G.

APPENDIX G: General Nyquist Criterion

G-0. APPENDIX PRELIMINARIES
FORMULATION OF NYQUIST CRITERION

ILLUSTRATIVE EXAMPLES: GENERAL NYQUIST CRITERION MINIMUM AND NONMINIMUM TRANSFER FUNCTIONS

STABILITY ANALYSIS OF MULTILOOP SYSTEMS

REFERENCE

PROBLEMS

G-1. FORMULATION OF NYQUIST CRITERION

The original Nyquist criterion presented in Chap. 8 is cumbersome to apply when the loop transfer function is of the
nonminimum-phase type; that is, L(s) has either poles and/or zeros in the right-half s-plane. We shall show that if the loop
transfer function is of the nonminimum-phase type, then plotting the Nyquist plot of L(s) only for s = jω to s = j0 and not
enclosing the (−1, j0) point in the L(s)-plane is only a necessary, but not sufficient, condition for closed-loop stability. For a
system with a nonminimum-phase loop transfer function L(s), the original Nyquist criterion requires that the L(s) plot that
corresponds to the entire Nyquist path in Fig. 8-20 be made. If the loop transfer function L(s) has poles or zeros on the jω-axis,
then the Nyquist path of Fig. 8-20 must have small indentations around them on the jω-axis. This adds even more complexity to
the construction of the L(s) plot. Our MATLAB toolbox (ACSYS) or other computer software can be used to construct the plots
of only functions that correspond to the positive jω-axis of the s-plane. The rest of the Nyquist plot that corresponds to the
small indentations and the large semicircle on the Nyquist path have to be plotted manually. With modern computer facilities
and software, the analyst should not be burdened with the chores of manual plotting. Therefore, we are introducing a simplified
Nyquist criterion that can be applied by using only the positive jω-axis of the Nyquist path and then observing its behavior with
reference to the (−1, j0) point.

Yeung [1] introduced a general, yet simplified version of the Nyquist criterion that allows the determination of stability of closed-
loop systems of minimum- as well as nonminimum-phase loop transfer functions by using only the positive part of the jω-axis
of the Nyquist path. However, if the system is of the minimum-phase type, the test of whether the (−1, j0) point is enclosed is
still simpler to apply. We shall show that, for nonminimum-phase systems, if the (−1, j0) point is enclosed, the system is still
unstable. However, if the (−1, j0) point is not enclosed, then an additional angle condition is all that must be satisfied by the
Nyquist plot of L(s) for the system to be stable.

Let us consider the two Nyquist paths shown inFig. G-1a and b. Apparently, the Nyquist path Γs1 in Fig. G-1a is the original one
shown in Fig. 8-20, whereas the path Γs2 in Fig. G-1b encircles not only the entire right-half s-plane but also all the poles and
zeros of L(s) on the jω-axis, if there are any. Let us define the following quantities.

Z = number of zeros of 1 + L(s) that are in the right-half s-plane.

P = number of poles of L(s), or of 1 + L(s), that are in the right-half s-plane.

Pω = number of poles of L(s), or of 1 + L(s), that are on the jω-axis, including the origin.

N1 = number of times the (−1, j0) point of the L(s)-plane that is encircled by the Nyquist plot of L(s) corresponding to Γs1 .

N2 = number of times the (−1, j0) point of the L(s)-plane that is encircled by the Nyquist plot of L(s) corresponding to Γs2 .

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Figure G-1 (a) Nyquist path. (b) An alternative Nyquist path. (Source: K. S. Yeung, "A Reformulation
of Nyquist's Criterion," IEEE Trans. Educ., Vol. E-28, pp. 58–60, Feb. 1985.)

Then, with reference to the two Nyquist paths in Fig. G-1, and according to the Nyquist criterion,

N1 = Z − P

(G-1)

and

N2 = Z − P − Pω

(G-2)

Let Φ1 and Φ2 represent the net angles traversed by the Nyquist plot ofL(s) with respect to the (−1, j0) point, corresponding to
Γs1 and Γs2 , respectively. Then,

Φ 1 = N1 × 360° = (Z − P) 360°

(G-3)

Φ 2 = N2 × 360° = (Z − P − Pω) 360°

(G-4)

Let us consider that each of the Nyquist paths Γs1 and Γs2 is composed of three portions:

1. The portion from s = −j∞ to +j∞ along the semicircle with infinite radius

2. The portion along the jω-axis, excluding all the small indentations

3. All the small indentations on the jω-axis

Because the Nyquist paths in Fig. G-1 are symmetrical about the real axis in thes-plane, the angles traversed by the Nyquist
plots are identical for positive and negative values of ω. Thus, Φ1 and Φ2 are written as

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Φ 1 = 2Φ 11 + Φ 12 + Φ 13

(G-5)

Φ 2 = 2Φ 11 + Φ 12 + Φ 13

(G-6)

where Φ11 = angle traversed by the Nyquist plot of L(s) with respect to the (−1, j0) point, corresponding to the positive jω-
axis or the −jω-axis of the s-plane, excluding the small indentations.

Φ12 = angle traversed by the Nyquist plot of L(s) with respect to the (−1, j0) point, corresponding to the small indentations on
the jω-axis of Γs1 . Because on Γs2 the directions of the small indentations are opposite to that of Γs1 , the sign of Φ12 in Eq.
(G-5) is negative.

Φ13 = angle traversed by the Nyquist plot of L(s) with respect to the (−1, j0) point, corresponding to the semicircle with
infinite radius on the Nyquist paths.

For a transfer function L(s) that does not have more zeros than poles, the Nyquist plot of L(s) that corresponds to the infinite
semicircle must either be a point on the real axis or a trajectory around the origin of the L(s)-plane. Thus, the angle Φ13
traversed by the phasor drawn from the (−1, j0) point to the Nyquist plot along the semicircle with infinite radius is always zero.

Now adding Eq. (G-5) to Eq. (G-6) and using Eqs. (G-3) and (G-4), we get

Φ 1 + Φ 2 = 4Φ 11
= (2Z − 2P − Pω)360°

(G-7)

Solving for Φ11, we get

Φ 11 = (Z − P − 0.5Pω) 180°

(G-8)

The equation states:

The total angle traversed by the phasor drawn from the (−1, j0) point to the L(s) Nyquist plot that corresponds to the portion on
the positive jω-axis of the s-plane, excluding the small indentations, if any, equals

The number of zeros of 1 + L(s) in the right-half s-plane

The number of poles of L(s) in the right-half s-plane
0.5 (the number of poles of L(s) on the jω-axis)180°

(G-9)

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Thus, the Nyquist stability criterion can be carried out by constructing only the Nyquist plot that corresponds to the s= j∞ to s = 0
portion on the Nyquist path. Furthermore, if the closed-loop system is unstable, by knowing the values of Φ11, Pω, and P, Eq. (G-
8) gives the number of roots of the characteristic equation that are in the right-halfs-plane.

For the closed-loop system to be stable, Z must equal zero. Thus, the Nyquist criterion for stability of the closed-loop system is

Φ 11 = − (0.5Pω + P) 180°

(G-10)

Because Pω and P cannot be negative, the last equation indicates thatif the phase traversed by the Nyquist plot of L(jω) as ω
varies from ∞ to 0, Φ11, is positive with respect to the (−1, j0) point, the closed-loop system is unstable.

However, if Φ11 is negative, it still has to satisfy Eq. (G-9) for the system to be stable. With reference to the Nyquist plot of L(jω)
and the (−1, j0) point, we see that, when the angle variation Φ11 is positive, it corresponds to the (−1, j0) point being enclosed.
Thus, the condition that the Nyquist plot of L(jω) not enclosing the (−1, j0) point is a necessary condition for closed-loop stability
for nonminimum-phase systems. However, if the (−1, j0) point is not enclosed by the Nyquist plot of L(jω), for the nonminimum-
phase system to be closed-loop stable, the angle variation Φ 11 still has to satisfy Eq. (G-9).

G-1-1. System with Minimum-Phase Loop Transfer Functions

If L(s) is of the minimum-phase type, then P = 0 and Pω denotes the number of poles of L(s) that are at the origin. Equation (G-
8) becomes

Φ 11 = (Z − 0.5Pω) 180°

(G-11)

For closed-loop stability, Z = 0; Eq. (G-11) becomes

Φ 11 = −Pω × 90°

(G-12)

Because Pω denotes the number of poles of L(s) that are at the origin, it is easy to see that, if the (−1,j0) point is not enclosed
by the Nyquist plot of L(s), Φ11 will always be given by Eq. (G-12). Thus, when L(s) is of the minimum-phase type, the condition
that the (−1, j0) point not be enclosed by the Nyquist plot is a necessary and sufficient condition for closed-loop stability.

G-1-2. Systems with Improper Loop Transfer Functions

Equation (G-8) is derived based on the condition that Φ13 = 0, which is true only if L(s) is strictly proper; that is, it has more poles
than zeros. For improper transfer functions, we can again use the method discussed in Sec. 8-6 by plotting the Nyquist plot of
1/L(s).

G-2. ILLUSTRATIVE EXAMPLES: GENERAL NYQUIST

CRITERION MINIMUM AND NONMINIMUM TRANSFER

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
FUNCTIONS
In the following example, we shall show that the Nyquist plot of a nonminimum-phase transfer function does not enclose the
(−1, j0) point, and yet the system is unstable.

Example

EXAMPLE G-2-1

Consider that the loop transfer function of a control system is given by

s2 − s + 1
L (s) =
s (s 2 − 6s + 5)

(G-13)

Because L(s) has one pole at the origin and two poles in the right-halfs-plane, Pω = 1 and P = 2. From Eq. (G-9), the closed-
loop system is stable if the following condition is satisfied:

Φ 11 = − (0.5Pω + P) 180° = −450°

(G-14)

The Nyquist plot of L(jω) for ω = ∞ to ω = 0 is plotted as shown in Fig. G-2. Apparently, in this case, the (−1, j0) point is not
enclosed by the Nyquist plot. However, because Φ11 is −90°, and not −450°, the system is unstable. Substituting Φ11 = 90°
into Eq. (8-82) and solving for Z, we have Z = 2, which means that there are two closed-loop poles in the right-halfs-plane.

s2 − s + 1
Figure G-2 Nyquist plot of L (s) = .
s (s 2 − 6s + 5)

Example

EXAMPLE G-2-2

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Consider the system described in Example 8-1-1. The loop transfer function of the system, L(s), is given in Eq. (9-59) and is
repeated below.

K
L (jω) =
s (s + 2) (s + 10)

(G-15)

The Nyquist plot of L(jω) is shown in Fig. 8-25. It is shown in Example 8-1-1 that, for closed-loop stability, the (−1,j0) point in
the L(jω)-plane must be to the left of the intersect of the Nyquist plot with the real axis.

Now we shall apply the generalized Nyquist criterion to the system. Because L(s) is of the minimum-phase type, P = 0, and it
has one pole at the origin, thus, Pω = 1. Substituting these quantities into Eq. (G-11), we have

Φ 11 = (Z − 0.5) 180°

(G-16)

For closed-loop stability, Z must equal zero; thus, the last equation gives Φ11 = −90°. This means that the phasor drawn from
the (−1, j0) point to the Nyquist plot, from ω = ∞ to ω = 0, must equal −90°, or 90° in the CW direction.Figure G-3a shows
that, if the (−1, j0) point is to the left of the intersect of L(jω) with the real axis, Φ11 is indeed −90°. On the other hand, if the
(−1, j0) point is to the right of the intersect, as shown in Fig. G-3b, when the value of K is greater than 240, then Φ11 is +270°,
or (more easily observed when the critical point is enclosed) the system would be unstable. Substituting Φ 11 = 270° into Eq.
(G-16), we get Z = 2, which means that the characteristic equation has two roots in the right-halfs-plane. Thus, for systems
with minimum-phase loop transfer functions, the Nyquist criterion with the "enclosure" test is easier to observe, but when
the system is unstable, it does not tell how many characteristic equation roots are in the right-half plane; the general
Nyquist criterion does.

K
Figure G-3 Nyquist plot of L (s) = . (a) K < 240. (b) K > 240.
s (s + 2) (s + 10)

Example

EXAMPLE G-2-3

Consider that a control system has the loop transfer function

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 K (s − 1)
L (s) =
s (s + 1)

(G-17)

We observed from the last equation that Pω = 1, and P = 0. The function L(s) is of the nonminimum-phase type, because it
has a zero at s = 1. Thus, the Nyquist criterion on enclosure cannot be used adequately in this case. From Eq. (G-8), the
requirement for closed-loop stability is

Φ 11 = − (0.5Pω + P) 180° = −90°

(G-18)

Thus, the stability criterion requires that the phasor drawn from the (−1, j0) point to the Nyquist plot of L(jω) should traverse
−90° as ω varies from ∞ to 0.

To sketch the Nyquist plot of L(s) that corresponds to the positive portion of the jω-axis of the s-plane, we set s = jω in Eq.
(G-17). We get

K (jω − 1) K (jω − 1)
L (jω) = =
jω (jω + 1) −ω2 + jω

(G-19)

When ω = ∞,

K
L (j∞) = = 0∠ − 90°
jωω= ∞

(G-20)

When ω = 0,

K
L (j0) = = ∞∠90°
jωω= 0

(G-21)

To find the intersect of the L(jω) plot on the real axis, we rationalize the function by multiplying the numerator and the
denominator of Eq. (G-19) by −ω2 − jω. We have

K (jω − 1) (−ω2 − jω) K [2ω + j (1 − ω2)]

L (jω) = =
ω4 + ω2 ω (ω2 + 1)

(G-22)

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Setting the imaginary part of L(jω) to zero and solving for ω2 , we have

ω2 = ±1 rad/s

(G-23)

For ω = 1,

L (j1) = K

(G-24)

G-2-1. Nyquist Plot for K > 0

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
G-2-1. Nyquist Plot for K > 0
Based on the preceding information, the Nyquist plot of L(jω) that corresponds to the positive portion of the jω-axis is sketched
as shown in Fig. G-4 for K > 0. Figure G-4 shows that, as ω varies from ∞ to 0 along the Nyquist plot, the net angle Φ11
traversed by the phasor drawn from the (−1, j0) point to the Nyquist plot is +90°. Thus, the system is unstable because Φ11 is
positive. We can also readily see that (−1, j0) is enclosed by the Nyquist plot, so the same conclusion on closed-loop stability
can be drawn.

K (s − 1)
Figure G-4 Nyquist plot of the system in Example G-2-3. L (s) = .
s (s + 1)

From Eq. (G-7),

Φ 11 = (Z − 0.5Pω − P) 180° = (Z − 0.5) 180° = 90°

(G-25)

Thus, Z = 1, which means that the characteristic equation of the closed-loop system has one root in the right-halfs-plane. The
characteristic equation of the system is

s 2 + (1 + K) s − K = 0

(G-26)

We can easily verify that stability requires

0 > K > −1

(G-27)

G-2-2. Nyquist Plot for K < 0

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Figure G-5a shows the Nyquist plot of L(jω) when K lies between 0 and −1. Notice that the plot is obtained by rotating the L(jω)
plot of Fig. G-4 by 180° about the origin. As ω is varied from ∞ to 0, the angle Φ11 in Fig. G-5a has a net rotation of −90°, which
agrees with the stability requirement in Eq. (G-18), and the closed-loop system is stable. It should be reminded that since L(jω)
is of the minimum-phase type, the fact that the Nyquist plot of L(jω) of Fig. G-5a does not enclose the (−1, j0) is not the reason
that the system is stable.

K (s − 1)
Figure G-5 Nyquist plots of the system in Example G-2-3. L (s) = . (a) −1 < K < 0. (b) K
s (s + 1)
<−1.

Figure G-5b shows the Nyquist plot when K < −1. Now we see that the (−1, j0) point is enclosed, so the system is unstable.
Checking the value of Φ11, we have Φ11 = 270°, which differs from the required −90°. Using Eq. (G-8),

Φ 11 = (Z − 0.5Pω − P) 180° = (Z − 0.5) 180° = 270°

(G-28)

Thus, Z = 2, which means that the characteristic equation has two roots in the right-halfs-plane.

In general, when K changes signs, it is not necessary to redraw the Nyquist plot as shown in Fig. G-5. Equation (8-40) can be
written as

1 + L (s) = 1 + KL1 (s) = 0

(G-29)

where K is positive. For negative K, the last equation can be written as

1 − KL1 (s) = 0

(G-30)

or

KL1 (s) = 1

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
(G-31)

where K is now positive. Thus, Eq. (G-31) shows that, when K is negative, we can still use the L(jω) plot for positive K but
designate the (+1, j0) point as the critical point for stability analysis.

Figure G-6 shows the Nyquist plot of Eq. (G-17) for K > 0. When K is negative, the (+1, j0) point is regarded as the critical point.
As shown in Fig. G-6, for −1 < K < 0, Φ11 is −90°, which is the required value, and the system is stable. When K < −1, the (+1, j0)
point is enclosed by the Nyquist plot, and the system is unstable. These results agree with those obtained from Fig. G-5 when
the Nyquist plots for K < 0 were actually constructed.

K (s − 1)
Figure G-6 Nyquist plot of L (s) = . The (+1, j0) point is the critical point.
s (s + 1)

It is of interest to compare the Nyquist stability analysis with the root-locus analysis. Figure G-7 shows the root loci of the
characteristic equation of the system with the loop transfer function given in Eq. (G-17). The stability condition of the system as
a function of K is clearly indicated by the root-locus diagram. The RL between 0 and +1 on the real axis indicates that the
system is unstable for 0 < K < ∞. The RL indicates that, for negative values of K, the system is unstable for − ∞ < K < −1, and
the system is stable only for the range of −1 < K < 0. The root loci cross the jω-axis at ω = ±1 rad/s, which are the values ofω at
which the Nyquist plot of L(jω) intersects the negative real axis.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Figure G-7 Complete root loci of system in Example G-2-3.

Example

EXAMPLE G-2-4

Consider the control system shown in Fig. G-8. It is desired to determine the range of K for which the system is stable. The
loop transfer function of the system is

Y (s) 10K (s + 2)
L (s) = = 3
E (s) s + 3s 2 + 10

(G-32)

Figure G-8 Block diagram of the control system in Example G-2-4.

The poles of L(s) are found to be at s = −3.72, 0.361 + j1.6, and 0.361 − j1.6. Or, we can use the Routh-Hurwitz criterion to
verify that L(s) has two poles in the right-half s-plane. Thus, P = 2, and Pω = 0. The transfer function L(s) is of the
nonminimum-phase type.

From Eq. (G-9), the requirement for the closed-loop system to be stable is

Φ 11 = −1 (0.5Pω + P) 180° = −360°

(G-33)

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Setting s = jω, Eq. (G-32) becomes

10K (jω + 2)
L (jω) =
(10 − 3ω2) − jω3

(G-34)

At ω = ∞,

L (j∞) = 0∠180°

(G-35)

At ω = 0,

L (j0) = 2K

(G-36)

To find the intersect on the real axis of theL(jω)-plane, we rationalize L(jω) as

10K {2 (10 − 3ω2) − ω4 + j [ω (10 − 3ω2) + 2ω2]}

L (jω) =
(10 − 3ω2)2 + ω6

(G-37)

Setting the imaginary part of L(jω) to zero, we have

ω (10 − 3ω2) + 2ω3 = 0

(G-38)

The solutions of the last equation are ω = 0 and ω = ±√10 = 3.16 rad/sec, which are the frequencies at which the L(jω)
plot intersects the real axis of the L(jω)-plane. When ω = 0, we already have L(j0) = 2K in Eq. (G-36). When ω = 3.16 rad/s,

L (j3.16) = −K

(G-39)

Figure G-9a shows the Nyquist plot of L(jω) for 0 < K < 1. Because the (−1, j0) point is enclosed by the Nyquist plot, the
closed-loop system is unstable. We can also show that the angle traversed by Φ11 is 0°, not −360°, as required in Eq. (G-33).
Figure G-9b shows the Nyquist plot of L(jω) when K is greater than unity. In this case, the angle Φ11 rotates a total of −360°;
thus, the system is stable.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Figure G-9 Nyquist plots of the system in Example G-2-4. (a) 1 > K > 0. (b) K > 1.

When K is negative, we can use the plots in Fig. G-9 and regard the (+1, j0) point as the critical point. The following stability
conditions are observed:

2K < −1: The (+1, j0) point lies between 0 and 2K and is not enclosed by L(jω), but Φ11 = −180°; the system is unstable.
For stability, Φ11 must equal −360°.

−1 < 2K < 0: The (+1, j0) point is to the right of the point 2K and is enclosed by the Nyquist plot; the system is unstable. In
this case, Φ11 = 0°.

The conclusion is that the system is stable for K > 1. The root loci of the system are shown inFig. G-10. Clearly, when K is
negative, one branch of the RL will always stay in the right-half plane, and the system is unstable. The system is stable only
for K > 1, and the root loci cross the jω-axis at ω = ±3.16 rad/s, which corresponds to the frequency at which theL(jω) plot
intersects the negative real axis. The value of K at the crossing point on the jω-axis is 1.

Figure G-10 Complete root loci of the system in Example G-2-4.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
Example

EXAMPLE G-2-5

Consider a control system with the loop transfer function

K
L (s) =
(s + 2) (s 2 + 4)

(G-40)

which has a pair of imaginary poles at s = j2 and −j2. Thus, Pω = 2, and P = 0. To apply the Nyquist criterion in the original
form, we would have to define the Nyquist path with small indentations around these poles.

Instead of constructing the entire Nyquist plot, the portion that corresponds to s = j∞ to j0 is plotted as shown in Fig. G-11.
The data for this Nyquist plot are easily obtained using any of the frequency-domain programs mentioned earlier.

Figure G-11 Nyquist plot of the control system in Example G-2-5.

From Eq. (G-9), the value of Φ11 required for stability is

Φ 11 = − (0.5Pω + P) 180° = −180°

(G-41)

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 As seen from Fig. G-11, the magnitude of L(jω) goes to infinity when ω = 2 rad/s. When K is positive, the critical point (−1, j0)
is not enclosed by the Nyquist plot, and the system is unstable. For the angle check, when ω varies from ∞ to 2, the angle
Φ11 is +135°, and for the portion of ω = 2 to 0, Φ11 is +45°. Thus, the total Φ11 is +180°, not −180°. The system is unstable for
all positive values of K.

When K is negative, the critical point in Fig. G-11 is at (+1, j0). Figure G-11 shows that, if the (+1, j0) point lies between 0 and
K/8, it is enclosed by the Nyquist path, and the system is unstable. Thus, the system is unstable forK < −8. When the (+1, j0)
point is to the right of the K/8 point, Φ11 from ω = ∞ to ω = 2 is −45°, and from ω = 2 to ω = 0 is −135°. Thus the total Φ11 as
ω varies from ∞ to 0 is −180°, which agrees with the value required in Eq. (G-41). The system is stable for 0 > K > −8. The
summary of the Nyquist criterion application to this system is as follows:

Range of K Φ11 (deg) for ω = 2 to 0 Φ11 (deg) for ω = 2 to ∞ Total Φ11 (deg) Critical Point Stability Condition

K>0 +135 +45 +180 −1 point enclosed Unstable

K < −8 −45 +45 0 +1 point enclosed Unstable

−8 < K < 0 −45 −135 −180 Stable

The complete root loci of the characteristic equation of the system are constructed inFig. G-12 using the pole–zero
configuration of Eq. (G-40). The stability condition of −8 < K < 0 is easily viewed from the root loci.

Figure G-12 Complete root loci of the system in Example G-2-5.

G-3. STABILITY ANALYSIS OF MULTILOOP SYSTEMS

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
The Nyquist stability analyses conducted in the preceding sections are all directed toward the loop transfer functionL(s). It
does not matter whether the system is with single loop or multiple loops because once the loop transfer function is obtained,
stability analysis can be conducted using the Routh-Hurwitz criterion, root loci, or the Nyquist criterion.

For multiloop-feedback systems, it may be advantageous to analyze the stability of the system by working from the inner loop
toward the outer loop, one at a time. This way, more insight may be gained on the stability of the individual loops of the system.
The following example will illustrate this approach.

Example

EXAMPLE G-3-1

Figure G-13 shows the block diagram of a system that controls the gun turret of a tank. During servicing of the turret control
system, the mechanic accidentally opened the outer loop of the system. With the power turned on, the gun turret went out of
control and finally flew apart. The purpose of this example is to show that it is inadequate to investigate just the stability of
the overall system. In general, for a multiloop control system, one should conduct a systematic stability analysis of all the
inner loops of the system. It is admissible to have unstable inner loops, as long as the overall system is stable. However, if
such a situation exists, it is important to forewarn or take precautionary measures to prevent opening the loops during
operation.

Figure G-13 Multiloop feedback control system for tank turret control.

The loop transfer function of the inner loop is

6
G (s) =
s (s + 1) (s + 2)

(G-42)

Figure G-14 shows the Nyquist plot of G(s). Because the plot intersects the real axis at the −1 point atω = 1.414 rad/s, the
inner loop is marginally stable. Therefore, if the outer loop of the system is opened, the system will oscillate continuously
with a frequency of 1.414 rad/s. The loop transfer function of the overall system is

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 G c (s) G (s) 100 (s + 0.1)
L (s) = =
1 + G (s) (s + 10) (s 3 + 3s 2 + 2s + 6)

(G-43)

6
Figure G-14 Nyquist plot of G (s) = .
s (s + 1) (s + 2)

Because L(s) has two poles on the jω-axis and the rest are in the left-half s-plane, Pω = 2 and P = 0. The function is also of
the nonminimum-phase type, so we must use Eq. (G-9) for the stability test of the overall system. Thus,

Φ 11 = − (0.5Pω + P) 180° = −180°

(G-44)

The Nyquist plot of L(s) is plotted as shown in Fig. G-15. The angle Φ11 for ω = ∞ to ω = 1.414 rad/s is −90°, and from ω =
1.414 rad/s to ω = 0 is −90°. Thus, the total value of Φ11 for ω = ∞ to ω = 0 is −180°, and the overall system is stable.

100 ( + 0.1)
© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 100 (s + 0.1)
Figure G-15 Nyquist plot of L (s) = .
(s + 10) (s 3 + 3s 2 + 2s + 6)

In general, when more than two loops are involved, the proper way is to start with the stability of the innermost loop by
opening all the outer loops, and then add one loop at a time, until the outermost loop is closed.

G-4. REFERENCE
1. K. S. Yeung, "A Reformulation of Nyquist's Criterion," IEEE Trans. Educ., Vol. E-28, pp. 58–60, Feb. 1985.

G-5. PROBLEMS
G-1. The loop transfer functions L(s) of single-feedback-loop systems are given in the following equations. Sketch the
Nyquist plot of L(jω) for ω = 0 to ∞. Determine the stability of the closed-loop system. If the system is unstable, find the
number of poles of the closed-loop transfer function that are in the right-half s-plane. Solve for the intersect of L(jω) on the
negative real axis of the L(jω)-plane analytically. You may construct the Nyquist plot of L(jω) using any computer program.

5 (s − 2)
a. L (s) =
s (s + 1) (s − 1)
50
b. L (s) =
s (s + 5) (s − 1)
3 (s + 2)
c. L (s) =
s (s 3 + 3s + 1)
100
d. L (s) =
s (s + 1) (s 2 + 2)
s 2 − 5s + 2
e. L (s) =
s (s 3 + 2s 2 + 2s + 10)

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 −0.1 (s 2 − 1) (s + 2)
f. L (s) =
s (s 2 + s + 1)
G-2. The loop transfer functions of single-feedback-loop control systems are given in the following equations. Apply the
Nyquist criterion and determine the values of K for the system to be stable. Sketch the Nyquist plot ofL(jω) with K = 1 for ω
= 0 to ω = ∞. You may use a computer program to plot the Nyquist plots.

K (s − 1)
a. L (s) =
s (s 2 − 1)
K
b. L (s) =
s (s + 10) (s − 2)
K (s + 1)
c. L (s) =
s (s 3 + 3s + 1)
K (s 2 − 5s + 2)
d. L (s) =
s (s 3 + 2s 2 + 2s + 10)
K (s 2 − 1) (s + 2)
e. L (s) =
s (s 2 + s + 1)
K (s 2 − 5s + 1)
f. L (s) =
s (s + 1) (s 2 + 4)
G-3. Figure GP-3 shows the Nyquist plots of the loop transfer function L(jω) for ω = 0 to ω = ∞ for single-feedback-loop
control systems. The number of poles of L(jω) that are on the jω-axis, Pω, and in the right-half s-plane, P, are indicated for
each case. Determine the stability of the closed-loop system by applying the Nyquist criterion. For the unstable systems,
give the number of zeros of 1 + L(s) that are in the right-half s-plane.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Figure GP-3

G-4. It was mentioned in the text that, when the function L(jω) has more zeros than poles, it is necessary to plot the
Nyquist plot of 1/L(jω) to apply the simplified Nyquist criterion. Determine the stability of the systems described by the
function 1/L(jω) shown in Fig. GP-4. For each case, the values of Pω and P for the function 1/L(jω) are given, where Pω
refers to the number of poles of 1/L(jω) that are on the jω-axis, and P refers to the number of poles of 1 + 1/L(jω) that are in
the right-half s-plane.

Figure GP-4

G-5. Figure GP-5 shows the Nyquist plots of the loop transfer function L(jω) for ω = 0 to ω = ∞ for single-feedback-loop
control systems. The gain K appears as a multiplying factor in L(s). The number of poles of L(jω) that are on the jω-axis and
in the right-half s-plane are indicated in each case. Determine the range(s) of K for closed-loop system stability.

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.
 Figure GP-5

G-6. The characteristic equations of linear control systems are given below. Apply the Nyquist criterion to determine the
values of K for system stability. Check the answers by means of the Routh-Hurwitz criterion.

a. s 3 + 4Ks 2 + (K + 5)s + 10 = 0

b. s 3 + K(s 3 + 2s 2 + 1) = 0

c. s(s + 1)(s 2 + 4) + K(s 2 + 1) = 0

d. s 3 + 2s 2 + 20s + 10K = 0

e. s(s 3 + 3s + 3) + K(s + 2) = 0

© McGraw-Hill Education. All rights reserved. Any use is subject to the Terms of Use, Privacy Notice and copyright information.