Int. 1. Elect. Enging Educ., Vol. 15, pp. 21-39. Manchester V.P., 1978.

Printed in Great Britain

THE BALL AND BEAM CONTROL EXPERIMENT

P. E. WELLSTEAD, V. CHRIMES, P. R. FLETCHER, R. MOOD Y and A. J. ROBINS

Control Systems Centre, University ofManchester Institute ofScience and
Technology, England

I INTRODUCTION
This document describes our experiences in the construction and use of a simple
but flexible electromechanical control system. The basic system, shown in Fig. I,
consists of a steel ball which is free to roll along a groove in the top of a rigid beam,
which is pivoted about the centre. The control task is to use the angle of tilt of the
beam o (r) (see Fig. I for a defmition of variables) in order to control the position
x (r) of the ball. The essential dynamical characteristics of the system may be
appreciated by noticing that the angle a(t) determines the fraction of gravitational
force which is parallel to the direction of motion of the ball. Thus if the ball is
thought of as a point mass m sliding on an inclined plane (Fig. 2), then by the
laws of motion

x=gsina (l)

When this expression is linearized and Laplace transformed with zero initial
conditions, the following transfer function is obtained

x(s) =gz a(s) (2)

s

Equation (2) is arrived at by a gross simplification of the actual system dynamics;

.f
nevertheless it serves to indicate the double integration which is the essential

mass, m

mg s~n a a

~~ ;. \
\

~ mg cos a

Beam mg

FIG.] FIG. 2

21

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

22

dynamical property of the system. Indeed it is just this property which makes the
ball and beam a difficult yet tractable and instructive control exercise. Specifically,
the ball is open-loop unstable and therefore requires active stabilizing feedback. A
useful feature of the simple form of equation (2) is that it allows the modern
control algorithms of optimal control and pole assignment to be applied in a form
suitable for hand-calculation.

2 CONSTRUCTION
In our design, the ball and beam equipment consists of a light plexiglass beam of
length 90 em and with a 2.5 em square cross-section. As shown in Fig. 3, the beam
is mounted in a cradle made from an aluminium angle-piece which in turn is
mounted via a bearing block to a rigid back plate. The beam is pivoted about the
axis of rotation and driven via a flexible connecting rod by means of a vertically
mounted moving coil actuate r. The angle of the beam is measured by means of a
vertically mounted linear displacement transducer which is again connected by a
flexible rod. Perhaps it would have seemed more logical to use a torque-motor
drive for the beam angle control in place of the push-pull control of the moving
coil actuator. However, the large torque requirements and small angular excursions
of the beam mitigate in favour of the moving coil system. For, in addition to
generating the required torques with only a low input power, the mechanical
limits on the travel of the actuator introduces a constraint on the maximum
angular displacement. A natural limitation such as this on control excursions is
desirable for operational reasons, and in addition allows the time-optimal control
problem to be posed in a natural manner (cf. section 4.3).
The position of the steel ball on the beam is measured by a potentiometric
method in which the ball replaces the wiper blade. Specifically the plexiglass beam

!90cm ----'~
I"
Back Plate

Plexi-GlasB Beam

Hinged Connecting Rods

2.5cm

Linear Displacement
Transducer

FIG.3a Front elevation.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 23

~6cm~
Bearing
Block Beam Cradle

___ ~ Axis of rotation

Back Beam
Plate

Pcv i.ng Coil

Actuator

:i//,I;
FIG.3b Side elevation (displacement transducer in line behind actuator, and not
shown).

has a channel cut in the top face and along the length of the beam. (see Fig. 4).
Mounted in grooves on the edge of the channel are a pair of parallel 1.6 mm diameter
nichrome wires. A small balanced voltage is developed across the ends AB of one
wire, and as indicated in Fig. 5, a voltage vit) proportional to the position of the
ball is measured by connecting one end (c) of the free wire to an operational
amplifier. A particular problem is the disturbance introduced into the measurement
scheme by the intermittent contact made by the ball as it rolls along the channel
between the two wires (Fig. 6). The contact noise cannot be completely removed,

Wire

f
2.5cm

~
~2.5cm
"-i
FIG. 4 Beam detail.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

24

A .... - -..... B

FIG. 5 Ball position measurement.

but it can be reduced by inserting a smoothing capacitor across the input to the
operational amplifier. When the ball makes good contact with the wire the capacitor
charges rapidly, with time constants determined by the small con tact resistance. If
the ball temporarily loses electrical contact with the wires, then the capacitor tends
to retain its potential, discharging slowly through the operational amplifier input
resistance. Operated in this manner, a fairly noise-free measurement of ball position
is possible, provided of course that the ball and wires are kept reasonably clean.
The remaining constructional details are summarized in appendix 1, together
with schemas of the actuator drive amplifier and the angle transducer output
circuitry. As a guide to potential constructors, the principal physical dimensions
used by us are given in Figs. 3 through 6. All dimensions are in centimetres.

3 MODELLING
The double integrator transfer function mentioned in the introduction establishes
the basic dynamical behaviour of the ball and beam. Indeed, most of the control
studies developed in subsequent sections rely upon this dynamical model. It is
important, however, to be aware of the approximations which are involved in
arriving at this model. Such insight can best be obtained through the development
of a complete dynamical model of the system. To this end the experimental ball
and beam can be represented in a form suitable for mathematical modelling by
Fig. 7. In this diagram the friction associated with the rotation of the beam is
lumped into a single linear coefficient b. The stiffness of the spring mounting of
the moving coil is assumed to be k, and the force exerted by the actuator is denoted
F(t). We can work directly in terms of F(t) as the system input since, over the
frequency range of interest, the moving coil actuator can be regarded as an ideal trans-
ducer. Thus F(t) is directly proportional to the actuator drive amplifier input voltage.

FIG. 6 The ball in position.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 25

k b

F(t)

moment of inertia
about pivot. I
a

ball rolling radius r.

moment of inertia lb'

FIG. 7 Depicting the ball and beam in a form suitable for dynamical modelling.

The most straightforward means of modelling the ball and beam is by variational
methods.' To do this we pick the ball position x, and the beam angle a as a com-
plete independent set of generalized coordinates. Since the ball position and the
beam angle are not externally constrained, the pair (ox, oa) also form a complete
independent set of variational coordinates.

The system Lagrangian L is defined thus

L=u*-T (3)

where U*, the kinetic co-energy, is composed of the translational and rotational
kinetic co-energy of the ball, plus the kinetic co-energy of the beam.

(4)

where the translational velocity of the ball v, and its angular velocity w must be
rewritten in terms of the generalized velocities, (x, Q). Consider first the angular
velocity; the angular distance 0 rolled by the ball with respect to the fixed datum is
given(from Fig. 8) by

o= l/J + a =-rx + a

dO x .
Hence w=- =-+a (5)
dt r

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

26

a.
The distance rolled is AB and related
to the peripheral distance on the ball
AC by
AB = AC
or rlji =x
FIGS. 8 and 9 Kinematic relations for ballangular/translational velocity.

The velocity of the ball can be obtained from kinematic relations shown in
Fig. 9 as

v={(xi+(xaY}! (6)

The potential energy T of the system is associated with the energy stored in the
spring. If only small angular excursions are considered and the spring is linear with
stiffness k, then T is given by

1
T=-k(lexY (7)
2

Combining equations (3) through (7) gives the system Lagrangian as

(8)

A further quantity required is the system co-content, J. This incorporates a measure

of the energy dissipation in the system into the variational solution. The co-content
is therefore associated with the energy dissipator which, if it is linear with
coefficient b, has co-content

1 • 2
J=-b(lcx) (9)
2

Lagranges equations of motion for the ball and beam are

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 27

A
dt
(aL) aL + aa
aa - aa aJ = 7
2
(lO)

where F 1 is the generalized force associated with forces resolved in the direction of
the generalised co-ordinate x. In this case F 1 consists of the components of the
gravitational force mg acting in the x-direction.

F 1 =mgsina (11)

The generalized torque 72 is composed of the components of torque contributed by

the external forces acting about the beam pivot. One component is due to the input
force and another opposing component stems from the gravitational force on the
ball.

T2 = cos a (mg x - F(t)!) (l2)

The equations of motion of the system can now be formulated by substituting the
appropriate quantities in equation (lO) and differentiating to yield

(13a)

(mx 2 + f b + fa)&. + (2mxx + bP)a + k/ 2a + f b i - x mg cos a = -F(t)/

r

(13b)

Equations (13a) and (l3b) are the equations of motion of the ball and beam.
They constitute a non-linear coupled set of differential equations which as such
are unsuitable for linear control studies. The equations reduce to a more tractable
form if the following assumptions are made
(i) The moment of inertia of the ball I b and its mass m are small and have relatively
little influence on the beam behaviour as expressed by equation (l3).
a
(ii) The influence of terms in &. and in equation (13a) can be neglected.
The validity of these assumptions are open to question, depending, as they do,
upon the relative physical features of the system. For the equipment described
here the approximations are felt to be quite justifiable. Indeed, the errors involved
in invoking them are small in comparison to certain other phenomena to be
mentioned later.
Using approximation (i), (ii) and remembering that angular excursions a are
small, the equations of motion reduce to the following linearized set

(m + ~)x= mga (l4a)

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

28

(14b)

The moment of inertia of a solid uniform sphere of radius R is

(15)

which, when substituted into equation (14a) gives

(14a)
repeat
The system equations can now be Laplace transformed to give an approximate
transfer-function model of the ball and beam. As illustrated in Fig. 10, this reduces
the system to its essential linear form of a beam transfer function in cascade with a
ball transfer function. Such a representation, albeit approximate, is convenient
since it separates the dynamical functions of the beam and the ball. For example,
two-loop feedback can be employed with an inner loop to control the beam angle
Q, and an outer loop to achieve ball position control. Actually, the reason for

including a beam angle sensor was to permit such an exercise. It has been found,
however, that with the beam drive described in section 2 the beam dynamics are
sufficiently fast to be negligible in comparison with those of the ball. This final
simplification yields the double integrator transfer function mentioned previously
_ g F(s)
( )
x s - ( 2
kl 1+--
R2)82
(16)

5 r2
Notice that the gain of the double integrator is determined, among other things, by
the ratio of ball radius R to rolling radius r. (see Fig. 6). Thus the dynamic response
of the ball may be conveniently modulated by varying this ratio.
To conclude, the ball and beam can be modelled by a pair of coupled non-linear
equations. Under certain conditions these can be linearized to give a second-order
beam transfer function in cascade with a double-integrator ball transfer function. In
most cases the beam dynamics are so fast that the entire system is adequately
modelled by the double integrator. This simplifies the control problem since it

beam transfer ball transfer

function function

F(s) a(s) x(s)

FIG. 10

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 29

avoids a detailed study of the beam dynamics which actually contain many modes
which cannot be readily incorporated into a modelling and control study. For
example, the flexurable mode of the backplates presents a resonance which is
coupled to the basic oscillatory frequency of the beam. In addition, backlash in the
actuator and transducer linkages, contributes a constant phase lag to the beam
frequency response which is often misinterpreted.'
For the practical studies outlined in the next section, the system transfer
function was determined by applying a known voltage to the actuator and record-
ing the time taken for the ball to roll the length of the beam. The gain of the ball
transfer function is then obtained from the standard equations of motion.

4 CONTROL

4.1 Classical phase advance compensation

Stabilizing feedback compensation of the ball and beam transfer function (as
approximated by equation (16)) can be provided by one stage of phase advance
compensation. Moreover because of the nominal 'double integrator' transfer
function, the stability of the closed loop system does not hinge upon the precise
design compensator parameters being adhered to. For example, Fig. 11 shows the
classically compensated ball and beam system together with closed loop step
responses for two sets of compensator pole/zero locations.
Notice, however, that the underlying nature of the system will introduce certain
practical constraints upon the controller design. Specifically, the closed loop system
is not stable for all positive gains as suggested by the double integrator model. If the
beam dynamics are taken into account, the root locus of Fig. 12 is obtained,
indicating that the system is unstable for sufficiently large loop gain. Actually, the
limitations on the angular displacement Q caused by the restricted actuator travel
puts a natural limit on loop gain. Nevertheless, for sufficiently large gains, the
system breaks into limit-cycle oscillations.
The non-linear nature of the system dynamics also introduces a restriction upon
the minimum loop gain which can be tolerated. Specifically, the out-of-balance
torque caused by the mass of the ball (and corresponding to the term x mg cos Q in
equation (l3b)) introduces a self-reinforcing load disturbance which cannot be
rejected at low loop gains. Even if the ball torque is kept in balance, then except
when the ball is in the centre of the beam (x = 0), there will be a non-zero actuator
input to counter the ball torque.

G (1 + Ts)
(1 + a.lTs)

measured
demanded ball
ball position
position

FIG. Ila Classically compensated system.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

30

ball
position step response for T = 3,

I
/\. step response for T = 2

,,
I

I
I
I

f£--~-~-~--,----.- time (seconds)

FIG. 11 b Step responses for classically compensated system.

beam poles

-10 -5

- 5

FIG. 12 Root locus for classically compensated system including beam dynamics.

4.2 Pole-assignment by state feedback 3,4

For a controllable single-input system with the state space description

x=Ax+ bu (17)

A state feedback control law of the form

u=-kx+r (18)

can be designed such that the closed loop characteristic polynomial has specified
zeroes. Assuming that the ball and beam is adequately represented by the double
integrator model with ball position x(t) related to actuator input u(t) by

x(t) =b u(t) (19)

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 31

a suitable state space model may be written in the form of equation (17) as

(20)

where the system states are x I , the ball position and x 2, the ball velocity.
If XI and X2 are measurable, the state feedback control law can be applied with
feedback vector

(21)

selected to give the required closed loop transient behaviour.

A direct measure of the ball velocity is not available. However, because of the
special form of the system, a crude state observer is obtained by integrating the
beam angle (or the actuator input) to give a signal which is proportional to X2' In
practice, care must be exercised when employing the integrated beam angle as the
velocity signal, since any offset in the transducer output is integrated up by the
integrator/observer. For this reason it is often better to employ the integrated
actuator input as a state generator. A nice feature of this is that it gives a controller
structure which is close to the classical phase advance realization.
Combining equations (21) and (20), the closed loop transfer function is as
illustrated in Fig. 13

(22)

In this simple case the state vector can be designed by classical root locus methods.
Specifically, k, is selected to give the required undamped natural frequency, w n '
while k 2 provides velocity feedback which governs the damping factor.j , of the
closed loop system.
A cautionary note is required which again stems from the presence of the beam
dynamics. When k 2 is zero, the root locus for k I does not lie on the jw axis.
Instead, as shown in Fig. 14a, the beam poles cause the locus branches associated
with the ball dynamics to move into the right half plane. In practice this means that

FIG. 13 State feedback controller.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

32

jw
jW

t
20
20

15

10

10 -10 -5 10
-5

-10

-20

-25

(a) (b)

FIG. 14a and 14b Root locus for (a) k 2 = O. k 1 varying, (b) k 2 varying and k, at
design value.

jw

1.0

00
-1.0 -5 00

-5

-1.0

FIG. 14c Close-up oforigin for the locus ofFig. 14b, showing locus of the ball
poles.

ball
positio

time(seconds)

FIG. 15 Step response ofball under state feedback.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 33

a small, but non-zero, coefficient k 2 is required to make the closed-loop system into
a bounded amplitude simple harmonic oscillator. A similar degree of caution is
required when employing large values of k 2 , since as indicated in the root locus for
k 2 (Fig. 14b) the beam poles migrate into the right half plane for k 2 sufficiently
large.
Despite these restrictions the magnified diagram (Fig. 14c) of the root locus of
k 2 shows that the branches of the locus associated with the ball poles closely follow
the semi-circular contour which one would expect if the beam poles were absent.
Moreover, the actual step response of the closed-loop system designed for critical
damping and undamped natural frequency of 0.7 radians/sec corresponds qualitatively
to that which would theoretically be expected (Fig. 15).
4.3 Optimal control
Two types of optimal control strategy can be demonstrated using the ball and
beam: linear quadratic regulationv" and time-optimal control." The former
is a tractable laboratory exercise if the infinite target time is considered. In this
case the optimal strategy can be synthesized, via the steady state Ricatti equation,
as a linear combination of the system states. As such linear quadratic regulation
of the ball and beam is similar in implementation to the pole-shifting experiment
described previously.
In particular, if the performance index (PI) is

ce

PI = f (x t Qx + zu 2 )dt (23)

where z is a scalar control weighting and Q is a positive semidefinite symmetric

matrix defmed by

(24)

The solution of the algebraic Riccati equation gives the optimal state feedback
gain vector as

(25)

where, for simplicity, the control gain b in equation (19) has been set to unity.
As in the case of pole shifting, the elements of the state weighting matrix can
be related to the classical transient response criteria of undamped natural
frequency w n and damping factor ~. Specifically,

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

34

~ =1(2 «;q22 + b~.)! (26)

Time-optimal control of linear systems involves a strategy whereby the control

variables are switched sequentially between their maximum and minimum values.
Usingthe ball and beam, such a strategy can be readily demonstrated since the limited
travel of the moving coil actuator introduces a natural constraint on the achievable
beam angle. In addition the simple form of the system dynamics means that the
optimal switching strategy can be obtained by direct reasoning. Specifically, for
an nth order single input system with real poles and bounded input (lul:S;;; 1) the
time optimal strategy has at most (n - 1) switches between control extrema.
Assume as before that the ball and beam is a double integrator with state equations.

X2 = bu(t) (27)

where x, is the ball position.x, is the ball velocity. The limits on the control
u(t) can be assumed to be ±I, with b scaled accordingly.
The time optimal control which transfers the ball from an initial state
{ x I (0), X 2 (D)} to a target state will involve one switch of control. From the system
equations the possible equations of motion are:

For u = + I

(28a)

Foru=-I

(28b)

If the origin of the state space is the desired final state, then the final part of the
control satisfies.

1 2
X I = - 2b x 2 x I negative
(29)
1
Xl = 2b x ~ Xl positive

The time-optimal rule thus consists of an initial period with the appropriately
signed maximum control with a switch of control sign when

(30)

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 35

Fig. 16 (curve a) shows the time optimal trajectory for the ball position (xd
transferred from an initial state Xl (0) = a, X2 (0) = 0, to the origin.
An interesting feature of time optimal control of the ball and beam is the
intuitively satisfying way in which the results of premature and retarded switching
of control are exhibited. Specifically, if the control is switched too soon the tra-
jectory bounces down the switching curve (equation (30)), rapidly alternating
between control extrema. Fig. 16 (curve b) shows the ball position trajectory when
this occurs. If the control is switched too late, the trajectory over-shoots the target
state and converges in an oscillatory manner. Fig. 16 (curve c) shows the ball
position trajectory for this case.
The appealing aspect of the ball and beam when it exhibits this behaviour is the
graphic way in which all the system states and control are exposed, so that one
actually sees the ball bouncing down the switching curve, or oscillating around the
target state.

5 CONCLUSION
The ball and beam is a simple, robust system which can be used for a variety of
control studies, some of which have been indicated here. The use of a flat beam is,
however, a special case of a more general problem and by way of conclusion, it is
interesting to speculate upon the uses of a curved beam. As an example, consider
a concave beam with uniform radius of curvature lit (Fig. 17). The linearized
dynamic equations for this system are

&+ ga 2R 2 =0 (31)
(lIt-r)(l +5~)

The system is thus a simple harmonic oscillator with natural frequency inversely
proportional to the concave radius of curvature ll. A root locus of equation (31)

ball
positio ~(c) retarded switch of control
/~-~ -
7
target
position
(a) optimal
control

1 2 t imef seconds )

FIG. 16 Response of ball under time-optimal control.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

36

ball mass m, rolling

radius r

FIG. 17 Showing a ball and beam system with a curved beam.

can be drawn (Fig. 18) to show that by selecting a concave radius of curvature the
twin poles at the origin are split and shifted as a complex conjugate pair along the
jw axis. Conversely, if the beam is convex (lR negative), the system poles split and
shift along the real axis.
Thus by a simple modification of the beam shape two variants on the basic
double integrator system are possible.

ACKNOWLEDGEMENTS
The ball and beam experiment did not originate with us. The idea was brought to
our attention in the laboratories of the Division of Automatic Control, Lund
Institute of Technology, Lund, Sweden.
We acknowledge the advice and help given us by Professor K. J. Astrom and
his colleagues of Lund Institute of Technology. In particular the comments made
by Johan Wieslander were most helpful. During the design, construction and
testing of the equipment many helpful suggestions were made by Mr. P. Bowler
and R. Shelley of the Electrical Engineering Department, U.MJ.S.T. Postgraduate
students from the Control Systems Centre who contributed significantly are
J. M. Edmunds and B. H. Thiang.

root locus for~ positive and

decreasing (eg , concave beam)

root locus for

fP.. negative and
increasing (eg , pole positiona for fiat
convex beam) beam (eg.1{ infinite)

FIG. 18 Root locus for the ball dynamics and a curved beam.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 37

APPENDIX - additional structural details

Actuator. The beam is actuated by a moving coil device of the type used in
vibration testing. The type used by us is the Ling Dynamic Systems 200 series
vibrator. The maximum stroke of this device is 5 mm, and in our configuration
this allows a maximum beam angle range of ±3.5 degrees. The maximum force
rating is 27 N for which a 4.5 A armature current is required.
Angle transducer. The angle of the beam is sensed by a linear displacement
transducer. In the equipment described here, a linear capacitative transducer (type
19250) supplied by Jackson Brothers (London) Ltd. was used. This type of
sensing system was used in order to achieve accuracy and good resolution. On
reflection, however, a potentiometer geared to the pivot shaft would probably
have been adequate.
Drive amplifiers and signal conditioning. The policy adopted by us has hinged upon
the use of proprietary analogue computers to implement control strategies. For
this reason the ball and beam is equipped with circuitry to drive the actuator and
condition the ball position and beam angle signals for input to the control com-
puter. Fig. Al shows the overall electrical layout of the system, while Figs. A2 to
A4 depict the relevant circuits.

Pivot

Beam

Actuator
Input
Measured
o-----l Ball
Position
Drive
Amplifier

Measur-ed
Beam Beam
Angle Position
Output
Buffer

FIG. Al Overall schema.

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

38

o---------------p---~----o+9v

loon

2N3055

Inputs Output ~
To Actuator T

PNP3055

2N3055

0 - - - - - - - 4 - - - - - - - - + - - - - - - + - - - - - - - 0 -9v

FIG. A2 Driveamplifier for actuator.

10m
+15v

loom

Input From Output

Angle
Transducer

FIG. A3 Angle transducer output buffer.

-15v
,----00

Input Output
From
Wire
1 2 2 IlF

•
FIG. A4 Ball position transducer output buffer.

REFERENCES
[1) Crandell, S. H. et ai, Dynamics ofMechanical and Electro-Mechanical Systems.
McGraw-Hill, (1968).
[2) Thiang, B. M.,Design Exercise, Control Systems Centre, (1976).

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016

 39

[3] D'Azzo, J. J. and Houpis, C. H., Linear Control System Analysis and Design - Con-
ventional and Modern. McGraw-Hill, (1975).
(4) Robins, A. J., Design Exercise, Control Systems Centre, (1976).
(5) Fletcher, P. E., Design Exercise, Control Systems Centre, (1976).
[6] West, J. C., Analytical Techniques for Non-Linear Systems. English Universities Press,
(1960).

ABSTRACTS-ENGLISH, FRENCH, GERMAN, SPANISH

The ball and beam control experiment

The construction and use of an electromechanical laboratory control experiment is described.
The system, which involves a steel ball rolling on a pivoted beam, presents a mechanically
simple system which is nevertheless an illuminating exercise in feedback control.

L'experience du controle de la bille sur une poutre

La construction et l'utilisation d'une experience de laboratoire de controle electro-mecanique
sont decrites dans cet article. Le systeme, qui utilise une bille d'acier roulant sur une poutre
pivotante, presente un systerne mecanique simple mais neanmoins illustrant tres clairement la
theorie du controle par centre-reaction.

Der Kugel-und-Balken-Regelungsversuch
Die Konstruktion und Anwendung eines elektromechanischen Regelungsversuch wird
beschrieben. Das System, das eine auf einem drehbaren Balken rollende Stahlkugelbenutzt,
repriisentiert ein einfaches mechanisches System, das jedoch eine aufschlussreiche Ubung in
der Regeltechnik darstellt.

El experimento de control de la bola y la barra

Se describe la construccion y utilizacion de un experimento electrornecanico de control,
compuesto por una bola de acero girando sobre una barra como pivote. Es un sistema
mecanico muy simple, que, sin embargo, resulta ser un ejercicio que arroja mucha luz sobre el
control realimentado.

BOOK REVIEW

Fundamental Electrical Technology: MARVIN H. KLAYTON

(Addison-Wesley, 1977 ,700 pp., £13.60)
This is a comprehensive volume covering 'a broad spectrum of topics'. The text was initially
developed to support a foundation course, at Ocean County College, New Jersey, leading to
more specialized courses in electronics, power, computer technology and other specific areas. Its
style and terminlogy is therefore probably more appealing to the American than the British
market.
The author states that M.K.S. units have been used in the text with exceptions where
common usage makes it appropriate to employ C.G.S. or the English system of units.
Unfortunately these exceptions appear far too frequently and today's students will be deterred
by such unfamiliar units as gauss, oersteds, kilolines/in? etc.
The book will make a useful addition to any technical library, but its cost will be pro-
hibitive in so far as its use as a students textbook is concerned.
J. W. DRAIN, Department ofElectrical and Electronic Engineering, Oldham College of
Technology

Downloaded from ije.sagepub.com at The University of Iowa Libraries on June 5, 2016