264 Poles, zeros and system stability

Root locus design

20

15
Position of poles
10 for k = ~ 570

5
Imag. axis

–5

–10

–15

–20
–30 –25 –20 –15 –10 –5 0 5 10
Real axis

Figure 10.20 Position of closed-loop poles as gain k increases.

The definition of system stability can be related to the poles of a system. However, its
strict interpretation also relies on our knowledge of the system’s response to bounded
signals. We now look at what is meant by bounded and unbounded signals before we
address the question of system stability.

10.7 What do we mean by bounded signals?

A time domain signal, x(t), is assessed by the behaviour of its magnitude over an infinite
time interval. As time tends to infinity, the absolute value of the signal magnitude can
either:

(a) continuously decrease and/or increase (or stay constant) but remain within a
bounded range

(b) continuously increase to very large values without any bound

Figures 10.21 and 10.22 show examples of bounded exponential signals and bounded sinu-
soidal signals. We can see that the magnitude of an exponential function, eat, with a < 0,
will decrease to zero as time tends to infinity. The magnitude of a unit step function is
finite since its value is 1, even when time tends to infinity. We call these types of signals
bounded.
 10.7 What do we mean by bounded signals? 265

0.8 All these

x1(t) = 1.0
signals are
0.6 x3(t) = 1–e–t bounded
within ±1
0.4

0.2 x2(t ) = e – t
Time
0
0 2 4 6 8 10

Figure 10.21 Examples of bounded exponential signals.

0.8

0.6
x4( t ) = 05
. sin(t )

0.4

0.2 All these

signals
0
are
–0.2 bounded
within ±1
–0.4

–0.6

–0.8 x 5( t ) = e – t sin(t )
Time
–1
0 10 20 30 40 50 60 70 80 90 100

Figure 10.22 Example of bounded sinusoidal signals.

Figures 10.23 and 10.24 show examples of unbounded signals. If the exponent in the
exponential signal eat is positive, a > 0, the signal will increase to infinity as time tends to
infinity. We categorise these signals as unbounded signals.
×10 4
2.5

2
x7( t ) = e 0.1t
1.5

0.5

–0.5

–1 x8( t ) = –e0.1t

–1.5

–2
Time
–2.5
0 10 20 30 40 50 60 70 80 90 100

Figure 10.23 Unbounded exponential signal.

266 Poles, zeros and system stability

×10 4
2.5

1.5

1 x6(t ) = t 2
0.5

–0.5

–1

–1.5
x9( t ) = e01
. t sin( t )

–2
Time
–2.5
0 10 20 30 40 50 60 70 80 90 100

Figure 10.24 Unbounded parabolic and sinusoidal signal.

Table 10.3 examines the exponential signals and their boundedness by studying the
location of their poles.

Table 10.3 Location of poles and boundedness of signals.

Exponential signal Laplace transform Poles Bounded/unbounded?

x1(t ) = e−10t X1 (s) =

1 s = –10 Bounded
s + 10

x 2(t ) = e− 6t X2 (s) =
1 s = –6 Bounded
s+6

x 3(t ) = e− at X3 (s) =
1 s = –a Bounded
s+ a a>0

x 4(t ) = e− 05
. t
X4 (s) =
1 s = –0.5 Bounded
s + 0.5

x 5(t ) = e− 0t = 1 X5 (s) =
1 s=0 Bounded
s

x 6(t ) = e15
. t
X6 (s) =
1 s = +1.5 Unbounded
s − 15
.

x 7(t ) = e at X7 (s) =
1 s = +a Unbounded
s− a a>0

x 8(t ) = e2. 8t X8 (s) =

1 s = +2.8 Unbounded
s − 2.8

x 9(t ) = e20t X9 (s) =

1 s = +20 Unbounded
s − 20

Although variable s is a complex variable given by s = σ+ jω , for these examples the pole
values are real, and the values are the same as the rate constant in the exponential func-
tion. For example, x7(t) = eat, and the pole is at s = a.
 10.8 System stability 267

Remark Decaying exponential signals have Laplace transforms with poles in the LHP. Growing or increasing
exponential signals have Laplace transforms with poles in the RHP. We can generalise this observation
in the Key Result that follows.

Key result: Poles in LHP, RHP and on j axis

Signals whose transforms have all the poles in the LHP are bounded.
Signals whose transforms have any one pole in the RHP are unbounded.

However, we must still consider the poles on the jω axis.

Poles on j axis
Signals whose transforms have poles in the LHP and no multiple poles on the jω axis are
bounded, otherwise they are unbounded.

This is intuitively so; we recall that a step function has the transfer function R(s) = 1/s
which is a bounded (and constant!) signal and has one pole at p = 0. A ramp function has
the transform R(s) = 1/s2 which has two poles at the origin on the jωaxis. The ramp func-
tion is unbounded, since its magnitude increases without bound as time progresses.

10.8 System stability

Stability is one of the most important considerations in any system. It is thus essential to
develop simple tools to examine the stability of systems. We term unstable systems as
those that will, for example, move off position and not return to a stable equilibrium
position after some initial excitation. A simple example of this is shown in Figure 10.25
(a) and (b), where we can see a pendulum and an inverted pendulum.

(a) (b)

Figure 10.25 Pendulum and inverted pendulum.

The pendulum shown in Figure 10.25(a) is a stable system. If we hold the rod horizon-
tally and release it, it goes to its rest position after some initial oscillation. An inverted
pendulum, as shown in Figure 25(b), is an unstable system. If we hold the rod vertically
and release it, it stays vertical unless perturbed, and then it falls from its unstable posi-
tion to the horizontal, stable, position. A real example of a system which is like an
inverted pendulum is a mountain bike. This is an unstable system where the rider stabi-
lises the bike using pedal force and rider balance to keep the bike upright and moving
(most of the time!).
268 Poles, zeros and system stability

Another example of an unstable system is an exothermic reactor. In this case, the

exothermic reaction process gives out heat which must be removed by a coolant if the
process is to remain under control or stable. If the heat removal balance is lost, then a
runaway situation can develop, causing a spiralling or unbounded temperature rise.

10.8.1 What is a stable system?

We call a system stable if its output signal is bounded for any bounded input signal. We
call this type of system stability ‘bounded-input bounded-output stability’.
A simple method to check the system stability is to examine the poles of the systems,
since there is a relationship between the system poles and the system stability, which we
explore now.
Consider a general transfer function of the form:
n( s) bm s m + bm− 1s m− 1 + … + b1s1 + b0
G( s) = =
d( s) an s n + an− 1s n− 1 + …a1s1 + a0

Set the denominator polynomial equal to zero to find the poles:

an s n + an− 1s n− 1 + … + a1s1 + a0 = 0
We can factorise this polynomial into real (first order) and complex poles (second order).

sr(s + p1) ... (s + pi) ... ( s 2 + 2ζωn1s + ω2n1) … ( s 2 + 2ζωni s + ω2ni )

The system poles will therefore consist of some real poles and some imaginary poles.
From the superposition principle (Chapter 2) the response of the general system is the
summation of the responses from all the individual poles. If only one of these poles lies in
the RHP, the total response will be dominated by this pole and the system becomes
unstable. A system is therefore stable if all the poles lie in the LHP.

Key result: Stability test

A system is stable if all its poles lie in the Left Half Plane.

What about poles on the j axis?

To answer this question we give two examples: Figures 10.26(a) and (b).

System 1 System 2
1 s
U(s) = s Y(s) U(s) = 2 Y(s)
1 s +w12 s
s s2+w12

Figure 10.26 Two examples of systems with bounded input signal

and unbounded output signal.

In the two examples, we can see that the input signal is a step and a sine wave. Both of
these are therefore bounded input signals. If we examine the poles of each system we find
that:

(a) For system 1, the pole lies at the origin, on the jω axis.

(b) For system 2, the poles lies at ± ω 1, both on the jω axis.

 10.8 System stability 269

However, the output signals are given by:

(a) For system 1,

1 1 1
Y(s) = × =
s s s2

which represents a ramp signal (an unbounded signal)

(b) For system 2,

2
s s  s 
Y ( s) = 2 × 2 =  2 
s + ω1 s + ω1  s + ω12 
2 2

The output is shown in Figure 10.27, where we see increasing unbounded oscillations.
Although the input signal to the system is bounded, the output signal is unbounded;
hence the system is unstable.

Output response
15

10

–5

–10

–15
0 5 10 15 20 25 30
Time

Figure 10.27 Unstable response from system with poles on jω axis.

Although we might think that these are specific examples, the statement about
stability tells us that for a system to be stable, it must produce a bounded output signal
for any bounded input signal. In both the above cases, we have chosen a bounded input
signal, but the output is unbounded; hence the systems are unstable. What we should
remember from this is that poles on the jω axis are not in the LHP, and therefore our
stability test is correct provided we do not include the jω axis in the LHP.

Example Figures 10.28 and 10.29 show the relationship between the stability of a first-order system and
its pole position. The first-order system is given by
1
G(s) =
s+a

We can see that the first-order system is stable if its pole lies in the LHP. If the pole lies in the RHP
including the origin the system is unstable.
270 Poles, zeros and system stability

Pole–zero map
1

0.8

0.6 Position of system

pole as a > 0
0.4 increases
(STABLE)
0.2
Imag. axis

0
–10 –7 –3 0 2 6 10
–0.2
1

–0.4
0.5

–0.6
0
0 50 100 150 200 250
–0.8 Typical step response
–1
–10 –8 –6 –4 –2 0 2 4 6 8 10
Real axis

Figure 10.28 The loci of poles for the first-order system a > 0.
Pole zero map
1

0.8

0.6 Position of system

pole; for a < 0,
0.4 |a| increases
(UNSTABLE)
0.2

0
–10 –7 –3 0 2 6 10
–0.2
1

–0.4
0.5

–0.6
0
0 50 100 150 200 250

–0.8 Typical step response

–1
–10 –8 –6 –4 –2 0 2 4 6 8 1
Real axis

Figure 10.29 The loci of poles for the first-order system a < 0.

Using MATLAB to check the stability of a system

Enter the system transfer function using:

s = tf('s');
g = ...

then either
 10.8 System stability 271

(i) run:

pzmap(g)

If all the poles are located in the LHP the system is stable. Otherwise the system is unstable.

or

(ii) use

pole(g)

If all the poles have negative real parts then the system is stable.

To find the damping ratio and natural frequency contours run sgrid and read the values
from the contours

Problem The following transfer function represents a large sea-going tanker where U(s) is the input signal
from the rudder and Y(s) is the heading angle of the ship. Investigate the stability of the system.
00192
. (1+ 46s )
Y(s) = U(s)
(1− 1230
. s )(1+ 160
. s)

Solution Using MATLAB we enter the transfer function as follows

s = tf('s');
g = 0.0192*(1+46*s)/((1–123*s)*(1+16*s));

The function pzmap(g) produces the output shown in Figure 10.30.

Pole zero ma
0.2

0.15

0.1

0.05
Imag. axis

–0.05

–0.1

–0.15

–0.2
–0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15 0.2
Real axis

Figure 10.30 Pole–zero map for the tanker system transfer function.
Figure 10.30 shows that both the poles and the zero lie on the real axis, but one pole lies in the
RHP. Since one pole is in the RHP the system is therefore unstable.
272 Poles, zeros and system stability

What we have learnt

3 To find the poles and zeros from the transfer function representation of a system:
System poles are the roots of the denominator polynomial equation of the
system transfer function

System zeros are the roots of the numerator polynomial equation of the system
transfer function

3 To recognise the link between the first-order time constant, τ , and the associated
pole position:

As the time constant increases the pole approaches the origin

3 To understand the relationship between the second-order system parameters (ζ and

ω n), and the associated contours on the pole–zero map:
Fixing the damping ratio fixes the angle of the second-order poles from the real
axis

Fixing the natural frequency ensures the poles lie a fixed distance from the origin

3 To understand how the zeros in a system arise and the blocking effect they have on
signals.

3 To understand that by closing the loop, the poles of the closed-loop system will be
different from the system poles.

3 To identify whether a signal is bounded or unbounded.

3 To establish whether a system is stable or unstable by using the stability test:

The system G(s) is stable if all the poles of G(s) lie in the Left Half Plane, other-
wise the system is unstable

Multiple choice

M10.1 What are the pole, p, and zero, z, of the M10.3 Three different first-order systems have the
transfer function G(s) = (s + 2)/(s + 3)? following poles: p1 = –0.2, p2 = –0.4, p3 =
(a) p = 2, z = 3 –0.6. Which system has the fastest step
(b) p = 3, z = 2 response?
(c) p = –2, z = –3 (a) the system with p1 = –0.2
(d) p = –3, z = –2 (b) the system with p1 = –0.4
(c) the system with p1 = –0.6
M10.2 If a first-order system has a pole at p = –2,
(d) they all respond the same since they are all
what is the system time constant?
first-order systems
(a) τ = 0.5
(b) τ = 2 M10.4 Purely oscillatory systems have poles that
(c) τ = –0.5 lie:
(d) τ = –2 (a) at the same place on the negative real axis
(b) at the same place on the positive real axis
(c) in the LHP, but not on the real axis
(d) on the imaginary axis