You are on page 1of 12

Nyquist stability criterion

Valeri Ougrinovski
August 9, 2006
Abstract
This note gives a brief introduction to the closed loop system stability analysis based on
the frequency response of the system open loop gain.
1 Nyquist polar plots
A polar plot of the frequency response produces a curve with frequency as a parameter relating
the points on the polar plot to the appropriate points on the Bode plots, see Fig. 1.
The Nyquist plot contains the same information as the Bode plot. Thus, one can use ap-
proximate Bode plots to reconstruct the shape of the Nyquist plot. For instance, the Bode and
Nyquist plots of
G(s) =
5(s + 1)
s
2
+ 2s + 10)
are shown below in Figure 1. Arrows on the Nyquist plot indicate the direction in which the
plot changes as the frequency increases from to + (or equivalently, j sweeps from the
j to +j along the imaginary axis).
Note the two symmetric branches of the Nyquist plot corresponding to positive and negative
frequencies. Unlike Bode plots, both positive and negative frequencies are used for plotting
Nyquist curves; the reasons for this will become clear later.
The Matlab command to produce accurate Nyquist plots is as follows:
>> nyquist(G,w);
The second argument is optional; it allows you to specify an array of frequencies of interest to
capture some details of the plot. For example, to produce the Nyquist plot in Fig 1, you need
to type
>> G=tf(5*[1 1],[1 2 10]);
>> nyquist(G);
The command nyquist can also be used for computing the frequency response of a system loop
gain for positive frequencies only. No plot is produced, hence the command plot must be used:
>> G=tf(5*[1 1],[1 2 10]);
>> [reG,imG]=nyquist(G,{0.01,1000});
>> plot(squeeze(reG),squeeze(imG))
1
Bode Diagram
Frequency (rad/sec)
P
h
a
s
e

(
d
e
g
)
M
a
g
n
i
t
u
d
e

(
d
B
)
10
0
10
1
10
2
90
45
0
45
System: G
Frequency (rad/sec): 2.83
Phase (deg): 1.66e14
30
20
10
0
10
System: G
Frequency (rad/sec): 2.83
Magnitude (dB): 7.96
Nyquist Diagram
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
1 0.5 0 0.5 1 1.5 2 2.5
1.5
1
0.5
0
0.5
1
1.5
System: G
Real: 5e06
Imag: 0.005
Freq (rad/sec): 1e+03
System: G
Real: 0.5
Imag: 0
Freq (rad/sec): 0
System: G
Real: 2.5
Imag: 0.00186
Freq (rad/sec): 2.83
System: G
Real: 5e06
Imag: 0.005
Freq (rad/sec): 1e+03
Figure 1: Bode and Nyquist plots of the loop gain transfer function G(s) =
5(s+1)
s
2
+2s+10
2
The graphical method can be used to obtain an approximate Nyquist plot. Consider an
example. Let
G(s) =
100(s + 10)
(s + 0.1)
2
(s
2
+ 0.1s + 100)
.
This transfer function has the single nonminumum phase zero at s = 10, the double pole at
s = 0.1 (real), and the pair of complex conjugate poles at s 0.05 10j. First let us sketch
the magnitude Bode plot.
PSfrag replacements
dB
60
40
20
0
-20
-40
-60
-80
0.1 1
10
100

Note that zero at 10 and the pair of complex conjugate poles at 0.05 10j yield the
same corner frequency = 10 rad/sec. Hence the net change of slope at this corner frequency
is +20 40 = 20 dB/decade.
Also, note that the poles 0.0510j are very underdamped, hence it is appropriate to draw
a peak at the corresponding corner frequency.
We now proceed to phase calculations. Write the transfer function in a standard form:
G(s) =
100(s 10)
(s + 0.1)
2
(s
2
+ 0.1s + 100)
.
Note the gain factor of 100. The sign means that the +180

or 180

term should be
taken into account, lets use 180

. The phase of G(j) can be expressed as


G(j) = 180

+
1
2
2

4
;
the angles are shown in the Figure 2 below. Note the factor of 2 in front of
2
, this factor is to
reect the contribution of the double pole at s = 0.1.
3
PSfrag replacements
j

4
Figure 2:
, rad/sec
1

2

3

4
G(j) = 180

+
1
2
2

4
0 180

-90

90

0.1 180

45

-90

90

-90

1 180

90

-90

90

-180

9.5 135

90

-90

90

-225

9.95 135

90

-45

90

-270

10 135

90

90

-315

10.05 135

90

45

90

-360

10.5 135

90

90

90

-405

100 90

90

90

90

-450

90

90

90

90

-450

This agrees with the true Bode plot shown in Figure 5. Note the rapid change of 180

around
the resonant frequency (natural undamped frequency) = 10rad/sec.
From the above analysis, we can sketch the Nyquist plot; see Figure 3. The true plot is
shown in Figure 4 (positive frequencies only). It was obtained using nyquist command:
>> num=100*[-1 100];
>> den=conv([1 0.2 0.01],[1 0.1 100]);
>> [re,im]=nyquist(num,den,{0.01,100});
>> plot(squeeze(re),squeeze(im));
>> grid
>>
On Figure 4 (left plot) the general shape of the plot can be seen but the loop near the origin
is invisible. In Figure 4 on the right, the axis limits have been set so that the behavior of the
polar plot near the frequency = 10 can be observed. One method of zooming in the picture
is to apply the commands
>> a=[-10 20 -10 20];
>> axis(a);
In the rst command, a zoom window is dened as [Xmin Xmax Ymin Ymax]. The second
command scales the axes to show the part of the gure within the desired window.
4
Figure 3: Nyquist sketch for the example
200 0 200 400 600 800 1000
700
600
500
400
300
200
100
0
100
10 5 0 5 10 15 20
10
5
0
5
10
15
20
Figure 4: True Nyquist plot: complete and detailed
5
10
2
10
1
10
0
10
1
10
2
80
60
40
20
0
20
40
60
10
2
10
1
10
0
10
1
10
2
500
400
300
200
100
0
, rad/sec
Figure 5: Bode plots
2 The Nyquist stability criterion
The Nyquist criterion is essentially a graphical method and allows to establish the stabilityof
the closed loop system from the Nyquist plot of the open loop gain transfer function G(s).
PSfrag replacements
R(s) Y (s)
G(s) =
M(s)
D(s)
+

Figure 6:
Consider for simplicity the system shown in Figure 6. The closed loop system has the transfer
function
G
cl
(s) =
G(s)
1 +G(s)
=
M(s)
D(s) +M(s)
.
Hence the poles of the closed loop system are the roots of D(s) +M(s):
D(s) +M(s) = 0, or, after dividing by D(s) (1)
1 +G(s) = 0. (2)
Hence, stability of the closed loop system can be determined from the analysis of the right
half-plane zeros of the function 1 +G(s).
6
Consider a contour
s
in the right half-plane of the s-plane; see gure 7. For s
s
, the set
of complex values of the function G(s) forms another contour
G
in the complex G(s) plane.
a)
Re G(s)
b)
Im G(s)
1
PSfrag replacements
Re s
Im s
Re G(s)
Im 1 +G(s)

G
Figure 7: The s plane and G(s) plane.
The Principle of the Argument states:
If a contour
s
in the complex s plane encircles Z zeros and P poles of 1+G(s) and
the traversal is in the clockwise direction along the contour, then the corresponding
contour
G
encircles the point 1+j0 in the complex G(s) plane exactly N

= ZP
times in the clockwise direction.
The Principle of the Argument allows us to establish the number of poles of the closed loop
system by simply looking at the Nyquist plot of G(s). Select a contour
s
to be large enough
to include the entire right half of the complex plane:
PSfrag replacements

j

s
R =
This contour is often called D-contour because of its shape. Note that the D-contour must avoid
imaginary axis poles of G(s), as shown in the gure. For stability, 1 + G(s) should not have
roots in the right half plane, i.e, we need Z = 0. Also, since 1 + G(s) has the same poles as
G(s), then P equals the number of unstable poles of G(s). This leads to the following principle,
called the Nyquist stability criterion:
The Nyquist stability criterion
The closed loop system is stable if and only if the net number of anticlockwise encirclements of
7
the point 1 +j0 by the Nyquist plot of the system loop gain transfer function G(s) is equal to
the number of poles of G(s) in the right half-plane:
N

= P.
The equivalent formulation of the Nyquist stability criterion
The closed loop system is stable if and only if the net number of clockwise encirclements of
the point 1 + j0 by the Nyquist plot of the system loop gain transfer function G(s) plus the
number of poles of G(s) in the right half-plane is zero:
N

+P = 0.
3 Examples
3.1 Example 1
Consider the feedback system in Figure 6, in which G(s) =
K
s+5
. The Nyquist D-contour for this
system is shown in Figure 8.
-5
PSfrag replacements

s
R =

j

Figure 8:
To sketch the Nyquist plot of the loop gain transfer function, use the graphical method:
s , rad/sec Gain |G(j)| Phase G(j) = , deg.
j0 0
K
5
0

5j 5
K
5

2
45

j +j + |G(j)| 0 G(j) 90

e
j
N/A 0 N/A
The Nyquist diagram of G(s) is shown in Figure 9. The Nyquist diagarm never encircles the
critical point 1 + j0 for any K > 0. I.e., the number of anti-clockwise encirclements (none)
= the number of unstable poles of G(s) (none). Thus, the closed loop system is stable for any
K > 0.
Now let G(s) =
K
s5
. We use the same Nyquist D-contour. Use the graphical method to
sketch the Nyquist plot of the loop gain transfer function:
s , rad/sec Gain |G(j)| Phase G(j) = , deg.
j0 0
K
5
180

5j 5
K
5

2
135

j +j + |G(j)| 0 G(j) 90

e
j
N/A 0 N/A
8
1 0.8 0.6 0.4 0.2 0 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
Nyquist Diagram
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
K=5
K/5
1+j0
=
=
(=0)
PSfrag replacements

s
R =

Figure 9:
The Nyquist plot of G(s) for K = 4 and K = 6 is shown in Figure 10. The loop gain transfer
function G(s) has one unstable pole, P = 1.
When K = 4, the Nyquist plot does not encircle the critical point 1 +j0, N

= 0; i.e, the
number of counter-clockwise encirclements = the number of unstable poles of G(s): N

= P.
Thus, the closed loop system is unstable for K = 4.
When K = 6, the Nyquist diagarm makes one circle around the critical point 1 +j0 in the
anti-clockwise direction, N

= 1, i.e., the number of anti-clockwise encirclements = the number


unstable poles of G(s), N

= P. Thus, the closed loop system is stable for K = 6.


Indeed,
G
cl
(s) =
G(s)
1 +G(s)
=
K
s 5 +K
.
Hence if K = 4 the closed loop system has the unstable pole at s = 5 4 = 1. If K = 6 the
closed loop system has the stable pole at s = 5 6 = 1.
1.4 1.2 1 0.8 0.6 0.4 0.2 0
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
Nyquist Diagram
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
1+j0
K/(5)
(=0)
=
=+
K=6
K=4
PSfrag replacements

s
R =
Figure 10:
9
3.2 Example 2
Consider the feedback system in Figure 6, with G(s) =
1
s(s+1)(s+2)
. The Nyquist D-contour for
this system is shown in Figure 11a. Note that the pole at 0 is avoided.
Since this pole is exluded from the right half plane by the chosen Nyquist contour, for
stability we need the Nyquist plot not to encircle 1 +j0. To sketch the Nyquist plot mapping
1 2
pole at 0
1
3
4
2
PSfrag replacements

s
R =

1
2
3
r
= +
=
+0
0
= 0
1
3/4
1/6
PSfrag replacements

s
R =

3
r
= +
=
+0
0
= 0
(a) (b)
Figure 11:
the D contour onto the G plane, consider four regions denoted 1,2, 3 and 4 as labeled in the
gure.
Image of region 1. In this region, s = j, r < < +, where r 0.
s , rad/sec Gain |G(j)| Phase G(j) =
1

3
, deg.
jr +0 r +0 90

1j 1 1/

10 165

2j 2 1/(4

10) 195

j +j + |G(j)| 0 G(j) 270

Since the G(j) changes from 90

to 270

, then the plot may or may not encircle the


point 1 +j. To nd out for sure, we need to calculate the location of the intersection
of the plot and the real axis. The intersection is characterized by the condition
Im
1
j(j + 1)(j + 2)
= 0.
Calculate Im
1
j(j+1)(j+2)
:
1
j(j + 1)(j + 2)
=
j(j + 1)(j + 2)
(
2
+ 1)(
2
+ 4)
=
3 +j(
2
2)
(
2
+ 1)(
2
+ 4)
10
Im
1
j(j + 1)(j + 2)
=

2
2
(
2
+ 1)(
2
+ 4)
= 0
at
2
= 2. Hence there are only two such frequencies =

2 rad/sec (not in region 1),


and =

2 rad/sec. At frequency =

2 rad/sec,
Re
1
j(j + 1)(j + 2)
=
3
(
2
+ 1)(
2
+ 4)
=
3
3 6
=
1
6
.
Thus, the plot crosses the real axis on the right of 1 + j0. Note that at =

2,
G(j) = 180

; i.e, we could determine the crossover point from the Bode plot of G(j)
by looking at the gain at the phase crossover frequency. Note also that as 0, then
Re G(j)
3
4
.
Image of region 2. This region maps into the origin on the G plane.
Image of region 3, s = j. The plot is a mirror image of the plot corresponding to region 1.
Image of region 4, s = re
j
, r 0. Note that goes from 90

to 90

anti-clockwise. Also,
G(s) =
1
re
j
(re
j
+ 1)(re
j
+ 3)

1
re
j
=
1
r
e
j
Thus, G(s) + as r 0, and G(s) changes from 90

to 90

clockwise. Hence the


traversal along the arc maps to a curve of Nyquist plot which has the radius 1/r +
and swings from
3
4
+jto
3
4
jin the clockwise direction. It is worthwhile to check
ourselves by computing G(re
j
) at = 0:
G(r) =
1
r(r + 1)(r + 2)
> 0.
The approximate Nyquist diagram is shown in Fig. 11b. The Nyquist plot does not circle around
1 +j0, i.e, N

= 0 and is equal to the number of poles of G inside the D-contour, hence the
closed loop system is stable.
3.3 Example 3
Consider the feedback system in Figure 6, with G(s) =
K(s+1)
(s1)(s+2)
. Note that the loop gain
transfer function has one pole in the right half of the complex plane at s = 1. Thus, for the
closed-loop system to be stable, the Nyquist diagram of G(s) must make exactly one turn around
1 +j0 in the anti-clockwise direction.
The Nyquist D-contour for this problem is the standard Nyquist D-contour; see Figure 8.
= 0, G(j0) = K/2, G(j) = 180

.
> 0, G(j) = (1 +j) (j 1) (j + 2).
, |G(j)| 0, argG(j) = 90

.
The Nyquist diagram of the frequency response function G(j) is shown in Figure 12 (plot (a) is
the approximate, plot (b) is the exact plot for K = 1). From the plot (a), we conclude that the
closed loop system is stable if and only if K > 2 because in this case the Nyquist plot encircles
the point 1 +j.
The plot in gure (b) illustrates the Nyquist plot of G(s) corresponding to the unstable
system. In the plot (b), the Nyquist plot does not encircle the point 1 +j. Since G(s) has
one unstable pole, then the function 1+G(s) has one zero in the right half of the complex plane.
11
-1
PSfrag replacements

s
R =
=
=
= 0

k
2
Real Axis
I
m
a
g
i
n
a
r
y

A
x
i
s
Nyquist Diagrams
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5

PSfrag replacements

s
R =
=
=
= 0

k
2
(a) (b)
Figure 12:
4 Further reading
[1] Dorf & Bishop. Modern Control Systems, Chapter 9. Sections 9.1-9.3, 9.10, 9.12 are a must.
Section 9.12 includes a table of most common transfer functions and their Nyquist and Bode
plots; you may nd it useful. Sections 9.5, 9.8, 9.11 are optional.
12

You might also like