Frequency Response Techniques

The results of frequency response

compensation techniques are not new
or different from the results of root
locus techniques

1
 The Concept of Frequency Response
In the steady state:
• Sinusoidal inputs to a linear system generate sinusoidal responses of
the same frequency.
• Even though these responses are of the same frequency as the input,
they differ in amplitude and phase angle from the input. These
differences are functions of frequency.

• Sinusoids can be represented as complex numbers called phasors.

M1cos (t + 1) can be represented as M11 where the

frequency, , is implicit.

2
Consider the mechanical system in the figure shown below

M o ( )o ( )  M ( ) ( )  M i ( )i ( )

M o ( )o ( )  M i ( ) M ( )[i ( )   ( )]

M o ( )
M ( )  ;
M i ( )
 ( )  o ( )  i ( )

M(): The magnitude frequency response

(): The phase frequency response.
Sinusoidal frequency response:
M() (): The frequency response
a. system;
b. transfer function;
Both responses are a function of frequency and apply only to the
c. input and output waveforms 3
steady state sinusoidal response of the system
 Analytical Expressions for Frequency Response
We will use this analytical expression to determine stability, transient
response, and steady-state error

r (t )  A cos t  B sin t  A2  B 2 cos[t  tan 1 ( B / A)]

As  B
R( s)  We can represent the input as
s2   2
a phasor in three ways:

We now solve for the forced response portion (1) in polar form,
of C(s),
M i  i; M i  A2  B 2 ;
As  B
C ( s )  R ( s )G ( s )  G( s)  i  tan 1 ( B / A)
s2   2
As  B
C ( s)  G( s) (2) in rectangular form,
( s  j )( s  j )
K1 K2
A  jB
   Partial fraction terms from G(s)
( s  j ) ( s  j ) (3) using Euler’s formula,

M i e ji
4
 K1 K2
C ( s)    Partial fraction terms from G(s)
( s  j ) ( s  j )
As  B As  B 1 1 1
K1  lim G( s)  G( s) [ A B ]G (  j )  [ A  jB ]G (  j )
s   j ( s  j ) ( s  j ) s   j
2 2j 2
1 MM
 M i e  ji M G e  jG  i G e  j (i G )
2 2
As  B As  B 1 1 1
K 2  lim G( s)  G( s) [ A B ]G ( j )  [ A  jB ]G ( j )
s   j ( s  j ) ( s  j ) 2 2j 2
s  j

1 MM
 M i e ji M G e jG  i G e j (i G )  K1
*

2 2
in which,
M G  G ( j ) ; G  angle of G(j )

5
The steady-state response is that portion of the partial-fraction expansion that
comes from the input waveform’s poles, or just the first two terms, the
sinusoidal steady-state output, Css(s), is

K1 K2
Css ( s )  
( s  j ) ( s  j )
M i M G  j (i G ) M i M G j (i G )
e e
 2  2
( s  j ) ( s  j )
M i M G  j (i G )  jt M i M G j (i G ) jt
css (t )  e e  e e
2 2
MM
css (t )  i G [e  j (t i G )  e j (t i G ) ]
2
css (t )  M i M G cos(t  i  G )
M oo  ( M i i )( M G G )
which can be represented in phasor form as M G G  G ( j )
css (t )  M i M G cos(t  i  G ) G ( j )  G ( s ) s j

MGG is the frequency response function 6

 Plotting Frequency Response

G(j)=MG( )G ( ) can be plotted in several ways; two of

them are
• As a function of frequency, with separate magnitude and
phase plots. When plotting separate magnitude and phase
plots, the magnitude curve can be plotted in decibels (dB)
vs. log , where dB = 20 log M. The phase curve is plotted
as phase angle vs. log .

• and as a polar plot, where the phasor length is the

magnitude and the phasor angle is the phase.

7
 Frequency Response from The Transfer Function
Find the analytical expression for the magnitude frequency
response and the phase frequency response for a system
G(s) =1/(s+ 2̃). Also, plot both the separate magnitude and
phase diagrams and the polar plot (Nyquist plot).
1 2  j 2 j
G ( j )    
2  j 4   2 4   2 4   2
2 2  2 1
G ( j )  [ ] [ ] 
4 2
4 2
4  2
1
20 log( G ( j ) )  20 log  20 log 4   2
4  2

1 4   1 
2
G ( j )   tan [ ]   tan [ ]
2 2
4  2
8
 w=0.1:0.1:100; logw=log10(w); m_G=-20*log10(sqrt(4+w.^2));
p_G=-atand(w/2); plot(w,m_G); figure; plot(w,p_G);

-10
-15
G=1/(s+2);
20 log M

-20
-25 bode(G);
-30
-35
-40
-1 0 1 2
10 10 10 10
Frequency (rad/s)

-20
Phase (degrees)

-40

-60

-80

-100
-1 0 1 2
10 10 10 10
Frequency (rad/s) 9
 1 2  j 2 j G=1/(s+2);w=0:0.1:100; >> nyquist(G,w)
polar plot G ( j )    
2  j 4   2 4   2 4   2 Nyquist Diagram
0.25
Re_G=2./(4+w.^2); 0.2
Im_G= -w./(4+w.^2); 0.15
plot(Re_G,Im_G); 0.1
for i=1:1:10

Imaginary Axis
0.05
x = Re_G(i); 0
y = Im_G(i);
-0.05
txt1 = strcat('\leftarrow w = ', …
-0.1
num2str(w(i)));
-0.15
txt= strcat(txt1,' r/s');
-0.2
text(x,y,txt)
-0.25
end -0.1 0 0.1 0.2 0.3 0.4 0.5
Real Axis
Re G
0 0.1 0.2 0.3 0.4 0.5
0

 w =0.1 r/s

-0.05  w =0.2 r/s

 w =0.3 r/s

-0.1  w =0.4 r/s

 w =0.5 r/s
Im G

 w =0.6 r/s
-0.15  w =0.7 r/s
 w =0.8 r/s
 w =0.9 r/s
-0.2  w =1 r/s

-0.25

10
 Asymptotic Approximations: Bode Plots
Sketching Bode plots can be simplified because they can be approximated as a
sequence of straight lines
K ( s  z1 )( s  z2 )...( s  zk )
Consider the following transfer function: G( s) 
s m ( s  p1 )( s  p2 )...( s  pn )
K ( s  z1 ) ( s  z2 ) ... ( s  zk )
The magnitude frequency response G( s) 
s m ( s  p1 ) ( s  p2 ) ... ( s  pn )
s  j
Converting the magnitude response into dB

20 log( K )  20 log ( s  z1 )  20 log ( s  z2 ) 

20 log( G ( s ) ) 
 ...  20 log s m  20 log ( s  p1 )  20 log ( s  p2 )  ...
s  j

Thus, if we knew the response of each term, the algebraic sum would yield the total
response in dB.

The phase frequency response is the sum of the phase frequency response curves of the
zero terms minus the sum of the phase frequency response curves of the pole terms. 11
 Bode Plots for G(s) = (s + a)
Consider a function, G(s) = (s+a). Letting s = j,
j
G ( j )  ( j  a )  a (
 1) where y = 20 log M; and x = log . The line has
a
when  approaches zero, a slope of 20 when plotted as dB vs. log .
G ( j )  a; 20 log G ( j )  20 log a
j
where   a G(j )  a ( )  90o
a
The magnitude response in dB is

20log G ( j )  20 log a  20 log( )  20 log  (*)
a

Notice from the middle term that

the high-frequency approximation
20 log(2a )  20 log(a )  20 log(2)  6dB
is equal to the low-frequency
approximation when = a, and
20 log(10a )  20 log(a )  20 log(10)  20dB
increases for  > a. • The low-frequency approximation is called
the low-frequency asymptote,
If we plot dB, 20 log M, against log • The high-frequency approximation is called
, Eq. (*) becomes a straight line: the high-frequency asymptote.
y=20x • The frequency, a, is called the break
12
Frequency.
 j
G ( j )  ( j  a )  a (
 1) To draw the curve,
a
when  approaches zero, • start one decade (1/10̃) below the
G ( j )  a; G ( j )  0o
break frequency, 0.1a, with 0 phase,
• and draw a line of slope +45o /decade
  a G ( j )  45o
passing through 45o at the break
j
where   a G(j )  a ( )  90o frequency
a
• and continuing to 90o one decade
above the break frequency

13
It is often convenient to normalize the magnitude and scale the
frequency so that the log-magnitude plot will be 0 dB at a break frequency of
unity. Normalizing and scaling helps in the following applications:
1. When comparing different first- or second-order frequency response plots, each
plot will have the same low-frequency asymptote after normalization and the
same break frequency after scaling.
2. When sketching the frequency response of a function such as
K ( s  z1 )( s  z2 )...( s  zk )
G( s) 
s m ( s  p1 )( s  p2 )...( s  pn )

Each factor in the numerator and denominator will have the same low-frequency
asymptote after normalization. This common low-frequency asymptote makes it
easier to add components to obtain the Bode plot.
To normalize (s + a), we factor out the quantity a and form a(s/a +1). The frequency is
scaled by defining a new frequency variable, s1 = s/a. Then the magnitude is divided by
the quantity a to yield 0 dB at the break frequency. Hence, the normalized and scaled
function is (s1 + 1̃). To obtain the original frequency response, the
magnitude and frequency are multiplied by the quantity a. 14
Asymptotic and actual normalized and scaled frequency response data for s+a

15
Asymptotic and actual normalized
and scaled magnitude response of
(s+a)

Asymptotic and actual

normalized and scaled phase
response of (s +a)

16
 1 1
G( s)  
( s  a ) a ( s  1)
a
Bode Plots for G(s) =1/(s+a)
1 The Bode log magnitude diagram will decrease
G ( j ) 
j
a(  1) at a rate of -20 dB/decade after the break
a
frequency.
when  approaches zero,
1 1 The Bode normalized and scaled log-magnitude
G ( j ) ; 20 log G ( j )  20 log
a a and phase plot
1 1
where   a G(j )     90o
j
a( ) 
a
The magnitude response in dB is
1 
20log G ( j )  20 log  20 log( )  20 log 
a a
1
G ( j ) 
j
a(  1)
a
1
when  approaches zero, G ( j )  ; G ( j )  0o
a
  a G ( j )  45o
1
where   a G(j )     90o
j
a( )
a 17
 Bode Plots for G(s) = s
G(j)= j, has only a high-frequency asymptote. The magnitude is 20 log .
Hence, the Bode magnitude plot is a straight line drawn with a +20 dB/decade
slope passing through zero dB when  = 1. The phase plot, which is a constant
90o

18
 Bode Plots for G(s)=1/s
The frequency response of function, G(s)= 1/s is a straight line with a
-20 dB/decade slope passing through zero dB at = 1. The Bode phase
plot is equal to a constant -90o.

19
Bode plot and Nyquist plot were proposed
to determine the closed-loop stability by
examining G(s)H(s) instead of 1+G(s)H(s)

20
 Bode Plots for Ratio of First-Order Factors
where :
Draw the Bode plots for the system K ( s  3)
G( s) 
s s( s  1)( s  2)
(  1)
K ( s  3) 3 3
G( s)  K The break frequencies are at 1, 2, and 3. The
s( s  1)( s  2) 2 s( s  1)( s  1) magnitude plot should begin a decade below
2
the lowest break frequency and extend a
3K
  0.1; G ( j 0.1)   15K decade above the highest break frequency.
2  0 .1
20 log G ( j 0.1)  20 log 15  20 log K If we choose K = 1, the magnitude plot can be
20 log G ( j 0.1)  23.52dB  20 log K denormalized later for any value of K that is
calculated or known

21
 K ( s  3)
G( s) 
s( s  1)( s  2)

22
 components
Bode
phase
plot

composite

23
 Bode Plots for G(s) = s2 + 2ns + n2
s2 s
G ( s )  s  2n s  n  n (  2  1)
2 2 2

n 2 n
when  approaches zero, n is the break frequency for
2 (*) the second order Polynomial
G ( j )  n ; 20 log G ( j )  20 log n
2
(*)&(**)
( j ) 2
where   n G(j )  n ( )   2   2180o
2

n 2

The magnitude response in dB is

20log G ( j )  20 log  2  40 log  (**)

The asymptotes for the normalized and scaled magnitude plot

24
 s2 s
G ( s )  s  2n s  n  n (  2  1)
2 2 2

n 2
n
when  approaches zero,
G ( j )  n ; G ( j )  0o
2
The phase plot increases at a rate
( j ) 2
where   n G(j )  n (
2
)   2   2180o of 90o/decade from 0.1 to 10 and
n 2

passes through 90o at 1.

G ( j )  180o
when   n
G ( j )  s 2  2n s  n  (n   2 )  2nj
2 2
s  j

G ( j )    2nn j; G ( j )    900
n n

The phase plotted with frequency scaled by n.

25
 Corrections to Second-Order Bode Plots
Whereas the first-order polynomial has a disparity of no more than 3.01 dB
magnitude and 5.71 phase, the second-order function may have a greater
disparity, which depends upon the value of .
G ( j )  s 2  2n s  n  (n   2 )  2nj; M  G ( j )  (n   2 ) 2  ( 2n ) 2 ;
2 2 2
s  j

2n
G ( j )  tan 1 ( )
(n   2 )
2
Normalized and scaled log-magnitude response

26
 Corrections to Second-Order Bode Plots
Whereas the first-order polynomial has a disparity of no more than 3.01 dB
magnitude and 5.71 phase, the second-order function may have a greater
disparity, which depends upon the value of .
G ( j )  s 2  2n s  n  (n   2 )  2nj; M  G ( j )  (n   2 ) 2  ( 2n ) 2 ;
2 2 2
s  j

2n
G ( j )  tan 1 ( )
(n   2 )
2
Scaled phase response

27
 Bode Plots for G(s) =1/( s2 + 2ns + n2)
The magnitude curve breaks at the natural frequency and decreases at a rate
of -40dB/decade.
The phase plot is 0 at low frequencies. At 0.1n it begins a decrease of
-90o/decade and continues until = 10 n, where it levels off at -180o.

Normalized and scaled log-magnitude response

1
G( s) 
s 2  2 n s  n
2

28
 1
G( s) 
s 2  2n s  n
2
Scaled phase response for ̃

29
 Bode Plots for Ratio of First- and Second-Order Factors
Draw the Bode log-magnitude and phase plots of G(s) for the unity feedback system

K ( s  3)
G( s) 
( s  2)( s 2  2 s  25)

Rewrite G(s) showing each term normalized to a low-frequency gain of unity

K ( s  3)
s
3 K ( 1)
G( s)   3
( s  2)( s 2  2 s  25) 50 ( 2s 1)( 25
s2 2 s
 1)
25

Bode magnitude plot components;

20*log10(3/50)
ans =
-24.4370

30
 K ( s  3)
s
3 K ( 1)
G( s)   3
( s  2)( s  2 s  25) 50 ( 2 1)( 25 251)
2 s s2 2 s

Bode magnitude plot composite

3
20 log( )  24.44
50

31
 O

Bode phase components

plot
K ( s  3)
G( s) 
( s  2)( s 2  2 s  25)

composite

32
 Nyquist: Introduction
G(j) WHAT IS A NYQUIST DIAGRAM?

A Nyquist diagram is a plot of G(j), in the

complex plane, for all real values of .

A simple example is given below:

1
G ( j ) 
j  1

33
 Nyquist Diagram
0.5

Imaginary Axis
System: Fs System: Fs
Real: 0.0111 Real: 0.99
Imag: -0.105 Imag: -0.0985
0
You can check your answers using the Nyquist Frequency (rad/s): 10 Frequency (rad/s): 0.0995

System: Fs System: Fs
command in MATLAB Real: 0.196 Real: 0.804
Imag: -0.397 Imag: -0.397
Frequency (rad/s): 1.99 Frequency (rad/s)
>> nyquist(G)
>>[re,im]=nyquist(G,[0.1,0.5,1,2,10]); -0.5
-1 -0.5 System:0 Fs 0.5 1
Real: 0.468
Real Axis
Imag: -0.499
Frequency (rad/s): 1

SUMMARY:
The tabulation method is tedious so
a more efficient method is needed. 34
 Nyquist: Using gain and phase
REMARKS:
1. On easy way to sketch a Nyquist diagram is to transcribe gain and phase
information.
2. Students may realize that gain/phase is already available in the Bode diagram!
From hereon, assume the Bode diagram is already known.

GUIDELINES FOR QUICK SKETCHES (beginning from Bode)

1. Phase is reducing (or becoming more negative) – plot is moving clockwise.

2. Phase is increasing – plot is moving anticlockwise.
3. Gain is reducing – plot is moving towards origin.
4. Gain is increasing – plot is moving away from the origin.

6
G( s) 
( s  1)( s  3)
6
G ( j ) 
( 2  1)(( 2  9)

G ( j )   tan 1 ( )  tan 1 ( )
3 35
36
 Introduction to the Nyquist Criterion
• Knowledge of the open loop system’s frequency response yields information
about the stability of the closed-loop system.
• Frequency response techniques are an alternate approach to the root locus.

Letting
NG N Consider the system
G(s)  ; H ( s)  H
DG DH
NG N H
G(s) H ( s )  ;
( DG DH )
Four important concepts:
N N [D D  NG N H ]
1  G(s) H ( s )  1  G H  G H (1) the relationship between the poles of 1 + G(s)H(s)
DG DH (DG DH )
and the poles of G(s)H(s);
NG
(2) the relationship between the zeros of 1 + G(s)H(s)
G( s) DG
T ( s)   and the poles of the closed-loop transfer
1  G ( s ) H ( s ) [ DG DH  N G N H ]
(DG DH ) function, T(s);
N G DH (3) The concept of mapping points;
T ( s) 
[ DG DH  N G N H ] (4) and the concept of mapping contours.

• (1) the poles of 1 + G(s)H(s) are the same as the poles of G(s)H(s), the open-
loop system,
• and (2) the zeros of 1 + G(s)H(s) ̃are the same as the poles of T(s), the
37
closed-loop system.
(3) mapping. If we take a complex number on the s-plane and substitute it into a
function, F(s), another complex number results. This process is called mapping.
• For example, substituting s = 4 + j3 into the function (s2 + 2s + 1̃) yields 16 + j30. We
say that 4 + j3 maps into 16 + j30 through the function (s2 + 2s + 1̃).

(4) Consider the collection of points, called a contour, shown in the figure as
contour A. Also, assume that:
( s  z1 )( s  z2 )...
F ( s)  Contour A can be mapped through F(s) into contour B by
( s  p1 )( s  p2 )... substituting each point of contour A into the function
F(s) and plotting the resulting complex numbers.
• For example, point Q maps into point Q’ through the
function F(s).

38
Mapping contour A through function F(s) to contour B
 Assume a clockwise direction for mapping the points on contour A,
• then contour B maps in a clockwise direction if F(s) in the figure has just zeros
or has just poles that are not encircled by the contour. Fig. (a)
Contour A

y=[1:-0.1:-1]'; 2
x=[-1:0.1:1]';
sAB = 1 + y*i; 1.5
Contour B
sBC = y - i; 1
sCD = -1 + x*i;
sDA = x + i; 0.5

FsAB = sAB - 2; 0
FsBC = sBC - 2;
FsCD = sCD - 2; -0.5

FsDA = sDA - 2; -1
plot(real(FsAB),imag(FsAB),'g');
hold on -1.5

plot(real(FsBC),imag(FsBC),'b--o'); -2
plot(real(FsCD),imag(FsCD),'c*'); -4 -3 -2 -1 0 1 2

plot(real(FsDA),imag(FsDA),'r');

Examples of contour mapping 39

 Assume a clockwise direction for mapping the points on contour A,
• then contour B maps in a clockwise direction if F(s) in the figure has just zeros
or has just poles that are not encircled by the contour. Fig. (b)

y=[1:-0.1:-1]'; 1

x=[-1:0.1:1]'; 0.8
sAB = 1 + y*i;
0.6
sBC = y - i;
sCD = -1 + x*i; 0.4

sDA = x + i; 0.2
FsAB = 1./(sAB - 2); 0
FsBC = 1./(sBC - 2);
-0.2
FsCD = 1./(sCD - 2);
FsDA = 1./(sDA - 2); -0.4

plot(real(FsAB),imag(FsAB),'g'); -0.6
hold on
-0.8
plot(real(FsBC),imag(FsBC),'b--o');
plot(real(FsCD),imag(FsCD),'c*'); -1
-1.5 -1 -0.5 0 0.5
plot(real(FsDA),imag(FsDA),'r');
Examples of contour mapping 40
 Assume a clockwise direction for mapping the points on contour A,
• If the pole or zero of F(s) is enclosed by contour A, the mapping encircles the
origin. Fig. (c)

y=[1:-0.1:-1]'; 2

x=[-1:0.1:1]';
1.5
sAB = 1 + y*i;
sBC = y - i; 1
sCD = -1 + x*i;
sDA = x + i; 0.5

FsAB = sAB - 0.5;

0
FsBC = sBC - 0.5;
FsCD = sCD - 0.5; -0.5
FsDA = sDA - 0.5;
plot(real(FsAB),imag(FsAB),'g'); -1
hold on
plot(real(FsBC),imag(FsBC),'b--o'); -1.5

plot(real(FsCD),imag(FsCD),'c*');
-2
plot(real(FsDA),imag(FsDA),'r'); -2 -1.5 -1 -0.5 0 0.5 1 1.5 2

41
 Assume a clockwise direction for mapping the points on contour A,
• The contour B maps in a counterclockwise direction if F(s) has just poles that
are encircled by the contour. Fig. (d)
• If the pole or zero of F(s) is enclosed by contour A, the mapping encircles the
origin. Fig. (d)

2
y=[1:-0.1:-1]'; x=[-1:0.1:1]';
sAB = 1 + y*i; sBC = y - i; 1.5

sCD = -1 + x*i;
1
sDA = x + i;
FsAB = 1./(sAB - 0.5); 0.5

FsBC = 1./(sBC - 0.5);

0
FsCD = 1./(sCD - 0.5);
FsDA = 1./(sDA - 0.5); -0.5

plot(real(FsAB),imag(FsAB),'g');
-1
hold on
plot(real(FsBC),imag(FsBC),'b--o'); -1.5

plot(real(FsCD),imag(FsCD),'c*');
-2
plot(real(FsDA),imag(FsDA),'r'); -2 -1.5 -1 -0.5 0 0.5 1 1.5
42
2
 Assume a clockwise direction for mapping the points on contour A,
• The pole and zero rotation cancel, and the mapping does not encircle the origin.
Fig. (e)

y=[1:-0.1:-1]'; 0.4
x=[-1:0.1:1]';
sAB = 1 + y*i; 0.3

sBC = y - i; 0.2
sCD = -1 + x*i;
sDA = x + i; 0.1
FsAB = (sAB-0.2)./(sAB - 0.5);
0
FsBC = (sBC-0.2)./(sBC - 0.5);
FsCD = (sCD-0.2)./(sCD - 0.5); -0.1
FsDA = (sDA-0.2)./(sDA - 0.5);
plot(real(FsAB),imag(FsAB),'g'); -0.2
hold on
-0.3
plot(real(FsBC),imag(FsBC),'b--o');
plot(real(FsCD),imag(FsCD),'c*'); -0.4
-0.5 0 0.5 1 1.5
plot(real(FsDA),imag(FsDA),'r');
43
 1
F ( s) 
( s  0.2)( s  0.3)

y=[1:-0.1:-1]'; 1.5
x=[-1:0.1:1]';
sAB = 1 + y*i;
1
sBC = y - i;
sCD = -1 + x*i;
sDA = x + i; 0.5
FsAB = 1./((sAB - 0.2).*(sAB - 0.3));
FsBC = 1./((sBC - 0.2).*(sBC - 0.3));
0
FsCD = 1./((sCD - 0.2).*(sCD - 0.3));
FsDA = 1./((sDA - 0.2).*(sDA - 0.3));
plot(real(FsAB),imag(FsAB),'g'); -0.5
hold on
plot(real(FsBC),imag(FsBC),'b--o');
-1
plot(real(FsCD),imag(FsCD),'c*');
plot(real(FsDA),imag(FsDA),'r');
-1.5
-1 -0.5 0 0.5 1 1.5 2

44
• A unique relationship exists between the number of poles of
F(s) contained inside contour A, the number of zeros of F(s)
contained inside contour A, and the number of
counterclockwise encirclements of the origin for the
mapping of contour B.
• This interrelationship can be used to determine the stability
of closed-loop systems. This method of determining stability
is called the Nyquist criterion

45
 Vector representation of mapping
( s  z1 )( s  z2 )...
F ( s) 
( s  p1 )( s  p2 )...

Assuming that
F(s)=1+G(s)H(s) has two
zeros and three poles

• Each vector drawn from the poles and zeros of 1+G(s)H(s) that exist outside
contour A will appear to oscillate and return to its previous position, undergoing a
net angular change of 0.
• Each zero inside contour A yields a rotation in the clockwise direction
• Each pole inside contour A yields a rotation in the counterclockwise
direction
N  PZ
Where:
• N equals the number of counterclockwise rotations of contour B about the origin;
• P equals the number of poles of 1+G(s)H(s) inside contour A.
• Z equals the number of zeros of 1+G(s)H(s) inside contour A.
46
• (1) the poles of 1 + G(s)H(s) are the same as the poles of G(s)H(s), the open-
loop system,
• and (2) the zeros of 1 + G(s)H(s) ̃are the same as the poles of T(s), the
closed-loop system.

N  PZ
Thus, P equals the number of enclosed open-loop poles, and Z equals the

Z  PN
number of enclosed closed loop poles.

tells us that the number of closed-loop poles inside the contour

equals the number of open-loop poles of G(s)H(s) inside the contour minus the
number of counterclockwise rotations of the mapping about the origin.

If we extend the contour to include the entire right half-plane

we can count the number of right–half-plane, closed-loop
poles inside contour A and determine a system’s stability

The only problem remaining is how to obtain

the mapping and find N. 47
what if we map through G(s)H(s) instead of 1+G(s)H(s)? The resulting contour is the
same as a mapping through 1+G(s)H(s), except that it is translated one unit to the left;
thus, we count rotations about -1 instead of rotations about the origin.

If a contour, A, that encircles the entire right half-plane is mapped through

G(s)H(s), then the number of closed-loop poles, Z, in the right
half-plane equals the number of open-loop poles, P, that are in
the right half-plane minus the number of counterclockwise
revolutions, N, around -1 of the mapping; that is, Z = P - N. The
mapping is called the Nyquist diagram, or Nyquist plot, of G(s)H(s).

48
 Applying the Nyquist Criterion to Determine Stability

Z = P – N = 0. Since
Z is the number of closed-loop
poles inside contour A, which
encircles the right half-plane,
this system has no right–half-
plane poles and is stable.

P = 0; N = 2, and the system is

unstable. Z = P - N = 2.

The number of encirclements can be determined by drawing a test radius from -1 in

any convenient direction and counting the number of times the Nyquist diagram
crosses the test radius. Counterclockwise crossings are positive, and clockwise crossings
are negative.
49
 Sketching a Nyquist Diagram
Sketch the Nyquist diagram for the system

500
G( s) 
( s  1)( s  3)( s  10)

50
 500 500
G ( j )   3 
( s  1)( s  3)( s  10) s  j ( s  14 s 2  43s  30) s  j
500 500
 
( j 3  14 2  43 j  30) (14 2  30)  j (43   3 )
(14 2  30)  j (43   3 )
G ( j )  500
(14 2  30) 2  (43   3 ) 2
(30) 50
ACA’C’ when   0 G ( j )  500
(30) 2

3
As  increases the real part remains positive,
and the imaginary part remains negative.

At  = sqrt(30/14) the real part becomes negative

( 14 2  30)  j ( 43   3 )

G( s)  30  500 
14 ( 14 2  30) 2  ( 43   3 ) 2 
30
14

500 j
  8.36 j
30 30 30
( 43  )
14 14 14

51
 At   43
( 14 2  30)  j ( 43   3 )
G( s)   500 
43
( 14 2  30) 2  ( 43   3 ) 2   43

500
 2
 0.8741; (Q ' )
( 14 43  30)
Continuing toward , the real part is
negative, and the imaginary part is positive. At
infinite frequency C
500 j
G ( j )    3  090o (C ' )
ACA’C’ 

CDC’D’
Nyquist Diagram
500 15
G (s)   0  ( 1   3   10 )
( 1   3   10 ) 10

 1 , 3 , 10  30o ,30o ,90o  3  30o ,3  30o ,3  90o 5

Imaginary Axis
0
The resultant undergoes a counterclockwise rotation of 3x 180o,
starting at point C’ and ending at point D’ -5

The real part of G(s) is an even function, whereas the -10

imaginary part of G(s) is an odd function. The mapping of the -15

-5 0 5 10 15 20
section of the contour from points D to A is drawn as a mirror Real Axis

image about the real axis of the mapping of points A to C. 52

Nyquist Diagram for Open-Loop Function
with Poles on Contour

Detouring around open-loop poles:

a. poles on contour;
b. detour right;
c. detour left

53
 Nyquist Diagram for Open-Loop Function with Poles on Contour
Sketch the Nyquist diagram of the unity feedback system, where G(s)=(s + 2̃)/s2

sys = (s + 2)/(s^2);
nyquist(sys)

nyquist1.m takes poles on the imaginary axis into account when

creating the Nyquist plot, and plots around them.
The function lnyquist.m plots
(log2(1+abs(G(jw))),angle(G(jw))

nyquist1(sys) lnyquist(sys)

54
 Stability via the Nyquist Diagram
For what range of gain is the system stable?
The general approach is to set the loop gain equal to unity and draw the Nyquist
diagram.
For this system, since P = 2, the critical
point must be encircled by the
Nyquist diagram to yield N = 2 and a
stable system
G=zpk([-3,-5],[2, 4],1)
In summary, then, if the open-loop system nyquist(G)
contains a variable gain, K, set K = 1 and sketch
the Nyquist diagram. Consider the critical point
to be at -1/K rather than at -1. Adjust the value
of K to yield stability, based upon the Nyquist
criterion.

The system is marginally stable if

the Nyquist diagram intersects the
real axis at 1. 55
 Example:
Apply the Nyquist criterion to determine the stability of the following unit-feedback
systems with s3
(i ) G ( s ) 
( s  2)( s 2  2 s  25)
Both systems (i) and (ii) are stable s  20
(ii ) G ( s ) 
since there are no close-loop poles ( s  2)( s  7)( s  50)
in the right half plane. 500( s  2)
(i ) G ( s ) 
( s  2)( s  7)( s  50)

numg=500* [1 -2];
deng=conv([1 2],[1 7]); Nyquist Diagram
1.5
deng=conv(deng,[1 50]); 2 dB 0 dB -2 dB

G=tf(numg,deng); 1
4 dB -4 dB

nyquist(G); 6 dB -6 dB

grid on; 0.5

10 dB -10 dB
Imaginary Axis

20 dB -20 dB
0
We have P = 0 (open loop stable
system), but N = −1, so System (iii) is -0.5

unstable with one closed loop pole in

-1
the right half plane
-1.5
-1.5 -1 -0.5 0 0.5 1 1.5
Real Axis 56
 Range of Gain for Stability via The Nyquist Criterion
For the unity feedback system, where G(s) = K/[s(s+3)(s+5)], find the range of gain, K, for
stability, instability, and the value of gain for marginal stability. For marginal stability
also find the frequency of oscillation. Use the Nyquist criterion.
K  8 2  j (15   3 ) • First set K = 1 and sketch the Nyquist
G ( j )  
s( s  3)( s  5) K 1 64 4   2 (15   2 ) 2
s  j diagram for the system
• Next find the point where the Nyquist
s=tf('s') diagram intersects the negative real axis.
G=1/(s*(s+3)*(s+5))
(imaginary part equal to zero)
nyquist(G)
  15
Nyquist Diagram The real part
0.12
 8 2 1
   0.0083
64 4   2 (15   2 ) 2 8  15
0.1
  15
0.08

For stability N must then be equal to zero

Imaginary Axis

0.06

0.04 (P=0). Z = P - N = 0. Thus, K can be increased

0.02 by 1/0.0083 = 120.5. Hence, for stability, K <
0 120.5. For marginal stability K = 120.5. At
-0.02 this gain the Nyquist diagram intersects 1,
-0.04 and the frequency of oscillation is
-0.06 -0.05 -0.04 -0.03 -0.02 -0.01 0 Sqrt(15) rad/s.
Real Axis 57
 Stability via Mapping Only the Positive j-Axis
Consider the system shown in
figure, which is stable at low
values of gain and unstable at high
values of gain.
G=(0.5*(s-2)*(s-3))/((s+4)*(s+5))
nyquist(G)

This system is stable for the range of loop gain, K, that ensures that the open-loop
magnitude is less than unity at that frequency where the phase angle is 180o (or,
equivalently, -180o).
The system is stable if the open-loop
magnitude is greater than unity at
that frequency where the phase angle
is 180o (or, equivalently, -180o).
First determine stability from the Nyquist criterion
and the Nyquist diagram. Next interpret the Nyquist
criterion and determine whether the mapping of
just the positive imaginary axis should have a gain of
less than or greater than unity at 180. If the Nyquist
diagram crosses 180 at multiple frequencies,
determine the interpretation from the Nyquist
criterion. 58
 Stability Design via Mapping Positive j-Axis
Find the range of gain for stability and instability, and the gain for marginal stability, for
the unity feedback system, where G(s)= K/[(s2 + 2s +2̃)(s + 2̃)]. For marginal stability find
the radian frequency of oscillation. Use the Nyquist criterion and the mapping of only
the positive imaginary axis.

P=0, we want no encirclements of -1 for stability. Hence, a gain less than unity at
180 o is required. Begin by letting K = 1 and draw the portion of the contour along
the positive imaginary axis as shown in Figure (a).
4(1   2 )  j (6   2 )
G ( j ) 
16(1   2 ) 2   2 (6   2 ) 2
(6   2 )    6
4(1   2 ) 4
   0.05
16(1   2 ) 2  6
16  5
K  0.05  1; K  20

The system is stable for K < 20, unstable

for K > 20, and marginally stable for K = 20.
When the system is marginally stable, the
radian frequency of oscillation is sqrt(6)59
Gain Margin and Phase Margin via the Nyquist Diagram
Define two quantitative measures of how stable a system is. These quantities are
called gain margin and phase margin.
• Gain margin, GM. The gain
margin is the change in
open-loop gain, expressed
in decibels (dB), required at
180o of phase shift to
make the closed-loop system
unstable.
• Phase margin, FM. The
phase margin is the change
in open-loop phase shift if the gain of the system were multiplied by a units, the Nyquist
diagram would intersect the critical point. We then say that the gain
required at unity gain to margin is a units, or, expressed in dB, GM = 20 log a.
make the closed-loop system At point Q’, where the gain is unity, a represents the system’s proximity
unstable. to instability. That is, at unity gain, if a phase shift of  degrees
occurs, the system becomes unstable.
60
Stability, Gain Margin, and Phase Margin via Bode Plots
The calculations of gain and phase margins are more convenient if Bode
plots are used rather than a Nyquist diagram
Use Bode plots to determine the range of K within which the unity
feedback system is stable. Let G(s)= K/[(s+2)(s+4)(s+5).
The Bode magnitude plot starts at K/40. For convenience, let K = 40 so that the log
magnitude plot starts at 0 dB.
At a frequency of 7 rad/s, when
the phase plot is 180o, the
magnitude plot is -20 dB.
Therefore, an increase in gain of
+20 dB is possible before the
system becomes unstable. Since
the gain plot was scaled for a
gain of 40, +20 dB (a gain of 10)
represents the required increase
in gain above 40.
Hence, the gain for instability is
40 x 10 = 400. The final result is
0 < K < 400 for stability.
61
 Evaluating Gain and Phase Margins

• By using the phase plot to find the

frequency, GM, where the phase
angle is 180o. At this frequency, we
look at the magnitude plot to
determine the gain margin, GM, which
is the gain required to raise the
magnitude curve to 0 dB.

• The phase margin is found by using the magnitude curve to find the frequency,
M, where the gain is 0 dB. On the phase curve at that frequency, the phase
margin, M, is the difference between the phase value and 180o.

62
 Gain and Phase Margins from Bode Plots
If K = 200 in the system of previous example, find the gain margin and the phase margin.

The Bode plot in the figure is scaled

to a gain of 40. If K=200), the
magnitude plot would be 20 log 5 =
13.98 dB higher.

The phase angle is 180o at

approximately 7 rad/s. On the
magnitude plot, the gain is
-20 + 13.98 = -6.02 dB.
Thus, the gain margin is 6.02 dB.

The magnitude plot is 13.98 dB lower than the actual plot, the 0 dB crossing (-13.98
dB for the normalized plot ) occurs at 5.5 rad/s. At this frequency the phase angle is
165o. Thus, the phase margin is -165o – (-180o) = 15o.

63
Previous example:
• The gain margin is 6.02 dB MATLAB
• The phase margin is 15o G(s)= 200/[(s+2)(s+4)(s+5).

G=zpk([],[-2,-4,-5],200) 20
Bode Diagram

bode(G)
0
grid on

Magnitude (dB)
System: G
-20 Gain Margin (dB): 5.53
At frequency (rad/s): 6.16
1. Right-click in the graph area. -40 Closed loop stable? Yes

2. Select Characteristics. -60

3. Select All Stability Margins.
-80
4. Let the mouse rest on the margin points to 0
read the gain and phase margins.

Phase (deg)
-90

-180 System: G
Phase Margin (deg): 23.5
Delay Margin (sec): 0.091
-270 At frequency (rad/s): 4.51
Closed loop stable? Yes
>> [Gm,Pm,Wgm,Wpm] = margin(G) 10
-2
10
-1
10
0
10
1
10
2

Frequency (rad/s)
Gm = 1.8900;
%20*log10(1.89)=5.5292
Pm = 23.4905
Wgm = 6.1644
Wpm = 4.5068
64
 Relation Between Closed-Loop Transient and
Closed-Loop Frequency Responses
Damping Ratio and Closed-Loop
Frequency Response

n2 n2 n2

T ( j )  2  ; M  T ( j ) 
s  2n s  n2 s  j
(n2   2 )  2nj (n2   2 ) 2  4 2n2 2
n4
dM 2  2(n2   2 )  4 2n2
M  2
2
;  n
4
;
(n   2 ) 2  4 2n2 2 d 2 [(n2   2 ) 2  4 2n2 2 ]2
dM 2
 0; 2 ( 2
  2
)  4 
2 2
 0   2
  2
(1  2 2
)
d 2
n n n

n4
M 2

 2 n2 (1 2 2 ) (n2  n2 (1  2 2 )) 2  4 2n2 [n2 (1  2 2 )]
n4 1 1
M 2
  
 2 n2 (1 2 2 ) ( 2n2 2 ) 2  4 2n4  8 4n4 4 2  4 4 4 2 (1   2 )
1
Mm  ; at a frequency m  n 1  2 2
2 1   2
65
 1
Mm  ; at a frequency m  n 1  2 2
2 1   2

(a) Plots of M versus ω/ωn for a second-order system.

(b) Corresponding time plots for a step input.

66
T=1/(s^2+0.2*s+1) Representative log-magnitude plot
bode(T)
grid on
Bode Diagram Mm
20

10
Magnitude (dB)

-10

-20

-30

-40
0

-45
Phase (deg)

-90
m
-135

-180
10
-1
10
0
10
1
BW: bandwidth of the closed-loop frequency response.
Frequency (rad/s)

Notice that:
• The resonant frequency, m, is not the natural frequency n.
• There will not be a peak at frequencies above zero if  > 0.707.
This limiting value of  for peaking on the magnitude response
curve should not be confused with overshoot Mp on the step
response, where there is overshoot for 0 <  < 1. 67
 1
Mm  ; at a frequency m  n 1  2 2 p_o_shoot=1:1:80;
2 1   2 Mm=-1./(2.*(log(p_o_shoot./100)./ …
sqrt(3.1416^2+log(p_o_shoot./100).^2)).* …
ln(% M p / 100%) (sqrt(3.1416^2./(3.1416^2+log(p_o_shoot./…
 
 2  ln 2 (% M p / 100%) 100).^2))));
plot( p_o_shoot,Mm);
1
Mm 
ln(% M p / 100%) ln(% M p / 100%)
2( ) 1[ ]2
 2  ln 2 (% M p / 100%)  2  ln 2 (% M p / 100%)

Closed-loop frequency response peak vs. percent

overshoot for a two pole system
8

Mp 5
Mm
4

1
Polar plots of G( jω) for different damping 0 10 20 30 40 50 60 70 80
68
ratios for the system Percent overshoot
Response Speed and Closed-Loop Frequency Response
Bandwidth of the closed-loop frequency response is defined as the frequency, BW,
at which the magnitude response curve is 3 dB down from its value at zero frequency

The -3dB, come from 20 Log (0.707) to determine the bandwidth of signal, when
decrease the voltage from maximum to 0.707Max or decreasing the power from
max to half power.

1 V2
20 log(0.707)  20 log( )  3dB; P  ; assume R  1 
2 R
1 1 1
Pi  Vi ; Vo  0.707Vi  Vi ; Po  [ Vi ]2  Pi
2

2 2 2
Vi  6V; Pi  36W
1 1 1
Vo  0.707Vi  Vi ; Po  [ Vi ]2  Pi  18W
2 2 2

Another relationship between the speed of the time response (as measured by
settling time, peak time, and rise time) and the bandwidth of the closed-loop
frequency response.
69
The bandwidth of a two-pole system BW  n (1  2 2 )  4 4  4 2  2

n2
M  T ( j ) 
(n2   2 ) 2  4 2n2 2
n2 1
M    
(   )  4  BW
2
BW 2 2 2 2 2 2
n BW n

n4 1
 ; BW  2n [1  2 ]BW  n4  0
2
4 2 2

(   )  4  BW
2
2 2 2 2 2
n BW n
2
'  n4 [1  2 2 ]2  n4  n4 (1  4 2  4 4 )  n4  n4 ( 4 4  4 2  2)
 '  n2 ( 4 4  4 2  2)
BW  n2 [1  2 2 ]  n2 ( 4 4  4 2  2)  n2 [(1  2 2 )  ( 4 4  4 2  2) ]
2

BW  n (1  2 2 )  ( 4 4  4 2  2)

70
Relate  to settling time Ts
BW

zeta=0:0.05:1;
wbw_ts=(4./zeta).*sqrt(1- BW  n (1  2 2 )  ( 4 4  4 2  2)
2*zeta.^2+sqrt(4*zeta.^4-4*zeta.^2+2));
plot (zeta,wbw_ts); 4 4
Ts  ; n 
n Ts
4
BW  (1  2 2 )  ( 4 4  4 2  2)
Ts
140

120

100

80
BWTs

60

40

20

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Damping ratio 71
BW  n (1  2 2 )  ( 4 4  4 2  2) Relate  to BW

  settling time Tp
Tp  ; n  ;
d Tp 1   2


BW  (1  2 2 )  ( 4 4  4 2  2)
Tp 1   2

zeta=0:0.05:1;
6.5
wbw_tp=(3.1416./…
sqrt(1-zeta.^2)).*…
6
sqrt(1-2*zeta.^2+…
wBWTp

5.5 sqrt(4*zeta.^4-4*zeta.^2+2));
plot (zeta,wbw_tp);
5

4.5

4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Damping ratio 72
BW  n (1  2 2 )  ( 4 4  4 2  2)
nTr  1.76 3  0.417 2  1.039  1
BW Tr  (1.76 3  0.417 2  1.039  1) * (1  2 2 )  ( 4 4  4 2  2)
2.3

2.2

2.1

wBWTr
1.9

1.8

1.7

1.6

1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Damping ratio

zeta=0:0.05:1;
wbw_tr=(1.76*zeta.^3-0.417*zeta.^2+1.039*zeta+1).*…
sqrt(1-2*zeta.^2+sqrt(4*zeta.^4-4*zeta.^2+2));
plot (zeta, wbw_tr);
73
 Relation Between Closed- and Open-Loop Frequency Responses
The frequency at which the maximum value of C( jω)/R( jω) occurs is referred to as the (peak) resonant
frequency ωm. The maximum value is labeled Mm. compensation to improve system performance is based
upon knowledge of ωm and Mm.

A rapid means of determining the values of Mm and ωm and

the value of gain required to achieve a desired value Mm.
The constant-magnitude loci and constant-phase-angle loci are convenient in
determining the closed-loop frequency response from the polar plot or Nyquist plot.
Constant Magnitude Loci (M circles)
Consider a unity feedback system, to obtain the constant-magnitude loci, let
us first note that G(jω) is a complex quantity and can be written as follows:
G ( j )  P ( )  jQ ( )
G( s) G ( j ) P ( )  jQ ( ) P 2 ( )  Q 2 ( )
T ( s)    ;M  2
2
;
1  G ( s ) 1  G ( j ) ( P ( )  1)  jQ ( ) ( P ( )  2 P ( )  1)  Q 2 ( )
M 2 [( P 2 ( )  2 P ( )  1)  Q 2 ( )]  P 2 ( )  Q 2 ( ); P 2 ( )[ M 2  1]  2 M 2 P ( )  M 2  Q 2 ( )[ M 2  1]  0;

If M=1 then, 2 P( )  1  0; P( )  0.5

This is the equation of straight line parallel to y-axis and passing
through (-0.5,0) point. 74
 If M≠1 then,
M2
P ( )[ M  1]  2 M P ( )  M  Q ( )[ M  1]  0; devide by [ M  1] and add
2 2 2 2 2 2 2
to both sides
( M 2  1) 2
2M 2 M2 M2 M2
P ( )  2
2
P ( )  Q ( )  [ 2
2
 ] ;
M 1 M  1 ( M  1)
2 2
( M  1)
2 2

M2 M2 M4 M2 2 M2
[ 2  ] ; [ P ( )  2 ]  Q ( ) 
2
;
M  1 ( M 2  1) 2 ( M 2  1) 2 M 1 ( M 2  1) 2

This is the equation of a circle with  M2  M

( xo , yo )    2 ,0 ; ro  2
 M 1  M 1
• The constant M loci on the G(s) plane are thus a family of circles.
• The centre and radius of the circle for a given value of M can be easily calculated.
• For example, for M=1.3, the centre is at (–2.45, 0) and the radius is 1.88.

1. For M → ∞, which represents a condition of oscillation (ζ → 0), the center of the M

circle x0 → −1 + j0 and the radius r0 → 0. As the G( jω) plot comes closer to the −1 + j0
point, the system’s effective ζ becomes smaller and the degree of its stability
decreases.
2. For M(ω) = 1, which represents the condition where C( jω) = R( jω), r0 → ∞ and the
M contour becomes a straight line perpendicular to the real axis at x = −1/2.
3. For M → 0, the center of the M circle x0 → 0 and the radius r0 → 0.
4. For M > 1 the centers of the circles lie to the left of x = −1 + j0, and for M < 1 the
centers of the circles lie to the right of x = 0. All centers are on the real axis. 75
76
 3
function h = my_circle(xc,yc,r)
% xc x-center of circle
2 % yc y-center of circle
xc=-0.5; % r radius
yc=-1:0.001:1;
plot(xc,yc); 1
th = 0:pi/50:2*pi;
for M = [0.4 0.6 0.8… x = r * cos(th) + xc;
1.2 1.3 1.4 1.6 2 3 5] y = r * sin(th) + yc;
0
xc=-M.^2/(M.^2-1); hold on
yc=0; h = plot(x, y);
r=M./(M.^2-1); -1 hold off;
my_circle(xc,yc,r);
end
-2

-3
-6 -4 -2 0 2 4 6

• Constant M circles are the locus of

the closed-loop
magnitude frequency response for unity feedback systems.
If the polar frequency response of an open-loop function (Nyquist
diagram), G(s), is plotted and superimposed on top of the constant M
circles, the closed-loop magnitude frequency response is
determined by each intersection of this polar plot with the
constant M circles. 77
A plot of the transfer function G( jω) and the Ma circle.
G ( j ) A( j )
M ( )  
1  G ( j ) B ( j )
The ratio of magnitudes of the phasors A( jω)
and B( jω) drawn to any point on the Ma circle
has the same value.
A( j1 ) A( j2 )
Ma  
B ( j1 ) B ( j2 )
A given point (x1, y1) is simultaneously a point on
a particular transfer function G( jω)
and a point on the M circle passing through it.

The circle M = Mb is just tangent to

the G( jω) plot. Maximum value of M
equal to Mm = Mb.

78
 Constant Phase Loci (N circles)
C ( j ) P ( )  jQ ( )
T ( j )  
R ( j ) 1  P ( )  jQ ( )
• The phase angle of closed loop transfer function is
j P ( )  jQ ( )
e 
1  P ( )  jQ ( )
• The phase angle α is

   
Q ( ) Q ( )
  tan ( 1 1
)  tan ( )
P ( ) 1  P ( )
79
 Q ( ) Q ( )
  tan (1 1
)  tan ( )
P ( ) 1  P ( )

• If we define
tan  N
• then
 1 Q ( ) 1 Q ( ) 
N  tan  tan ( )  tan ( )
 P ( ) 1  P ( ) 

• We obtain Q Q

N  P 1 P
Q Q
1
P 1 P 80
 Q Q

N  P 1 P  2
Q Q
; N ( P 2
 P  Q 2
)  Q ; P 2
 P  Q 2
 0
1
Q Q P  P Q 2
N
P 1 P
1 1
Adding  2 to both sides
4 4N

1 1 1 1 1
P  P  Q  Q 
2 2
2
 
4 N 4N 4 4N 2
2 2
 1  1  1 1
 P    Q    
   
2
2 2 N 4 4 N
This is an equation of circle with
 1 1 
( xq , y q )    , 
 2 2N 
1 1
rq  
4 4N 2 81
 tan  N The constant α contour is only an arc of the circle.
As an example, the αq+ = −310° and αq- =−130° arcs are part
 1 1 
( xq , y q )    ,  of the same circle. (tan αq+= tan αq-=N)
 2 2 N 
1 1
rq  
4 4N 2 For all points on the αq arc, yields the same phase angle αq, that is,
A( j )  B( j )   q

82
Arcs of constant α. (a) Arcs for α = –310° and –130°. (b) Constant α contours.
 3

-1

-2

-3
-2 -1.5 -1 -0.5 0 0.5 1

for alpha = [-20 -30 -40 -60 -80 -120 20 30 40 60 80 120]

N=tand(alpha);
xc=-1/2;
yc=1/2/N;
r=sqrt(1/4+1/4/N^2);
my_circle(xc,yc,r);
end

Superimposing a unity feedback, open-loop frequency response over the

constant N circles yields the closed-loop phase response of the system. 83
 Closed-Loop Frequency Response from Open-Loop
Frequency Response
Find the closed-loop frequency response of the unity feedback system, where
G(s)= 50/[s(s+3)(s+6)], using the constant M circles, N circles, and the open-
loop polar frequency response curve. The Polar plot of the open-loop freq response (
Nyquist diagram) is shown superimposed over the
50 50 50 M and N circles in the Figure
G( s)   2  3
s( s  3)( s  6) ( s  3s )( s  6) s  9 s 2  18s
50 50
G ( s ) s j  
 j  9  18 j  9  j (18   3 )
3 2 2

-1

-2

-3

-4
-5 -4 -3 -2 -1 0 1 2 3 84
The closed-loop magnitude
frequency response can now be
obtained by finding the intersection
of each point of the Nyquist plot
with the M circles.
While the closed-loop phase
response can be obtained by finding
the intersection of each point of the
Nyquist plot with the N circles.
The result is shown in the Figure.

Closed-Loop Frequency Response

2

1.5

numg=50; 0.5
Imaginary Axis

deng=poly([0 -3 -6]); 0
System: T
Real: 1.04
Imag: -0.44
G=tf(numg,deng); Frequency (rad/s): 1.01

-0.5
T=feedback(G,1); System: T
System: T
Real: 0.711
-1
nyquist(T); Real: -0.345
Imag: -1.63
Imag: -1.43
Frequency (rad/s): 2
Frequency (rad/s): 2.5
-1.5

-2
-1 -0.5 0 0.5 1 1.5 85
Real Axis
 Tangents to the M Circles

 M2  M
( xo , yo )    2 ,0 ; ro  2
 M 1  M 1

ob  xo ; bc  ro ;
oa 2  oc 2  ac 2  oc 2  (oc sin ) 2  oc 2 (1  sin 2  )
M
bc M 2  1 1
sin   
ob M2 M
M 2 1
M4 M2 M2
oc  ob  bc 
2 2 2
  2
( M  1) ( M  1)
2 2 2 2
M 1
M2 1 M 2 M 2 1
oa  oc (1  sin  )  2 [1  2 ]  2 [
2 2 2
] 1
M 1 M M 1 M 2

86
CONSTANT 1/M AND α CONTOURS (UNITY FEEDBACK) IN THE INVERSE POLAR PLANE

The closed-loop unity-feedback system can be C ( j ) G ( j )

  Me j
represented by either of the following equations: R ( j ) 1  G ( j )
R ( j ) 1  G ( j ) 1 1  j
  1  e
R( jω)/C( jω) can be obtained graphically by C ( j ) G ( j ) G ( j ) M
plotting the complex quantities 1/G( jω) and 1.
In the inverse plane it can be seen that

1. Contours of constant values of M

are circles whose centers are at the
−1 + j0 point and the radii are equal
to 1/M.
2. Contours of constant values of −α
are radial lines that pass through the
−1 + j0 point.

87
Contour for a particular magnitude Typical M and α contours on the inverse
R( jω)/C( jω) = 1/Mq. polar plane for unity feedback

1
ab M 1
sin   
ob 1 M

88
GAIN ADJUSTMENT OF A UNITY-FEEDBACK SYSTEM FOR A DESIRED Mm: DIRECT POLAR PLOT

Gx ( j )  x  jy  K xGx' ( j )  K x ( x '  jy ' )

x y (a) Plot of G’x(jω) with the respective Mm circle.
x '  jy '  j
Kx Kx (b) Circle drawn tangent both to the plot of G’x (jω) and
to the line representing the angle ψ = sin−1(1/Mm).

89
It is possible to determine the required gain to achieve a desired Mm for a given system by
the following graphical procedure:
Step 1. If the original system has a transfer function
K x (1  jT1 )(1  jT2 )...
Gx ( j )   K x x ( j )
G '

( j ) (1  jTa )(1  jTb )(1  jTc )...

m

with an original gain Kx, only the frequency-sensitive portion G’x ( j) is plotted.
Step 2. Draw the straight line at the angle ψ = sin−1 (1/Mm), measured from the negative
real axis.

Step 3. By trial and error, find a circle whose center lies on the negative real axis and is
simultaneously tangent to both the G’x ( j) plot and the line drawn at the angle ψ.

Step 4. Having found this circle, locate the point of tangency on the ψ-angle line. Draw a
line from the point of the tangency perpendicular to the real axis. Label the point where
this line intersects the real axis as a′.

Step 5. For this circle to be an M circle representing Mm, the point a′ must be the −1 + j0
point. Thus, the x′, y′ coordinates must be multiplied by a gain factor Km in order to
convert this plot into a plot of G( jω). From the graphical construction the value Km is
1/oa′.

Step 6. The original gain must be changed by a factor A = Km/Kx. 90

 Example
It is desired that the closed-loop system that has the open-loop transfer function
has an Mm = 1.3. The problem is to determine 1.47
the actual gain K1 needed and the amount by Gx ( j ) 
j (1  0.25 )(1  0.1 j )
which the original gain Kx must be changed to
obtain this Mm.

1 1
K1    2.94
oa ' 0.34
The additional gain required is

K1 2.94
A   2 .0
K x 1.47

The original gain of G( jω)

must be doubled to obtain
Mm of 1.3 for C( jω)/R( jω).

91
w=0:0.1:100;
G=(1)./(j*w.*(1+0.25*j*w).*(1+0.1*j*w));
plot(real(G),imag(G));
for r = [0.2 0.4 0.6 0.8 1]
my_circle(0,0,r);
end
hold on;
psi=asind(1/1.3);
x=-1:0.1:0;
plot(x,x*tand(psi));
M=1.3; 1
xc=-M.^2/(M.^2-1); 0.8
yc=0;
r=M./(M.^2-1); 0.6
my_circle(xc,yc,r); 0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

92
 w=0:0.1:100;
G=(1)./(j*w.*(1+0.25*j*w).*(1+0.1*j*w));
plot(real(G),imag(G));
for r = [0.2 0.4 0.6 0.8 1]
my_circle(0,0,r);
end
hold on; psi=asind(1/1.3);
x=-1:0.1:0; plot(x,x*tand(psi));
M=1.3; xc=(-M.^2/(M.^2-1))/(1.47*2);
yc=0; r=M./(M.^2-1)/(1.47*2);
my_circle(xc,yc,r);

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
93
w=0:0.1:100;
G=(1.47*2)./(j*w.*(1+0.25*j*w).*(1+0.1*j*w));
plot(real(G),imag(G));
for r = [0.2 0.4 0.6 0.8 1]
my_circle(0,0,r);
end
hold on;
psi=asind(1/1.3);
x=-1:0.1:0; 1
plot(x,x*tand(psi));
M=1.3; 0.8
xc=-M.^2/(M.^2-1);
yc=0;
r=M./(M.^2-1); 0.6
my_circle(xc,yc,r);
0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

94
CONSTANT M AND α CURVES ON THE LOG MAGNITUDE–ANGLE DIAGRAM (NICHOLS CHART)

A constant M circle is shown in the Figure on the inverse polar plot. The magnitude ρ and
angle λ drawn to any point on this circle are shown. The equation for this M circle is
1
y 2  (1  x ) 2 
M2
x   cos  ; y   sin 
M 2  2  2 M 2 cos   M 2  1  0
M 2
1
   cos   cos2  
M2
1 1  M  M 
2 2 2
  cos ( )
2 M 2

These equations are derived from the inverse polar plot 1/G( jω), a change in the
equations must be made by substituting
1
r ;   


95
Consider a unity feedback system with open- 10
loop transfer function G( s) 
s( s  5)
Nichols Chart
20 20

10 10

0 0
Open-Loop Gain (dB)

-10 -10

-20 -20

-30 -30

-40 -40

-50 -50
-180 -150 -120 -90 -180 -170 -160 -150 -140 -130 -120 -110 -100 -90
Open-Loop Phase (deg)

s = tf('s'); w=0.5:0.1:50;
G = 10/(s*(s+5)); G=10./((j*w).*(j*w+5));
w = 0.5:0.1:50; lm_mag=20*log10(abs(G));
nichols(G,w); phase_deg=angle(G)*180/pi;
figure
plot(phase_deg,lm_mag);

96
 ngrid;
Nichols Chart
40 40
0 dB
30 0.25 dB 30
0.5 dB
20 1 dB 20
-1 dB
10 3 dB 10
Open-Loop Gain (dB)

6 dB -3 dB
0 -6 dB 0

-10 -12 dB -10

-20 -20 dB
-20
-30
-30
-40 -40 dB
-40
-50
-50
-60 dB
-60
-360 -315 -270 -225 -180 -135 -90 -45 0 -60
-350 -300 -250 -200 -150 -100 -50 0
Open-Loop Phase (deg)
lamda=0:1:360;
w=0.5:0.1:50; G=10./((j*w).*(j*w+5)); for M=[0.891 0.708 0.501 0.251 0.100 0.010 0.001]
lm_mag=20*log10(abs(G)); phase_deg=angle(G)*180/pi; rho_t=-cosd(lamda)+sqrt((cosd(lamda)).^2-((M^2-1)/M^2));
Figure; plot(phase_deg,lm_mag); rho=abs(rho_t); rho=20*log10(rho); hold on; plot(-lamda,-rho);
lamda=90:1:270; %M=2 lmM=6dB end
for M=[2 1.41 1.122 1.059 1.029 1] for alpha=[0 -20 -40 -60 -80 -100 -120 -140 -160] ;
rho_t=-cosd(lamda)+sqrt((cosd(lamda)).^2-((M^2-1)/M^2)); lamda=180:-1:-alpha;
rho=abs(rho_t); rho=20*log10(rho); rho=-tand(alpha)./(sind(lamda)+cosd(lamda)*tand(alpha));
hold on; plot(-lamda,-rho); rho=20*log10(abs(rho)); hold on; plot(-lamda,-rho);
rho_t=-cosd(lamda)-sqrt((cosd(lamda)).^2-((M^2-1)/M^2)); end
rho=abs(rho_t); rho=20*log10(rho); for alpha=[-180 -200 -220 -240 -260 -280 -300 -320 -340] ;
plot(-lamda,-rho); lamda=180:1:-alpha;
end rho=-tand(alpha)./(sind(lamda)+cosd(lamda)*tand(alpha));
rho=20*log10(abs(rho)); hold on; plot(-lamda,-rho);
end 97
Nichols chart with frequency response for G= K/s(s=1)(s+2)
superimposed. Values for K = 1 and K = 3:16 are shown

If the gain is increased by 10 dB, simply raise the curve for K = 1 by 10 dB and obtain
the curve for K = 3.16(10 dB).

98
G=1/(s*(s+1)*(s+2));
ltiview

99
G=3.16/(s*(s+1)*(s+2));
ltiview

100