of the points on the unit circle
in z-plane, z  1 , onto the GH (z ) plane.
Example:
Consider the loop transfer function of a digital control
system given by
1.57
G (s) H ( s) 
s ( s  1)
 Let us say  s = 4 rad/s (T = /2). The z-transform of
G(s)H(s) is 1.243 z
GH ( z ) 
( z  1)( z  0.208 )
By substituting z   jT ,
jT 1.243 jT
GH ( ) jT
(  1)( jT  0.208 )
Since GH ( jT )is a complex quantity, it can be written as:
jT jT jT
GH ( )  GH ( ) GH ( )

= Re[ GH ( jT ) ]+ j Im[ GH ( jT )]

The graphical interpretation of magnitude and phase
of GH ( jT ) in z-plane when z takes any value z1 j1T
on the unit circle is shown.
 jT 1.243  A
GH ( )
BC
where A is the length of phasor drawn from the zero at
z = 0 to z1 , and B and C are the lengths of the phasors
drawn from the poles at z = 0.208 and z=1,
respectively, to z1 .
• The angle relationship is given by:
jT
GH ( )   A   B  C

Since the unit circle is traversed once for every   n s

for n  1,2, the polar plot of GH (z ) repeats over the
same frequency range.

Polar plot of G(s)H(s)

and GH(z)
 • The polar plot of GH (z ) when z takes values on unit
circle is shown.
• The polar plot of G ( j ) H ( j ) for positive values of  is
also shown.

• The polar plot of GH (z ) repeats for    s  4 . Thus,

only the portion for   0 to   4 rad/s is shown; the
portions corresponding to rest of the frequency range
are all identical.
• Since the portion of the unit circle for   0 to   s / 2 is
symmetrical to that from    s / 2 to    s , the
corresponding portions of the polar plot of GH (z ) will
also be symmetrical.
 Bode Plot
• Given a transfer function G (z ) , the Bode plot
consists of a magnitude plot and a phase plot.
jT
• The magnitude plot is the plot of 20 G ( ) in
decibels (dB) vs.  . The phase plot is G ( jT )
in degrees vs.  .
j T
• As known, the frequency response of G ( )
repeats every  s , and the polar plot is symmetrical
about real axis. Thus, the magnitude of the Bode
diagram is identical for every  s .
Example: GH ( z )  0.0952 Kz {K = 1, T = 0.1sec}
( z  1)( z  0.905 )

Bode- (magnitude) plot

 Bode- (phase) plot

Due to logarithmic  axis, magnitude plot of G ( jT ) is

distorted, but it repeats for every  s . The shape of the
phase plot also repeats for the same frequency range.
Nyquist plot and Bode plot show that the absolute and
relative stability of the closed-loop system are governed
only by the Bode plot from   0 to  s 2 .

• Thus, for all practical purposes, it is only necessary

to plot the Bode plot over this frequency range.

Example:

0.758( z  1)
G( z)  K = 1, T = 0.1sec
( z  1)( z  0.242 )

Bode plot is shown for the frequency range   0 to

 s 2  31.4 rad/s.
• One of the advantages of Bode plot is that, since the
magnitude curve is in decibels, changing the gain
factor of G (z ) simply causes the curve to move up or
down without distortion and by multiplying G (z ) by
another transfer function H (z ) simply involves addition
of the magnitude curve of H (z ) to that of G (z ) .
20log10 G( jT ) H ( jT )  20log10 G( jT )  20log10 H ( jT )

Similarly, for the phase curve,

jT jT jT jT
[G ( ) H ( )]  G ( )  H ( )

• This means when G (z ) has a time delay nT,

magnitude curve is not affected, but phase is
reduced by nT.
• Another advantage of Bode plot is that, in s-domain,
the magnitude curve can be approximated by straight
lines. However, this convenient feature is lost in z-
domain, since z is related to j through  jT .
• But, one can use a bilinear transformation to
transform the unit circle of z-plane, to the imaginary
axis of another complex plane. The transform
introduced in stability studies (Chapter 4) may be
used for this purpose.
• There is another bilinear transformation which
transforms z into, say, w, such that the imaginary
axis of the w-plane closely resembles the j of the
s-plane.
• Let w be related to z through the z-transform relation,
i.e., 1
w ln z
T
Taking the first term from power series of ln z,
2 z 1 1  (T / 2) w
w z
T z 1 1  (T 2) w
These are defined as w-transformation.
jT
• Substituting z    cos T  j sin T :
2 sin T 2 T
w j  j tan
T 1  cos T T 2
Thus, when z takes values on the unit circle, w is
imaginary and can be written as
s 
w  j w  j tan
 s
• This equation establishes the relationship between
the transformed “frequency”  w and the real
frequency , and the sampling frequency  s .

• For frequency-domain analysis, this transformation

can be substituted into the transfer function G (z ) .
• If G (z ) is a rational function of z, consisting of first-
order and second-order factors in the numerator and
denominator, then the transformed function G (w) will
also be a rational function with first-order and/or
second-order factors.
• Advantage is: Bode plot of G (w), or rather G( j w ) , can
now be sketched using straight-line approximation
method, similar to s-domain techniques.
Example:
0.0952 Kz
GH ( z )  , T = 0.1sec
( z  1)( z  0.905 )

Applying w-transformation,
10.02 K (1  0.0025 w 2 )
GH ( w) 
w(1  1.0026 w)
Bode plot of GH (w) with K =1 is shown.