0% found this document useful (0 votes)
83 views278 pages

1 - Automatic Control Principles and Practice

The document is a comprehensive text on automatic control systems, authored by Werner G. Holzbock, aimed at providing practical understanding without delving into complex differential equations. It covers fundamental concepts, static and dynamic characteristics, hardware components, and various control strategies, emphasizing the importance of both static prerequisites and dynamic behavior. The book serves as a primer for students and a practical guide for engineers interested in control systems and their applications across various industries.

Uploaded by

Ashish Mahajan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
83 views278 pages

1 - Automatic Control Principles and Practice

The document is a comprehensive text on automatic control systems, authored by Werner G. Holzbock, aimed at providing practical understanding without delving into complex differential equations. It covers fundamental concepts, static and dynamic characteristics, hardware components, and various control strategies, emphasizing the importance of both static prerequisites and dynamic behavior. The book serves as a primer for students and a practical guide for engineers interested in control systems and their applications across various industries.

Uploaded by

Ashish Mahajan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 278

L.

c ]i74ba
Holzbock 1051646
somatic control:principles
i practice

-1-8 H748a
Holzbock 1061646
itoraatic control:principles
id practice

CARDS FROM POCKET


no NOT REMOVE

PUBLIC LIBRARY
ORT WAYNE AND ALLEN COUNTY, IND.

"** ' ■ r V ^L-N- ■

^ ' ■■ ■ A

' - 1
, "
OJ u Hill lllio —

ACPLITEM
DISCARDED
Digitized by the Internet Archive
in 2017

https://archive.org/details/autonnaticcontrolOOholz
V
1^

^9
t I. I

>» \
♦ k-
t
./
on

» I
4: . ,■ I
I
■f ‘ f

fr'ier

. V

. lf^
\MJ.
f

rjr
I f" .^,. •
A'A
AUTOMATIC CONTROL
PRINCIPLES AND PRACTICE

WERNER G. HOLZBOCK
Chief Engineer
Askania Regulator Company
Chicago, Illinois

REINHOLD PUBLISHING CORPORATION


NEW YORK
CHAPMAN & HALL, LTD., LONDON
Copyright 1958 by
REINHOLD PUBLISHING CORPORATION

All rights reserved

Library of Congress Catalog Card Number 58-9742

REINHOLD PUBLISHING CORPORATION


Publishers of Chemical Engineering Catalog, Chemical
Materials Catalog, “Automatic Control,” “Materials in
Design Engineering,” “Progressive Architecture”; Ad-
vertising Management of the American Chemical Society

Printed in U.S.A. by
THE GUINN CO., INC.
New York 1, N. Y.
PREFACE
1051646
This book tries to answer the need for a comprehensive text on auto-
matic control without entering into differential equations.
After clarifying some basic concepts and terms in the first chapter,
static characteristics are discussed in Chapter 2. It is hoped that sufficient
emphasis is given to static prerequisites which are frequently neglected
in concentrating on the dynamic aspects. This does not impair the
importance of dynamics and this subject is extensively treated in Chap-
ters 3 to 5.
After this introduction to operational behavior, the hardware of con-
trol systems is dealt with in Chapters 6 through 8. It is felt that the first
five chapters will prepare for the discussion on hardware and provide a
basis for the understanding of its decisive role.
Chapters 9 through 11 describe the elements of the control loop: the
measuring element, the controller itself, and the final control element.
Particular emphasis has been put on temperature control and control
valves, since both are elements whose behavior frequently limits the
potential benefit of automatic control.
The remaining two chapters are concerned with the use of controllers
in the system. Various controls—averaging, ratio, cascade, etc.—are
discussed and a number of systems from various industries are described
as examples.
This text is within the grasp of the intelligent technician. This does
not mean that it can be read without effort. The nature of the subject
material does not allow a facile text which can be digested by the reader
without exertion on his part.
The student interested in control theory will find this book a primer
which prepares him for more profound treatment of the subject in a
number of more theoretical books of excellent quality.
The engineer who wants to get a practical understanding of controllers
and control systems will appreciate this text which foregoes the refine-
ments of theoretical analysis and stresses physical realities.
The method of presentation which I have used in this book is the
IV PREFACE

result of many talks and publications. I have published articles in Air


Conditioning, Heating and Ventilating, Chemical Engineering, Control
Engineering, Regelungstechnik and other leading journals, and many a
thought which is presented in the following pages was first discussed in
these articles, as well as in talks to ISA and ASME groups.
WERNER G. HOLZBOCK
Evanston, Illinois
March 1958
CONTENTS

Chapter

PREFACE iii

1. THE AUTOMATIC CONTROL SYSTEM 1


The Control Loop 1
Time Lags 3
Speed of Corrective Action 6
Stability 7
2. STATIC CHARACTERISTICS 8
Accuracy 8
Calibration Errors 10
Resolution Sensitivity 10
Dead Band 11
Linearity 11
Hysteresis 12
Linearity and Hysteresis in the Control Systems 14
3. STEP FUNCTION RESPONSE OF A PROCESS 16
Methods and Results 16
Process without Self-regulation 18
Process with Self-regulation 19
Gain, Response Rate, and Time Constant 20
Various Definitions of Time Constant 21
Processes with Dead Time and Multiple Capacitances. 22

4. STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 25


Two-position Action 27
Single-speed Floating Action 33
Proportional-speed Floating Action 38
Proportional-position Action 43
Proportional plus Reset Action 51
Proportional plus Reset plus Rate Action 55
5. FREQUENCY RESPONSE 61
Sinusoidal Input 61
Phase Lag, Frequency, and Magnitude Ratio 63
Technique of Taking Frequency Response Data 63
VI CONTENTS

Chapter

Graphic Representation of Frequency Response Data. 65


Phase Lag and Phase Reversal 67
Stability Conditions 69
Dead Time as Instability Factor 70
Combination of Frequency Response Data 71
Block Diagram 71
Graphic Addition of Frequency Response Data 73
Unstable Components in Stable Loop 74
Proportional-speed Floating Controller 74
Proportional-position Controller 77
Proportional plus Reset Controller 79
Proportional plus Reset plus Rate Controller 82
Conclusions 86
6. MECHANICAL COMPONENTS 87
Springs 88
Diaphragms 92
Bellows 94
Flapper-nozzle 95
Pilot Valves 100
Jet Pipes 103
Four-way Valves 105
Cylinders Ill
7. ELECTRICAL COMPONENTS 112
Potentiometers 112
Potentiometers and Rheostats 114
Capacitors 115
Inductance Coils 116
Dilferential Transformer 117
Galvanometers 119
Oscillators 120
Phototubes... 121
Synchros 122
Electronic and Magnetic Amplifiers 126
Moving Coil 130
Servomotors 132
Torque Motors 134
8. BALANCES AND COMPUTING CIRCUITS 136
Force Balances with Deflection Systems 137
Force Balances with Null System 138
CONTENTS vii

Chapter page
The Null System in Pneumatic Circuits 140
Position Balances 144
Wheatstone Bridge 146
Computing with Differential Transformers 152
9. MEASURING ELEMENTS 154
Square-root Extraction 154
Temperature Control 159
10. CONTROLLERS 177
Pneumatic Controllers with Position Balance 177
Pneumatic Controllers with Eorce Balance 186
Hydraulic Controllers 192
Electric Controllers 196
Interaction 203

11. FINAL CONTROL ELEMENTS 204


Forces in a Spring-opposed Diaphragm Valve 205
Stroke Time of the Spring-opposed Diaphragm Valve. 205
Valve Positioners 206
Valve Characteristics 209
Rangeability 214
Process Gain 215
Conclusions on Valve Action in Control Loop 215
Other Control Valves and Actuators 216

12. CONTROL SYSTEMS 220


Averaging Control 223
Cascade Control 226
Ratio Control 231
2- and 3-Elements Control 234
Totalizing 237
Time Schedule Control 240

13. INDUSTRIAL APPLICATIONS 243


Steam Generation 243
Soaking Pit 245
Water Treatment 246
Air Conditioning 248
Solvent Recovery 248
Gasoline Production 250
INDEX 253
1. THE AUTOMATIC CONTROL SYSTEM

With few exceptions, and these will be pointed out, this text follows
the terminologies prepared by the American Society of Mechanical
Engineers and published in their Standards 105 and 107. Definitions
that are extracted from these standards are made with the permission
of the publisher.

The Control Loop


The elements of an automatic control system are combined in a closed
loop, as illustrated in Figure 1-1. The measuring means, automatic

REFERENCE
INPUT ACTUATING AUTOMATIC CONTROLLER ^ FINAL MANIPULATED
CONTROL
(SET POINT)' V J SIGNAL CONTROLLER OUTPUT SIGNAL
ELEMENT VARIABLE

PRIMARY MEASURING CONTROLLED


PROCESS
FEEDBACK* MEANS * VARIABLE

LOAD CHANGES

Figure 1-1. Automatic control system.

controller, and final control element are considered separate units. This
is in agreement with common usage but differs from the ASME termi-
nology, which combines all three units under the term automatic con-
troller and leaves no term to designate the unit which is generally known
as the controller proper.
The input to the automatic controller is the actuating signal which is
the difference at any time between the reference input and the primary
feedback. This difference is obtained in a component called the sum-
ming point which is symbolized by the circle in the illustration. The sum-
ming point is generally part of the controller.
The primary feedback is the magnitude of the controlled variable as
interpreted by the measuring means. Hence, the actuating signal can
also be defined as the deviation of the controlled variable from the set
point.
1
2 AUTOMATIC CONTROL (

The automatic control system includes the process. In a more specific


sense, however, the control system may be considered as the combination
of measuring means, controller, and final control element, disregarding
the process. Thus, in speaking of the hardware of a control system,
process elements like cracking towers, evaporators, etc., are usually
excluded.
The purpose of the automatic controller is to reduce the actuating
signal to a minimum, and preferably to eliminate it, with the least delay
possible. The means of effecting this is the controller output signal which
is transmitted to a final control element. This may be a valve, a pump
or any other device which directly changes the value of the manipulated
variable. The manipulated variable is a condition or characteristic of the
control agent. For example, when a final control element changes the
fuel-gas flow to a burner, the manipulated variable is flow, and the
control agent is fuel gas.
The controlled variable is a condition or characteristic of the con-
trolled medium. For example, where temperature of water in a tank is
automatically controlled, the controlled variable is temperature and the
controlled medium, such as heat, is utihzed or transformed by the primary
element to produce an effect which is a function of change in the value
of the controlled variable. The effect produced by the primary element
may be a change of pressure, force, position, electrical potential, or
resistance. The primary element is a portion of the measuring means
which consists of those elements involved in ascertaining and communi-
cating to the summing point of the controller the value of the controlled
variable. The signal which the measuring means transmits to the sum-
ming point is the primary feedback. It is compared with the reference
input signal which in turn is determined by the set point.
The set point is the position to which the control-point setting
mechanism is fixed. Where the automatic controller possesses a set-point
scale, the set point is the scale reading translated into units of the con-
trolled variable. Where a setting scale is not provided, the set point is
the position of the control-point setting mechanism translated into units
of the controlled variable. The set point may be varied manually or
automatically, as in time-schedule or ratio control systems.
In some types of automatic controllers—for example, those with two-
position differential-gap action or floating action with neutral zone—
the set point is related to the position of a range of values of the con-
trolled variable. The set point is often selected as the center of this range
of values.
A permanent deviation of the controlled variable from the set point
is called offset. It is convenient to speak of the value for which the
THE AUTOMATIC CONTROL SYSTEM 3

controller is adjusted as set point and of the value which the controller
maintains as control point. The control point is the set point plus or
minus offset.
The disturbances, called load changes, entering the process represent
the variable that requires adjustments by the controller. Without load
changes the control system could just as well be manual, since once it
is set no changes will take place.
Load changes have various causes. Figure 1-2 shows a reboiler into
which the process fluid enters as condensate and is brought to a boil by

Figure 1—2. Reboiler.

a steam heat exchanger. The process fluid leaves the reboiler as vapor.
A level controller (not indicated) maintains the level. The vapor tempera-
ture is the controlled variable and steam flow is the manipulated varia-
ble. The load changes of this process are affected by (a) temperature of
entering process fluid, (b) steam temperature, (c) steam pressure as re-
flected in rate of flow change for a given valve opening, and in condensa-
tion temperature, and (d) outside temperature, draft cooling, etc. In
addition to these load changes, there are other long-range but not less
important changes like scahng of tubes, deterioration of heat insulation,
etc.

Time Lags
If actions and reactions in a control system were to take place without
loss of time, no control problem would exist. For example, the steam
production of a boiler operating under such ideal conditions would
4 AUTOMATIC CONTROL

increase upon the slightest drop in steam pressure and balanced condi-
tions would be restored immediately and without difficulty.
These balanced conditions require that the input of steam-producing
elements correspond to the steam output as required by the load. Prac-
tical circumstances produce a time interval between the output change,
the necessary corrections, and the taking effect of these corrections.
The first delay occurs due to the fact that a load change is not sensed
immediately, but after it results in a change of the controlled variable,
i.e., steam pressure in the above example. This means that the control
system which is intended to keep the controlled variable constant is built
on the premise that the controlled variable changes before corrective
action takes place. This is an unavoidable contradiction inherent in all
automatic control systems. The art is to keep the change to a minimum.
If, for example, a controlled variable is to be kept within ± 1 per cent
of its set point value, it is necessary that a much smaller change than
this suffices to initiate correetive action.
The next delay occurs in converting the controller output signal into
a change of the magnitude of the manipulated variable by means of the
final control element.
Frequently, the most serious delay is due to the time lost in the process
before the controlled variable responds. Two main causes produce this
delay: resistances and capacitances.
Resistance is opposition to flow. It is expressed in units of potential
change required to produce a unit change in flow rate. For example, the
flow of heat from the steam to the water through the tubes of a heat
exchanger is delayed because of the thermal resistance caused by scales,
gaseous Aims, etc.
Thermal resistance is the temperature differential in degrees Fahren-
heit* required to produce a heat flow rate of one Btu (British thermal
unit) per hour through a given body.
Fluid resistance is the psi fluid pressure differential required to produce
a fluid flow rate of 1 cu ft/min. through a given passage.
Electrical resistance is the voltage necessary to produce a current of
one ampere (one coulomb per seeond) through a conductor.
Capacitance is the change in quantity contained per unit of change
in a reference variable; for example, the speed with which the liquid
level in a tank changes depends on the change of volume of stored liquid
per unit change of head. If a tank level drops one foot when 20 cu ft of
hquid are removed, then its capacitance is 20 cu ft/ft or 20 sq ft which

*Uiiits in the following definitions for resistances and capacitances are those most commonly
used in practice.
THE AUTOMATIC CONTROL SYSTEM 5

is equivalent to the area of the liquid surface. If the shape of the tank
causes the liquid surface area to vary with change of head, the capaci-
tance will likewise vary with head.
In controlling the pressure in a gas-filled tank, the capacitance may
be expressed as a function of the weight of gas. In this case the capaci-
tance is the change of weight of stored gas per unit change of pressure.
The most typical capacitance is probably the flywheel. All capaci-
tances act like flywheels which are storing up energy, reducing rapid
response to a change.
Thermal capacitance is the Btu’s absorbed by a body per degree
Fahrenheit rise in its temperature.
Volume capacitance is the cubic feet of solids or liquids that can be
stored in a container per foot of increase in level. In the case of gases
it is standard cubic feet*/psi of pressure change.
Weight capacitance is the pounds of solids or liquids that can be stored
in a container per foot of increase in level. In the case of gases it is
standard cubic feet under standard conditions/psi of pressure change.
Electrical capacitance is the change of electrical charge of a capacitor
expressed in microfarad/volt across its terminals.
Capacitances produce a delay which is expressed in the time constant
and which will be discussed in greater detail further below. In addition
to the time constant, a delay exists between two related actions in a
control loop; this is called the dead time.
Figure 1-3 illustrates a pH control system. The control valve regu-
lates a chemical which maintains the pH of the main flow. Before
corrective action of the control valve reaches the pH measuring means.

* Standard cubic feet refers to the volume of the gas at a pressure of 29.921 in. Hg (760 mm
Hg) and a temperature of 59°F (15°C).
6 AUTOMATIC CONTROL

some finite time elapses which is a function of the length of travel. The
time thus lost is dead time and plays an important part in the stability
of the system, as will be shown. It is paramount in any control system
that dead time be kept as low as possible.

Speed of Corrective Action


If the corrective action in a system with time lags would be just enough
to bring the system back to its control point after a load change, the
action would be too slow, the controlled variable would continue to
change in the meantime, and close control would be impossible in the
majority of cases.
On the other hand, intensifying the corrective action after a load
change—for example, admitting more fuel to a process than is actually
necessary, in order to gain faster action—produces overshooting. When
the controlled variable, e.g., pressure, is back at its control point, the
excessive fuel admission results in overcorrection. This changes the pres-
sure in a direction opposite to its initial deviation, and the resulting
control action will now produce undershooting. Oscillations around the
control point are the result.

Figure 1-4. Various process responses after a change in set point.


THE AUTOMATIC CONTROL SYSTEM 7

The entire problem of automatic control centers around maintaining


a control system within acceptable stability conditions while obtaining
the greatest possible speed in corrective action.

Stability
Figure 1-4 shows various forms of responses of a controlled variable
to a sudden change in set point of the controller. The controlled variable
may be pressure, level head, temperature, etc. The shape of these response
curves expresses the stability of the system. In curve a, the new value is
approached without overshoot; however, considerable time elapses until
reaching the new value. In curve b, the response is considerably faster.
There are a few rapidly subsiding oscillations. They are usually accept-
able and this form of response is generally considered the most desirable.
Curve c shows a response of almost continuous oscillations. In general,
such a response is not acceptable. Occasionally, however, such oscilla-
tions, provided they are within limited amplitudes, can be tolerated. In
such cases discontinuous controllers, described in a later chapter, would
probably give a satisfactory and economical solution.

s
2. STATIC CHARACTERISTICS

Accuracy, resolution sensitivity, dead band, hysteresis and linearity


are static characteristics of the components of an automatic control
system which determine its operational quality. Although these charac-
teristics must meet the requirements of the case, unnecessary refinements
must be avoided for reasons of economy. A thorough knowledge of
these static characteristics is necessary for realistic appraisal.
One of the difficulties in discussing static characteristics is the over-
lapping of terms. Accuracy includes scale and calibration errors as well
as the results of resolution sensitivity, dead band, hysteresis, and linearity.
Resolution sensitivity is due to static friction, frequently called break-
loose force, or to discontinuities such as the windings of a potenti-
ometer. Dead band includes static friction and discontinuities, as well as
lost motion. Hysteresis is, strictly speaking, only electrical or mechanical,
and is a material property that will be described later. However, since
in control system components hysteresis can generally not be separated
from dead band and nonlinearity, it may occasionally be convenient to
consider hysteresis as including hysteresis proper, dead band, and
nonlinearity.

Accuracy
Figure 2-1 is the block diagram of an active component in a control
system. Input and output are related by a mathematical function which
expresses the correct value.

INPUT OUTPUT

Figure 2-1. Component in control system.

For example, the component of Figure 2-1 may be a pneumatic


controller which includes the control-point setting mechanism. The input
may be a temperature, and the output, a pneumatic signal. At 60°F, the
controller output signal is 3 psi, and at 120°F, it is 15 psi. The mathe-
matical function that applies for the range from 60° to 120°F is
P = 02T - 9
8
STATIC CHARACTERISTICS 9

where P is the correct value of the controller output signal and T is the
input temperature.
The output value differs from the correct value because of inherent
imperfections of the controller. The degree to which the output value
approaches the correct value is the accuracy.
Accuracy may be expressed in units of the input, in per cent of the
output range, in per cent of the value at which the accuracy was deter-
mined, etc. A per cent specification which is not related to some specific
magnitude is meaningless.
In the case of the aforementioned temperature controller with an
output of 3 to 15 psi, the accuracy may be stated as 1 per cent of range.
If the controller has a range of 100°F, the output signal differs from the
correct value by not more than ±0.2 psi. This is derived by considering
that 1 per cent of 100° is equal to 1°; since the output signal changes
12 psi for 60° change, the equivalent of 1 degree is 0.2 psi. This is the
pressure by which the output may be above or below the correct value.
The specification of accuracy by a number leaves open several ques-
tions which relate mostly to the conditions under which the accuracy
was determined. For example, temperature and relative humidity may
have considerable influence on the accuracy. A pneumatic controller
can show inaccuracies with changes in air temperature, which may not
be reflected in an accuracy statement. Acceptable changes in supply volt-
age, line frequency, supply air pressure, and hydraulic pressures may
have to be specified before an accuracy rating becomes meaningful.
A Bourdon tube pressure gauge, for example, usually is accurate
within ±0.5 per cent of its range. This gauge responds to the difference
between process pressure and atmospheric pressure. In fact, it is an
instrument that measures differential pressure. If it is calibrated for 15
psig (Ib/sq in. gauge pressure), it really measures about 30 psia (Ib/sq
in. absolute pressure) minus the atmospheric pressure. Let this gauge be
calibrated for an atmospheric pressure of 15 psia. Since this pressure
may vary between ±5 per cent, it follows that the gauge pressure has
an accuracy of ±2.5 per cent of its range due to atmospheric pressure
conditions alone. To this must be added a temperature error which is
usually in the order of plus 2 per cent for a temperature rise of 100°F.
Accuracy specifications are further left open if the accuracy figure is
the result of a number of readings of which it is the average, and if it
refers to any value of the input or is only attainable for certain values.
In order to give a fair indication of accuracy, the worst reading over the
whole range should be the one reported. If this is not the case, then the
conditions to which the accuracy refers should be stated. This, of course,
is equally true of other specifications like dead band, resolution
sensitivity, etc.
10 AUTOMATIC CONTROL

Calibration Errors
Calibration errors belong to one or both of two categories: zero errors
and angular errors. The zero error is a linear shifting of the range. For
example, the output may read two units too high at any point through-
out its range. This is a zero error. The adjustment consists in linearly
shifting the range to obtain correct zero reading. The angular error
shows correct reading at an arbitrary point of the scale, e.g., the zero
point, and an error which increases in proportion to the distance from
this point. This requires adjusting the ratio of the output-input relation-
ship which is usually accomplished by changing a lever ratio in a
mechanical linkage or the gain in an electric circuit to increase or
decrease the output signal range for a given input range. Zero errors
and angular errors are frequently combined in the overall calibration
error, and require separate adjustments.

Resolution Sensitivity
Resolution sensitivity is the minimum change in the measured vari-
able which produces an effective response of the component in a control
system. Suppose the resolution sensitivity of a spring-opposed diaphragm
control valve is to be determined. The input is an air-pressure signal.
The output is stem motion. The air pressure is raised to an arbitrary
value and the stem is allowed to come to rest. Next, the air pressure is
further increased, but now by very small increments in order to deter-
mine the minimum increment required to start the stem moving. Let
this minimum increment be 0.012 psi. For a pressure range of 3 to 15
psi, this means a resolution sensitivity of 0.1 per cent.
Resolution sensitivity in this case expresses the force that is required
to induce a physical system to produce motion. It is due to static fric-
tion, sometimes called stiction.
Another cause of limited resolution sensitivity is due to a built-in
stepping arrangement, such as a wire wound rheostat. The resistance
can be changed only in discrete steps since one winding after another
is shunted out and each winding is of some finite resistance. The resist-
ance in one winding corresponds thus to the resolution sensitivity of
the rheostat.
In the case of the rheostat, static friction is probably also present and
the resolution sensitivity is caused by both static friction and winding
resistance. In no case is lost motion included in resolution sensitivity as
it is in dead band. Hence from the viewpoint of actual operating condi-
tions, the significance of resolution sensitivity becomes questionable.
The input signal may either increase or decrease and the essential ques-
STATIC CHARACTERISTICS ] ]

tion is how much the input can change without producing a response in
the output. However, this is dead band and not resolution sensitivity.
In the test for the latter, the signal is changed in one direction only and
is not reversed as in actual operation. The specification of resolution
sensitivity is of interest only if it is desirable to separate lost motion in
the analysis.

Dead Band
The range of values through which the input can be varied without
initiating output response is defined as dead band. This is not only the
limiting factor for accuracy in a component, but is also the most impor-
tant operational quality criterion of static characteristics for an auto-
matic control system. If a flow controller with a range of 100 gpm has
a dead band of zb 1 per cent of range, this means that if it is to control
at 60 gpm, the actual flow may vary between 59 and 61 gpm without
producing action.
If the flow has been at 59 gpm and now begins to increase, no action
may take place until the flow is 61 gpm. This means not only that the
controlled variable is able to change through 2 gpm without correction
taking place, but also that if the flow continues to increase, the control-
hng action is delayed. This is equivalent to dead time and contributes to
the instability of the control loop, as will be discussed in the following
chapter.
The cause of dead band is static friction and discontinuities, i.e., reso-
lution sensitivity plus lost motion. The most common cause of lost
motion is probably backlash, which refers to the looseness of mechanical
linkages, gears, etc. There are, however, other causes of lost motion,
such as the overlap between spool lands and ports in valves, as will be
discussed in the chapter on Hydraulic Control Elements.

Linearity
Linearity can be expressed in a number of ways as shown in the
following;
Normal or Independent Linearity. This is illustrated in Figure 2-2
and expresses the maximum deviation of the output values (full line)
from the best straight fine (dashed line) that can be drawn through these
values. Considering a scale of ten units, the maximum deviation in
Figure 2-2 is about one unit or ten per cent.
Zero-based Linearity. This is illustrated in Figure 2-3. The dashed
line to which the linearity is referred passes through zero. In the illus-
tration, the maximum deviation from this line is about 1.3 units or 13
per cent.
12 AUTOMATIC CONTROL

10

0 0 ^
01 23456789 10 0123456789 10
OUTPUT VALUE
OUTPUT VALUE

Figure 2-2. Normal linearity. Figure 2-3. Zero-based linearity.

Absolute Linearity. Neither the normal nor the zero linearity


expresses the accuracy of the output value as compared to the correct
value. The resulting deviation from linearity is considerably larger and
is shown in Figure 2-4. The dashed reference line is the idealized output
which corresponds with the correct value. The greatest deviation is now
about 1.6 units or 16 per cent.

0
0 123456789 10
OUTPUT VALUE

Figure 2-4. Absolute linearity.

In no case does linearity necessarily include backlash or hysteresis. It


consists of values obtained by changing the input signal in one direction
only. The causes of nonlinearity are general mechanical or electrical
imperfections.

Hysteresis
It was stated before that the hysteresis curve of a control component
generally can not be isolated from the effects of dead band and linearity,
STATIC CHARACTERISTICS 13

and that it may therefore be convenient to consider a hysteresis curve


as an agglomeration of these characteristics. Hysteresis by itself, however,
is part of the inherent physical characteristics of materials used. It is
distinguished by this from all other characteristics here discussed which
are based on the interrelation of the elements in a component. (The

Figure 2-5. Magnetic hysteresis.

jumps in discrete quanta as known in physical science are ignored in


this connection as being irrelevant in the practical behavior of control
systems.)
Hysteresis may be magnetic or mechanical, as shown in Figures 2-5
and 2-6, respectively. Both graphs show output-input relations of a

Figure 2-6. Mechanical hysteresis.


14 AUTOMATIC CONTROL

physical specimen. In the case of Figure 2-5, the input is magnetic


intensity H produced by and proportional to the current in the winding.
The output or result of the magnetic intensity is the magnetic flux B.
In Figure 2-6, the input is the stress F, a force applied to the elastic
specimen, which results in a strain or deformation S, equivalent to the
output.
In both cases, the output-input relation depends on the previous history
of the specimen. Initially, they are both at zero, but once the input signal
has been apphed, there remains a memory of this signal in the specimen,
called the retentivity or the remanence. The magnitude of this remanence
depends on the amplitude of the input signal. If the input signal is
cycled through smaller and smaller cycles, the hysteresis becomes less
and less noticeable. It is therefore necessary to specify the amplitude
of the input signal when giving hysteresis values.
Occasionally, dead band and hysteresis are used as equivalents. Dead
band, however, refers to a change of input without output response,
under which conditions hysteresis does not appear at all. The two terms
relate to entirely different characteristics.

Linearity and Hysteresis in the Control Systems


Wherever a reading of the controlled variable is required, both linearity
and hysteresis in the measuring means affect directly the accuracy of the
reading and are important criteria. The controlling function, however,
can tolerate considerable nonlinearity and hysteresis in any part of the
loop without deterioration of performance. Figure 2-7 shows a simphfi-

-SET POINT

CONTROLLER

. LOAD CHANGES

PROCESS

Figure 2-7. Basic control system.

cation of Figure 1-1 in which the control system is reduced to a con-


troller and a process. In this case, the controller is supposed to
incorporate both measuring means and final eontrol element. The con-
troller acts upon the process, and the process acts upon the controller. A
STATIC CHARACTERISTICS 15

change of input from the process to the controller results in a modifi-


cation of output from the controller. This produces a correction in the
process and a resultant modification of the input to the controller com-
ing from the process. If these steps were to be immediate ones and happen
in the indicated succession only once, the linearity and hysteresis of the
controller would be very important. The physical reahty however, is that
these changes occur gradually, though they may be fast, and because of
the interactions in the closed loop, a self-correcting system is set up in
which many nonlinearities can be absorbed.
A typical example is the industrial control valve which will be dis-
cussed in a separate chapter. Some pertinent facts are anticipated here.
The control valve is part of the controller in Figure 2-7 and adjusts the
manipulated variable, i.e. flow, which is part of the process. The rela-
tion between change of the manipulated variable and response of the
controlled variable is a characteristic of the process and generally not
subject to alteration. The relation is not a linear one and, furthermore,
depends on the nature of load changes. In order to compensate at least
partially for the nonhnearities in the process, the valve plug is generally
characterized; i.e., the flow versus stem travel characteristics follow a
certain pattern. This also is an approximation since these characteristics
deviate largely from any ideal curve. Thus, lack of accuracy exists be-
cause of non-ideal valve characteristics as well as because of non-ideal
and non-constant process characteristics. It is quite obvious that under
such conditions it is not realistic to look for excessive linearity and free-
dom from hysteresis in valve positioning or in any other control system
component.
Nonlinearities, however, change the ratio between corrective action
and deviation of the controlled variable. This is equivalent to changes in
the sensitivity of the controller. Conceivably, nonlinearities may become
so high that at certain valve positions, the system becomes unstable or
corrective action becomes too slow.
Characteristics have been described in this chapter that refer to the
output-input relations of components in control systems. The conditions
investigated were those that exist after the system has undergone a
change of input and has returned to the balanced steady-state condition.
No attempt has been made to describe the behavior while the change is
taking place. This will be discussed in the following chapters.
3 STEP FUNCTION RESPONSE
OF A PROCESS

To determine the characteristic dynamic behavior of a process or con-


trol component, an input is artificially applied, and the output response
that follows is observed. For the purposes of comparison, the input sig-
nal that is applied must follow a defined pattern. In Figure 3-1, three
typical patterns of input signals are shown: The step function, the ramp
function, and the sinusoidal function.

Figure 3-1. Patterns of input signals.

The ramp function, which corresponds to a gradually changing input


signal, is hardly used today in control system analysis, though it may
give interesting information regarding the response of an element.,The
analysis of a proportional plus reset plus rate controller in Chapter 4
does make use of it, however. The preferred approaches are step func-
tion and sinusoidal (i.e., frequency) methods which will be discussed in
this and the next two chapters.

Methods and Results


The equipment necessary for determining the step function response
of a process includes a means for recording the controlled variable and
16
STEP FUNCTION RESPONSE OF A PROCESS 17

a provision for manual loading of the final control element, together


with an automatic-to-manual transfer switch. If the test is made under
operating conditions, load changes must be eliminated for the duration
of the test and a limited change of the set point during the test must be
permissible.
An arrangement is illustrated in Figure 3-2, in which the controlled
variable is pressure. The controller is disconnected and the final control

MANIPULATED

CONTROLLED
VARIABLE

Figure 3-2. Arrangement of step-function response.

element is controlled manually. A sudden manual change of the air pres-


sure applied to the final control element results in a change of the proc-
ess pressure which is recorded and called the step function response.
The step change of the valve position should be as small as possible.
The reason is that the interpretation of the step function response is based
on the assumption of linear responses. Generally, this holds true only
for sufficiently small step changes. Furthermore, the nonlinearity over
the operating range may be so considerable, that the response for a valve
which is, e.g., VA open is entirely different from the response for the
same valve % open. Hence, the step change should be repeated at
various valve positions if the conditions of the installation so allow.
The step change varies the energy which enters the process through
the manipulated variable. This energy change shows up in the process
as variations in potential and kinetic energy, but also as a change
of energy losses which is inherent in the majority of processes. The step
function response may be considered a record of the method by which
additions and losses of energy in a process take place.
18 AUTOMATIC CONTROL

It is possible to read from the step function response of a process its


three basic characteristics;
1. The dead time.
2. The time constant as a result of resistances and capacitances.
3. The process gain as discussed further below.
With these data it can be determined what control action is best suited
and for what values the various adjustments of the controller should be
set.
Two basic forms of step function responses are illustrated in Figure
3-3. These are highly idealized patterns but they do contain the main
components of practical responses. Figure 3-3a represents the step input,

a - STEP INPUT

b - RESPONSE OF
PROCESS WITHOUT
SELF-REGULATION

c- RESPONSE OF
PROCESS WITH
SELF-REGULATION

Figure 3-3. Patterns of process response.

i.e., the sudden manual change of the signal which under normal operat-
ing conditions is the controller output signal. Figures 3-3b and 3-3c are
two basically different forms of responses to the step input as recorded
from the controlled variable.

Process without Self-regulation


The response illustrated in Figure 3-3b is typical of a process which
stores up potential energy at a constant rate without altering the energy
losses that may occur as an inherent characteristic. Figure 3-4 is an ex-
ample showing a tank with a liquid level control system. A pump main-
tains the output flow constant and a valve controls the inflow. Suppose
the level is initially at equilibrium. A step change at the valve increases
STEP FUNCTION RESPONSE OF A PROCESS 19

Figure 3-4. Process without self-regulation.

the input and the level rises. Because the output remains constant, the
level continues to rise and a new balance is never reached. A process
that responds in this form is said to have no self-regulation.
The rate of response is expressed by the ratio of the response to the
time in which this takes place. Since C is the magnitude of the controlled
variable, a change by a certain amount, e.g., from Q to Cg, is conveniently
expressed by c. The time interval between and which is required
to produce the increment c is expressed by t. The rate of response for a
process without self-regulation is therefore expressed by

Process with Self-regulation


The response in Figure 3-3c differs from that seen in 3-3b in that it gradu-
ally approaches a new value of C. Such a response would correspond with
the arrangement of Figure 3-5. Liquid enters the tank at a constant flow

Figure 3-5. Process with self-regulation.


20 AUTOMATIC CONTROL

rate. The rate at which the liquid leaves the tank depends on the valve
opening and the level of the liquid. Assume that Figure 3-3c represents
the change in level seen in Figure 3-5 after a step change in valve
position, introduced by manual means, closes the valve slightly. The in-
flow now exceeds the outflow and the level in the tank rises; but as the
level rises, the hydraulic head increases also. The increase in head
causes the outflow to increase. Eventually, at this new valve position, the
inflow again balances the outflow. Thus a new balance has been reached
even without the benefit of an automatic controller, although the offset
may be considerable. The ability to reach a balance in this manner is
known as self-regulation.
Another example is a temperature-controlled process. An increase of
heat input is balanced by an increased heat loss to the surrounding at-
mosphere. Similar self-regulating conditions exist in the majority of in-
dustrial processes.

Gain, Response Rate, and Time Constant


The response curve of the self-regulated process in Figure 3-3c is de-
termined by two conditions: the process gain and its response rate.
The gain is the total change in magnitude of the controlled variable
per unit corrective action of the final control element. Suppose a step in-
put applied to the final control element moves it by 0.1 in. and the pres-
sure changes thereupon gradually until it settles at 20 psi above the
initial value. The process gain in this case is 20/0.1 = 200 psi/in. valve
motion. As an algebraic expression this reads

where g is the process gain, c is the increment of the process variable for
a given step change of the final control element, and u is the magnitude
of the step change.
The response rate is the maximum change of the controlled variable
per unit time per unit corrective action of the final control element. Fig-
ure 3-6 illustrates the response of the same process for various magni-
tudes of step inputs. The maximum change per unit time is at the
beginning of the response in all three cases. The speed is expressed by a
line drawn tangent to the initial, i.e., maximum, slope of the curve. The
steeper the tangent line, the faster the response. Suppose the step input
moving the final control element by 0.1 in. results in a response with
maximum change of 50 psi per minute of the controlled variable. The
response rate in this case is 50/0.1 = 500 psi/min./in. valve motion.
Designating the process gain by g and the rate of response by v, the
STEP FUNCTION RESPONSE OF A PROCESS 21

Figure 3-6. Step function responses with various magnitudes of step inputs.

two are combined in what is commonly called the time constant by


writing

(3-1)

Inserting the numbers of the examples for process gain and response
rate, the result is a time constant of
, _200
= 0.4 min.
^ 500
Q
Since it was shown that g = —, equation (3-1) may also be written:
lA'

Returning for a moment to a process without self-regulation, it is quite


evident that its gain is infinite, hence its time constant is also infinite.

Various Definitions of Time Constant


The time constant is defined in many different ways. Its magnitude is
always the same. Thus, one definition of the time constant is that it ex-
presses the time which the controlled variable needs to pass through 63.2
per cent of its total change. This is illustrated in Figure 3-6. No matter
how big is the step input for a given process, the controlled variable al-
ways reaches 63.2 per cent of its final value within the same amount of
time which is given by the time constant i, .
22 AUTOMATIC CONTROL

The tangents which are drawn in Figure 3-6 to the respective maxi-
mum speeds of response show that the ratio of the total change of con-
trolled variable to is equal to the initial speed of response. It is only
necessary to multiply this ratio by the magnitude of the step input u, to
obtain the response rate of the process. Thus

(3-2)

and similarly
(3-3)

where u is used with suffixes 3 and 2 to indicate that different magni-


tudes of step change are involved.
However, the total change of the controlled variable divided by the
magnitude of the step change is equal to the gain. Hence
£3
1/3

Substituting this in either expression (3-2) or (3-3) gives

^ = V or
V

which is the same expression for the time constant as equation (3-1).
From Figure 3-6 it is also obvious that the time constant can be
obtained from a line which is drawn tangent to the initial speed of
response. The time at which this line reaches the final value of the con-
trolled variable is the time constant.

Processes with Dead Time and Multiple Capacitances


Figure 3-7 compares a step function input signal with a process
response in which a time interval elapses between input and the
beginning of the response. This time interval is the dead time of the
process.
Dead time is frequently used as an approximation in cases of multiple
capacitance processes. It was mentioned in Chapter 1 that the time
constant is the result of resistances and capacitances. There exists prac-
tically always a number of various resistances and capacitances in an
industrial process. Occasionally one capacitance is considerably larger
than the others, and for all practical purposes it is then justified to speak
of a single-capacitance process. The response of single-capacitance
processes corresponds to Figures 3-3 and 3-7. More frequently, however,
two or three, and occasionally even more capacitances combine. The
STEP FUNCTION RESPONSE OF A PROCESS 23

t
u
a-STEP INPUT

b-RESPONSE OF
PROCESS WITH
DEAD TIME

Figure 3-7. Step function response of process with dead time.

result is a composite ^'-shaped curve as shown in Figure 3-8. The sharp


distinction between time constant and dead time is lost. The concept
of these two time components, however, is so important for the evalua-
tion of control systems that an approximation is used which proves to
be sufficiently accurate. This approximation, together with some methods
which are used in this text in slightly modified form to determine
adjustments of proportional-position controllers, was first used by J. G.
Ziegler and N. B. Nichols in their classic papers on “Optimum Settings
for Automatic Controllers” and “Process Lags in Automatic Control
Circuits,” published in the ASME Transactions in 1942 and 1943,
respectively.

f
a-STEP INPUT ^

b-RESPONSE OF
t
C
MULTI-CAPACITANCE
PROCESS

Figure 3-8. Response of multi-capacitance process.


24 AUTOMATIC CONTROL

It was mentioned that a line drawn tangent to the initial speed of


response in Figure 3-6 intersects the final value of the process response
after time The initial speed of response in this case was equivalent
to the maximum speed of response. In the iS-shaped curve of Figure
3-8, the maximum slope or speed of response is in the bend of the S,
i.e., at the point of inflection. A line is drawn tangent to the response
curve at the point of inflection. The points of intersection with the
abscissas of the minimum and maximum values of the response curve
are marked a and b, respectively. The time between the two points is
considered the time constant t-^. The time between the step input and
the first of the two points is the dead time These are the approxima-
tions which have proven to be very successful in their practical usage
and will be used throughout this text.
Dead time is thus caused either by a transportation lag (see Figure
1-3) or by the additive effect of multiple capacitances.
4 . STEP FUNCTION RESPONSE AND
ADJUSTMENTS OF CONTROLLER

In order to utilize the data on process characteristics obtained by the


methods described in the previous chapter, it is necessary to analyze
the controller actions that may be combined with these processes. The
most commonly used controller actions are;
1. Two-position action
2. Single-speed floating action
3. Proportional-speed floating action
4. Proportional-position action
5. Proportional plus reset action
6. Proportional plus reset plus rate action
In order to obtain a step function response from any controller, it is
necessary to disconnect it from the final control element. In case of an
electronic controller, an ammeter is connected in series with the specified
load resistance across the controller output signal. For hydraulic control-
lers, a cylinder and piston of negligible inertia and friction replaces the
final control element. The piston motion is picked up by a differential
transformer, potentiometer, or similar device, to convert it into a signal
for suitable recording. In pneumatic controllers, the controller output
signal is applied to a pressure recorder. The medium measured by the
primary element must be kept at constant conditions during the test.
The step input is produced by a sudden small adjustment of the control-
point-setting mechanism.
Figure 4-1 compares the arrangement for the step function test of the
process with that of the controller. In Figure 4-la the closed loop is
shown connected for normal operating conditions. In Figure 4-lb, the
loop is opened and between the step input and step function recorder
are the final control element, process and measuring means. This means
that this test does not investigate the process by itself but in combina-
tions with the final control element and measuring means.
In Figure 4-lc, only the controller is inserted between step input and
step function recorder. The two tests together comprise all the compo-
nents of the loop. If the measuring means or final control element were
combined with the controller instead of with the process, a pure approach
25
26 AUTOMATIC CONTROL

O-CLOSED LOOP

b-STEP FUNCTION RESPONSE OF PROCESS

C-STEP FUNCTION RESPONSE OF CONTROLLER

Figure 4-1. Step function tests.

typical of the controller action could not be obtained. This is the par-
ticular reason for the arrangements as shown.
In discussing the step function responses for various controller actions,
two successive input steps as shown in Figure 4-2 will be considered.
At time the first step is applied to the initial input. The second step
follows at time Tg. This second step increases the input above the initial
input by twice the magnitude of the first step change. An exception are
two-position and single-speed floating controllers, for which the step
inputs are inherently of the same absolute magnitude.

Figure 4-2. Step inputs for controller tests.


STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 27

Two-position Action
Processes without Dead Time. Figure 4-3 illustrates a liquid-level
control system. A solenoid valve controls the flow into the tank. The
controller includes two metal rods m and n, which form part of an electric

FINAL
CONTROLLER CONTROL

Figure 4-3. Liquid-level control system.

circuit that is closed through the ground connections as indicated. The


liquid itself and the tank walls are electric conductors. In the condition
shown, the level will rise until it reaches metal rod m. At this point an
electric relay is energized and locked in, shutting down the solenoid
valve. The valve remains closed until the level drops far enough to open
the circuit through metal rod n. This unlocks the relay and another
relay is energized and locked to maintain the valve open until the level
rises again to m. There are only two input signals possible, either calling
for flow or not calling for flow. The output, i.e., the tank level, is simi-
larly limited to two alternative positions.
The process illustrated in Figure 4-3 has self-regulation since the flow
out of the tank increases with the height of the level due to head pres-
sure. With the same rate of hquid flowing in, the speed of response
becomes slower as the level rises. The process response is hence com-
parable to Figure 3-3c.
Let the process be started up with an empty tank. The solenoid valve
is fully open and the level rises along the process curve a illustrated in
Figure 4-4. On reaching level Ci, the solenoid valve closes. The level
now begins to drop along the inversed process response curve b. On
28 AUTOMATIC CONTROL

Figure 4-4. Process response for two-position control system


without dead time.

reaching level the valve opens again. The difference between and
C2 corresponds to the difference in lengths between electrodes m and n
in Figure 4-3 and is called the differential gap of the controller.
The smaller the differential gap of the controller, the shorter is the
period t in Figure 4-4, and the greater is the number of times per
minute of opening and closing the solenoid; that is, the frequency of its
operation. How large this frequency can be is largely a question of life
expectancy of the solenoid valve, contacts, and associate equipment.
The frequency can be reduced for a given process by widening the
differential gap. This means, however, that the level in the tank will
oscillate between wider limits.
Another factor which determines the period t is the process response.
A process with a large time constant, that is, a slow-responding process,
will have a longer period t; i.e., the slower the process, the smaller can
be the differential gap, and hence the closer are the limits within which
a process can be controlled, for the same frequency of solenoid valve
operation.
Since response curves a and b express the change of level for a fully
open and closed valve, respectively, it can be deduced that the response
would be slowed down by limiting the valve stroke to two extreme posi-
tions, e.g., V4 open and % open. This reduces the stroke to one half of full
stroke and consequently reduces the flow variations caused by the control-
ling action. High-low or bypass adjustments are usually available with
solenoid valves for this purpose. They allow the setting of any limits
between which the valve will operate and thus make it possible to
operate within a closer differential gap without changing the frequency
of operation. The limitations of this method are given by the magnitude
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 29

of load change. The maximum consumption of liquid from the tank,


for example, should not exceed the flow capacity at maximum opening,
while the minimum consumption should not be less than the flow
capacity at minimum valve opening.
The period t also depends on the set point of the controller. For
example, controlling between levels C\ and , Figure 4-4, results in a
longer period than between levels and C^. In general, the longest
period can be obtained by controlling toward either the upper or the
lower level of the tank.
Processes with Dead Time. The conclusions for the two-position
control system without dead time must be modified when dead time is
present. As has been shown in the previous chapter, the concept of dead
time also applies where the process consists of several capacitances con-
nected by resistances. Such a process is illustrated in Figure 4-5. This

Figure 4-5. Two-position temperature controller.

is a temperature-controlled process. The temperature controller has


electrical contacts like any thermostat that break when the temperature
reaches a certain predetermined point. If the temperature drops by more
than the differential gap, then the contacts connect again.
These contacts open and close the electric circuit that energizes the
heater element, and thereby regulate the heating of the process. The
electric current increases the temperature of the heater element which
represents the first capacitance. The heat then flows through the resist-
ance of the material in which the heater is imbedded and which consti-
tutes another capacitance. The bath itself and the enclosure are further
capacitances and resistances to be considered. All this results in a
process with dead time.
The response is schematized in the graph in Figure 4-6. As shown,
the process has been started up from a cold condition. If the contacts
of the temperature controller were short-circuited, the temperature
would rise along the dashed line up to a level which is determined by
the self-regulation of the system. If the thermostat cuts off the heating
30 AUTOMATIC CONTROL

system at C , the temperature will still continue to rise somewhat due


2

to the dead time which delays the effect.


If the time constant is sufficiently long, which means a more gradual
slope of the response curve, Figure 4-6, then the overshoot is correspond-

Figure 4-6. Process response for two-position control system


with dead time.

ingly reduced and it is hence the relative length of dead time, or the ratio
^ /^, that is responsible for the overshoot.
2

The overshoot of a two-position controller is approximately

Y=
It,
where and t\ are the dead time and time constant, respectively, and g
?2

is the process gain when the controller is switched from one position to
the other. Thus a heat-treating furnace with on-off control may have a
process gain of 2400 °F, which means that at a room temperature of
100°F, the furnace temperature will increase to 2500 °F when full heat
is supplied. Self-regulation, i.e., increased heat losses, limit the possible
temperature increase at this point and the process gain is hence 2400°F.
Assuming that this furnace has a dead time of 0.8 minutes and a time
constant of 16 min., the overshoot is
y ^ 2400 X 0.8 ^ ^Qop
32
If the furnace is to control at 1400°F, a recording thermometer will
probably show that the temperature oscillates between 1340° and
1460°F. The heat-treated material itself, however, because of its rela-
tively large capacitance, may maintain a comparatively constant tem-
perature even under such wide fluctuations of the furnace temperature.
The above formula does not consider the contribution of the differen-
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 31

tial gap to the overshoot amplitude. Where Y = 60°F it is quite obvious


that the differential gap can be neglected, but where the magnitudes of
Y are comparatively small and the differential gap can no longer be
neglected, the amplitude will also depend on the differential gap.
There are several methods to reduce the overshoot amplitude. Where
the differential gap contributes essentially to the overshoot, the most
obvious step is to reduce the differential gap. This has, however, mechan-
ical limitations due to backlash and lost motion. In a bimetallic ther-
mostat, for example, it is hardly possible to reduce the differential gap
below 1.5°F.
Another method is the accelerator, a heater element frequently built
into thermostats, which is energized whenever the latter calls for heat.
The primary element, usually consisting of a bimetallic spring, responds
to the additional heat of the accelerator and therefore cuts off the heat
source before the upper limit of the differential gap is actually reached.
This is equivalent to reducing the differential gap and hence limits
overshoot.
A third method, which is frequently used in heat-treating furnaces,
consists in operating at a slower heat-producing rate, hence at a longer
time constant, when approaching the set point. The higher rate is used
for starting up to avoid delays and the lower rate reduces overshoot. A
multi-position controller switch automatically changes from one rate to
another at a pre-set temperature.
Load Changes. Figure 4-7 shows the effect of load changes by
means of three curves, namely, the load curve, the curve that shows the

LOAD

FINAL CONTROL
ELEMENT

CONTROLLED
VARIABLE

Figure 4-7. The eflFect of load changes on two-position control systems


with dead time.
32 AUTOMATIC CONTROL

position of the final control element, and the curve with the fluctuations
of the controlled variable. The controlled variable is assumed to be tem-
perature and the process response is considered to include dead time.
The final control element, for example, a valve in a steam heating
system, is shut down when the controlled variable reaches the tempera-
ture C2 at T^\ however, due to dead time there is some overshoot. When
the temperature begins to drop, it has to reach Cj before the valve opens
and again there is some overshoot. The dilference between and Cj
is the differential gap.
At Tg the valve is closed again, but shortly after, a load change occurs
that calls for considerably more heat input. Since the valve is closed, the
temperature drops at a faster rate than before. This is equivalent to a
shorter time constant, and results in a larger ratio, hence larger
overshoot as expressed in the graph. The temperature finally builds up
again, at a slow rate since the load continues to be large and the conse-
quence is that the overshoot between and TQ is small. Nevertheless,
the amplitude of oscillations has increased considerably.
Another effect is that there is a shift of the center around which the
controlled variable oscillates. This separates the control point from the
set point. The center of the differential gap is considered the set point.
Initially this coincides with the center of the oscillations. After the load
change, as shown in Figure 4-7, the center of the oscillations has
dropped as indicated by the control point. Originally, set point and con-
trol point are identical. After the load change, they are separated by the
offset.
Conclusions. The following conclusions can be made about two-
position action:
1. The controlled variable oscillates within a certain amplitude above
and below the set point.
2. For a given process and load, the amplitude depends on the differ-
ential gap.
3. The minimum differential gap is generally limited by mechanical
conditions.
4. In a process without dead time, the minimum differential gap is
also limited by the maximum permissible frequency of operation
of the final control element. The frequency increases inversely to
the time constant of the process. Hence processes with short time
constants are not suited for two-position action.
5. In a process with dead time, the overshoot is the limiting factor.
For a given differential gap and load, the overshoot depends on
the ratio of dead time to time constant and on the gain g.
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 33

Hence processes with large ratio and large g are not suited for
two-position control.
6. A number of remedies are available to improve the response of a
two-position controller, such as high-low or bypass adjustments in
solenoid valves, accelerators and multi-position controllers.
7. The difficulties caused by overshooting increase with load changes.
In spite of all limitations, there are a great number of processes in
the industry that allow two-position action which results in the most
simple and generally the most economical controller.

Single-speed Floating Action


Figure 4-8 illustrates a single-speed floating controller. The controlled
variable is liquid level head which is applied as pressure to the bellows.
A B

Figure 4-8. Single-speed floating controller.

Increase of level, i.e., pressure, and hence expansion of the bellows, will
result in contact between B and C. Inversely, decrease of level produces
contact between A and C. A certain amount of level change may occur
around the midposition of C without making contact on either side.
This range of pressure change is the neutral zone.
The motor is a shaded-pole reversible single-phase motor. If winding
34 AUTOMATIC CONTROL

D is energized, the crank and cam rotate counterclockwise; if E is


energized, clockwise. The travel is limited, e.g., to 180 degrees, by means
of the cam which is mounted on the motor shaft and actuates hmit
switches F and G. For example, if contact C closes on A, winding
D energizes, and the motor rotates counterclockwise until it opens con-
tact F, deenergizing winding D. No further action occurs until contact
C closes on B and starts the motor in clockwise rotation. The crank
rotating on the motor shaft would be linked to a final control element.
Processes without Seif-regufation. Suppose the controller positions
a valve in the water line to a boiler drum. If the boiler drum is of equal
cross-sectional area throughout its height and the valve is suddenly
opened beyond the requirements of the process, the level rises contin-
uously. This characteristic defines a process without self-regulation.
Let the system be initially in balanced condition. This is shown
in Figure 4-9 for the case of a controller without neutral zone. Before

LOAD

SUPPLY

CONTROLLED
VARIABLE
(LEVEL) ^/|\/\/\/
T, T, r, T, T.
Figure 4-9. Process without self-regulation with floating controller
(no neutral zone).

time Ti, a given load corresponds with a certain valve opening which as-
sures enough water supply to maintain the controlled variable, i.e., the
level at the set point. At time a load change enters, the steam de-
mand increases and the level drops. From Figure 4-8 it can be seen that
as a result of this change, contact C moves on A and energizes the
motor, which opens the valve to increase the supply as shown in Figure
4-9. While the full line 5' corresponds with the supply flow into the tank,
the dashed line m represents the rate of liquid removed in form of
steam. The valve opens until the supply equals the demand m, which
occurs at time T^.
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 35

However, at this point it is not enough that supply equal demand. As


long as the deviation from the set point persists, added supply is re-
quired to make up for the deficit in liquid level. The valve will therefore
continue to open until T^, at which time the controlled variable recovers
the set point. At this point, the valve is too wide open, because the ex-
cess of supply to return the level to the set point is no longer needed.
The valve begins to close. Until however, the level continues to rise,
since only then has the valve obtained the opening m which would de-
liver the required quantity; however at this time the level is too high.
Hence the valve goes on reducing the flow until the required level
is reached at T^. At time Tg, however, the situation is the same as it was
at time T^, namely the supply is not sufficient for what the load demands.
The cycle starts again.
As soon as dead time is added to these conditions, the response gets
worse. Dead time may exist in the process as well as in the controller.
In the following example the controller is assumed to be the cause of
the dead time, e.g., because the motor of Figure 4-8 coasts slightly be-
fore it reverses. The effect of such coasting is illustrated in Figure 4-10.
At time T, the load increases and the controlled variable begins to drop.
1051646
LOAD

SUPPLY

CONTROLLED
VARIABLE

Figure 4-10. Process without self-regulation with floating controller


(actuating motor coasting).
36 AUTOMATIC CONTROL

Action of the controller gradually increases the supply 5 reaching the


m-line at T^, from which time the controlled variable begins to return to
the set point which it reaches at T^. At this point, the motor should re-
verse but due to coasting it continues its movement to T^. The m-line is
now crossed at T^, and an increase of amplitude in the response of the
controlled variable is already visible. From TQ to T-, the motor coasts
again before it reverses and this results in a further amplitude increment.
It is obvious that the controller will rapidly reach its maximum ampli-
tude, limited by mechanical conditions, and will continue to oscillate
within these amplitude limits. Such control is undesirable, and single
speed floating action can therefore not be used with processes without
self-regulation.
Processes with Self-regulation. The above described situation
changes completely when the process contains self-regulation. A typical
example is an air-conditioning system.
Suppose the motor actuator of Figure 4-8 positions the damper in a
duct admitting hot air to a recirculating drier. The drier temperature is
the controlled variable and the bellows in Figure 4-8 may be considered
gas-filled and sealed. On a drop in temperature the gas contracts, the
bellows moves with it, and contact C moves toward A.
If the controller were to be disconnected, and the damper were man-
ually changed, e.g., toward closure, then the temperature would drop.
This drop would, however, not continue indefinitely as in processes
without self-regulation, but would level out at a new temperature. This
is because with the decrease of the drier temperature, the heat losses
through the ducts and walls of the drier also become less.
Figure 4-11 illustrates the action of such a system. Dead time is
neglected. At time T^, a sudden increase in heat demand offsets the

LOAD

SUPPLY

CONTROLLED
VARIABLE
(TEMPERATURE)

Figure 4-1 1. Process with self-regulation with floating controller.


STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 37

balance. The temperature starts to drop and this leads to a gradual


opening of the damper and an increase of warm air supply.
As in previous examples, the 5-line expresses the position of the final
control element, and simultaneously the flow into the controlled system,
while the m-line interprets the flow leaving the system, which in the
present case is the heat loss from the drier system. This flow leaving the
system is no longer a constant as in Figures 4-9 and 4-10, but becomes
a function of the controlled variable.
When the temperature drops, the heat loss diminishes, which is ex-
pressed in the m-line. Thus, the m-line reflects the oscillating response of
the controlled variable. The crossover of the 5-line and m-line indicates
the point where the supply is equal to the heat loss. This happens at 2 T,
and from this moment on the temperature begins to rise. At the set
point is reached and the damper begins to elose. The fact that the m-line
is no longer straight but is deflected in a direction that shortens the time
until m equals 5, having thus a damping effect on the dynamic action,
leads to rapidly subsiding cycles. The control system is stable, which
proves that the process with self-regulation lends itself to control by
floating action.
Figure 4-11 also illustrates the fact that the greater the speed of the
damper motor is, the sooner will the m-line be reached, and the smaller
will be the amplitude of the temperature oscillations. This is true for an
ideal system, but dead time and time constant must again be taken into
account for practical applications. The magnitude of dead time produces
overshoots for reasons quite similar to those illustrated in Figure 4-10.
The difference is that in the damped response of Figure 4-11, the re-
sulting oscillations are not necessarily of increasing amplitude, but may
rapidly subside.
An offset as previously deseribed for the two-position controller can-
not exist with a floating controller. As long as a deviation of the con-
trolled variable exists which is larger than the neutral zone, the floating
controller will produce corrective action.
The floating action of the controller provides a gradual corrective ac-
tion. The faster the process response is, i.e., the shorter the time con-
stant, as compared to the floating speed of the controller, the less neces-
sary will it be to move the final control element in order to obtain a
measurable change in the controlled variable, and henee the less will be
the amplitude of the overshoot. For the same reasons, the time constant
of the measuring element should be small as compared with the floating
speed of the controller.
This implies rather serious hmitations on the floating speed and henee
on the speed of correction when load changes occur. As long as these
38 AUTOMATIC CONTROL

load changes proceed slowly, the slow corrective action is of minor im-
portance; however, if these changes are fast, then the time that elapses
before the controlled variable is returned to the set point may become
objectionably long.
Conclusions. The following conclusions might be made about single-
speed floating action;
1. Single-speed floating action can be used only with processes with
self-regulation.
2. The smaller the dead time and the time constants, the better the
control.
3. The floating speed, and hence the speed of corrective action, has to
be slow, relative to the time constants of the process and the
measuring means, to keep the amplitude of the oscillations within
acceptable limits.
4. The relative slow corrective action permissible with this type of
control action becomes more obvious and hence objectionable in
processes with fast load changes.

Proportional-speed Floating Action


With this action the speed is proportional to the deviation of the con-
trolled variable from the set point. The action of such a controller can
be faster by far than that of a single-speed floating controller, because in
reducing the deviation of the controlled variable and bringing it back
to zero, it automatically slows down the floating speed.
It is used most frequently with hydraulic controllers and, hence, in
those applications where not only the advantages of proportional-speed
floating action but also the high power output of hydraulic actuators is
desirable, such as in large dampers, heavy butterfly valves, etc.
Figure 4-12 illustrates the action of a proportional-speed floating con-
troller. Process pressure is the controlled variable. It is applied to a dia-
phragm. The deflection of the diaphragm depends on the magnitude of
the process pressure and the spring constant of the spring which opposes
the force exerted by the process pressure. The spool of a four-way valve
is connected to the diaphragm through a lever system. The result is that
the spool position is proportional to the process pressure. The spool con-
trols the oil flow to the actuating cylinder. The position shown persists
as long as the controlled variable is at the set point. If the process pres-
sure rises, the spool moves downward. This connects the lower part of
the cylinder with the oil supply and the upper part with the drain. As a
consequence the actuating piston moves upward. Inversely, a decrease
in process pressure results in a downward motion of the actuating piston.
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 39

PROCESS
PRESSURE

ACTUATING
CYLINDER

ACTUATINC
PISTON

Figure 4-12. Proportional-speed floating action.

The piston is linked to a final control element which it positions


accordingly.
The speed of the actuating piston depends on the rate of flow of oil
passing through the four-way valve. The rate of flow is a function of the
valve opening. Since the opening of the four-way valve increases with
the deviation of the controlled variable from the set point, it follows
that the actuating piston moves faster, the larger the deviation.
Thus the two characteristics of proportional-speed floating action are
illustrated. The motion of the final control element continues as long as
the deviation of the controlled variable from the set point persists. The
speed at which the final control element moves is proportional to the
deviation of the controlled variable from the set point. There are, of
course, limitations both to duration and speed of the motion of the final
control element, since the motion will discontinue when the end of the
stroke is reached, and the speed cannot be faster than the oil supply
allows.
Calling jc the displacement of the final control element and t the time
required for this displacement, then the ratio x/t expresses the floating
speed. It is proportional to the deviation e of the controlled variable
from the set point. Hence

where/is a constant which is called the floating rate of the controller.


Figure 4-13 compares the response between single-speed and propor-
tional-speed floating action, illustrating for the latter the gradual de-
40 AUTOMATIC CONTROL

Figure 4-13. Floating speed vs. deviation of single-speed and proportional-


speed floating controllers.

crease of floating speed as the controlled variable approaches the set


point, as compared with the abrupt change of the single-speed type.
The speed x/t of the final control element is expressed by n, and hence
n =fe

The floating speed n is limited by the maximum speed that can be ob-
tained from the controller. From previous considerations, it is obvious
that the floating rate should be as high as possible without sustained
oscillations. Adjustments on the controller allow attainment of the
optimum floating rate.
Adjustment of Floating Rate. A slow floating rate is identical with
low sensitivity of the controller, and vice versa. Figure 4-14 shows
several responses of a controlled variable, assuming that initially at
r = 0, the set point is changed from Cj to Cg. The slowest floating rate,
i.e., lowest sensitivity, would result in an overdamped response as shown.
The floating rate can be increased to a maximum, a point at which still
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 41

no overshooting of C2 occurs. This is called critical damping and ex-


pressed by the corresponding curve in Figure 4-14. Any further increase
of floating rate results in oscillations. The 14-decay ratio means that each
successive overshoot is VA the magnitude of the previous one. Similarly,
the V2-decay ratio reduces each overshoot to V2 of the previous one. In
the interest of fast corrective action and limited overshoot, the 14-decay
ratio is generally the preferred one. In a single-capacitance process, for
example, the V4-decay ratio allows operation more than 20 times faster
than under critically damped conditions.

Figure 4-14. Various degrees of damping.

The floating rate for *4-decay ratio for a process without dead time is
given by

where / = floating rate of final control element in in./min./unit change


of controlled variable
g = process gain in units of controlled variable/in. displacement
of final control element
= time constant in min.

If dead time ^2 is present, or if a multi-capacitance process is broken


down into a single time constant plus a dead time, allowance has to be
42 AUTOMATIC CONTROL

made correspondingly. A practical method with sufficient accuracy for


most applications is based on the equation

/=
g{ti + ^2)
where /c is a number chosen from the following table:

greater than 25 5
12-25 4
6-12 3
2-6 2
1-2 1.5
smaller than 1 1

For example, a temperature-controlled process has a gain of 5°F/in. of


valve motion, dead time of 1 min. and time constant of 3 min. Substi-
tuting values in the above equation gives a floating rate of

A method for determining the floating rate in an actually installed


system consists of the following steps:
1. Increase of the floating speed to the critical setting where the con-
trol system just begins to oscillate without changing amplitude.
2. Measurement of the time needed per unit length of stroke of
the final control element under these conditions.
3. Estimating the relation of time constant to dead time and
decreasing the floating rate so that the time measured in the second
step is increased by the factor listed in the following table:

Uh Increase of Stroking Time

25-100 10 times
10-25 5 times
4-10 3 times
smaller than 4 2 times

The table gives no values beyond tjtz = 100, since beyond this value it
is generally not possible to obtain continuous oscillations.
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 43

Limitations for critical setting. Rules for critical setting as given for
the different mechanisms in this chapter require certain qualifications.
The response of a process to a load change depends on the nature of the
load change, and where load changes of different origins are expected,
all of them should be investigated. On a feedwater heater, for example,
the temperature may be influenced by the amount of throughput as well
as by the steam pressure. Hence the critical setting should be repeated
by keeping the set point constant and changing the steam pressure
slightly. The setting should be determined by whichever method requires
less sensitivity.
Furthermore, nonlinearities may exist, e.g., between the change in
position of the final control element and the change in magnitude of the
controlled variable. Valve characteristics are used to make this relation
at least approximately linear. If the nonhnearities are too high, the con-
trol system may be stable at one valve position but unstable at another.
The answer is to change the valve characteristic or to reduce the sensi-
tivity of the controller sufficiently to accommodate a wide enough range
of valve openings.

Proportional-position Action
In the proportional-speed floating controller described above, the
position of the final control element is independent of the deviation of
the controlled variable from its set point. As long as the deviation per-
sists, the final control element continues to move—as far as its mechani-
cal limitations allow. In the proportional-position controller, on the
other hand, the response becomes a function of the deviation of the con-
trolled variable.
Figure 4-15 shows the responses of different controllers. A similarity
exists between the two-position and the proportional-position con-
troller insofar as the response is an immediate one in either case, and
does not correct gradually as in the floating controller.
The corrective action caused by the proportional-position controller
is modified by the final control element. A spring-opposed diaphragm
valve does not respond immediately, but a certain time is required to
move it from one position to another. Figure 4-16 compares the
proportional-position action of the controller with that of the controller
and the final control element combined. This shows that the final con-
trol element becomes the limiting factor in the fast corrective action of
a proportional-position controller.
Proportional-position action is used equally in electric, electronic, hy-
draulic, and pneumatic controllers. Figure 4-17 illustrates a pneumatic
application. The signal from the controlled variable is applied as pres-
44 AUTOMATIC CONTROL

sure to the signal bellows. An increase of pressure will tend to rotate


the flapper clockwise. This brings the flapper closer to the nozzle, de-
creasing the opening through which air flows out.
Supply air passes through a fixed restriction and then through the
nozzle to atmosphere. Due to the movable flapper the nozzle tip be-
comes a variable restriction. The air flow passes through two successive

a - STEP INPUTS

b-TWO-POSITION ACTION d- PROPORTtONAL-SPEED


FLOATING ACTION

c- SINGLE-SPEED «-PROPORTIONAL-POSITION
FLOATING ACTION ACTION

Figure 4-15. Step function responses of various controllers.

a-RESPONSE OF CONTROLLER ALONE

Figure 4-16. Response of proportional-position controller without and with


final control element.
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 45

SIGNAL FROM
CONTROLLED VARIABLE

Figure 4-1 7. Proportional-position controller (pneumatic application).

pressure drops, one in the fixed restriction, the other in the variable re-
striction. The pressure between the two restrictions is the nozzle back
pressure, and it will assume a value determined by the flapper position.
When the flapper approaches the nozzle, the nozzle back pressure in-
creases. When it moves away, it decreases.
Since relation between flapper position and nozzle back pressure is
linear only within relatively small pressure changes and since, further-
more, the flow capacity that can be handled by such an arrangement is
rather limited, it is customary to use a pilot valve which increases the
effect of the flapper-and-nozzle arrangement and which regulates the air
flow to and from the final control element.
The nozzle back pressure is applied to a diaphragm in the pilot valve.
Connected to the diaphragm is the plug of a three-way valve, which
connects the final control element through the supply port to the supply
air and through the exhaust port to atmosphere. When the supply port
tends to close, the exhaust port opens, and vice versa. Again there are
restrictions connected in series, the supply port and the exhaust port.
Both restrictions are variable, and the pressure that exists between them
is applied to the final control element.
Suppose the signal from the controlled variable rotates the flapper
clockwise, the nozzle back pressure increases, the exhaust port closes,
the supply port opens, and hence the pressure between the two ports in-
46 AUTOMATIC CONTROL

creases. This increased pressure is transmitted not only to the final con-
trol element, but also to the feedback bellows, which now exerts a
counter-clockwise force on the flapper. This is the picture of a force
balance between signal bellows and feedback bellows. Their balanced
position determines the deflection of the flapper and hence the signal
transmitted to the final control element.
If the final control element is a spring-opposed diaphragm valve, as it
usually is, this pressure is converted into a force and balanced with the
force of a spring. The result is a finite position of the final control
element which is proportional to the signal from the controlled variable.
The step function response of a proportional-position controller as il-
lustrated in Figure 4-15e is expressed by

where x is the change of the controller output signal, e is the magnitude


of the step input, and P is the proportional band in per cent of instru-
ment range. The meaning of the proportional band is discussed in detail
further below.
Offset. The most important limitation of a proportional-position con-
troller is its inability to return the controlled variable to its set point after
a load change has caused a deviation. This is illustrated in Figure 4-18.
The regulation of the supply, indicated by the center curve, corresponds
to the output of the controller and final control element combined. The
change in controlled variable expressed by the lower curve represents
the input to the controller. Since in the proportional-position controller
the output is inversely proportional to the input, their curves must cor-
respond with each other, one being the mirror image of the other.
After the load change, the controlled variable begins to decrease and
the supply rises. The purpose of the increase in supply is to bring the
controlled variable back to its set point. However, this is impossible for
the proportional-position controller, because its basis is that controlled
variable and supply maintain their proportion, and the mechanism does
not permit maintenance of one at the set point while the other operates
at a higher level. The proportional-position controller can only reduce
the deviation of the controlled variable from the set point, but cannot
eliminate it. The remaining deviation is called the offset, as shown in
Figure 4-18.
The extent to which the proportional-position controller reduces
the deviation that would occur without it, is expressed by the de-
viation reduction factor Q. For example, a feedwater heater is to raise
the feedwater temperature to 180°F. For a given steam admission
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 47

LOAD

SUPPLY

CONTROLLED
VARIABLE

Figure 4-18. Proportional-position action.

this temperature can be maintained as long as a certain average feed-


water demand from the boiler exists. However, at peak periods, the tem-
perature would drop to 160°F unless an automatic controller is used.
Let the controlled variable be the outlet temperature of the feedwater
and the control agent be the steam. Suppose the control system is to
prevent the feedwater temperature from dropping below 178°F. This
means that without the controller, the feedwater temperature could de-
crease by a maximum of 20°F, with the controller the maximum
decrease is reduced to 2°F. The allowable deviation reduction factor is
p ^ 180 - 160 ^ 20^
^ 180 - 178 2
Obviously, the higher Q is, the less is the permissible offset.
The deviation reduction factor of a process which is considered for
proportional-position control can be determined approximately for a
*4-decay ratio from knowing its time constant and dead time, by means
of the equation*
Q _ ^2 + 0.8/]
^2

If the time constant of the above feedwater heater is 3 minutes and the
dead time is 0.5 min., then

Q 0.5 -f- 0.8 X 3 CO

* Based upon an article by Dr. W. Oppelt, “Rules of Thumb for Adjustments of Control Sys-
,
tems,” (German), Chemie-Ingenieur-Technik, No. 8 pp. 190-193, 1951.
48 AUTOMATIC CONTROL

which shows that the desired quality of control of g = 10 would not be


obtained. The feedwater temperature may drop to 176.55°F, since
180 - 160 _ . g
180 - 176.55

which corresponds with the deviation reduction factor of the particular


process. Another way of improving Q is to increase the amplitude of
the oscillations. This method shall be ignored since it generally is not
acceptable.
If such a drop is not permissible, the proportional-position controller
cannot be used. The deviation reduction factor is therefore a means to
determine if a proportional-position controller is acceptable for a specific
control system.
Adjustment of Proportional Band. The proportional band is the
change of controlled variable necessary to move the final control ele-
ment through its full stroke. It is usually, as it is here, expressed in per
cent of the total measuring range of the controlling instrument. An in-
strument with total range from 0° to 200°F may have a 10 per cent pro-
portional band. This means the final control element moves from fully
open to fully closed for a temperature change of 20°F. The proportional
band may also be expressed in units of the controlled variable that cor-
respond to the full movement of the final control element. For example,
a proportional band of 10°F means that the final control element moves
through its full stroke when the temperature of the controlled variable
changes by 10°F. For smaller temperature changes the movement is cor-
respondingly smaller.
The narrower the proportional band, the faster the response, the
smaller the offset, but the greater the tendency to oscillate. Figure 4-19

Figure 4-1 9. Controlling action with various proportional bands.


STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 49

shows the typical response of a process with a proportional controller.


The pH of a chemical solution may be considered the controlled vari-
able. A sudden change of alkalinity occurs at time T^. With a propor-
tional band of P = 20 per cent, the controlled variable returns very
quickly to the set point, but continues to oscillate around it. Since these
oscillations subside gradually, they may be acceptable, particularly if a
mixing tank provides equalization of the solution. Assuming however,
that these oscillations are considered excessive, a proportional band of
50 per cent may be chosen. The response is slower, the overshooting is
reduced considerably, but offset is the consequence. A further increase
of the proportional band to 100 per cent results in an even larger offset
but overshooting is now eliminated completely. An adjustment for a
y4-decay ratio would presumably lie between 25 and 30 per cent.
The proportional band of the controller for a y4-decay ratio can be
determined either by experiment or by calculation on the basis of cer-
tain numbers that have to be known. The method of calculation shall
be discussed first.
Either one of the following equations may be used:

P = lOOvt^^
R
and
P = 100^^1-
?! R
where the symbols have the following meaning:
P is the proportional band in per cent of the measuring range of the
controlling instrument;
V is the process response rate (see Chapter 3) in units of the controlled
variable/min./in. of motion of the final control element;
is the dead time in minutes;
S is the stroke in inches which the final control element moves when
the controller output signal changes through its full range;
R is the measuring range of the controlling instrument in units of the
controlled variable;
g is the process gain in units of the controlled variable/in. of stroke of
the final control element;
C is the time constant in minutes.
The magnitudes of time constant and dead time do not refer to the
process alone, but to the results of a step response which includes final
control element, process and measuring element.
For example, the maximum response speed of a process to a valve
displacement of one inch is 20°F/min. The dead time is 10 min. The
50 AUTOMATIC CONTROL

Stroke is 2 in. and the measuring range of the controlling instrument is


200° F. The resulting proportional band is

P = 100 X 20 X 10 2^ ~

The experimental method assumes that the control system is in opera-


tion and that a limited amount of process disturbances are permissible.
The first step is to find the critical setting of the proportional band ad-
justment which means reducing the proportional band to the point
where continuous oscillations of constant amplitude begin to appear.
This must be done in successive steps. After each change of the propor-
tional band, a load change is introduced by a small change of the set
point, and the response is observed. If the response is still too stable, the
proportional band must be further reduced.
The second step consists only in adjusting the proportional band to
twice the critical setting. This is then the correct adjustment for 14-decay
ratio. For example, if the critical setting at which continuous oscillations
begin to appear is P = 10 per cent, then the suitable proportional band
for the process is P = 20 per cent.
Proportional-position vs. Proportional-speed Floating Action.
The factor that limits the application of floating action is the floating
rate. The stability of the control system may require a very slow floating
rate. In this case, correction for a deviation can become so slow that the
time which elapses before the controlled variable returns to the control
point is longer than permissible for satisfactory control.
A slow floating rate may be compared with a wide proportional band
of a proportional-position controller. In either case, speed of corrective
action is limited in the interest of stability. From the equation

P = lOOg ^ I-
ti R

follows that low tj/fa ratios (i.e. high ^3/^ ratios) diminish the response
speed of proportional-position action.
The equation for proportional-speed floating control

f= + b)
shows that the ti/t2 ratio affects floating control only to the extent of
the constant k. The table on p. 42 shows that the maximum effect of
k does not exceed the ratio 5:1.
The magnitude of the sum of b plus b, however, has a much more
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 51

pronounced effect on floating control than it would have on proportional-


position action.
The conclusion is that proportional-position control is particularly
valuable where large values of dead time and time constant are con-
cerned, but where the ratio t^/t^ is also large. For proportional-speed
floating control the reverse is true and processes with small values of
dead time and time constant and small ratios are best suited.
For example, a temperature-controlled process with a time constant
of 3 min. and a dead time of 1 min. is suitable for proportional-position
control, disregarding offset. On the other hand, for a pH control system
which may have a time constant of 0.02 min. and a dead time of 0.48
min., a proportional-speed floating controller would probably be the
best answer.

Proportional plus Reset Action


Figure 4-20 illustrates proportional plus reset action in a pneumatic
arrangement. The same mechanism as in Figure 4-17 is used for this

To final control element

Figure 4-20. Mechanism showing proportional plus reset action.

example; the pilot valve, however, is ignored for the sake of simpler
presentation. The reset bellows is added.
Suppose that initially the pressure in both the reset and the feedback
bellows is equal and that the pressure in the signal bellows corresponds
to the set point pressure. The torques exerted by the three bellows on
the flapper balance each other. If the signal increases, the flapper moves
clockwise, tending to close the nozzle. This results in an increase of
nozzle back pressure which is also applied to the feedback bellows. The
resulting position of the flapper, and hence the nozzle back pressure, is
52 AUTOMATIC CONTROL

thus determined by a new balance of forces exerted by the bellows.


However, the increased nozzle back pressure gradually leaks through
the restriction of a needle valve designated in the illustration as reset
rate adjustment. In doing so the reset bellows continues to move the
flapper in the clockwise direction which began with the action of the
signal bellows. The nozzle back pressure continues to rise in an effort
to increase the controller output signal to a point where the controlled
variable finally returns to the set point.
Proportional plus reset action is a combination of proportional-
position action and proportional-speed floating action. The result is
elimination of offset from proportional-position action.
Reset or Integrating Action. Figure 4-21 compares the step func-
tion responses of the three actions: proportional-speed floating, propor-

T, TJ T

PROPORTIONAL
PLUS RESET
ACTION

Figure 4-21. Step function response of different controllers.

tional-position, and proportional plus reset. Two step inputs have been
applied as actuating signals at times and 7; respectively. The second
input was twice as large and in an opposite direction to the first one.
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 53

The action of the proportional-speed floating controller is expressed by


w = fet
where w is the change of controller output signal,/is the floating rate, e
is the actuating signal and t is the time interval measured from / and
T2, respectively. This means that as long as there exists an actuating
signal, the controller output signal will continue to increase (or decrease)
since the actuating signal e is multiplied by the time t. This continuous
action which is a function of time is called integrating action. Offset
cannot exist with integrating action, since an actuating signal will always
result from offset. This is because offset is a deviation of the controlled
variable from the set point, and the actuating signal is the difference
between controlled variable and set point.
In the proportional-position controller, as shown in Figure 4-21, the
action is expressed by

where P is the proportional band of the controller. In this case the


actuating signal e produces a change in controller output signal w of
finite magnitude. If the change is not large enough, the result is offset,
which results because the proportional-position controller does not
provide integrating action.
The equation which corresponds with the proportional plus reset
controller reads
100 100 100
w e -I

The first term contains the proportional-position control only. The


second term corresponds with the floating controller, but the floating
rate /is now replaced by 100(//JP). The factor I is the reset rate of the
controller. It is multiplied by the elapsed time t. Due to this integrating
action the controller eliminates offset.
The step function response for proportional plus reset action as illus-
trated in Figure 4-21 shows clearly the distinction between the initial
sharp increase due to the proportional action and the integrating action
of the reset. A proportional plus reset controller may well be considered
a two-speed proportional-speed floating controller. First is the fast
change, followed by a more gradual change. This becomes particularly
obvious when the action of the final control element is considered; this
follows only gradually a vertical change of the controller output signal.
Thus the change from the action due to proportional control to that due
54 AUTOMATIC CONTROL

to reset action is only a change in speed in the response of the final con-
trol element.
The Reset Rate. The reset rate I in the above equation depends on
the opening of the needle valve of the reset rate adjustment (see Figure
4-20). It also depends on the pressure differential across it and hence
on the magnitude of change in controller output signal.
Figure 4-22 further illustrates the reset rate. In this case the initial
corrective action of a controller, in response to a given deviation of the

Figure 4-22. Proportional plus reset action.

controlled variable, is an increment of the transmitted pressure of 1.2


psi which is due to proportional-position action only. This corrective ac-
tion is increased through the effect of reset action by 4.8 psi/min.
Hence the reset rate / is equal to 4 repeats/min.
Suppose this controller has the usual output of 3 to 15 psi which is
the signal range to the final control element. The corrective action at
r = 0 is due to proportional-position action only. It is 1.2 psi or 10 per
cent of the output range (15 — 3 = 12 psi). Hence w = 10 per cent at
r = 0. If the proportional band P = 50 per cent, then at T = 0

e = or e = 5 per cent

which means that the response as illustrated was caused by a deviation


of the controlled variable from the set point of 5 per cent of the instru-
ment range. Since the reset rate I was stated to be 4 repeats/min. it
follows that after 1 min.
100
w = 5(1 -1- 4) = 50 per cent
50
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 55

which expressed in psi of controller output change is 6 psi, a figure


which corresponds with Figure 4-22.
Adjustments of Proportional Bond and Reset Rate. The adjust-
ments of the proportional plus reset controller may again be made either
by experiment or by calculations. Given the measuring range R of the
controller and the stroke 5' of the final control element, and knowing,
furthermore, either the process response rate and the dead time, or the
process gain, time constant, and dead time, the adjustments can be
made by following the expressions

P = lOOvt^^
R
or
P=100,i|

These equations are the same as those used before for the proportional-
position controller. The reset rate is given by

7 =^
^2

The example which was used for the proportional-position controller had
a dead time of 10 minutes. Adding reset to the controller would require
a reset rate of

I = ^ = 0.03 repeats/min.

The experimental method follows the steps outlined for proportional-


position control. In determining the critical setting of the proportional
band at which continuous oscillations begin to appear, the reset action
is cut out. The time in minutes, T, between two successive maxima of
oscillations at the critical setting is measured. The suitable final adjust-
ment of the proportional band is the same as for proportional action
alone, i.e, twice the critical setting, and the reset rate is

Proportional plus Reset plus Rate Action


The proportional band of a controller can only be reduced to a certain
point. Beyond this point the control system becomes unstable. This
limits the speed of correction. The wider the band, the less the sensi-
tivity and hence the slower the correction. This lag in correction be-
comes particularly noticeable where the time constant of the measuring
element is of noticeable magnitude.
56 AUTOMATIC CONTROL

A typical example is a multi-capacitance temperature-controlled


system. In this case the equivalent dead time is relatively large because
of the multiple capacitances. This requires a wide proportional band.
In addition, the measuring lag of the primary element is considerable.
This results in relatively slow corrections for the deviations of the con-
trolled variable from the set point.
It is desirable in such cases to create a response to a slow change of
the actuating signal by an acceleration of this signal in the controller.
This gives faster response, but one which slows down as the controlled
variable returns to the set point, thus preventing excessive overshooting
and hence instability. Controllers can be provided with rate action which
results in the described response. Occasionally, proportional action may
be provided with rate action only. However, this is rather rare. More
frequently the controller in this case includes also reset action and
becomes a proportional plus reset plus rate controller. The reason is
that a proportional controller that can operate with a narrow propor-
tional band does not generally require reset action, since the offset is
insignificant. In this case, however, the narrow band provides quick
enough corrective response and rate action is hardly warranted. As the
band becomes wider, the response is slower, the offset larger, and reset
action becomes necessary with rate action added for certain conditions.
Figure 4-23 is a schematic illustration of proportional plus reset plus
rate action. It differs from Figure 4-20 in that a fourth bellows—the
rate bellows—is added, and that the area of the feedback bellows is
smaller. When a change in signal occurs, the defiection of the flapper is
much larger, due to the reduced area of the feedback bellows. The con-
troller output signal as transmitted to the final control element is assumed

Figure 4-23. Proportional plus reset plus rate action.


STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 57

to be proportional to the flapper deflection. A change in controller


output signal is gradually transmitted to the rate bellows through the
restriction of the rate time adjustment. This reduces the initial deflec-
tion of the flapper.
If the pressure of the controlled variable, as applied to the signal
bellows, changes slowly, the pressure of the controller output signal will
also change slowly and there will be no large difference between the
pressures in the feedback and rate bellows. However, the faster the con-
trolled variable changes, the larger will be this pressure difference and
the greater will be the effect of the rate action.
The response of the proportional plus reset plus rate controller to a
step input would be maximum and only limited by the mechanical
conditions of the system. Hence, the inherent characteristics could not
be fully represented by a step function response. The change of the
controlled variable is therefore represented by a ramp function, i.e., an
input signal of a more gradual changing magnitude as illustrated in
Figure 4-24. At the input signal U begins to decrease. The controller

Figure 4-24. Response of proportional plus reset plus rate controller.


58 AUTOMATIC CONTROL

output signal rises correspondingly but at a faster rate. At the input


signal stops decreasing and remains constant at a new level. The effect
of the rate action subsides immediately to a point which is determined
by the proportional band. However, the action of the reset mechanism
also enters and continues to increase the controller output signal, though
at a slower rate. From time to T^, the input signal increases but this
time at a faster rate, and the magnitude of the change is also larger. The
action is correspondingly reflected in the controller output signal.
Rate Time. The mathematical expression for proportional plus reset
plus rate action is

w =
100 , 100 r, ,
^(se)D = + I,) + {se)D]

The last term stands for the rate action. The rate of change of the input
signal e is expressed by (^e). For example, if the input signal at a given
moment changes 3 psi/min., then = 3. The constant D expresses
the adjustable rate time of the controller. Rate time is generally given
in minutes and indicates the time difference in response between a pro-
portional and a proportional plus rate controller. For example, assume
that the controller has no reset action, i.e., 7 = 0, then

w= (1)
Let the controlled variable change at a certain rate (se), then after a
time the controlled variable will be
e = (se)Ta
Equation (1) can then be written

>*' = ^[(se)r. + (2)

Suppose the rate action to have been cut off, then it would have taken
a longer time—expressed by —to obtain the same magnitude w of
controller output signal, or

w = M(se)7; (3)

Equations (2) and (3) both express w of the same magnitude, hence

100
[ise)T^ -h {se)D] = 100 (sem
P P
STEP FUNCTION RESPONSE AND ADJUSTMENTS OF CONTROLLER 59

and this reduces to


D = n-T,

which shows that the rate time expresses the difference between time
due to proportional action only and the time that it takes for the combi-
nation of proportional plus rate action to obtain the same controller
output.
For example, if P, the proportional band, is 25 per cent of the measur-
ing range of the controller and if the controlled variable changes at the
rate of 50 per cent/min., i.e., (se) = 50, then with proportional action
only, the controller output after 1 min. is

w = 50 = 100 per cent of its range.

Adding rate action with a rate time of 0.25 min., the same change
should be obtained in 1 — 0.25 = 0.75 min. This can be checked by in-
serting values into equation (2), which gives

H; = M(50 X 0.75 -F 0.25 X 50) = 100 per cent

Adjustments of Proportional Band, Reset Rate, and Rate Time.


The equations which can be used to determine adjustments are: for the
proportional band;

P = SOvtz^
R
or
P = S0g
R
for the reset rate;
I=^
^2

and for the rate time;


D = 0.5 tz
using the same example as for proportional-position action, the propor-
tional band is
2
P = 80 X 20 X 10 160 per cent.
200
the reset rate is
/ = ^ = 0.05 repeats/min..
60 AUTOMATIC CONTROL

and the rate time is


Z) = 0.5 X 10 = 5 min.
The experimental method proceeds along the steps outlined for pro-
portional plus reset controllers. The adjustments are then as follows:
1. The proportional band should be 1.7 times the critical setting.
2. The reset rate should be

where T is the time between two successive maxima of oscillations


at the critical setting.
3. The rate time should be
5. FREQUENCY RESPONSE

Frequency response methods are more cumbersome than step func-


tion methods because they require special equipment and are more time
consuming. From the practical viewpoint, they are justified because they
allow division of a control loop into an arbitrary number of elements,
to obtain frequency response curves of each of the elements, and to pre-
dict their behavior in the loop or in combination with any other com-
ponents of which the frequency response is known.
Frequency response is similar to step function response in that it
compares the output of a component or of a group of components with
the input. It is similar also in its object which is to obtain a stable con-
trol system with maximum correction speed. The difference between the
two is the form of the input signal, which is a sudden change in one case,
and a continuously changing signal in the other.

Sinusoidal Input
The pattern which the continuously changing signal follows is that of
a sinusoidal curve, such as shown in Figure 5-1. The reason for choosing

such a signal is mainly to allow correlation of mathematical and experi-


mental results, which however is of more academic than practical interest.
There is, however, a deeper reason for choosing the continuously chang-
ing signal, in that many physical systems have responses of a simple
sinusoidal pattern. Wherever an elastic restoring force is involved, e.g.,
61
62 AUTOMATIC CONTROL

in the springs of a car, the pendulum of a clock, the mercury column in


a manometer, etc., the system will tend to restore its balanced condi-
tions by sinusoidal motions once its equilibrium is affected by an out-
side force. A periodic motion of sinusoidal pattern, repeating itself at
definite intervals of time, is called simple harmonic motion. The am-
plitude of a sinusoidal curve is defined as the distance from the mean
value to the point of maximum displacement. This is indicated by A in
Figure 5-1, curve a. The time required for the motion to complete one
cycle is called the period. Thus, in Figure 5-1, for either curve a or b,
the period Tis 12 min.; after this the pattern repeats itself.
Once the period and the amplitude are known, it is possible to draw
a sinusoidal curve with the help of a trigonometric table, or simply by
using the following values:

Angle Sine

0° 0.000
15° 0.259
o
o

0.500
45° 0.707
o^
oo

0.866
75° 0.966
90° 1.000

The period T is subdivided into 360 degrees. In Figure 5-1 this inter-
val coincides with 12 min. Hence each minute is equal to 30 degrees.
The amplitude is multiplied by the sine value of various points to obtain
the displacement at these particular points. Thus the following table is
established for curve a of Figure 5-1, which has an amplitude equal to
A:

Time in Minutes Angle in Degrees Displacement

0 0 0.000 A
0.5 15 0.259 A
1.0 30 0.500 A
1.5 45 0.707 A
2.0 60 0.866 A
2.5 75 0.966 A
3.0 90 1.000 T

Thus enough points can be established to draw a curve up to 90


degrees. From there the preceding table can be used up to 180 degrees
FREQUENCY RESPONSE 63

by taking the values in descending order, i.e., 0.966 A, 0.866 A, 0.707 A,


etc., for successive intervals of 0.5 min. It is obvious that by using nega-
tive values the curve can be completed.

Phase Lag, Frequency, and Magnitude Ratio


Figure 5-1 shows two curves—curve a and curve b. These can be
visualized as the input and the output of a process, e.g., in a gas-heated
steam generator.* Suppose the gas fuel control valve were to be cycled
according to curve a, then the steam pressure would follow with a similar
pattern, only delayed in action. This delay is equal to 2 minutes in
Figure 5-1. Such a delay in sinusoidal response is called a phase lag or
phase angle, and is expressed in degrees. As previously established, the
period T, 12 min. in the case under consideration, corresponds to 360
degrees, and hence the phase angle is — 60 degrees, where the minus sign
indicates that the output lags behind the input.
A period of 12 min. means that a complete cycle of the periodic mo-
tion is completed in this time interval. Conversely, this means that Vi 2
of the cycle is completed in 1 min. The frequency of a sinusoidal input
is expressed in those terms, i.e., Vii cycle/min. is the frequency of either
curve in Figure 5-1. Thus, the frequency is the inverse of the period.
The higher the frequency of the fuel valve, which means the faster the
valve is moved through its periodic motion, the more difficult it becomes
for the pressure to follow the input. Although the frequency of both
input and output is always the same, the phase angle will increase and
the amplitude of the pressure curve will get smaller. Finally a frequency
is reached which is too fast for the pressure to follow and it will stay
constant at some average value. The ratio between amplitude of output
signal to the amplitude of input signal is called the magnitude ratio.

Technique of Taking Frequency Response Data


Suppose it is desirable to take the frequency response of a steam gen-
erator control system. The controller is assumed to be air operated and
with proportional-position action. Fuel is admitted to the generator,
which produces steam. The pressure of the steam is measured and the
controller puts out a signal as a function of the deviation of the steam
pressure from the set point. The control signal is applied to a control
valve which regulates the fuel flow. The interaction of the closed loop is
thus established.

* Linear relations are assumed throughout this chapter. This is generally permissible. As
pointed out, however, at the end of this chapter, the interpretation of frequency response data
has its limitations largely because of this assumption.
64 AUTOMATIC CONTROL

In order to obtain the frequency response of such a system it is neces-


sary to open the loop at some arbitrary point, e.g., between controller
and valve, as shown in Figure 5-2. The closed loop thus becomes an
open loop.

An input signal is applied to the control valve. This requires a pneu-


matic sine-wave generator, which contains a variable-speed electrical
motor. The rotation of the motor shaft is converted into simple harmonic
motion by means of a cam, linkage, or slider. A pneumatic relay is
operated by this linkage, thus providing an air pressure that cycles
sinusoidally. By increasing the supply pressure or varying lever ratios,
the amplitude of the signal can be changed. The frequency is regulated
by the motor-speed control. The arrangement is basically the same for
hydraulic and mechanical sine-wave generators. Where electrical signals
can be used, a number of sine-wave generators are commercially
available.
Another piece of equipment is the recorder illustrated in Figure 5-2,
which records graphically both the input and the output of the open
loop. It is important that this recorder does not add to the attenuation
and phase shift of the input signal; i.e., the frequency response of the
testing equipment must be considerably above the frequency applied in
the test of the open loop. Frequently, oscilloscopes and other electronic
equipment are used for this purpose.
A certain amplitude is chosen for the frequency test. For reference
purposes, it is quite necessary to always note the amplitude at which re-
sponse data were taken. The amplitude is limited by the speed of the
components. For example, a valve may move 1 in. in 5 sec. Choosing
an amplitude which corresponds to a valve motion of one inch limits the
FREQUENCY RESPONSE 65

frequency to less than 3 cycles/sec. Thus the speed and not the fre-
quency may become the limiting factor.
The difference between speed and frequency response is an important
one. A system may be rather slow, but follow the input signal without
attenuation. With a small enough amplitude, this system will respond
satisfactorily even at high frequency. For larger amplitudes, however,
the slowness of the system becomes the controlling factor. Whether
speed or high frequency response or both are required depends entirely
on the application.
Once the amplitude is chosen, a number of runs are taken, changing
only the frequency of the input signal. It is assumed in the following, un-
less otherwise stated, that the proportional band of the controller is such
that at low enough frequencies the magnitude ratio is 1. The frequencies
are then increased, the data logged and put into graphic form.

Graphic Representation of Frequency Response Data


The frequency response of a system coordinates three variables: fre-
quency, magnitude ratio, and phase lag. Figure 5-1 illustrated magnitude
ratio and phase lag at one specific frequency, namely at Vn cycle/min.
This information does not suffice to draw any conclusions about the be-
haviour of the control system. For that purpose, it is necessary to graph-
ically represent the change of magnitude ratio and phase lag at various
frequencies. Several methods exist. The Nyquist diagram has been used
for many years. It has the advantage of combining all the information
in a single curve, but it is generally considered more difficult to interpret
than the so-called Bode diagram, a method which conveys the same in-
formation as the Nyquist diagram, but in two separate diagrams. The
American Society of Mechanical Engineers has accepted the Bode
diagram as the preferred standard for the presentation of frequency re-
sponse data.
Nyquist Diagrams. Let the control valve in Figure 5-2 have a total
stroke of 2 in. The controller output signal has a range of 12 psi. Cycling
the valve through an amplitude of 0.2 in. corresponds then to cycling
it through an amplitude of 1.2 psi. Suppose at a certain frequency
the controller output signal responds to this input with a maximum am-
plitude of 0.96 psi, then the magnitude ratio of output to input is
0.96/1.2 = 0.8.
Figure 5-3 represents a Nyquist diagram corresponding to the control
system of Figure 5-2. At 0.01 cycle/min. the magnitude ratio between
output and input is 1.0. The length of the arrow or vector a in Figure
5-3 is proportional to the magnitude ratio. The phase angle is zero.
66 AUTOMATIC CONTROL

At 0.03 cycle/min., the phase angle is —30 degrees, although the mag-
nitude ratio remains the same. The vector b is hence equal in length to
vector a, but it is displaced by —30 degrees as indicated in Figure 5-3.

-270®

At 0.05 cycle/min., the phase angle is —50 degrees and the magnitude
ratio is 0.9, which is expressed by making vector c correspondingly
shorter.
Vectors, d, e, and / are of increasing negative phase angle and decreas-
ing magnitude ratio. Actually at 0.7 cycle/min. the magnitude ratio de-
creases to 0.16 psi/ in. which is too small a value to enter on the plot.
By connecting the end points of the vectors, a curve is obtained which
can be used for the interpretation of the control system. The Nyquist
criterion states that a system will be stable as long as the — 1 point (on
the horizontal axis of Figure 5-3) is to the left of the curve.
Figure 5-4 shows several curves obtained with proportional-position
control systems. Curve 1 is typical for a single capacitance system and
curve 2 for a two-capacitance system. Curve 3 represents a three-
capacitance system. These three systems are stable, although curve 3
gets close to the — 1 point, which indicates that some oscillations will be
unavoidable. In the case of curves 4 and 5, however, the systems—also
of three capacitances—will be definitely unstable since the — 1 point is
no longer to the left of the curve.
The number of capacitances in a proportional-position control system
can be read from the Nyquist diagram by dividing it into four quadrants,
as indicated by numbers I, II, III and IV in Figure 5-4. Curve 1 remains
FREQUENCY RESPONSE 67

- 90“
Figure 5-4. Nyquist diagrams for various processes.

in the first quadrant, hence it represents a single capacitance system,


curve 2 enters the second quadrant indicating a two capacitance system,
and curves 3, 4, and 5 stand for a three capacitance system since they
enter into the third quadrant.
The Bode Diagram. Figure 5-5 gives the information of Figure 5-3,
but in the form of a Bode diagram. The upper curve illustrates the
change of magnitude ratio with increasing frequency. The scales used
are not linear, but are so-called logarithmic scales. The distance from
0.01 to 0.1 is the same as the distance from 0.1 to 1.0. Similarly,
the steps from 0.2 to 0.4 to 0.6 to 0.8 get progressively shorter. Thus the
scale decreases with increasing values. This has the advantage that the
lower values are easier to read than the higher ones. It has the further
advantage of replacing the multiplication of frequency response data by
graphic addition, as will be discussed later.
In the lower diagram, the phase angle is shown on a linear scale while
the logarithmic scale continues to be used for the frequencies.

Phase Lag and Phase Reversal


Figure 5-2 illustrated the method of taking two simultaneous graphs
in a frequency response test. As long as the output does not lag behind
the input, the corresponding graphs would appear as in Figure 5-6,
where one is the mirror image of the other. While the amplitudes of input
and output signals are equal, the output is at maximum when the input
68 AUTOMATIC CONTROL

is at a minimum, and vice versa. Since a full period corresponds to 360


degrees, the distance between maximum and minimum of a sinusoidal
curve is 180 degrees, and the output is said to lag behind the input by
180 degrees.

0.06
0.08

rsj
o
0.8
0.4

0.6

O
0.2
0.1

O
1.

d d d
Cydes per minute

Figure 5-5. Bode diagram.

Actually this is not a phase lag but a phase reversal which is inherent
in controller action. The controller has to counteract any disturbance in
the loop. For example, an increase in steam pressure must result in such
action of the controller as to reduce the steam pressure. This is equiva-
FREQUENCY RESPONSE 69

lent to saying that a 180 degree phase reversal as illustrated in Figure


5-6 is required. For the graphie representation of the frequeney response,
the phase reversal is ignored and only the additional phase shift is rep-
resented as phase lag.

Stability Conditions
Suppose there is an additional phase shift of 180 degrees while the
magnitude ratio is still unity. In this case the maxima of the output coin-
cide with the maxima of the input. The controller actually repeats the
disturbance, sustaining instead of counteracting it. The system is unstable.
The condition of unity magnitude ratio and 180 degree phase lag must
therefore be avoided. It is common practice to accept safety margins ac-
cording to the following rules:
1. The phase lag should not be more than 150 degrees when the mag-
nitude ratio is one or more. The 30 degree difference between the ac-
ceptable and the unstable condition is called the phase margin.
2. At a 180 degree phase lag the magnitude ratio should be equal or
less than 0.5. For a magnitude ratio of 0.5, it would be necessary to in-
crease it by a factor of two in order to make it unity and hence make
the system unstable. The factor by which the magnitude ratio has to be
increased to obtain instability is called the gain margin. A gain margin
of 2 is therefore desirable for process control.
70 AUTOMATIC CONTROL

Acting according to these stability criteria results in obtaining the


fastest practical control response under stable conditions. Figure 5-7
shows frequency response curves which fulfill these conditions. The
phase margin is 30 degrees and the gain margin is 2.

Tigure 5-7. Frequency response with minimum conditions for stable system.

Dead Time as Instability Factor


It can be concluded from the foregoing that an increase of phase angle
without a simultaneous decrease of magnitude ratio will lead to insta-
bility, excepting the case where the magnitude ratio is sufficiently below
unity at aU times. This condition, i.e., increase of phase angle at constant
magnitude, is the result of dead time, since it represents the time it takes
before a change in the loop becomes noticeable. Figure 5-8 illustrates
the frequency response as it results from dead time in the system. This
response indicates instability.
FREQUENCY RESPONSE 71

Cycles per minute

Figure 5-8. Frequency response of dead time.

Combination of Frequency Response Data


The frequency response of a controller or any particular component
is of interest only if the frequency responses of all other components in
the control loop are equally known. It is then possible to combine these
data and obtain a complete interpretation of the behavior of the control
system.

Block Diagram
Block diagrams illustrate how the frequency responses of various com-
ponents in a control loop can be combined. Figure 5-9 shows a typical
application of this method for a pressure control system. The pressure
is applied to a diaphragm-and-spring combination. When, for example,
the pressure increases, the jet pipe is deflected to the right. The deflec-
tion is proportional to the process pressure change. Oil is pumped
through the jet pipe and leaves at its tip to be directed into two orifices
which connect the two sides of an actuating piston. As the jet pipe is de-
flected toward the right, more oil passes into the right-hand than into
the left-hand orifice. The piston moves toward the left, opening the con-
trol valve, thereby reducing the pressure in the process line.
In the block diagram, the control system is divided into its most
obvious parts. This is quite arbitrary and different breakdowns are pos-
72 AUTOMATIC CONTROL

Process

Figure 5-9. Block diagram of control system.

sible. The system is now represented by five blocks, namely; process,


diaphragm-and-spring combination, jet pipe, piston, and valve.
The process changes its pressure per inch of valve motion (linear re-
lations are assumed throughout). Hence the “process” block reads
“psi/in.” The valve motion expressed in inches is shown as input and
the pressure in psi as output. This pressure is directed toward the
diaphragm-and-spring combination, which produces a deflection of
in./psi pressure change. This is converted into cu. in./sec. of hydraulic
flow by means of the jet pipe, and the piston transforms the hydraulic
flow into in./sec. of piston motion. The valve moves with the piston, and
the final outcome is in./sec.
The important point is that a chain of multiplications has been per-
formed by each block around the loop, namely
Process: in. X psi/in. = psi
Diaphragm-and-Spring; psi X in./psi = in., etc.
Each block acts as a multiplier upon its input. Each input is the out-
put of another block. It is possible to lump together two or more of the
blocks into one by a simple multiplication. For example, the “diaphragm-
FREQUENCY RESPONSE 73

and-spring” block can be combined with the “jet pipe” block by giving
the new block the dimensions of:
in. ^ cu. in./sec. _ cu. in./sec.
psi in. psi
which is obviously in line with the input of psi and the output of cu.
in./sec., since the multiplying of the new block with the input results in
the output.
Each of these blocks could be separately submitted to a frequency re-
sponse test. For example, the process could be analyzed by cycling the
valve and measuring the pressure. The relation of output to input both
in magnitude ratio and phase lag could thus be obtained. Similar data
could be obtained from each block.
The output/input ratios (the magnitude ratios) can be multiplied to
obtain the overall magnitude ratio, since they represent exactly the re-
lationship represented in the block diagram. The phase angles, however,
must be added since they represent delays in each block which simply
add up as a signal passes through the loop.

Graphic Addition of Frequency Response Data


The magnitude ratios of frequency response data are represented in
logarithmic scales, which resemble slide rule scales where the adding of
two distances cahbrated for the magnitude of the factors is equivalent to
multiplication. The result is that the Bode diagram allows the graphic
addition of magnitude ratios.
Since phase angles are represented on linear scales, their graphic ad-
dition corresponds with the physical addition required for phase angles.
The addition of magnitude ratios and phase angles is illustrated in
Figure 5-10, where the frequency response data of two components A and
B are combined and the result of this combination corresponds with C.
In order to add graphically magnitude ratio B to A, the ratio of 1 is used
as a reference line. This is obvious when one considers that if the output
equals the input, i.e., the ratio equals 1, then no matter how many com-
ponents are connected together, their output to iflput ratio will still
equal 1. It is also obvious because multiplication by 1 does not change
the multiplicand.
The illustration shows three different arrows, a, d, and g which have
been added to b, e, and h, respectively. For example, arrows a and b,
which are taken at the same frequency, are of equal length but opposite
polarity. Hence they cancel each other, which is expressed in C, by point
c being located at unity magnitude ratio. Adding d and e, the opposite
polarity has again to be taken into account. Therefore, the resultant/
74 AUTOMATIC CONTROL

corresponds to e diminished by d. The arrow i is simply the sum of g


and h, both being of equal polarity. This method of addition results in
the magnitude ratio C, which corresponds with the combined character-
istics of components A and B. The addition of phase angles shown in
Figure 5-10 uses a 0-degree phase angle as reference. When A lags by

A I B C

A I + I B I C I

Figure 5-10. Addition of magnitude ratios and phase angles.

110 degrees at a certain frequency and B leads by 60 degrees at the


same frequency, then the combined phase angle is — 50 degrees. The il-
lustrated addition is based on this method.

Unstable Components in a Stable Loop


It is interesting to observe that component A in Figure 5-10 is
unstable, since the phase angle is — 180 degrees at a frequency at which
the magnitude ratio is equal to unity. By adding B to A, the result is C,
and the combined response shows stability. This proves that a stable
control loop may contain components which, considered by themselves,
are unstable.

Proportional-Speed Floating Controller


Figure 5-11 shows the response of a proportional-speed floating con-
troller. As long as an input signal is applied, the output of such a con-
troller continues changing at a certain speed. When the change of the
FREQUENCY RESPONSE 75

Cycl«» per minute

Figure 5-1 1. Frequency response of proportional-speed floating action.

input signal is slower than the floating speed of the controller, then the
amplitude of the output becomes larger than that of the input. Hence
the slower the frequency of the signal, the greater the magnitude ratio.
The output of a proportional-speed floating controller reverses direc-
tion when the input cycle passes through zero. This is equivalent to a
phase lag of 90 degrees and accounts for the steady phase angle in
Figure 5-11.
Two graphs, (a) and (b), are shown in the illustration for the magni-
tude ratio. They correspond to dilferent floating rates of the controller.
The magnitude ratio at any given frequency is directly proportional to
the floating rate. For example, graph (b) corresponds to a floating rate
ten times that of graph (a). The phase angle remains unaffected.
A process without self-regulation behaves essentially like a propor-
tional-speed floating controller. For example, water in a tank with a con-
stant flow leaving the tank will continue changing in one direction for
a given flow into the tank. The phase lag would be likewise 90 degrees.
76 AUTOMATIC CONTROL

If a process without self-regulation is combined with a proportional-


speed floating controller the phase lag of the combination will be
90 -f 90 = 180 degrees, and at some point of the frequency response
curve the magnitude ratio will be 1.0. Hence, this is an unstable combi-
nation. The proportional-speed floating controller cannot be used for
processes without self-regulation.
Figure 5-12 illustrates the action of a proportional-speed floating con-
troller in a control loop. In this case, the loop is considered to consist

r-»— Process

1
i ^

Controller

Controller

Process

Control loop

o o O OOr-<_ CSI ^ lOOpp


o d d odd d d dd-r-i
Cycles per minute

Figure 5-12. Proportional-speed floating action in control loop.

only of controller and process. Graphs (a) and (c) are controller re-
sponses; graph (m) is the process response; graphs (b) and (d) are the
control loop responses of process and controller combined. Initially, the
floating rate of the controller is set for condition (a) and the response of
FREQUENCY RESPONSE 77

the control loop (b) results from combining it with the frequency re-
sponse of the process.
The phase angle graph (e) is the sum of the phase angles of process
and controller. At approximately 0.17 cycle/min., the phase angle is
—150 degrees. Comparison with the magnitude ratio curve shows that it
is possible to increase the floating rate of the controller considerably and
still remain within the conditions of stability.
Graph (c) corresponds to a floating rate about ten times that of (a).
The new control loop response is illustrated by graph (d). This response
has a magnitude ratio of 1.0 at 0.17 cycle/min., i.e., at the frequency at
which the phase angle is —150 degrees. At —180 degrees the magnitude
ratio is 0.5. Stability is thus still assured and the speed of response of
the controller has improved considerably.

Proportional-Position Controller
The frequency response curve of a proportional-position controller
would be a straight line, both for the magnitude ratio and the phase
angle up to relatively high frequencies. This is illustrated in Figure 5-13.

Figure 5-1 3. Frequency response of proportional-position action.

The principle of proportional-position action implies that a change of


the input signal is reflected in an immediate output signal of propor-
tional magnitude. However, when the input signal changes very rapidly,
the output signal can no longer follow, and attenuation of the signal as
78 AUTOMATIC CONTROL

well as delay occurs, which produces the decrease of magnitude ratio


and the increase of phase lag.
Graph (b) corresponds to a magnitude ratio ten times that of (a). This
is produced by a change in proportional band. Narrowing the propor-
tional band of the controller to one tenth increases the magnitude ratio
by ten. The phase lag is not altered.
Figure 5-14 is the response of a control loop containing a propor-
tional-position controller and a process with dead time only. Graph (a)

20

10
0.8
0.6
O

2 0.4
<u
\i)

0-1

Figure 5-14. Proportional-position action with dead time.

represents an unstable condition since at a 180 degree phase lag the


magnitude ratio is not 0.5 or less. If the dead time cannot be eliminated,
it is necessary to increase the proportional band.
Suppose the initial magnitude ratio is 1.0 for the process and the
controller, respectively. Hence the combined ratio is also 1.0. Doubling
the proportional band results in a magnitude ratio of 0.5 for both the
controller and the control loop. This corresponds to graph (b) in Figure
5-14 and assures stability.
The widening of the proportional band results in slower control
response. It is the compromise that any control system requires: greater
stability is equivalent to slower corrective action.
FREQUENCY RESPONSE 79

Steady-state Conditions. The condition which corresponds to a


frequency signal that approaches zero cycles is called the steady state.
It is equivalent to a control system that reestablishes balanced conditions
after a disturbance.
The magnitude ratio of a proportional-position controller at sufficiently
low frequencies is 1.0 with 0 phase angle. This implies that for zero
input the output will also be zero.
Offset Characteristics. Offset is an inherent steady-state character-
istic of the proportional-position controller. It cannot be overcome
because the output-input ratio of the proportional-position controller
does not increase at steady state and hence no additional impetus can
be expected. This is different from the proportional-speed floating
controller which shows an increase of magnitude ratio as the system
approaches steady-state condition. Comparison of Figures 5-11 and
5-13 illustrates the difference between the two control actions: one
graph is of negative slope, the other is of zero slope.
Proportional-position action, as compared with proportional-speed
floating action, offers the advantage of providing an inherently stable
component and permitting operation at larger magnitude ratios. This
increases the speed of response.
A combination of proportional-speed floating and proportional-posi-
tion actions, i.e., a proportional plus reset controller, becomes therefore
desirable, since it gives fast action and eliminates offset.

Proportional plus Reset Controller


Figure 5-15 shows the response of a proportional plus reset controller.
The reset rate of the controller is expressed by the so-called “break
frequency.” This is illustrated by the dashed lines in Figure 5-15. Both
the slanted line, which corresponds to the reset action, and the hori-
zontal line, which corresponds to the proportional-position action, are
prolonged as straight lines. The frequency at which these two lines
intersect is the break frequency. The phase lag is 45 degrees at the break
frequency.
The reset rate is obtained by multiplying the break frequency by lir
or 6.28. This means that
/ = 6.28/
where / is the reset rate in repeats/min. and/is the frequency in cycles/
min. Thus, in Figure 5-15, graph (a), the break frequency is 0.1, and
hence the reset rate is 6.28 x 0.1 = 0.628 repeats/min.
When the reset rate is changed the angle which the slanted line forms
80 AUTOMATIC CONTROL

Figure 5-15. Frequency response of proportional plus reset action.

with a horizontal line remains unchanged. The slope of this line always
expresses a decrease of the magnitude ratio by roughly Tio for an in-
crease of frequency by ten times.
The magnitude ratio graph (a) can be shifted in either the horizontal
or the vertical plane. Shifting it in the horizontal plane means changing
the reset rate. For example, graph (b) results from an increase of reset
rate to 1.25 repeats/min. Similarly, by narrowing the proportional band
to VA, i.e., increasing the magnitude ratio four times, a horizontal shift
of graph (a) is obtained and graph (c) results.
The phase angle is unaffected by changes of the proportional band,
but it shifts when the reset rate is adjusted. This has some effect on the
final adjustment of the controller. It is, however, of little practical con-
sequence, as will be illustrated in the following example, and can usually
be ignored.
Adjustment of Proportional plus Reset Controller. In the previous
chapter it was described how the reset action of a proportional plus
reset controller can be adjusted by using the relation;

where / is the reset rate, and T is the time between two successive
maxima of oscillations at the critical setting.
The critical setting which was defined as the widest proportional band
FREQUENCY RESPONSE 81

at which continuous oscillations occur, corresponds to a magnitude


ratio of 1.0 at a phase lag of 180 degrees. The time T is the period, i.e.,
the inverse of the frequency, at this condition.
On this basis, it is possible to determine the controller adjustments
from the frequency response data of a control loop. These data should
be taken with the reset action omitted.
Suppose the results are those of Figure 5-16 where the magnitude
ratio is 2.0 for frequencies up to about 0.04 cycles/min. In order to

Cycles per minute

Figure 5-16. Frequency response of control loop without reset.

determine the proportional band, the frequency at a phase angle of


— 150 degrees is determined. In this case, it is about 0.27 cycles per
minute. In accordance with the rule 1 of the stability conditions, the
82 AUTOMATIC CONTROL

graph of the magnitude ratio is shifted in a vertical plane until a ratio


of 1.0 coincides with this frequency. The dashed line corresponds to this
condition.
It will be found that the new amphtude ratio is four times the old one
at any frequency. This means that the controller sensitivity can be
increased by a factor of four, which is equivalent to reducing the pro-
portional band to VA its previous value.
To ascertain the reset rate, it is necessary to determine the frequency
at a 180 degree phase lag, which in the present case is 0.4 cycles/min.
Since the period is the inverse of the frequency this corresponds to
a period of 2.5 min. The reset rate is then
1 2
/= = 0.48 repeats/mm.

Figure 5-17 illustrates the change of the control loop response due to
the addition of reset action. The condition of the dashed line of Figure
5-16 is combined with the added reset. The phase angle is represented
in three graphs. One, (a), is the response without reset action; (b) is the
phase angle due to proportional plus reset action; and (c) is the result-
ing combination. Although the (c) graph differs considerably from the
original (a) graph, both show the phase lag of 150 degrees at practically
the same frequency. Hence the stability condition remains unaltered.

Proportional plus Reset plus Rate Controller


The addition of rate action provides corrective action in proportion
to the speed with which a controlled variable deviates from its set point.
It is used to compensate for large time constants in the process or in the
measuring element.
Since the effect of rate action is proportional to the speed of devia-
tion, it must also increase its effect with the frequency of a test signal.
Figure 5-18 shows the frequency response of a proportional plus reset
plus rate controller. The increase of the magnitude ratio after about 0.3
cycles/min. is due to rate action. The graph has two break frequencies:
one at 0.08 cycles/min., the other at 0.4 cycles. From the first break
frequency the reset rate can be determined as previously described. In
this case it is 0.5 repeats/min. The rate time can be determined by using
the equality:
1 0.16
D = or D =
2^f /
where D is the rate time in minutes and/is the frequency in cycles/min.
Thus in the present case
FREQUENCY RESPONSE 83

p. 0.16 A
D = —— = 0.4 min.
0.4
The slope which the slanted hne forms with the horizontal is the same
whatever the rate time is. Its magnitude is also the same as the slope of
the reset action, only it is positive instead of negative. Increasing, for
example, the rate time to 0.8 min., shifts the right-hand slanted portion
of the graph to the left until a break frequency of 0.2 cycles/min. is

Figure 5-1 7. Frequency response of control loop with reset.


84 AUTOMATIC CONTROL

By changing the proportional band, the entire graph is shifted in a


vertical plane. For example, by cutting the proportional band in half,
graph (b) is shifted to become graph (a).
The phase angle becomes modified by the added rate action, and the
output signal leads the input signal at frequencies that are under the in-
fluence of rate action. In Figure 5-18, this starts at about 0.2 cycles/min.

Figure 5-18. Frequency response of proportional plus reset plus rote action.

Adjustment of Proportional plus Reset plus Rate Controller. In


the previous chapter which discussed the adjustments on the basis of
step function response, it was pointed out that the proportional band
should be twice the critical setting for proportional-position controllers
with or without reset. It was mentioned that the proportional band
could be narrowed to 1.7 times instead of twice the critical setting when
rate action is added. This means that the magnitude ratio can be increased
because of rate action. The reason is that the effect of the phase shift
becomes noticeable, as will be shown further below.
For the same reasons, when adjusting from the frequency response
as in Figure 5-19, the magnitude ratio of 1.0 no longer refers to a phase
FREQUENCY RESPONSE 85

Figure 5-19. Frequency response of control loop without and with


rate and reset.

angle of —150 degrees, but to one of —160 degrees. Figure 5-19 shows
in graph (a) the same frequency response as in Figure 5-16. Reset and
rate actions are cut off. Adjusting the response of Figure 5-19 for mini-
mum proportional band at stable conditions results in the dashed line
(b). The magnitude ratio is now 1.0 at —160 degrees and at low fre-
quencies it is 10.0 instead of 2.0, which is 5 times the initial ratio or 0.2
times the initial proportional band.
To determine the reset and rate actions, the corresponding expres-
sions of the previous chapter are used. They were:

where / is the reset rate in repeats/min., D is the rate time in minutes


86 AUTOMATIC CONTROL

and T is the period in minutes referred to the critical setting, i.e., to a


— 180 degree phase angle. In the present case, the period at —180
degrees is 2.5 min. Hence the reset rate is 0.80 repeats/min. and the rate
time is 0.31 min.
Adjusting reset and rate actions accordingly, and then repeating the
frequency response test of the control loop, results in graph (c) of Figure
5-19 both for magnitude ratio and phase angle. The phase angle in
graph (c) is the result of adding the process phase angles (a) to the
controller phase angles (d). It can be seen that at the frequency at which
curve (a) showed —160 degrees, the phase angle now shows only —150
degrees. This justifies the initial assumption of 160 degrees in case of
rate action instead of the conventional 150 degrees.

Conclusions
The preceding description was illustrated by graphs which were more
or less idealized. Actual readings frequently show considerable random
variations for which a smooth curve only indicates a trend. In view of
these conditions a differentiation between 150 and 160 degrees may
become impractical.
The question arises as to the accuracy of this method. This points to
the limitation of frequency response data used for the setting of auto-
matic controllers. For a system which is actually installed, the step
function response method is simpler and more reliable than the fre-
quency response method. The advantage remaining is that frequency
response data allow the theoretical combination of several components
without their being installed, and the estimation of what approximate
magnitude their settings will be.
6 MECHANICAL COMPONENTS

The measuring means of a controller converts the magnitude of the


controlled variable into either a mechanical or an electrical signal. In
the majority of cases, the mechanical input signal is a pressure which
is applied to a diaphragm, bellows, or Bourdon tube to convert it into
a force or position. Occasionally force or position is obtained without
conversion, as in the deflection of a bimetallic element or the displace-
ment of a float in a level controller. However, even then, a secondary
conversion into a mechanical or electrical signal is usually desirable.
Mechanical systems generally use hydraulic or pneumatic fluids.
Figure 6-1 shows the most common components of such control sys-

Figure 6-1. Components of pneumatic and hydraulic control systems.

terns. In a pneumatic system, for example, the deflection of the bellows


positions a flapper of a flapper-nozzle combination. This results in a
change of the pressure of a secondary air supply which is applied to the
87
88 AUTOMATIC CONTROL

diaphragm of a final control element. In order to amplify the output


from the flapper-nozzle, a pilot valve may be connected between it and
the final diaphragm. In some cases the pressure which is applied to the
bellows as input signal to the controller can be used directly at the
diaphragm of the final control element. The resulting simplification is
desirable. However, in many cases this primary signal would not be
powerful enough. It would furthermore limit the possibilities of control
actions such as proportional-position, reset, and rate controls.
In a hydraulic system, the flapper-nozzle may be used in connection
with a four-way valve or, if the input force is high enough, the flapper-
nozzle stage may be eliminated and the four-way valve be positioned
directly. The cylinder’s actuating piston, which is linked to the final
control element, is positioned through the four-way valve. Another
device is the jet pipe which either positions an actuating piston directly
or acts through a four-way valve.
Internal feedback arrangements produce numerous interconnections
and repetitions of all these components. Thus a flapper-nozzle may be
used in the measuring means, again in the controller, and again in the
positioning of the final control element.
The description of various components in the succeeding pages follows
largely the order outlined in Figure 6-1. An exception is made with one
element which is probably the most universally used. This is the spring
and it will be discussed first.

Springs
One of the most universal elements in mechanical control elements
is the spring. The helical spring has a deflection which is proportional
to the force applied. Written as an equation, this means that
F = kx
where F is the force in lb, k is the spring rate in Ib/in. deflection, and
X is the deflection in inches. This equation is known as Hooke’s law. It
is valid within the elastic limits of the spring. The practical behavior of
the spring does not necessarily follow completely the idealized Hooke’s
law. Deviations from linearity of as much as 5 per cent of the total
deflection must be expected.
The spring rate is given by

k=-^
8T)W
where G is the torsional modulus of rigidity, which for steel wire is
approximately 11,500,000; d is the diameter of the spring wire in inches.
MECHANICAL COMPONENTS 89

as shown in Figure 6-2; D is the coil diameter in inches, as shown; and


N is the number of active coil. The equation shows that minute changes
of coil diameter and even smaller changes in spring wire diameter will

produce large changes in the spring rate. This is the main reason for
the difficulty of controlling spring rates in production. The usual method
is to select from a production run those that come closest to a specified
spring rate, and even then deviations of 5 per cent of the spring rate are
generally accepted.
The above equation also shows that the spring rate is inversely pro-
portional to the number of coils. It is therefore possible to increase
spring rates by cutting out coils. For example, decreasing the number
of coils by 50 per cent increases the spring rate twofold.
Helical springs are either compression springs or tension springs, as
shown in Figure 6-3. The above mentioned equations are valid for either
form. The deflection x refers to compression as well as tension.
Figure 6-4 illustrates the difference between spring rate and spring
force. Graph (1) illustrates a spring rate of 100 Ib/in. If this corresponds
to a spring with 7 coils, then the same spring with 10 coils would have
a spring rate of 70 Ib/in., as illustrated in graph (2). Using, however.
90 AUTOMATIC CONTROL

Figure 6-3. Tension and compression springs.

Figure 6-4. Spring rote and spring force.

the 7-coil spring and precompressing the spring by 0.3 in. increases the
spring force correspondingly, as shown in graph (3).
Springs may be combined either in series or in parallel arrangement.
The basic combinations are those seen in Figure 6-5. Case a is a series
connection of two coils with spring rates and /ca- The combined spring
rate k is smaller than that of either /ci or k^. It can be calculated by the
equation

ki k2

Cases b and c are parallel connections. Since each coil acts on the weight
MECHANICAL COMPONENTS 91

(a) (b) (c)


Figure 6-5. Spring combinations.

directly, the total spring rate is the sum of the individual spring rates,
or
k = k-^ + k^.
Frequently, the springs are used in connection with levers, as shown
in Figure 6-6. The force which is exerted by the weight on the spring in case

ais W, which is balanced by the spring with a force equal to kx. Increas-
ing the lever of the spring force as in case b increases the acting force
92 AUTOMATIC CONTROL

M
by and the deflection of the spring for the same motion of the lever
also increases by —. Hence the weight now becomes balanced by a spring
force which is equal to

or F = —kx
m m nF
which shows that the effect of the lever is to increase the spring rate by
a factor which is equal to the square of the lever ratio.

Diaphragms
Pressure applied to an area is force, i.e., P • A = F. This means that
once the pressure is given, the force can be derived by knowing the ef-
fective area to which the pressure is applied. If a diaphragm is used, the
effective area is less than the actual area of the diaphragm. The pressure
that acts on the diaphragm, as shown in Figure 6-7, produces a force

I PRESSURE

that is absorbed in part by the clamping of the periphery of the dia-


phragm. The remaining force is available to compress the spring.
Slack Diaphragms. Diaphragms of sheep or goatskin leather are
probably the most sensitive. However, synthetic materials, generally
with a fabric layer, have largely superseded the leather diaphragms. The
convolution is either molded into the material, or sufficient slackness is
provided so that the convolution is formed by the pressure loading.
Considering the lowest part of the convolution of the diaphragm, the
effective diameter of the diaphragm changes as illustrated in Figure 6-8.
MECHANICAL COMPONENTS 93

Figure 6-8. Various deflections of diaphragm.

The mid position is illustrated in Figure 6-8a, while in 6-8b the pressure
has decreased. This moves the diaphragm upward due to the force of a
spring which is not shown, and changes the shape of the convolution. As
a consequence of the new shape of the convolution, the effective diam-
eter increases. Conversely, in Figure 6-8c the pressure has increased,
the diaphragm has moved downward and the convolution is now of
such shape that the effective diameter has decreased.
Area changes with the square of the diameter, hence a relatively small
change in effective diameter will have a considerably more pronounced
effect on the area and hence on the force that is produced by the pres-
sure applied to the diaphragm.
The change in effective diameter and hence of the effective area has
the result that the force is no longer proportional to the pressure. A non-
linear element is the result. In order to minimize this nonlinear effect, the
stroke should be reduced to a minimum. This is not always possible, par-
ticularly in spring-opposed diaphragm valves where strokes are frequently
2 or more inches. In such cases, the convolutions are made relatively
deep to reduce the shifting of the effective diameter. This has a side ef-
fect insofar as it produces a slightly different shape of convolution de-
pending on whether the valve moves upward or downward. The conse-
quence is the equivalent of backlash.
“Belioframs.” In order to increase the linear effect of diaphragms,
the Bellofram Corporation developed a particular shape of diaphragm,
called the “Bellofram,” in which the deep convolution of the molded
synthetic diaphragm rolls between two surfaces, as illustrated in Figure
6-9. The effective area is maintained constant for all practical purposes.
94 AUTOMATIC CONTROL

Figure 6-9. ‘‘Bellofram.”

Metal Diaphragms. Slack diaphragms have a pressure rating that


generally limits them to pressures of 20 psi or less. Metal diaphragms
are less limited in this respect. Their structural rigidity is of advantage
in designing components. Generally, they can be used without counter-
acting springs.
Metallic diaphragms may cause nonlinearities because of their rapid
change of effective area with deflection. This can be made negligible by
reducing the deflection. Flat plate diaphragms must not be deflected by
more than about 16 of their thickness. This means a motion in the order
of a thousandth of an inch, which is quite acceptable in many compon-
ents. In order to improve the linearity over somewhat longer strokes,
corrugated metal diaphragms are frequently used. The corrugations are
generally concentric.

Bellows
Further increase of deflection, relative rigidity, and comparatively
high pressure rating are characteristic of bellows. The amount of deflec-
tion per unit change of pressure is a function of the number of corruga-
tions, the spring rate, and the area of the bellows. Thus a typical bellows
may have a spring rate of 600 lb/in. corrugation. Providing ten convolu-
tions, a spring rate of 60 Ib/in. results. Let the area of the bellows be
0.3 sq in.; with a pressure change of 20 psi, the bellows will then deflect
0.1 in.
Hysteresis and nonlinearities of several per cent of maximum stroke
may have to be expected with bellows. To improve performance, a
MECHANICAL COMPONENTS 95

helical spring is used parallel with the bellows. The spring rate of the
helical spring should be ten times that of the bellows.

Flapper-Nozzle
The flapper-nozzle as illustrated in Figure 6-10 is an adjustable area
in a flow passage. In this respect it behaves like a valve. The advantage

0
Figure 6-10. Flapper-nozzle.

over a valve is the simplicity of construction and the smallness of force


required to position it. The adjustable flow restriction of the flapper-
nozzle is used in series with a fixed restriction. Supply air enters the
fixed restriction at an absolute pressure p^. It drops from to pt in
passing through the fixed restriction, and fromptiopr in passing through
the nozzle. A restriction with a passage of negligible length is considered
an orifice. This applies to both fixed and adjustable restrictions. The
flow which comes through the fixed restriction leaves through the nozzle
and the rate of air flow Q can be approximately expressed by

Q = kA, V(TS - Pt)Pt = kA^ ^{pt - pr)pr


All pressures are absolute pressures, A i and A 2 are the areas of fixed re-
striction and nozzle, respectively, and /c is a constant which depends on
the flow pattern and the conditions of the fluid. The above equation can
be written

V(/^S - Pt)Pt = A^^{pt - Pr)Pr (6-1)

After squaring and rearranging, the following equation results:


96 AUTOMATIC CONTROL

Solving this equation forpt gives

Only one solution, however, gives positive results and these are the only
ones of practical significance. The equation, therefore, becomes

Suppose the supply pressure is 30 psia and the reference pressure is


atmospheric, i.e., approximately 15 psia. Inserting these values in the
above equation and simplifying gives

p, = (l5 - 7.5 + 15 yo.25^ + 1 (6-2)

The area of ^2 is changed by means of the flapper, hence the ratio between
A and
2 can assume any value within certain limits. Suppose that A 2

varies from O.IA^ to 2^41, then a number of ratios can be assumed


within these limits, and the corresponding magnitudes of pt can be cal-
culated. The nozzle back pressure expressed in gauge pressure is ap-
proximately equal to — 15).
The following is a table of computed values for the above conditions:

A,/A, Pt . Nozzle Back Pressure A,/A, Pt Nozzle Back Pressure


in psi in psi

0.2 29.3 14.3 1.2 22.7 7.7


0.4 28.8 13.8 1.4 21.4 6.4
0.6 27.5 12.5 1.6 20.3 5.3
0.8 26.0 11.0 1.8 19.4 4.4
1.0 24.3 9.3 2.0 18.5 3.5

These values are plotted as curve in Figure 6-11 and show how the
nozzle back pressure varies with the position of the flapper. The curve
illustrates the nonlinearity in this relation, but also makes it clear that
the linearity can be improved considerably by operating only over a
limited range of area ratios, e.g., from ^43/^1 = 1.2 to = 1-4. This
results in correspondingly small pressure changes and is one of the rea-
sons that pilot valves, as described later, are used to multiply the small
nozzle back pressure change into a 3 to 15 psi controller output signal
range.
Nozzle diameters are in the order of magnitude of V64 in. This cor-
MECHANICAL COMPONENTS 97

responds to an area of the nozzle opening about 0.0002 sq in. Since the
force of the air stream impinging on the flapper is approximately equal
to the maximum force that may result when the flapper com-
pletely closes the nozzle is 0.003 lb. Actually, even this minute force is
not obtained, because the nozzle is always partially open. On the other
hand the simple fact that a force exists points to certain limitations. For
example, if the nozzle opening should be increased to Vs in., the force
would increase 64 times and would no longer be negligible. Similarly,
the air pressures that can be handled cannot be increased indefinitely.
Another reason for operating with small nozzle diameters and low
nozzle back pressures is to reduce the air consumption. In general, the
air consumption is in the order of 1 to 2 standard cu ft/hr. For this
purpose the fixed orifice is generally about Vi 28 in. in diameter.
The flow passes through the nozzle, impinges on the flapper and is
deflected at right angles through an area which is given by the cir-
cumference of the nozzle opening and the distance .s’ between flapper
and nozzle (Figure 6-10). Considering a fixed orifice of Viis in. in
diameter, this corresponds with an area of approximately 0.00005 sq in.
For a nozzle V64 in. in diameter the circumference is about 0.05 in. For
an area ratio of A2/A ^ = 2, the distance .s would be given by
0.00005 0.002 in.
s X 2
0.05
This clearance is, in fact, so small that it is not practical to close the nozzle
tighter than this. Previous calculations have shown that for a supply
98 AUTOMATIC CONTROL

pressure of 15 psi (30 psia), the nozzle back pressure for this area ratio
is 3.5 psi, which consequently is usually the maximum back pressure in
the flapper-nozzle.
In order to obtain 15 psi as maximum controller output signal, an
amplification factor of 15/3.5 = 4.3 is required in the pilot valve. For a
minimum output of 3 psi, a nozzle back pressure of 0.7 psi would then
be required. From equation (6-2) it can be computed that this nozzle
back pressure is the result of an area ratio of approximately AJA^ = 3.
This needs a distance between flapper and nozzle of

0.00005
s = X 3 = 0.003 in.
0.05

In other words, the flapper motion of 0.003 — 0.002 = 0.001 in. regulates
the controller output signal over its entire range. Smaller ranges of
flapper motion, e.g., 0.0006 in., are by no means uncommon.
The fluid for the flapper-nozzle as described here was air or any other
gas. The flapper-nozzle can be and is used also for liquids, specifically,
hydraulic oil. The characteristics are similar, even though the flow equa-
tions as given above do not apply. Hydraulic operation of flapper-nozzles
has, however, its difficulties. High pressures occurring with hydraulics
produce forces on the flapper which are no longer negligible. Further-
more, it is more difficult to keep oil clean than air, and the extremely small
diameters of the flow resistances clog easily, unless all impurities are
continuously removed.
Other Forms of Flapper-Nozzles. Figure 6-12 shows a push-pull
arrangement of a flapper-nozzle. When the flapper moves toward one

PILOT VALVE

t
SUPPLY

Figure 6-12. Push-pull flapper-nozzle.

nozzle it moves away from the other. The effect is therefore to increase
the nozzle back pressure on one and simultaneously to decrease it on
MECHANICAL COMPONENTS 99

the other. The pilot valve must be designed so that it operates on a pres-
sure differential. The push-pull flapper nozzle is occasionally used in
hydraulic circuits where the differential pressure developed is applied
across the spool of a four-way valve.
Figure 6-13 shows the free-vane principle used by the Bristol Co. The
vane takes the place of the flapper. The main advantage is that flow

SUPPLY

Figure 6-13. Free vane principle. (The Bristol Co.)

forces on the vane are at a minimum because the motion of the vane is
vertical to the air stream, letting the nozzles and sidewise forces balance
each other because of the twin nozzle design. The edge of the vane is
between the two nozzles. It operates with a motion of 0.005 in., changing
the nozzle back pressure from 4.5 to 5.5 psi.
Other Pneumatic Flow Restrictions. The orifice is only one way of
providing a restriction in a pneumatic circuit. Another method consists
of inserting a length of tubing of considerably reduced diameter, a so-
called capillary. Flow through a capillary follows the equation

Q = 10,142,220^-^^^-;"^?^
r]L

where Q is the flow in cu in./min., D is the inside diameter of the capil-


lary in inches, and are the pressures in psi before and after the
capillary, t] is the viscosity of the fluid in centipoise and L is the length
of the capillary in inches.
Since in this equation, the diameter appears in the fourth power, it
is readily seen that by decreasing the diameter the rate of flow dimin-
ishes rapidly. It is also interesting that the flow is inversely proportional
to the length of the capillary. This provides the possibility of adjusting
the flow rate by lengthening or shortening the capillary.
100 AUTOMATIC CONTROL

Another means for providing a flow restriction is the needle valve. Its
flow is expressed by
ma^U (d/Ly + (1 + d/Ly
Q = 1900 (Pi -P2)
V 1 - 2 (d/L)
where m, L, and d are dimensions shown in Figure 6-14 expressed in
inches, the angle a is also shown in Figure 6-14 and is expressed in

degrees. The viscosity TJ is in centipoises and the flow Q in cu in./min.


Adjustments are usually made with one or two turns of the needle valve
handle and cover ranges of 500:1.

Pilot Valves
The pilot valve is a pneumatic amplifier. The flapper-nozzle is designed
for small flow capacities, but for fast action of final control elements,
larger capacities are needed. Furthermore, the limited pressure range of
the flapper-nozzle also requires amphfication.
Two basicTypes of pilot valves are used: the continuous-bleed and
the non-bleed type. The continuous-bleed type, illustrated in Figure
6-15, has the advantage of simplicity of construction. The nozzle back
pressure is admitted to a pressure-sensitive element; in this case, a dia-
phragm. A spherical valve plug is connected to the diaphragm by means
of a valve stem. In the illustrated position, the exit, which connects to
the atmosphere, is closed by the plug and the controller output signal
consists of the undiminished air supply pressure. In this position, the
nozzle, to which this pilot valve is connected, is obviously wide open
since the diaphragm in the pilot valve is not deflected at all. Should the
MECHANICAL COMPONENTS 101

Supply Atmosphere
Supply Exhaust
port port

Controller
output signal

Figure 6-15. Continuous-bleed pilot valve.

nozzle back pressure increase as the flapper moves closer toward the
nozzle, the diaphragm would deflect and the valve plug would, at least
partially, open the port to atmosphere and close by a corresponding
amount the port that admits air from the supply line. The maximum
deflection of the pilot valve diaphragm would block olf the air supply
and release all excess air in the controller output signal line to reduce
the signal to atmospheric pressure. All intermediate positions of the pilot
valve diaphragm produce intermediate signal pressures in the output.
As shown in the diagram of Figure 6-15, the supply port and the
exhaust port are two resistances through which the air may flow. The
pressure between the two resistances is the controller output pressure.
Supply port resistance and exhaust port resistance may be visualized as
being coupled together so that when one resistance increases, the other
diminishes automatically, and vice versa.
The nonlinearity between valve plug motion and controller output
signal is considerable as shown in Figure 6-15. Hence, in general it is
used in this form only for on-off control, when the valve plug assumes
either a fully open or a fully closed position. Satisfactory linearity, how-
ever, can be obtained by feedback methods. The general procedure is
to convert the controller output signal into a position by deflecting a
102 AUTOMATIC CONTROL

bellows in proportion to the pressure. This position is then compared


with the flapper position as will be described in Chapter 8. The non-
linearity of the pilot valve has no effect in this case.
Figure 6-16 illustrates the action of the non-bleed type. Suppose the
nozzle back pressure increases due to the flapper approaching the nozzle.
To nozzle

^Vpper diaphragm To
atmosphere
M'' *■
^ ("""I
Valve plug <—*• Controller
output
signal
Air
Supt>l/

Spring

Figure 6-1 6. Non-bleed pilot valve.

This will deflect the upper diaphragm of the pilot valve and with it the
lower diaphragm. It is assumed that no amplification of the change in
nozzle back pressure takes place in the pilot valve, although all that
would be required for this purpose is a lower diaphragm of smaller area
than the upper one.
As the diaphragm assembly moves downward, the valve plug is also
pushed down. This action results in an opening of the valve port and
admission of supply air to the signal line. The signal pressure which is
now rising pushes against the lower diaphragm. Once the signal pres-
sure has increased sufflciently to equal the nozzle back pressure apphed
to the upper diaphragm, the forces across the diaphragm are in balance,
the valve is closed again, and all conditions are the same as shown in
Figure 6-16; but the signal pressure of the controller output is now at
a higher value. Nonlinearity in the diaphragm response and in the
characteristic of the pilot valve does not influence the action.
When the nozzle back pressure decreases, the diaphragm deflects
upward. This opens the outlet port to atmosphere and the signal line
releases air through this port until the force balance across the diaphragm
is reestablished, and the outlet port to atmosphere is again closed.
The non-bleed pilot valve requires a precompressed spring to press
the valve plug against the ports as illustrated in Figure 6-16. The force
exerted by the nozzle back pressure to deflect the diaphragm downward
must be equal to the signal pressure times the effective area of the dia-
phragm plus the force that results from the precompression of the spring.
If, however, the nozzle back pressure decreases and the diaphragm
MECHANICAL COMPONENTS 103

moves upward, the acting force is only that of the controller output
signal applied to the lower diaphragm. The spring force does not enter.
In addition to the spring force there are pressures active across the
supply and the exhaust plugs. Calling the supply pressure the con-
troller output signal pressure p^, the atmospheric pressure p„ the area of
the diaphragm Aa, the area of the supply plug A^, of the exhaust plug
Ag, and the spring force due to precompression F^, the total force active
in downward direction in case of a small increase in nozzle back pres-
sure pn is
Pn - [pcAd + {Ps - pMs + iPc -pMe + T’J (6-3)
while in the case of a small decrease in nozzle back pressure, the forces
in an upward direction are
-PcAa -f Pn (6-4)
Hence the nozzle back pressure may change by an amount equal to
the difference between (6-4) and (6-3) or
ip, -pc)A, -h (pc -pMe + F,
without moving the diaphragm, i.e., without changing the controller out-
put signal. A dead zone is thus created which can be minimized by
design, but not eliminated.
The capacity of a non-bleed pilot valve is generally larger than that of
a continuous-bleed pilot valve. A good design of the non-bleed type
takes about 15 seconds to build up the pressure from 3 to 15 psi in
a 350 cu in. volume. The continuous-bleed type may need 3 to 4 times
that much.

Jet Pipes
The jet pipe is widely used with hydraulic control systems. Figure
6-17 is an illustration of jet pipe action in its conventional form. The jet
pipe swings in a vertical plane around a horizontal axis, supported on its
left by a ball bearing pivot, and on its right by a sleeve bearing through
which a minute amount of oil is allowed to leak into the common sump
(not shown). This method all but eliminates the friction in the suspen-
sion. The pressure at which the oil is pumped through the jet pipe de-
pends on the application and is usually somewhere between 100 and 400
psi. It is kept constant at the predetermined pressure. From the jet pipe
tip the oil is directed to the two closely adjacent receiving orifices in the
distributor block, which are in turn connected with the opposite ends of
the actuating cyhnder.
With the jet pipe in its midposition, the same pressure exists in either
104 AUTOMATIC CONTROL
SIGNAL
FORCE

Figure 6-17. Schematic drawing of jet pipe action.

receiving orifice, consequently the pressures on either side of the actuat-


ing piston are equal. With the slightest deflection of the jet pipe,
one orifice will receive more oil than the other, thus creating a difference
in pressure between the two ends of the actuating cylinder and causing
the piston to move in response to the jet pipe deflection.
As the oil leaves the converging jet pipe tip, its potential energy
is changed into kinetic energy. As it enters the receiving orifices, kinetic
energy is again changed into potential energy. This conversion of the oil
pressure into velocity pressure in the jet stream between jet pipe tip and
receiving orifice, and the subsequent reconversion into the static recovery
pressure, results in a certain pressure loss. The recovery is generally be-
tween 85 and 96 per cent, which means that with a 100 psi supply pres-
sure, the pressure at the actuating piston would be 85 to 96 psi when the
jet pipe is fully deflected.
The relation between jet pipe displacement and static recovery pres-
sure in the receiving orifices is shown in Figure 6-18. Under normal
operation, the jet pipe deflection is limited by mechanical stops to about
0.035 in. on either side from center. The action is linear around the mid-
position, though nonlinearities creep in toward the extremes.
MECHANICAL COMPONENTS 105

-0,040 -0.030 -0.020 -0.010 0 +0.010 +0.020 +0.030 +0.040


JET PIPE DEFLECTION -INCH

Figure 6-18. Static recovery pressure as function of


jet pipe deflection.

Various jet pipe tips are available. Typical sizes and corresponding
flows in cu in./min. at various supply pressures are as follows:

Jet Pipe
Tip Diameter 100 psi 200 psi 300 psi 400 psi
in Inch

0.050 115 cim 162 cim 200 cim 230 cim


0.065 194 275 336 388
0.080 294 415 510 588
0.100 460 650 800 920

Flow rates as shown are applicable when the jet pipe is fully deflected
and the actuating piston does not deliver work. In the measure as load
is applied to the piston, the load flow and hence the speed of the piston
decrease. This is illustrated in Figure 6-19 which shows the decreasing
load flow due to load pressure as well as to different deflections of the
jet pipe.
One of the essential advantages of the jet pipe is that flow forces are
practically non-existent, hence it can be operated with very weak input
signals. On the other hand it is limited in its power output.

Foor-way Valves
Where high power outputs are needed, the four-way valve is the most
frequently used component in hydraulic circuits. It is frequently posi-
tioned by a jet pipe or a flapper-nozzle. If the input signal is strong
enough, the spool of the four-way valve can also be positioned directly.
Figure 6-20 shows the basic concept of a four-way valve. In position
106 AUTOMATIC CONTROL

20 40 60 80 100
LOAD PRESSURE IN PSI

Figure 6-19. Typical pressure-flow characteristic of jet pipe.

Figure 6-20. Schematic drawing of four-way valve.


MECHANICAL COMPONENTS 107

a a certain hydraulic volume is trapped on both sides of the actuating


piston. When the input signal displaces the spool to the left, as in posi-
tion b, a connection between supply and the right side of the actuating
cyhnder is established, as well as a connection between left side and
drain. The resulting pressure difference moves the piston to the left.
Ideally speaking, the spool in its neutral position should close the
ports to the actuating cylinder and the slightest displacement should
open them. Under manufacturing conditions this can hardly be obtained,
and either overlap or underlap is the result as illustrated in Figure 6-21.

Figure 6-21. Underlap and overlap of four-way valve.

With overlap the lands on the spool are wider than the ports and there
must be a certain displacement of the spool before the ports actually
open. This approaches the conditions of a dead band which is obviously
undesirable, and the underlapped eondition is preferred.
The underlapped condition produces leakage flow around the valve
ports. This results in a basic change in the behavior of the valve when
load is applied to the actuating piston. With overlap and at a fixed piston
position, the full hydraulie pressure is available across the load. The load
sensitivity as illustrated in Figure 6-19 for the jet pipe does not exist for
the jet pipe under these conditions. With underlapped conditions,
however, a bypass around the load exists, and as long as the displace-
ment is not larger than the underlap, this bypass produces a pressure
drop in the supply port which increases with the load.
This condition is illustrated in the two graphs of Figure 6-22. These
graphs were published by J. L. Shearer.* The symbol x designates the
spool displacement. For a valve with zero underlap, corresponding to the

* Shearer, J. L., “Dynamic Characteristics of Valve-Controlled Hydrauhc Servomotors,”


AS ME Transactions, (Aug. 1954).
108 AUTOMATIC CONTROL

LOAD PRESSURE IN P S I

X= DISPLACEMENT OF SPOOL FROM CENTER POSITION

Figure 6-22. Pressure-flow characteristics of typical four-way valve.

Upper diagram, the smallest displacement provides the full break-loose


pressure of 950 psi across the actuating piston. With an underlap, as il-
lustrated in the lower diagram, the characteristic changes completely.
With a displacement of 0.001 in., the maximum obtainable pressure is
only 250 psi.
The fluid that passes through the ports of the four-way valve causes
flow forces along the spool axis. Figure 6-23 illustrates the flow through
the valve ports. The flow enters at a certain pressure and velocity. At the
metering port, i.e., at the outlet port, the velocity is greatest because the
cross-sectional flow area is smallest. Hence the static pressure at the
metering port is smallest. On the other hand, the velocity at the inlet
MECHANICAL COMPONENTS 109

port is comparatively low and the static pressure is therefore high. The
result is that the pressure components along the axis of the spool are
larger toward the left than the right and tend to close the valve. Exami-
nation of Figure 6-23 shows that the flow may be reversed without alter-

Figure 6-23. Flow through ports of four-way valve.

ing the direction of the flow force. The general statement holds therefore
true that flow forees oppose the opening of such a valve. Their magni-
tude can be expressed within certain assumptions by
F = 0.0000750 - P2 (6-5)
where Tis the flow force in lb, Q is the flow in cu in./min., and (p^ — pz)
is the pressure drop through the port in psi.
In a practical case, the pressure drop may be 900 psi, the flow 500 cu
in./min., and the resulting flow force would be about IVs lb.
Equation (6-5) is based on the assumption that diameter d (Figure
6-23) is small in comparison with the diameter D of the piston spool.
By making d as large as practical, the velocity at the inlet port can be
increased, and the flow forces will decrease correspondingly.
Special valve configurations have also been developed, such as the
one shown in Figure 6-24, providing streamlined surfaces which cancel
most of the flow forces. The relatively high cost of machining the neces-
sary parts is the main disadvantage of such construction.
Hole-and-plug Valve. This is a special configuration of the four-way
valve which has been developed at the Massachusetts Institute of Tech-
nology. The purpose of this design is to make a valve that approaches
zero lap at a lower cost than is possible with the spool type discussed so
far. As illustrated in Figure 6-25, the valve consists of a plate and a body,
both made of rectangular stock and their flat, ground surfaces facing each
other with a minimum of clearance.
no AUTOMATIC CONTROL

Figure 6-24. Flow through force-compensating ports of four-way valve.

Essentially, the operation is the same as in any four-way valve, i.e., it


consists of a relative displacement of lands and ports. The difference in
the two approaches lies in the production method. The blackened parts
in Figure 6-25 are inserts. To prepare the holes into which they are in-

VALVE PLATE

serted, the valve plate is aligned with the valve body and a hole is drilled
straight through. This is considerably easier than making square ports as
is customary in conventional spool valves to obtain linear flow versus
displacement relation. Since the inserts in the hole-and-plug valve are
round, the displacement of the plate and the corresponding opening of
the circular hole result in hnear relations.
Assuming that the outer edge of the lower cylindrical insert is perfectly
square, that the surfaces are flat and the interspace between plate and
MECHANICAL COMPONENTS 111

body negligible, then a perfect zero-lap valve is obtained. Such ideal


conditions can, however, only be approached but not attained. The re-
sult is a behavior that resembles that of an underlapped valve.

Cylinders
Figure 6-26a shows a single-acting cylinder, and Figure 6-26b, a
double-acting cylinder. Flapper-nozzles are used for positioning in this
example but other relays could be substituted. In the single-acting
SUPPLY

(al SINGLE-ACTING CYLINDER (b) DOUBLE-ACTING CYLINDER

Figure 6-26. Single-acting and double-acting cylinders.

cyhnder the nozzle back pressure is balanced by the spring in the actuat-
ing cylinder and the load on the piston. The disadvantage of this ar-
rangement is the loss of energy spent in compressing the spring. An ad-
vantage, however, for a number of applications is the fact that a spring
provides a safety feature against failure. In case of supply-pressure
failure, the spring will drive the actuating piston to the left.
The double-acting cylinder requires a feedback arrangement in order
to obtain a position proportional to the nozzle back pressure. The ar-
rangement shown is that of a proportional-speed floating controller. The
major advantage of the double-acting cyhnder is that it operates without
spring and that the full nozzle back pressure is available in any position
to actuate against the load.
ELECTRICAL COMPONENTS

A rather large number of components are available to convert the


magnitude of a controlled variable into an electrical signal. Input sig-
nals to these components are the emf of a thermocouple or of a pH
electrode, the resistance change of a resistance temperature detector or
of a strain gauge, etc. These signals are applied to an amplifier and its
circuit or to a galvanometer. Diaphragms and bellows are other fre-
quently used input signal systems which are then combined with a
variety of electrical components to convert the mechanical into an
electrical signal.
Figure 7-1 shows the most common of these components. They are by
no means the only ones, but are those most frequently found in control
systems. The following pages describe the components as listed in this
illustration.

Potentiometers
The potentiometer consists essentially of a sliding contact and a slide-
wire. The sliding contact is mechanically positioned, generally by rotary

112
ELECTRICAL COMPONENTS 113

but sometimes by translational motion. Thus a device is obtained that


can convert mechanical position into an electrical signal. Figure 7-2
shows the general arrangement. The supply voltage is kept constant

Figure 7-2. Potentiometer circuit.

and the load voltage Ei^ is varied by means of the sliding contact. The
equation that expresses E^ is

E, kR, X E,
kRp + Ri^ — k^Rp
where k defines the position of the sliding contact. R^ is the resistance
of the load and R^ is the resistance of the potentiometer or slidewire. If
all of the slidewire R^ is parallel with the load resistance R^, then /c = 1;
if only half is parallel, then ^ = 0.5; etc.
The equation shows the nonlinearity of the relation between load and
supply voltages. However, in making large as compared with the
equation approaches the hnear relation E^ = kE^.
The linearity behavior of the potentiometer in the circuit differs from
that of the potentiometer as such. The latter refers to the relation of
sliding contact position to resistance between sliding contact and slide-
wire terminals. As a rule this linearity is specified as normal or independ-
ent linearity (see p. 11).
The resolution sensitivity of a potentiometer is largely determined by
the fact that the sliding contact passes from one slidewire turn to the
next. The wire length of this turn represents a certain magnitude of re-
sistance which is covered by the slidewire in one single step. Thus the
resistance changes in steps rather than continuously. The finer the wire
the more the turns and the smaller the steps, but at the same time, the
finer wire represents a larger resistance, and the actual change in re-
sistance is about the same. The best way to improve resolution sensi-
tivity is to increase the sliding contact travel. If the travel is rotary, then
an increase of radius will give best results. Multiple-turn potentiometers
provide another method to increase resolution sensitivity. In either case
the travel is increased and the resistance per turn is decreased. The total
114 AUTOMATIC CONTROL

resistance of the potentiometer remains the same, but the resolution


sensitivity is improved.
The disadvantage of the stepwise response is overcome by using a
single straight wire. This results generally in resistances which are too
low to be practical. The wire may be replaced by conducting plastics or
conducting films of high resistance. It can be expected that conducting
plastics will in the future widen considerably the usefulness of potenti-
ometers in control systems.
The resolution sensitivity of the potentiometer depends further on the
static friction between sliding contact and wire, and in the bearing of
the shaft which carries the sliding contact. A certain minimum torque
on the shaft has to be developed before the sliding contact is loosened
from the hold of this friction.
Frequently, the resolution—not the resolution sensitivity—of the
potentiometer is specified. This resolution refers only to the stepwise
arrangement of the windings, and not to the friction effects.
The contact between slider and wire constitutes in itself a resistance.
This is produced by a film that develops by oxidation or other causes
on the metallic surfaces. It causes local heating and the heat has fre-
quently a cleansing effect although it may even worsen the conditions
of the metallic surfaces. Resulting are minute changes in overall resist-
ance which may be harmful in their total effect in control circuits.
Vibration causing the contact pressure to vary may have similar effects.
There are a number of other causes that may produce spurious changes
in resistance. The resulting phenomenon is termed noise.
Most disadvantages of the potentiometer are due to the physical con-
tact between sliding contact and slidewire, which produces friction and
wear. A precision potentiometer has a life expectancy of about one
million cycles. Considering that in a control system a sliding contact
may easily move an average of 4 times per minute, and that this motion
may concentrate on a limited portion of the slidewire, the life expectancy
of such a potentiometer would be only one half year. This leads to the
conclusions that the potentiometer should only be used where the sliding
contact moves not more than occasionally, and that a noncontact device
offers great advantages for all other applications.

Potentiometers and Rheostats


Figure 7-3 illustrates the difference between a potentiometer and
rheostat. The potentiometer is an adjustable voltage divider, while the
rheostat is an adjustable resistor which regulates current. Frequently, a
choice between either one of the two elements has to be made. In
control circuits, the potentiometers will be generally preferred. This is
ELECTRICAL COMPONENTS 115

POTENTIOMETER

t .
RHEOSTAT
Figure 7-3. Potentiometer and rheostat.

because the accuracy of the output to input relationship of the potenti-


ometer is only determined by the linearity of the winding resistance. In
commercial precision potentiometers this is about 0.1 per cent. Com-
mercial tolerances of total resistances are usually ±5 per cent. This has
no influence on the overall accuracy of potentiometers since

El- ^
E2 kR
and variations of R do not alter the result. This is of particular impor-
tance where the magnitude of resistance is affected by temperature
variations. In rheostats, a change of 5 per cent will change the current
by the same percentage.
A further point to consider is that the potentiometer is connected
across the power source and if the circuit through the load is open, there
remains a shunt through the potentiometer. In the case of a rheostat the
power supply is left with an open circuit when the load is removed.

Capacitors
The capacitor is an essentially frictionless device that changes the
electrical characteristics of a circuit. Two metallic plates separated by
air and connected into an a.c. circuit result in a capacitance which is
expressed by

X 10-1^

where C is the capacitance in farad, A is the minimum sq in. area of


either plate which is directly opposite the other plate, and D is the dis-
tance between plates in inches. For example, a pressure-sensitive dia-
phragm can be arranged to deflect with the process pressure and at the
116 AUTOMATIC CONTROL

same time eonstitute one plate of a capacitor, while the other plate
is fixed. Since the distance D changes with the deflection, the capaci-
tance also changes.
The current through an electric circuit is given by
I = E/Z
where / is the current in amperes, E is the potential in volts, and Z is
the impedance, i.e., the combination of capacitance and resistance, in
ohms. Furthermore

where / is the frequency in cycles per second and R is the resistance in


ohms, as shown in Figure 7-4.

Figure 7-4. Resistance and capacitance circuit.

It is quite obvious from these equations, and can be verified by sub-


stituting numbers, that the capacitance has to be very large or the fre-
quency very high to produce a current of reasonable magnitude.

Inductance Coils
Another resistive element in an electric circuit with alternating cur-
rent is the inductance of a coil. When alternating current flows through
a coil, a magnetic field is produced around the coil which pulsates in
intensity and direction with the frequency of the current. This pulsating
field generates an emf in the coil which opposes that of the circuit. This
counter emf has effects similar to those of a resistor, and it is this effect
that determines the inductance.
Figure 7-5 shows a circuit which combines resistance and inductance.
The impedance of this circuit is

Z = + (6-28/L)^
where L is the inductance in henries.
ELECTRICAL COMPONENTS 117

INDUCTANCE

A.C.POWER
SOURCE

-A//WW-
RESISTANCE

Figure 7-5. Resistance and inductance circuit.

If an insulated wire is wound about an iron core the inductance is


given by

L = 316 X

where L is the inductance in henries, N is the number of turns on the


coil, D is the inside diameter of the coil in inches and / is the length of
the iron core in inches.
If the iron core is removed the inductance becomes
A/'2 n2
L = 2.5x 10-*^^

In other words, the inductance changes by about 150 times when an iron
core is introduced.
Since an iron core can be moved in a coil without physical contact,
the inductance, and with it the impedance, of the circuit can be changed
without friction.

Differential Transformer
Inductance as described above is the self-inductance of a circuit
component. When two coils, each belonging to another circuit, are
brought close together, mutual inductance results. This is the principle
of any transformer. The alternating current in the primary winding pro-
duces a magnetic field within the vicinity of the primary which changes
in intensity and direction with the frequency of the alternating current.
The flux lines of this magnetic field cut the windings of the secondary
coil. This induces an emf in the secondary. The magnitude of mutual
inductance is expressed in henry, as is self-inductance. The mathematical
relationships are also the same, and the difference between iron core
and air core is similarly pronounced.
In the differential transformer, the effect of the iron core becomes
further intensified because the secondary is split into two different coils
which are arranged in various ways. One typical arrangement is illus-
trated in Figure 7-6. The two secondary windings are connected in such
118 AUTOMATIC CONTROL

a way that their polarities buck each other at any given moment. In the
position shown, the iron core is in its center position, and due to the
bucking effect the resulting emf in the secondary is zero. As the core is

Figure 7-6. Differential transformer.

displaced upward, the emf in the upper winding increases, and simul-
taneously decreases in the lower winding. The result is an output equal
to the difference between the emf’s induced in the upper and the lower
windings. Phase shift between the secondary and primary is caused by
the magnetic circuit. Disregarding this phase shift, the polarity would
be such that for any given instant when the current in the a wire of the
primary has minus polarity, the current in the A wire of the secondary
is also minus. If the core is displaced downward the output is equal to
the difference between the emf’s induced in the lower and upper wind-
ing, but phases are now reversed which means that for any given instant
when the polarity of the a wire is minus, the polarity of the A wire is
plus. The addition of phase shift between secondary and primary does
not alter the basic phase reversal characteristic as described.
The differential transformer has perfect resolution sensitivity. Normal
linearity is in the order of 0.05 per cent of maximum output over the
specified stroke. The power of its output is low. This is due largely to
the imperfect coupling between the primary and secondary and to the
restriction in size that is generally imposed on the design of the differen-
tial transformer. As it stands, the output of differential transformers re-
quires amplification to be utilized in control applications.
ELECTRICAL COMPONENTS 119

Galvanometers
The phenomenon of a magnetic field which is established by an alter-
nating current fiowing through a coil has been discussed under Induct-
ance coils. If the alternating current is replaced by direct current, the mag-
netic field still exists. Since, however, it does not pulsate, no counter emf
is produced and the current flow through the coil is determined only by
its ohmic resistance. This steady magnetic field has a north and south
pole and the intensity of the magnetic field is a function of the magni-
tude of the current flowing through the coil. If the coil is free to move,
the coil will orient itself with respect to any other magnetic field that
surrounds it because equal poles repel and unequal poles attract each
other.
As seen in Figure 7-7, a small cylindrical airgap is left in a magnet
circuit which is established by a permanent magnet with its pole pieces

Figure 7-7. Galvanometer.

and a stationary core. With a d.c. current flowing through the movable
coil, the magnetic field will interact with the flux of the permanent mag-
net. The resulting torque on the coil is restrained by the hair spring.
Thus, the magnitude of the current flowing through the coil determines
its deflection.
The torque of the galvanometer is comparatively small. The motion
120 AUTOMATIC CONTROL

of the pointer has to be picked up by other devices. Mechanisms that


would periodically feel the position and relay this position into a more
powerful action have been used for many years, but are now being
gradually abandoned in favor of vanes mounted on the pointer moving
between oscillation coils or interfering between a light source and a
phototube.

Oscillators
The principle of mutual inductance is employed not only in the differ-
ential transformer, but also in the oscillator. Figure 7-8 shows a typical

Figure 7-8. Oscillator control circuit.

oscillator control circuit. The plate circuit consists of the network com-
bination of impedance L^, relay coil, and capacitance Cj, in series with
the power supply and plate and cathode of the tube. The grid circuit
combines capacitor Cg, impedance and resistor R, which are in series
with grid and plate of the tube.
Oscillations are produced because plate and grid circuit are coupled
by impedances and La. As long as this coupling persists, a current
will flow which keeps the relay energized. If a metal vane is inserted
between the impedances, the vane acts as shield in the magnetic field,
disturbing the mutual inductance. The result is that the current flow is
interrupted and the relay deenergizes.
The combination of and Cj is a so-called tank circuit. With the
capacitor charged but no plate current flowing, the capacitor discharges
into the inductance. The current flow induces a counter emf in the
inductance which discharges into the capacitance. In case of pure
inductances and capacitances this current pulsation could continue
indefinitely. Since, however, some resistance is always present, the mag-
nitude of this circulating current will rapidly decay, unless plate current
begins to flow and supplies new energy.
ELECTRICAL COMPONENTS 121

When the vane prevents mutual inductance, the grid will be negatively
charged because of the action in the grid bias circuit consisting of R
and Cg. It is the principle of the grid controlled tube that a negative grid
prevents flow of current in the plate circuit excepting the circulating
current between and Lg. The grid bias circuit also contributes to hm-
iting the amplitude as will be shown.
If the vane is removed, an even minute current flowing through Lg
will induce some voltage in L^. The polarity of this voltage is such that
it makes the grid less negative. This results in more current flowing through
La and more induced voltage in L^, making the grid bias more positive and
increasing the current through L^, etc. The action is clearly a feedback
action but with positive rather than negative feedback. The circuit is
purposely made unstable. However, the instability, i.e., the amplitude of
the oscillations, is controlled. This is largely due to the grid bias network.
Suppose the voltage induced in L^ is large enough to make the grid bias
voltage positive for a part of the cycle. This results not only in an increase
of the plate current, but also produces a considerable grid current. This
charges capacitor C^. As the grid becomes negative again, discharges
through resistor R. The consequence is that the grid becomes more
negative than it would without capacitor. This means that the more posi-
tive the grid may be at one moment the more negative will it be the
next moment. This limits the plate current and hence the amplitude of
the oscillations.
The oscillator circuit may also be so designed that the frequency,
rather than the amplitude, of the oscillations changes with change in
inductance. The principle is used in a Bristol controller.
Oscillators have the advantage of highest sensitivity. A vane motion
of 0.002 in. suffices to change the oscillation in the circuit enough to
energize or deenergize a relay. Since it is difficult to transmit high fre-
quencies over any appreciable distances, it is therefore necessary to
build the moving element with the vane and the complete oscillator in
a single unit.

Phototubes
An ordinary vacuum tube emits electrons when the cathode is heated.
The resulting flow of electrons is equivalent to an electric current, the
plate current. The phototube differs insofar as the cathode does not
need to be heated, but is light sensitive rather than heat sensitive. The
current through a phototube depends on the voltage across it, the light
intensity and also the light color.
There are three types of phototubes, namely, vacuum, gas-filled, and
photomultiplier tubes. The current which passes through a vacuum
122 AUTOMATIC CONTROL

phototube at a given voltage potential is directly proportional to the


light intensity. With the gas-filled phototube, the relation between cur-
rent and light intensity is nonlinear. The gas-filled phototube has, how-
ever, the advantage of more change of current per unit change of light
intensity.
Both tubes have generally an upper limit of 20 microamperes. The
photomultiplier tube is usually rated for a maximum current of 20 milli-
amperes, i.e., one thousand times more than the other types. It is linear
in its characteristics and has more change of current per unit change of
light intensity than any other phototube. Gas-filled and vacuum tubes
are, however, frequently preferred because of their lower cost. Besides,
vacuum phototubes are of greater stability than any other.
Figure 7-9 shows a typical circuit as used with a phototube. The
phototube is connected into the grid circuit of a triode. The grid current

is assumed to be zero. The current flowing through resistor R is then


the same which flows through the phototube. The result is a voltage drop
across the resistor with a polarity which is opposed to that of voltage E^.
For example, the phototube current is 10 microamperes and the resist-
ance R is 2,000,000 ohms. The resulting voltage drop, E = I X R, is 20
volts. If voltage E^ is 45 volts, then the grid voltage is 20 — 45 = —25
volts. The current through the phototube is determined by the light
intensity. The current through the triode and the relay coil depends on
the grid voltage. Hence the current through the relay becomes a func-
tion of the light impinging on the phototube.

Synchros
There are different synchro components, such as the synchro trans-
mitter, synchro receiver, synchro control transformer, etc., which are
combined in control circuits in various ways. In the following pages the
ELECTRICAL COMPONENTS 123

most important synchro components are described and typieal combina-


tions are shown.
Synchro Transmitter. The prineiple of a synchro transmitter is that
of any transformer. An alternating current in the primary winding
results in a magnetic field within the vicinity of the primary which
changes in intensity and direction with the frequency of the alternating
current. If a secondary winding is placed into the same magnetic field,
a voltage is induced in the secondary and a current will flow through a
load connected to it. This is the result of electromagnetic induction pro-
duced by an alternately building up and decaying of a magnetic field.
If primary and secondary can be moved with respect to each other, then
the maximum voltage is indueed when the secondary is located so that
a maximum flux from the primary passes through its winding. Depend-
ing on the direction of the flux lines with respeet to the secondary, the
polarity of the voltage in the secondary will be either in one direction
or another.
The primary of the synehro transmitter is wound on a rotor, while the
secondary is distributed in three separate windings on a stator as illus-
trated in Figure 7-10. Depending on the position of the rotor, more or

Figure 7-10. Synchro transmitter.

less of magnetic flux lines will cut through any one of the three wind-
ings. In the position shown, which is known as the zero position, voltages
induced in windings A and C are equal and in the same direetion.
Hence the voltage between terminals S-^ and 5*3 is zero.
124 AUTOMATIC CONTROL

The voltage induced in B is of opposite polarity to either winding A


or C. Thus in a given moment, the terminal is of positive polarity, as
indicated in Figure 7-10, while at the same time the polarities at S-^, as
well as S^, are negative. With the expanding and decaying magnetic field,
polarities reverse, but as shown by an a.c. voltmeter, a voltage will be
continuously indicated between terminals and S2, as well as between
terminals ^'3 and but not between and S^. Rotating the primary
winding will change this condition gradually, since the number and
direction of flux lines which cut through the secondary windings are
altered. The result is illustrated in Figure 7-11, Each rotor position has

a corresponding pattern of voltages between the secondary windings.


After rotating the primary 90 degrees from its neutral position, the volt-
age between and is the maximum obtainable in any position.
Synchro Receiver. The most basic synchro arrangement may be
considered the combination of a synchro transmitter and a synchro
receiver. The secondary voltages of the synchro transmitter are applied
to the stator windings of the synchro receiver as shown in Figure 7-12.
Essentially, the receiver is of the same construction as the transmitter.
However, the input signal of the transmitter is a shaft position and its
output is corresponding voltage. The input signal of the receiver is volt-
age and its output is position. The stator windings of the receiver set up
a magnetic field in which the rotor, which is itself magnetized because
of its windings, acts like the needle of a compass. The pattern of the
field in the receiver follows the position of the rotor in the transmitter.
Hence the receiver rotor—and with it the shaft—reproduces continually
the position of the synchro transmitter shaft. The power which is avail-
able at the receiver shaft is small and suffices at best to position the needle
of an indicator.
ELECTRICAL COMPONENTS 125

Figure 7-12. Schematic drawing of synchro transmitter and receiver.

Synchro Control Transformer. For closed-loop control purposes,


the synchro receiver is replaced by the synchro control transformer. The
control transformer is of essentially the same construction as either
transmitter or receiver. The major difference is that the rotor shape is
round, to prevent the magnetic field from producing a torque on the
rotor. The only electric power connected to the control transformers is
the signal voltage from the transmitter. This is applied to the rotor
windings, while the stator windings represent the secondary of a trans-
former. The voltage induced in the secondary depends on the position
of the rotor with respect to the magnetic field. For a field that corre-
sponds to the zero position in the transmitter, the voltage induced in the
secondary of the control transformer will be zero when the transformer
rotor is displaced by 90 degrees from the transmitter rotor. This 90
degrees displacement applies for any other position of the rotor in the
synchro transmitter.
Synchro Differential Transmitter. By arranging the rotor so that it
carries three different windings distributed like those on the stator, a
synchro differential transmitter is connected between the transmitter
and the control transformer. The stator winding receives the signal from
the transmitter. The output is modified by the position of the rotor.
Thus the synchro differential transformer puts out a signal which is the
sum of the shaft positions of the synchro transmitter and the synchro
differential transmitter.
Combination of Synchros in a Control System. Figure 7-13 shows
a typical arrangement of synchros. The controlled variable positions the
rotor of the synchro transmitter. The output of the transmitter is con-
nected to a synchro differential transmitter, where the set point is
cranked in by positioning a rotor. The resulting actuating signal is
126 AUTOMATIC CONTROL

Figure 7-1 3. Synchros as part of control system.

applied to a synchro control transformer, the rotor of which is nor-


mally positioned so that the output signal to the amplifier is zero. How-
ever, it will deviate from zero when the actuating signal changes. This
signal is amplified and transmitted to the servomotor, a device which
will be described further below. The servomotor responds to the signal
by rotation. Its shaft is linked with the rotor of the synchro control
transformer. The rotation zeroes the output from the synchro control
transformer, which in turn stops the servomotor. The same steady-state
conditions as before are then obtained, but with a new position of the
servomotor shaft. The shaft position which is the controller output sig-
nal can be used to position directly a final control element, or some
intermediary element such as a slide wire potentiometer.
Other Usages of Synchros. The above descriptions do not intend
to cover all modifications and apphcations of synchros. For example, a
receiver similar to the differential transmitter may be built with triple
stator windings. This permits applying to the stator as well as to the
rotor signals which correspond with the angular displacements of two
different shafts. The resulting rotor position of this differential receiver
is then proportional to the sine of the difference between the angles of
the other two shaft positions.
Another possibility is to obtain a primary feedback signal from small
angular deviations, using for this purpose the synchro transmitter with-
out any other synchros. Figure 7-11 showed that for limited deviations
from zero, the voltage change is practically linear with the change in
angle.

Electronic and Magnetic Amplifiers


Electronic amplifiers use either vacuum tubes or transistors or both.
Gas-filled tubes—thyratrons—are used in the output stages when the
required output power exceeds about 25 watts.
Transistors. Of the various forms of transistors—point-contact,
junction, etc.—the junction transistor is used in the control field most
frequently. The typical construction of a p-n-p transistor is illustrated
ELECTRICAL COMPONENTS 127

in Figure 7-14. It consists of a single crystal made from germanium or


some other equivalent semiconductor. The «-type area conducts because
excess electrons exist, while the p-type area has an electron deficiency

Figure 7-14. p-n-p Type junction transistor.

or so-called “holes,” and conducts because electrons shift continuously


to fill up the holes, only to produce holes where they leave. The center
area which constitutes the base is very narrow, usually in the order of
0.001 inch. A n-p-n transistor differs from the p-n-p type in that the base
is /7-type, and emitter and collector are both n-type.
Transistors have replaced vacuum tubes in many electronic circuits.
Their advantages are longer life expectancy, smaller size and less heat
dissipation. The latter two characteristics are usually of lesser impor-
tance for industrial control systems. Some of the disadvantages how-
ever, are their temperature limitations, the variations of characteristics
between transistors of the same type and manufacturer, and the voltage
restrictions.
Ambient temperatures of 120°F are usually the operating limit for
most transistors. In addition, the output changes with temperature and
unless ambient conditions can be kept constant, the circuit design must
consider these variations. Voltages in excess of about 30 volts can gen-
erally not be used. This compares with about 250 volts for vacuum
tubes. The restriction is of particular importance where rate and reset
actions are provided. Practically always, a resistor-capacitor combination
is used for this purpose. The extent to which a capacitor can be charged
depends on the voltage, and the reduced voltage of the transistor limits
the usefulness of the resistor-capacitor network. Therefore, a vacuum
tube will generally be used with rate and reset networks.
Electronic Amplifiers. Figure 7-15 illustrates two methods of using
an amplifier. Accuracy requirements for the amplifier may depend on
which method is used. In case a the amplifier output must be exactly
proportional to the input since otherwise the recorder can not provide
128 AUTOMATIC CONTROL

(b) AMPLIFIER AS NULL DETECTOR

Figure 7-15, DifFerent connections of amplifier.

a faithful reproduction of the controlled variable. In case b, however, the


amplifier is a null detector. An output appears only when the controlled
variable deviates from the set point. This allows relaxing the require-
ments for linearity and saturation. The term saturation means that the
output signal changes with the input signal only to a certain point, at
which the amplifier saturates and the output signal remains constant,
though the input signal may change beyond the saturation point.
Drift and noise are characteristics which are disturbing in any appli-
cation. For example, if the output of the amplifier starts to drift because
of aging of parts, temperature changes, etc., a voltage will appear where
zero is expected. Similarly, noise, which means spurious changes in out-
put due to the thermal agitation of molecules in circuit components, or
to electrons impinging on tube plates, may deviate the amplifier output
from zero.
There is generally a preference of a.c. over d.c. amplifiers, because the
latter have a greater tendency to drift. A d.c. input signal may be con-
verted into a.c. by means of choppers, and then applied to a standard
a.c. amplifier.
Magnetic Amplifiers. The magnetic amplifier is essentially a d.c.
amplifier but, in general, does not have the tendency to drift like its
electronic counterpart. It is by far the most rugged of all amplifiers.
Since, however, even a vacuum-tube amplifier is hardly a fragile device,
the higher initial cost of the magnetic amplifier is frequently its most
serious handicap.
The ability of the magnetic amplifiers to convert small d.c. voltages
into high-power output has led to their application in the temperature
control of electric heat-treating furnaces. The power input to the heater
elements of the furnace is directly regulated by the magnetic amplifier.
ELECTRICAL COMPONENTS 129

The d.c. input is from a thermocouple which measures the controlled


variable.
Figure 7-16 illustrates the principle of the magnetic amplifier. The
load current supplied from a.c. voltage passes through a rectifier. The

RECTIFIER
CONTROL . / OUTPUT WINDINGS /
WINDIN&S \

D.C. SIGNAL A.C.


SUPPLY
VOLTAGE

LOAD

Figure 7-16. Principle of magnetic amplifier.

result is that current flows through load and output windings only dur-
ing the positive half-cycles of the supply voltage.
The output windings act as an inductance in the load circuit. This in-
ductance is, as stated before, the result of a changing magnetic field.
Since the output windings are wound about an iron core, it is the mag-
netic field in the core that controls the inductance. The magnetization of
iron can be increased only to its saturation point. Further increase of
current does not change the magnetic field. In other words, once an iron
core is saturated, the magnetic field remains constant, hence no counter
e.m.f. is induced and the inductance is reduced to its minimum.
A magnetic amplifier is designed so that saturation is reached at ap-
proximately the positive peak of the supply voltage, while the control
signal is zero. This means that before the peak is reached the inductance
is at its maximum, hence only a minute current will flow. At the peak,
however, the situation changes, the inductance vanishes, and the cur-
rent obtains its maximum value. This voltage peak, however, does not
coincide with the load current peak, since as in any inductive circuit the
current lags behind the voltage. The saturation thus occurs before the
actual current peak would be obtained.
The current flow keeps the core saturated. Hence, current will con-
tinue to flow throughout the remaining half cycle. After this, current
flow is stopped because of the rectifier, and when the next positive half
cycle starts, the same sequence as outlined above is repeated.
Signal current through the control winding either adds or subtracts to
the saturation effect of the output winding. When the signal is positive,
saturation^—and hence current flow—appears before peak supply voltage
is reached.
The time in which current flows can be shortened by making the sig-
130 AUTOMATIC CONTROL

nal negative. This is possible because the load current peak lags behind
the supply voltage peak.
Thus the supplementary signal voltage controls the length of the cur-
rent pulses through the load. This is illustrated in Figure 7-17 for various

values of load current. During the negative cycle of the supply voltage,
current does not flow. This pulsating nature of the magnetic amplifier
is overcome by the use of a bridge circuit combining two cores in a single
unit, as shown in Figure 7-18. If A is positive, current passes through B

TO LOAD

Figure 7-18. Full wave magnetic amplifier.

and C to the load, and back from there through D and E. As the phase
reverses, and E becomes positive, current passes through E and C to the
load, and back through D and G to A.

Moving Coil
Figure 7-19 is a schematic representation of the moving coil. It con-
sists of a permanent ring magnet and an assembly made of iron to pro-
ELECTRICAL COMPONENTS 131

A-A
Figure 7-19. Schematic drawing of moving coil system.

vide a magnetic flux path as indicated by arrows in the illustration. The


flux passes through an air gap where the moving coil is free to move. A
signal current flowing through the coil results in a force of attraction or
repulsion which is expressed by

F =9 X lO-'^BLi
where F is the force in ounces, B is the magnetic flux density in the air
gap, in gauss, L is the length of wire that makes up the coil, in inches,
and i is the current in milliampere.
The above equality shows that even a weak current could be made to
produce a force by increasing either the flux density or the wire length,
or both. This of course has practical limitations. For example, 4 milliam-
peres d.c. is a common output signal range of electronic controllers. For
a force of 10 oz. the product of flux density and wire length would have
to be
BL = 2.8 X 10«
Even with a coil that has 5000 ft. of wire the required flux density is still
some 4700 gauss. Considering an air gap of 2 in. mean diameter and 0.5
132 AUTOMATIC CONTROL

in. height, the required cross-sectional area of an Alnico V ring magnet


is about 4 sq in. This illustrates that bulk becomes a limiting factor
when the force available from a moving coil is considered.
The permanent magnet can be replaced by an electromagnet. This is
illustrated in Figure 7-20. There are two windings, one for the moving

Figure 7-20. Electromagnet moving coil.

coil, the other for the electromagnet. The resulting force is now propor-
tional to the product of the currents in the two windings. Since they are
connected in series, the current in both windings is the same and the force
is proportional to the square of the current. An important computing
element is thus obtained. Its apphcation will be described in Chapter 9.

Servomotors
Figure 7-21 illustrates the principle of operation of a two-phase induc-
tion motor. The rotor is represented by an iron bar. No electrical
connection exists between the rotor and the stationary poles and
windings. The windings are connected to a two-phase a.c. electric
power supply, so that a and b are connected in series with one phase and
c and d in series with the other phase. The windings are arranged so that
with the direction of current flow as assumed in the illustration, there
will be a north pole at a and a south pole at b.
The phase angle between the phases of a two-phase system is 90
degrees, which means that when the current in phase 1 is maximum, the
current in phase 2 is zero, and vice versa. As the phase of the power supply
proceeds, pole d will gradually magnetize and become a north pole,
while the current magnetization in a decays. After this pole b will be-
come north and so on. A revolving held is thus established and the iron
rod will rotate with it.
If either phase were to be cut olf, the torque available from the rotor
would be correspondingly reduced. Furthermore, the motor could not
start by itself.
ELECTRICAL COMPONENTS 133

PHASE 2 PHASE 1

Figure 7-21. Schematic drawing of two-phase induction motor.

A servomotor is generally a two-phase induction motor. The so-called


reference voltage obtained directly from the line is applied across wind-
ings a and b, while windings c and d are connected across the control
voltage from a servo amplifier. The control voltage is reversible in phase,
i.e., it either leads or lags the reference voltage by 90 degrees. This de-
termines the direction of rotation.
The magnitude of control voltage depends on the actuating signal fed
into the servo amplifier. The torque on the servomotor shaft is approxi-
mately proportional to the magnitude of the control voltage.
The rotor, shown as an iron rod in Figure 7-21 for the purpose of
illustration, is replaced in an actual servomotor either by a squirrel-cage,
sofid cylindrical or drag-cup rotor. The important consideration is that
the rotor be of light weight in order to reduce its inertia as much as pos-
sible. This permits it to obtain speed rapidly, and to be easily stopped
and reversed.
In general the servo amplifier will combine an a.c. and a d.c. output
signal. The d.c. signal is constant and is present even when the a.c. com-
ponent is zero. This keeps the corresponding poles at constant polarity,
providing an efficient braking action.
A significant difference between a servomotor and a conventional
two-phase induction motor is that the latter will rotate when power is
applied, while the servomotor will frequently stall though the reference
134 AUTOMATIC CONTROL

winding is energized. This produces considerable heating. As a result


servomotors can be used with limited power output only, from 0.5 to
100 watts. Where the power output is of 25 watts or more, a blower is
usually needed for cooling. Stalling torques from 0.5 to 95 oz/in. are
available. This value can be increased by gears, since the high rotational
speed is generally less a requirement than the torque output. Gears,
however, must be used with caution, since backlash can readily cause
instability of an entire control system.

Torque Motors
Restraining the motion of a servomotor by a spring allows it to move
through a limited angle only. When the torque due to control voltage
equals the force of the spring, the motion stops. Thus the angle is roughly
proportional to the control voltage. A servomotor used in this form is
called a torque motor.
Another type of torque motor is illustrated in Figure 7-22. Though the
designation motor is hardly justified, it is this device which is more fre-

13 2 4

n“ "A|=
(b) COIL CONNECTIONS

Figure 7-22. Magnetic circuit and coil connections of torque motor.

quently referred to as torque motor than the spring-restrained servo-


motor. This torque motor is basically a high-speed polarized type of
relay. Coils 1 and 3 in the illustration are connected in series on one
side of a push-pull amplifier, and coils 2 and 4 are in series on the other
ELECTRICAL COMPONENTS 135

side. The current flowing through these coils consists of a quiescent cur-
rent /o and the signal current i.
When the signal current is zero, magnetic fluxes/i,/ ,/ , and are
2 3

equal and the resulting torque on the rotor is zero. However, if the sig-
nal current departs from zero it is added in coils 1 and 3 and subtracted
in 2 and 4, or vice versa. In this case fluxes /a and remain about the
same, but /a and change, so that the magnetic flux in one pair of
diagonally opposite air gaps increases while it decreases in the other
pair. This produces a torque on the spring-restrained armature. The re-
sulting torque is proportional to the signal, but to obtain this pro-
portionality the armature motion must be limited to a very small angle.
The comparatively high hysteresis of these torque motors, which
amounts to two or three per cent, is generally acceptable in control cir-
cuits. A somewhat more serious consideration is the differential connec-
tion as illustrated. Commercial electronic controllers do not generally
provide this kind of signal, and an additional amplifier may therefore be
required.
8 . BALANCES AND COMPUTING
CIRCUITS

Measuring means, controlling means and final control elements have


in common elements, which convert input signals into output signals.
The relation of output to input may be proportional, inversely pro-
portional, a function of time, etc. If several inputs are combined, the out-
put may be the sum, the ratio or some other mathematical relationship.
This chapter will discuss principles and examples of proportional
elements (implying in this expression also inversely proportional ele-
ments), and computing elements which combine several inputs. Those of
specific control function, particularly reset and rate elements, will be
described in Chapter 10. One other class, namely those elements which
produce an output proportional to the square root of the input, is dis-
cussed also in Chapter 9.
Figure 8-1 shows a block representing a proportional element. Within
this block a transformation takes place that changes an input into an

INPUT OUTPUT
PROPORTIONAL
ELEMENT

Figure 8-1. Function of a proportional element.

t)utput signal. Some kind of balance system produces this transforma-


tion. Balances are specific forms of feedback arrangements. They are
divided into force balances and position balances. A force balance
operates either as deflection or a null system.
The force balance is well represented in weighing scales. A beam scale
indicates the weight when the deflection is zero. This is a null system.
In the spring scale the deflection is proportional to the force. This is a
deflection system.
The deflection system always involves a spring. This is quite obvious
when Hooke’s law, F = kx, is applied. In this equation, Fis the force ex-
erted by a deflected spring, k is the spring rate, a constant, and x is the
spring deflection. If W is the weight of the suspended object, then
W = F, which is the expression of the null system; but it is also true
that W = kx, which is a deflection system, because x is a deflection.
136
BALANCES AND COMPUTING CIRCUITS 137

Since, however, W = kx, this also means that the deflection system pred-
icates a spring because of the factor k which stands for the spring rate.
The equilibrium of a position balance is obtained by displacement and
not by forces as in the force balance. A typical example of a position
balance is the flapper-and-nozzle arrangement of a pneumatic controller
as illustrated in Figure 8-2. The flapper is displaced to obtain a balanced

Figure 8-2. Flapper-and-nozzle system.

position in responding to two counteracting motions; one from the


measuring signal, the other from the output signal. The flapper is not sub-
mitted to strain at any moment. When the bourdon tube responds to a
change in controlled variable, it rotates the flapper around pivot A. This
changes the nozzle back pressure and hence the controller output signal.
The feedback bellows responds to this change and rotates the flapper
around pivot B in such a direction that the initial clearance between
flapper and nozzle is approached.

Force Balances with Deflection Systems


A typical example of a force balance deflection system is the gal-
vanometer described in the previous chapter. The torque of the coil is
produced by the current flowing through it. The motion is restrained by
the hair spring. As the coil rotates, the counterforce of the hair spring
changes until an equilibrium between the two forces is obtained. This is
138 AUTOMATIC CONTROL

a force balance. The deflection system is characterized by the fact that


deflection is required to obtain equihbrium.
Another application of a deflection system in force balances is found
in pneumatic systems, e.g., in the pressure-duplicator method, frequently
used in converting liquid level into an equivalent air pressure. Figure
8-3 illustrates the principle involved. The top side of the diaphragm is

LIQUID HEAD

Figure 8-3. Pressure duplicator—pneumatic force balance


with deflection system.

exposed to the liquid head. Air pressure is admitted to the bottom side,
through a fixed restriction. The bottom chamber communicates directly
with the receiver. A nozzle opening decreases when the diaphragm moves
toward it. The further open the nozzle outlet is, the more air bleeds
through the nozzle to the atmosphere and the lower is the pressure in
the chamber. Suppose the liquid level rises, exerting an additional pres-
sure on top of the diaphragm. It is assumed that the air supply pressure
is higher than the maximum pressure due to liquid level head. The dia-
phragm will then tend to close the nozzle which will make the pressure
in the bottom chamber rise until a balance is reached where the forces
on the top and bottom of the diaphragm are equal. The arrangement is
therefore a force balance, and it is a deflection system in that the new
air pressure requires that a certain deflection of the diaphragm be main-
tained to partially close the nozzle opening.

Force Balances with Null System


The galvanometer is widely used in pyrometers—in millivoltmeter
types and in some potentiometer types. When a potentiometer uses a
galvanometer, the latter no longer exercises the function of quantitative
measurement. Figure 8-4 shows the typical potentiometer circuit. There
are two sources of emf: the battery and the thermocouple. Their re-
spective circuits have one portion of the slidewire in common. The rel-
ative length of this common portion can be either increased or decreased
BALANCES AND COMPUTING CIRCUITS 139

Figure 8-4. Potentiometer—electric force balance with null system.

by means of the slider. In balanced conditions no current flows through


the thermocouple because of the bucking effect of the two emf’s in the
common slidewire portion. If the thermocouple emf changes, it is neces-
sary only to readjust the slider in order to reduce to zero any current
flow due to this new condition. Since a scale can be made to indicate
the position of the slider, it can be cahbrated in temperature units which
express the reading of the thermocouple. The sole function of the gal-
vanometer is to detect an unbalance in the circuit, but not the magnitude
of the change.
The balancing action in the potentiometer can be compared with the
beam scale mentioned before. One dish of the scale carries an object of
unknown weight, the “input,” and known weights are added in the
opposite dish until the deflection of the balance is zero. The magnitude
of the known weights that were added expresses the measurement, i.e.,
the “output.” The input of the potentiometer is the emf of a thermo-
couple which produces a deflection of the movable coil in the galva-
nometer. Balance will be obtained by a known emf—in lieu of weights
—which is mechanically adjusted until the coil deflection is zero. The
amount of mechanical adjustment required is read in terms of the vari-
able measured by the thermocouple.
As far as force balance is concerned, the potentiometer represents it
as much as the millivoltmeter. The potentiometer, however, does not
require the deflection of a spring in order to obtain balanced conditions.
It is an outstanding characteristic of the null system, as compared
with deflection systems, that the former requires an outside medium
activated by some added power source to make it work, while the latter
can be self-contained and respond on its own accord. The added power
source may be the emf of a battery, weights, air pressure, etc., and the
140 AUTOMATIC CONTROL

outside medium may be either a human operator or some automatically


operated mechanism, commonly called feedback. Since the deflection
system does not require such additional provisions, it appears that it has
the advantage of greater simplicity. The null system does have advan-
tages, however.
The beam balance always returns to the same balanced conditions.
At the time when it indicates the measurement, there are no mechanical
deflections in the system. In the spring scale this is different; for each
measurement there is another configuration of mechanical parts. These
movements must be free from hysteresis and nonlinearities—and in this
respect there is a limitation in any spring. A further limitation is the
influence of temperature changes on spring members, which may pro-
duce inaccuracies in their response to changing loads.
The null system has the advantage of not depending upon the deflec-
tion of spring members. It also allows greater sensitivity. When a milli-
voltmeter, i.e., a deflection system, is to operate over a certain range,
e.g., 0 to 500°F, the movable coil will have its maximum deflection at
500 °F, and any smaller values will produce correspondingly smaller
deflections. It is quite feasible that the galvanometer in a potentiometer,
i.e., a null system device, may obtain full deflection at, say, 10°F, pro-
vided that restraining members will allow the stress, though not the
movement, for sudden changes beyond that range. The response of the
galvanometer in a potentiometer to a given change can thus be made
much greater than in a millivoltmeter. A smaller change of temperature
can therefore be sensed. However, this assumes at the same time that the
dead band in the system is negligible. If the system is unable to respond to
a change of 1°F, neither the null system nor deflection system will im-
prove the situation.
Since the galvanometer when used in a potentiometer is part of a closed
loop and hence reduced to the function of an unbalance detector, it
becomes of relatively little importance whether or not its response is
linear. It is only necessary that it detects the unbalance, while the deter-
mination of the magnitude is obtained by auxiliary means.

The Null System in Pneumatic Circuits


This characteristic of permitting nonlinearities makes the null system
particularly desirable in pneumatic flapper-and-nozzle circuits because
the degree of nozzle opening is not entirely proportional to the transmitted
pressure.
The pressure duplicator in Figure 8-3 operates very well as long as
the spring constant of the diaphragm is low enough to be ignored. If
this is not the case, the deflection of the diaphragm which is necessary
to change the nozzle opening results in an added force and the trans-
BALANCES AND COMPUTING CIRCUITS 141

mitted air pressure is no longer proportional to the pressure of the liquid


head. For a metallic diaphragm, for example, the spring constant may
be quite high and the accuracy of the device would suffer. One method
of reducing this error would be to keep the required motion of the dia-
phragm as small as possible, i.e., to approach a null system.
While this can be done by making the nozzle diameter relatively large
as compared with the motion of the diaphragm, or by reducing the size
of the fixed restriction, one may also apply the principle of the pneu-
matic null balance, which was introduced by C. B. Moore. This uses an
ingenious feedback principle which calls for maintenance of a relatively
constant pressure drop across the nozzle and thereby reduces the deflec-
tion of the diaphragm considerably.
In order to clarify what is involved in converting the arrangement of
Figure 8-3 into a null balance, equation (6-1) from Chapter 6 is repeated
here, which equally applies for Figure 8-3:

- Pt)Pt = - Pr)Pr (8-1)

where is the flow area of the fixed resistance, the flow area of the
nozzle,/7s is the supply pressure,/?( is the transmitted pressure, andpr
is the reference pressure which in Figure 8-3 is the atmosphere. All
pressures are absolute pressures.
In maintaining the diaphragm position, A^ and A 2 are to be constants
and may be combined into a single constant K, so that

This can be substituted in equation (8-1) and makes

K= ~ Pt)Pt (•g_2)
(Pt - Pr)Pr
In this equation pt, the transmitted pressure, is necessarily a variable.
In order to make the expression constant as a whole, both p^ and pr
would have to be varied in direct proportion to/?;, so that

Pr = apt
Ps = bpt

where a and b are proportionality constants. Substituting this in equa-


tion (8-2) gives
^ {bpt -Pt)Pt ^ b - 1
(/?; - apt)apt a(\ - a)
thus ehminating all variables, and obtaining a null system.
142 AUTOMATIC CONTROL

The methods, however, of changing and in proportion to pt are


imperfect and complicated. Nevertheless the ideal null system can be
closely approached. In view of the mechanical complexities, however,
it is frequently preferred to leave the supply pressure constant and vary
only the reference pressure. Even under these conditions it is possible
to reduce considerably the diaphragm deflection.
Kirk* gives figures in one particular case for the diaphragm deflec-
tion required to change the transmitted pressure from 3 to 18 psig. For
constant supply and reference pressures the diaphragm motion was
0.00209 in. By changing the reference pressure in proportion to the
transmitted pressure, this motion was reduced to 0.00171 in. By chang-
ing both, the reference and the supply pressure, in proportion to the
transmitted pressure, the motion was only 0.00011 in.
The principle which is used in a large variety of components of the
Moore Products Company and which employs a variable reference
pressure is illustrated schematically in Figure 8-5. It combines the pres-

INPUT SI6NAL

AIR SUPPLY

Figure 8-5. Pilot relay with variable reference pressure.

sure duphcator of Figure 8-3 with a pneumatic relay to incfease the


flow capacity for the transmitted pressure. The input pressure A is con-
verted into an output pressure B by either admitting air pressure from
the supply line C or by releasing it from B through D to the atmosphere.
When the input signal increases, diaphragm E moves downward, de-
creasing the nozzle opening and thereby increasing the nozzle back
pressure F. The diaphragm assembly G, consisting of two diaphragms
rigidly connected together, moves downward in response to the increased

* Kirk, D. B., “Nozzle Flow Characteristics in Pneumatic Force-Balance Circuits.” ASME


Transactions, (Feb. 1948).
BALANCES AND COMPUTING CIRCUITS 143

nozzle back pressure. This opens supply port H, admitting air and thus
increasing the output pressure. This pressure is fed back into the chamber
/ and forms the reference pressure of the nozzle system. It balances the
increased signal pressure and drives the diaphragm E back nearly to its
initial position. The increased pressure B also counteracts pressure F
and restores the initial position of the diaphragm assembly.
Compression spring L adds to pressure B in balancing the diaphragm
assembly against pressure F. It is therefore necessary that the nozzle
back pressure is at all times above the output pressure by a fixed amount,
which is equivalent to the spring force. This is the constant pressure
differential across the nozzle.
This method assures that the increased input signal results in increases
at B, F, and I with little displacement of original positions, and thus
approaches the null system.
The schema of Figure 8-6 adds the mechanism required to also vary
the supply pressure to the nozzle system with changes of the input sig-

INPUT SI6NAL

nal. The air supply C does not pass directly to the fixed restriction, but
flows first through the pressure regulator /. The diaphragm K controls
the flow and pressure drop across the regulator. The position of the dia-
phragm is controlled by the nozzle back pressure F. As described above,
the nozzle back pressure is proportional to the output pressure B. This
makes p, vary in proportion with />(.
The fact that diaphragm K is now being deflected while the deflection
of the E diaphragm is minimized should not be overlooked. The deflec-
tion is placed at a point where its effects are negligible.
The conclusions which can be drawn and are of rather general validity
144 AUTOMATIC CONTROL

are that (a) a null system can be designed for greater accuracy than a
deflection system, (b) the deflection system is inherently a simpler
arrangement than the null system, (c) the advantage of the null system
may be approached by reducing the deflection. The foregoing also
implies a generalized statement: Since a simpler system is always to be
preferred, the deflection system should be chosen unless the greater
accuracy obtainable with the null system is desired.

Position Balances
The choice between force balance and position balance is determined
by the nature of input and output. If they both are forces, the force
balance will be chosen. If they both are positions, the position balance
will be preferred. If one is a force and the other a position, the decision
becomes generally arbitrary.
The primary feedback from a controlled variable is practically always
a force, either mechanical or electrical. Frequently, this signal is used
not only for controlling, but for recording and indicating as well. In a
recorder, for example, a pen displacement indicates the magnitude of
the controlled variable. The output of a pneumatic recorder-controller
is a pressure which applied to a certain area is a force. In the case of an
electric controller it is generally an electromotive force (emf).
Ignoring for the moment the control-point-setting mechanism, the
problem that arises is to make a choice among three possible arrange-
ments. One is using the force of the primary feedback and comparing it
with the force of the output signal. The other consists of using the posi-
tioning mechanism of the pen and comparing the position with the
controller output signal by converting the latter from force into position.
The third arrangement possible is to convert the displacement of the
pen mechanism into a force and compare it with the force of the con-
troller output signal. The first possibility is rarely, if ever, used in a
recorder-controller. The main reason is that the force of the primary
feedback depends on the nature and range of the controlled variable.
It therefore does not allow a simple universal balancing mechanism.
Furthermore, the force level is frequently extremely low and produces
practical design difficulties. This makes the positioning mechanism the
preferred input source for the balancing system.
In electric recorder-controllers, the third arrangement—converting
the displacement of the positioning mechanism—is frequently preferred.
For example, the displacement may be linked to the wiper of a potenti-
ometer and thereby regulate an emf which is the controller output signal.
In pneumatic recorder-controllers the most widely used methods are
those illustrated in Figure 8-7 where the controller output signal is con-
BALANCES AND COMPUTING CIRCUITS 145
INPUT
SIGNAL

SIGNAL

OUTPUT
SIGNAL

Figure 8-7. Three typical forms of pneumatic position balances.

verted into a bellows displacement; this is done with a position input


signal derived from the measuring means which may simultaneously be
used to position an indicator or pen.
Pneumatic Position Balances. Figure 8-7 illustrates three typical
forms of pneumatic position balances. The flapper in case (a) is positioned
by the input signal linkage which is connected with the pen or indicator
positioning mechanism and by the linkage to the feedback bellows. This
is identical with Figure 8-2.
Case (b) is a modification of (a) which allows, by means of the adjust-
able fulcrum m, the change of the proportional relation between input
and output signal. The spring shown in the illustration has no balancing
function but serves only to hold the flapper against the fulcrum.
Case (c) differs insofar as both flapper and nozzle positions are
changed. The input signal positions the flapper and the feedback signal,
the nozzle. The relative position of flapper and nozzle is the position
balance which determines the pressure of the output signal.
Hydraulic Position Balances. The output signal of a hydraulic con-
troller is essentially flow which is converted into change of position of
an actuating piston. Hence a position balance is generally applied. Fig-
146 AUTOMATIC CONTROL

ure 8-8 illustrates a hydraulic controller, the main components of which


are a four-way valve and an actuating cylinder.
Spool and sleeve of the four-way valve are both movable. Thus the
spool may be displaced upward and if the sleeve is displaced upward by

FEEDBACK

Figure 8-8. Hydraulic system position balance.

the same amount, their relative position is as before. The input signal
displaces the spool. Suppose the change of the controlled variable is
such that the spool moves upward. This will open the port to the top
of the actuating piston connecting it with the oil supply. The port to
the lower part of the actuating cylinder also opens, connecting from
underneath the piston to the drain. As a consequence the actuating
piston moves downward. The feedback lever holds on its extreme left
the sleeve of the four-way valve against the force of a spring. As the
actuating piston moves downward, the feedback lever follows the
motion by rotating clockwise. This moves the feedback sleeve upward,
restoring the initial relative position between spool and sleeve, i.e., clos-
ing the ports to the actuating cylinder. This stops the motion of the
actuating piston.
The position balance between spool and sleeve thus assures accurate
positioning of the actuating piston in proportion to the input signal.

Wheatstone Bridge
The most frequently used bridge circuit is the Wheatstone bridge.
The schema of Figure 8-9 illustrates the principle.
BALANCES AND COMPUTING CIRCUITS 147

Figure 8-9. Wheatstone bridge.

The signal current / through resistor e is given by

I = E
be — ad (8-3)
ab{c + d) + {a b)(ed de + ce)
The equation shows that for be = ad the current through resistor e is
zero. If, for example, resistor a is a resistance temperature detector, the
resistance of which changes with temperature, then the signal current
can be measured by an ammeter with a scale calibrated in degrees
Fahrenheit.
To obtain the maximum sensitivity of the bridge, all five resistors
should be of approximately the same magnitude. If resistors e, d, and e
are of value equal to b, then the equation (8-3) can be written
b — a
I = E
Sab + 36^
The nonlinearity between change of resistance in resistor a and signal
current flow is illustrated in Figure 8-10.
The nonlinearities of the Wheatstone bridge disappear when it is used
as a null balance. The signal current that flows through resistor e is then
the unbalance which is applied to a servo amplifier. A servomotor ener-
gized by the amplifier positions the sliding contact of a slidewire resist-
ance which is one of the legs in the bridge. The null condition of
be = ad is thus re-established after each disturbance of the bridge. If
resistor a is the primary element which responds to changes in the con-
trolled variable, then resistor e is readjusted by the servomotor when a
changes.
The disadvantage of the described arrangement is that contact resist-
ance of the sliding contact produces an erroneous signal. The method
which prevents this is illustrated in Figure 8-11. The contact resistance
148 AUTOMATIC CONTROL

Figure 8-10. Current flow in Wheatstone bridge.

of slidewire A appears in the connection to the servo amplifier, and has


no influence, since no current flows under balanced conditions.
The illustration shows a second slidewire. The positioning shafts of
both slidewires A and B are coupled together and controlled by the same
servomotor. The arrangement is used with resistance temperature detec-
tors. Resistor a is the primary element which is exposed to the controlled
temperature. It is located at a distance from the remaining bridge cir-
cuit. The connecting wires are therefore also exposed to atmospheric
connections which may be subject to change and hence influence the
wire resistance, producing an erroneous signal. In this case a three-wire
connection is used, i.e., wires m, p, and n, are brought to the location of
a. Resistance changes in wire p are again of no influence in a null-
balanced bridge. Changes in m and n affect two different legs in the bridge
and compensate each other. Since a changes with temperature, b must
BALANCES AND COMPUTING CIRCUITS 149

Figure 8-11. Self-balancing double-slidewire bridge.

be made to change correspondingly, otherwise the compensation is not


complete. This change of b is obtained by means of slidewire B. The
coupling between A and B provides that for any position
c + J =/-f g

from which it follows that

/+ S ^ 1
c + d (8-4)
For null condition in the bridge it is necessary that

1 = f+ ^
b c + d
Substituting equation (8-4) yields

^
- = 1\ 01 a = bu
b
which is the condition for complete compensation.
In strain gauges, for example, temperature changes which alfect the
stress-sensitive resistor a must not be detected. In this case, a second
resistor of the same material but not exposed to stress is mounted close
to a. This resistor then represents a compensating leg in the bridge
circuit.
150 AUTOMATIC CONTROL

Where slide wires are objectionable because of limited life expectancy


other elements may be used. The Wheatstone bridge is not limited to
resistance changes; capacitors or inductances may replace the resistors.
The null condition be = ad shows the adaptability of the null-balanced
Wheatstone bridge for computing purposes. For example, let J be a
fixed resistor where d = 1. In this case a = be. If the servomotor posi-
tions resistor a, then its corresponding angular shaft position is in null
position always equal to the product of b and e. Thus, resistances b and
e may represent two controlled variables which require an output equal
to their product.
Similarly, if b — e, which means both resistors are of the same mag-
nitude and exposed to the same controlled variable, then a = b^. In this
case the servomotor shaft position is proportional to the square of the
input.
Inversely, if b = e, and both resistors are adjusted by the servomotor,
while a is exposed to the controlled variable, then b'^ = a, or b = \fa.
This means that the shaft position is proportional to the square root of
the input.
Finally, if resistor b is fixed and made equal to unity, then a = e/d. If
a is positioned by the servomotor then its shaft position is proportional
to the ratio of inputs e and d.
The usage of bridge circuits for computation is a fine example of the
possibilities inherent in a null balance. The Wheatstone bridge is used
as position balance for many purposes, but for computation it is practi-
cally a prerequisite that it be a null balance.
Sorteberg Bridge. Equivalents to the Wheatstone bridge are used in
pneumatic and hydraulic circuits. In fact, the push-pull flapper nozzle
in Figure 6-12 is really a Wheatstone bridge.
A specific pneumatic bridge arrangement is the Sorteberg bridge, the
principle of which is illustrated in Figure 8-12. Connections are shown
for square root extraction. The input signal is applied at bellows D. An
increase of the signal rotates lever E clockwise about roller F. This in-
creases the back pressure of nozzle G and with it the output signal. This
signal is also applied to bellows A, which results in an increase of the
back pressure of nozzle H. The back pressure is applied to a Bellofram
or equivalent, which pushes rod M downward and with it rollers K and
F. This changes the fulcrums for levers E and P. The result is that both
flappers tend to move away from their nozzles, counteracting the action
of the input signal which tends to close the nozzles. A null balance is
thus obtained.
The active forces are those due to the pneumatic pressures multiplied
BALANCES AND COMPUTING CIRCUITS 151

by the bellows areas and by the length of the lever as determined by the
position of the fulcrum. An additional force is produced by the spring.
Since the deflection in the null balance is negligible, the spring force is
practically a constant. Force balance is established when
A X a = B X b and D X d = C X c
where a, b, c and d are the lever lengths associated with A, B, C, and D,
respectively. Displacement of rollers K and F changes simultaneously
the lengths of all four levers. However, at all times b = c and a = d.
Hence
A X d = B X c and D X d = C X c
Dividing the first of these two equations through the second, yields
A^B or A
D C
Furthermore, B is the constant spring force which may be assumed to
152 AUTOMATIC CONTROL

have unity. Since A and C are interconnected, they have the same pres-
sure and A = C. Therefore,

A = may be written as

A = D— or A^ = D or A = y/W
A
The output signal is equal to A and the input signal is equal to D. Hence
the output is proportional to the square root of the input.
It is quite obvious that the algebraic relationship is the same as for
the previously described Wheatstone bridge. The same variations apply
therefore. For example, multiplying of signals requires only that the air
supply—via the restriction as shown—together with nozzle G, be cut off
from A and C and connected to D. Applying one input signal to A and
the other to C results in an output signal from D which is equal to the
product of the inputs. If the same signal is applied at A and C, the out-
put squares the signal. Likewise, connections for dividing and propor-
tioning are possible.

Computing with Differential Transformers


Differential transformers can be arranged so that their output is equal
to the square, the product or other function of the input. Figure 8-13

OUTPUT
POWER
SUPPLY

Figure 8-13. Multiplication with differential transformers.

illustrates the use of two differential transformers for multiplication.


Voltage is proportional to and the displacement of the core.
Thus

E^ = k,E^D^
The voltage E^ is amplified to E^, hence

Finally, E^ = k^E^D^
BALANCES AND COMPUTING CIRCUITS 153

By substitution

or by combining the constants k-^, /cg, and k^ into the single constant k,

E^ = kE^D^D^

which expresses the multiplication of the inputs and D^. By making


Z)i = D2, the resulting output is the square of the input.
MEASURING ELEMENTS

Problems of measuring elements that have specific bearing on control


are primarily square root extraction and temperature measurement. The
square root extraction becomes important in flow measurement. Since in
most cases flow is determined by measuring the pressure differential
across an orifice plate, nozzle, or similar element, the magnitude is ex-
pressed as the square of the flow rate, since
Q = kA VTI -P2
and (/?! — P2) is measured. If the flow controller gain is expressed as
change of output signal per unit change of flow, then the gain will
change with the magnitude of the flow rate, because the flow rate is a
nonlinear relation of the differential pressure. This change in controller
gain is equivalent to changing the proportional band in a controller with
proportional-position action. This has consequences in response speed
and stability and is undesirable for optimum control although quite
frequently it can be tolerated. More serious, however, is the nonlinearity
in multiplications with other signals which are linear. This is necessary,
for example, where gas flow is measured and corrections for changes in
the gas pressure and temperature are required. In order to avoid all
these difficulties, the square-root is extracted from the differential-
pressure measurement.
Of particular importance in temperature measurement for automatic
control is the time lag that occurs in the primary element. In practically
all other cases the response time of the primary element is negligible in
comparison to the response time in the process or occasionally, the final
control element. In temperature control, however, the response time of
the primary element has sufficient bearing on the dynamic response of
the system to warrant separate discussion.

Square-Root Extraction
Three basic methods for square-root extraction are in general use:
Ledoux bell, mechanical linkage, and electrical force balance.
Ledoux Bell. Figure 9-1 illustrates the principle of the Ledoux bell.
The bell is floating, semi-immersed in mercury or some other suitable
liquid. The mercury surface inside the bell is exposed to the high-pressure
154
MEASURING ELEMENTS 155

Figure 9-1. Principle of Ledoux bell.

side of a differential-pressure primary element. The surface outside the


bell is exposed to the low pressure side. Changing the differential pres-
sure will displace the bell correspondingly. The parabolic shape on the
inside of the bell makes this displacement proportional to the square
root of the differential pressure. The differential transformer thus picks
up a signal which changes in linear relation with flow.
Limitations of this arrangement are the relatively high machining cost
of such a bell and the fact that each bell is suitable for one range only.
To change the measuring range, it is necessary to change bells. Further-
more, it requires mercury or some other suitable liquid and is not ap-
plicable to diaphragm and bellows type flow meters.
Mechanical Linkage. A cam can be shaped so that it converts a
motion into practically any mathematical function. However, such cams
must generally be made with high accuracy to render useful results. The
relatively small motions which are usually available in flow meters result
in additional difflculties. Furthermore, a cam introduces friction which
is undesirable with the small forces available. However, cams are used
for square-root extraction, particularly in ring-balance meters where mo-
tion and available forces are larger than in most other differential-
pressure devices.
Figure 9-2 shows an arrangement in which the input force is balanced
by a counterweight with the lever effect of the counterweight modified
by a cam. The deflection of the beam is the result of two opposing
156 AUTOMATIC CONTROL

torques, one from the input force, the other from the weight. In the po-
sition shown in the upper diagram, the torque is proportional to Lj. In
the lower diagram the input force has increased. The resulting torque

Figure 9-2. Cam and counterweight for square-root extraction.

rotates the lever counterclockwise. This increases the effect of the


counterweight in proportion to L2 — L^. A new balance is obtained with
a deflection as shown. The cam can be shaped to obtain a defleetion pro-
portional to the square root of the input force.
Figure 9-3 shows a lever system for square-root extraction. Levers A
and C both follow circular paths as outlined. The input and output sig-
nals are assumed to be angular displacement. If the input is a shaft at E,
the rotation of which moves lever C and hence levers B and A, then a
second shaft at D, connected to lever A, will provide an output signal
which can be made a square-root function of E, provided that certain
dimensions and angular displacements are observed.
The angles through which the input shaft D and the output shaft E
move, are called a and yS, respectively. The purpose of the arrangement
in Figure 9-3 is a relationship in which
a = k
MEASURING ELEMENTS 157

where /c is a proportionality constant. This relation holds true in Figure


9-3 for a maximum angle of 100° for either shaft. When shaft D moves
through the maximum angle a = 192.5 — 92.5 = 100, shaft E moves

92.5°

Figure 9-3. Lever system for square-root extraction.

through = 148.9 — 48.9 = 100 degrees. Since 100 = k \/100, it fol-


lows that = 10.
In the position shown in the illustration, a = 163.2 — 92.5 = 70.7 and
= 98.9 — 48.9 = 50. This corresponds with the desired relationship
because 70.7 = 10 \/50.
Certain dimensions must be maintained in the layout of such a mech-
anism in order to obtain a square-root relation. These dimensions are as
follows:
Distance between D and E 1 in.
Length of lever ^ 2.291 in.
Length of lever B 1.227 in.
Length of lever C 1.409 in.

Any multiple of these dimensions may, of course, be used.


The disadvantage of this and similar linkage mechanisms is that high
158 AUTOMATIC CONTROL

precision is required in manufacturing them in order to obtain reasonable


accuracy.
Electrical Force Balance. The method of obtaining square-root ex-
traction by electrical force balance is illustrated in Figure 9-4. The force

OUTPUT

Figure 9-4. Square-root extraction by electrical force balance.

due to the differential pressure across the diaphragm is balanced by the


force in the moving coil. The winding of the electromagnet is in series
with the moving coil, so that the force that the coil exerts on the beam
is equal to the square of the current.
The differential pressure is proportional to the square of the flow,
hence the corresponding force on the beam is
F, =

The moving coil force is


jp2 =

Since F, = and the square root of the proportional constants k-^ and
k2 can be expressed by k, it follows that
i = kQ
which means that the current flowing through the moving coil is propor-
tional to flow, i.e., the square root of the differential pressure.
The differential transformer picks up any motion in the beam. Its out-
put is connected to an electronic amplifier and a corresponding signal
from the amplifier is applied to the moving coil system. The combined
gain of the differential transformer and amplifier is so large that the
beam motion required to increase the signal from minimum to maximum
can be made negligible for all practical purposes. This has the further
MEASURING ELEMENTS 159

advantage that it minimizes those nonlinearities which increase with di-


aphragm deflection.
The output is taken as a voltage across a resistor in the feedback from
the amplifier to the moving coil system.

Temperature Control
Of the three most frequently controlled variables—temperature, pres-
sure, flow—temperature is generally the most difficult to control.
Inaccuracies due to structural and physical limitations of the measur-
ing means are disregarded here. The inherent difficulties of temperature
measurement are however discussed. Two groups of errors can be dis-
tinguished in temperature measurement:
1. Static errors due to inaccuracies of reproduction of the temperature
under measurement by the temperature of the primary element;
2. Dynamic errors due to thermal lags caused by thermal resistances
and capacitances in the primary element, its protective wells, etc.
Temperature reproduction. The temperature to which the measur-
ing means actually responds is that of the primary element. Only through
the primary element—hence indirectly—is the temperature of the con-
trolled medium measured. Static errors in the control system result from
lack of accuracy in temperature reproduction.
Heat is transmitted from the surroundings to the primary element by
conduction, convection, and radiation. A room thermostat, for example,
is generally supposed to respond to changes in room temperature
brought to it by heat conduction. The fact is that exposure to draft, i.e.,
convection, will result in faulty action of the thermostat. The reason is
that the thermostat is intended to respond to conduction alone, but can
obviously not distinguish between conduction and convection. Similarly,
in the measurement of hot gases, e.g., in furnaces, or in the uptakes and
regenerators of open hearth steel furnaces or glass tanks, the thermo-
couple or—more commonly—its protective well, will pick up radiant
heat from hot surroundings (walls, etc.) and thus produce erroneous
readings.
Primary elements are available that reduce the effects of undesirable
radiation to a minimum. Frequently, however, simple shielding of the
primary element suffices to eliminate much of the error that otherwise
results.
Another static error is caused by the loss of heat that flows along the
primary element into the surrounding atmosphere. Figure 9-5 shows the
installation of a well in a pipeline. In cases where the temperature dif-
ference between the controlled medium and the atmosphere is large,
160 AUTOMATIC CONTROL

Figure 9-5. Installation of a well in a pipeline.

heat flow along the well may actually lead to a reduction of temperature
even in those parts of the well that are immersed in the controlled
medium. This may lead to a temperature indication by the primary
element which is too low. In cases of gas and vapor measurements with
considerable temperature differences between atmosphere and controlled
medium the following precautions should be taken;
1. Lagging of the well exposed to the atmosphere;
2. Deep immersion of the primary element into the controlled me-
dium. The immersion depth should be at least 1 Vi to 2 times the
length of the sensitive part of the element when the fluid velocity
is high, and twice that much when it is low. Another rule is 6 to 8
times the well (or protection tube) diameter for high and 12 to 16
times for low velocities.
3. Heating of the well exposed to the atmosphere, possibly by
branching off some of the process fluid.
4. If oil, mercury, or similar material for better heat transmission is
used, as described further below, such filling should not go beyond
the temperature-sensing element, since it otherwise contributes to
heat conduction away from the element.
It happens occasionally in very fast flowing gases that temperature
MEASURING ELEMENTS 161

signals of the primary element are on the high side. This is mainly due
to friction. Air flowing at 10,000 feet per minute may result in an error
of about 2°F, which rises however to 40°F at 40,000 feet per minute.
The error of temperature-measuring elements due to their location in
air pockets, under conditions of impinging flames, etc., is well known.
The conclusion regarding temperature reproduction by the primary ele-
ment is that conditions and location of the installation must be con-
sidered in order to obtain a correct primary feedback signal.
Thermal lags. Thermal lags affect the dynamic response of the con-
trol system. Consider a fllled-system thermometer bulb as shown sche-
matically in Figure 9-6. Suppose the bulb has been moved from a cold

TO CONTROLLER

Figure 9-6. Filled-system thermometer bulb.

medium into the tank bath as shown, which is at an elevated tem-


perature. This corresponds to a step change. Heat flows then through
the walls of the bulb to the actuating medium until the temperature of
the actuating medium is equal to that of the bath. Only then will the
measuring means settle at the new temperature.
162 AUTOMATIC CONTROL

The following discussion investigates the rate and quantity of heat


flow that takes place until the thermal balance is reestablished.
Heat quantity is generally expressed in British thermal units—Btu—
defined as the heat required to raise the temperature of one pound of
water by one degree Fahrenheit. Heat capacity expresses how much
heat in Btu has to be transmitted to one pound of a substance in order
to raise its temperature by one degree Fahrenheit.*
The amount of heat required to raise the heat of the primary element
to that of the bath depends on the heat capacity of the primary element.
It is given by the equation
Q = c,W{Ta-T,) (9-1)
where
%

Q is the amount of heat in Btu to be stored in the primary element


Cp is the heat capacity of the primary element in Btu/(lb)(°F)
W is the weight of the primary element in pounds. Since the primary
element contains several substances of different weights and heat
capacities, it is necessary to establish the c^W product for each
substance and then add the various products to obtain a single
CpW.
{Ta — T„) is the temperature rise from Tj, to T^.
The heat flow into the capacities of the primary element in Figure 9-6
has to overcome a certain thermal resistance. This resistance exists not
only in the well of the bulb—in fact this is the smallest contribution—
but mainly in the film on both sides of the bulb’s metallic surface.
A fluid is an extremely poor heat conductor. For example, at room
temperature, the following figures illustrate the poor conductance of
water and air as compared with steel:
Steel 26 Btu/(hr) (sq ft) (°F/ft)
Water 0.36
Air 0.016
Steel conducts over 70 times as much heat as water and over 1600 times
as much as air.
However, fluids produce convection currents in partial compensation
for their poor conductance. This is due to the formation of layers of dif-

* The term specific heat is frequently used in the same connection. Specific heat compares the
heat capacity of a given substance to that of water and is expressed as a ratio between the
two. Since, however, the heat capacity of water is approximately unity, it follows that nu-
merically specific heat and heat capacity are identical for all practical purposes.
MEASURING ELEMENTS 163

ferent temperatures. If the warmer layers are in the lower regions then
they rise beeause of their lesser density. In rising they propagate heat
much faster than would be possible by conduction only. This kind of
heat propagation is referred to as natural convection, or simply convection.
Convection currents, however, slow down considerably as the fluid
gets nearer a wall because of the friction between the wall surface and
the fluid. Close to the surface there are no longer convection currents,
and a sluggish film clings to the surface, through which the heat has to
pass by conduction. In spite of the relative thinness of the film, its re-
sistance to heat flow is far greater than the resistance of the bulb wall.
Figure 9-7 shows a cross-section through films and bulb wall of a pri-
mary element similar to Figure 9-6. Suppose it were possible to measure

Figure 9-7. Cross-section through a piece of bulb wall with films


and temperature gradient.

the temperature at various points in the fluids, films, and bulb wall, and
that these measurements could be done shortly after the bulb has been
changed from a cold to a hot bath. The various temperatures thus
measured could then be represented by a graph showing how the tem-
perature changes through the various points of the cross-section. Such a
temperature gradient is illustrated in Figure 9-7, showing the sharp tem-
perature drop through the film, expressing its high resistance, and the
relatively small drop in the bulb wall itself.
164 AUTOMATIC CONTROL

The resistance of the film decreases with the velocity of the convec-
tion, the density of the fluid, the thermal conductivity of the fluid, and to
a lesser degree with the heat capacity of the fluid. It increases slightly
with the diameter of the bulb, and the viscosity of the fluid.
Natural convection is not the only form of convection. By means of
stirrers, etc., a fluid may be agitated and thus the speed of heat propaga-
tion can be increased by so-called forced convection. The possibility of
increasing the rate of heat flow by locating the primary element where
convection currents are greatest (excepting those extremes where the
very speed of those currents contributes to the error, as previously
described), or by introducing forced convection should be given first
consideration in any attempt to reduce the thermal lag in temperature
control systems.
The lag in heat transfer increases with the protection required for the
primary element. Thermocouples, for example, generally require wells
to protect the thermocouple and to allow its removal in service without
interrupting the process. In this case, the air pocket in the well, and the
poor convection within it, slows down the response.
Hornfeck* compared responses of two thermocouples. Both were of
No. 20 wire, iron-constantan couples, in brass wells of ys in. outside and
yi6 in. inside diameter. In one case the thermocouple wires were not
touching the wall, in the other they were silver-soldered to it. He found
that measuring the temperature of air at 1150°F would take less than
one half the time with the thermocouple silver-soldered to the wall and
in agitated water it would take less than 4 per cent of the time it would
take with the wire not touching the wall. This illustrates the effect of the
air pocket. Of course, with a thermocouple silver-soldered to the wall of
the well, one of the purposes of the well—to make the thermocouple easy
to remove—is lost. Some commercially available arrangements to im-
prove the heat transfer will be discussed farther below.
Occasionally two protecting tubes instead of one are required. This
adds considerably to the temperature lag. Similarly, in pressure-actuated
bulbs, a well is used over the bulb for extra protection and easy removal
of the bulb. In such cases, the interspace should be filled with oil,
mercury, or a similar liquid. Improved thermal contact may also be ob-
tained by inserting metallic foils in the air space.
The equation that expresses heat flow through a thermal resistance is
q = hA (T; - r.) (9-2)

* Hornfeck, A. J., “Response Characteristics of Thermometer Elements,” A .S'M^ Transactions,


(Feb. 1949).
MEASURING ELEMENTS 165

where
q is the rate of heat flow in Btu/hr,
h is the heat-transfer coefficient, comprising films, wall and other paths
of thermal conductance, in Btu/(hr) (sq ft) (°F),
A is the cross-sectional area through which the heat flows,
is the temperature dilference of the two endpoints between
which the heat flows.
The product hA is the total heat conductance between the controlled
variable and the primary element.
This equation shows that to increase the rate of heat flow for a given
temperature difference either the area or the heat-transfer coefficient
should be increased.
The heat-transfer coefficient is largely controlled by the film, as has
been discussed. Changing the bulb material or reducing its wall diameter
may therefore be of benefit for the heat capacity, but has little influence
on the overall heat-transfer coefficient.
Another recourse is, however, to use a bulb of large area. This is fre-
quently done in gas temperature control and particularly where convec-
tion currents are slow. Various types of large-area bulbs are available.
Time Constant and Dead Time. As has been defined before, the
time constant is the time required to produce 63.2 per cent of the total
change. This is a convenient figure by which to express the dynamic
action of a temperature-measuring primary element and which allows
comparing the performance of similar elements.
In connection with temperature measurement, the speed of response
is frequently expressed in response time instead of the time constant,
and is defined, for example, by the Scientific Apparatus Makers Associa-
tion (SAM A Tentative Standard RC3-12-1955) as follows: “The SAM A
response time is the time required for the instrument to achieve 90 per
cent of the change which the instrument is going to make as a result of
an abrupt change in the measured quantity.” Conversion from the time
constant to the 90 per cent-response time is readily made by multiply-
ing the time constant by 2.3. In the following discussion, the time con-
stant will be used exclusively.
A large number of measurements have been made by various
researchers. Unfortunately, their results are widely divergent. Figures
given in the following discussion are an attempt to correlate these
measurements, but while they may be valid for certain cases they can
at best be considered general behavior patterns and are not meant to be
specific data for universal application.
166 AUTOMATIC CONTROL

The smaller the time constant, T^, the faster can the controller be
made to respond to process deviations. The time constant is given by

T, = 60 (9-3)
hA
The coefficient 60 is required to convert time units, if the time constant
is to be expressed in minutes.
Tests to determine time constants are not necessarily based on step
changes. Occasionally, the response of a primary temperature element
is shown graphically as illustrated in Figure 9-8 where the temperature
of the test bath is gradually raised. The change of the bath temperature
is expressed by graph a and the response of the primary element by
graph b. The time constant can be obtained from such graphs by means
of the equation
Ti = E/m (9-4)
where E is the dynamic error in °F and m is the rate of change of the
temperature in the test bath in °F/min. For example, the dynamic error
in Figure 9-8 is 15°F and the rate at which the test bath temperature

0 20 +0 60 80 100 120 140 160 180


Seconds

Figure 9-8. Response of primary element to gradual increase of


test bath temperature.

changes is 30°F/min. Consequently, the time constant is 15/30 = 0.5


min.
This form of presentation illustrates the effect of thermal lags particu-
larly well, since the dynamic error expresses the quantity by which the
controller lags in its response to temperature changes, and hence may
contribute to instability in the control system. It should be noticed that
according to equation (9-4), the dynamic error increases with the rate
of change of the test bath temperature which in a control system corre-
sponds to the controlled variable.
MEASURING ELEMENTS 167

It follows that the dynamic error is proportional to the rate of change


of the controlled variable. If the controlled variable changes slowly
enough, then the thermal lag will be of no adverse consequence. It may
mean unnecessary expense to attempt reduction of thermal lags in a
control system for which only slow temperature changes are to be
expected.
Similarly, if the time constant of the process is considerably larger
than what can be expected even of a slow primary element, no benefit can
be derived from selecting a high-speed element. This is in general typical
for fractionating columns where the temperature control system may have
to cope with time constants of 40 to 60 min. Another example are
annealing furnaces with time constants of 10 to 20 min. For heating
systems, the time constant is usually between 1 and 5 min.
The weU, interspace, bulb, etc. are all capacitances. It has been pointed
out before that with multi-capacitance systems there results the assump-
tion of dead time in addition to the time constant. For primary tempera-
ture elements in wells this dead time amounts to about 6 to 7 per cent
of the time constant for water and to about 2 to 3 per cent for air.
In view of rather considerable dead time in most temperature-con-
trolled processes, the dead time of the primary element can usually be
neglected. Typical dead times in temperature-controlled processes are
as follows:

Process Dead Time in Min.

Ammonia absorbers 8-10


Annealing furnaces 1-3
Fractionating columns 1-10
Heating systems 1-5
Oil tube stills 3
Superheaters 2

Linahan* measured the time constants of gas-filled bulbs using two


configurations as illustrated in Figures 9-9a and b. The two gas-filled
bulbs are identical. The inside diameter of the well, however, is smaller
in case {a) than in {b). In one case, the interspace is filled with grease,
in the other, with mercury. Time constants were very nearly the same
so that for practical purposes, they can be considered identical. They

* Linahan, T. C., “The Dynamic Response of Industrial Thermometers in Wells,” ASME


Transactions, (May 1956).
168 AUTOMATIC CONTROL

were approximately 4 seconds for water with forced convection and 130
seconds for air flowing at 1400 ft/min., which corresponds with normal
air duct velocities.

(9) Inside diainelerof well 0.384 inch (b) Inside diameter of well 0.407 inch
Interspace filled with grease Interspace jilled with mercury

Figure 9-9. Dimensions of two bulbs used in Linohon’s measurements.

For arrangements which are similar to the one used in Figure 9-5, the
following time constants may be used as guide:

Controlled Medium Time Constant in Min.

Water and dilute solutions with


forced convection 0.04-0.12
Saturated high-pressure steam,
molten metals, water, and dilute
solutions with natural convection 0.12-0.5
Oils and saturated steam at
atmospheric pressure 0.3-1.2
Air at normal duct velocities 1-3
Heavy oils, syrups, etc. 2-7
Air with natural convection 6-30
MEASURING ELEMENTS 169

The difference in time constants for the various types of filled-system


thermometers, i.e., gas, vapor, liquid, and mercury, are small enough to
be ignored. Measurements taken by the Brown Instrument Company*
showed an average time constant of 0.1 min. for bulbs without well im-
mersed in water at natural convection.
Figure 9-10 shows an interesting result from the same series of
measurements. Curve (a) shows the response of a vapor-filled system

Figure 9-10. Response of a vopor-filled system at cross-ambient conditions.

which is transferred from a temperature lower to one higher than the


temperature to which the connecting tubing and the measuring spiral
are exposed. Curve (b) shows the reverse case, the response after a trans-
fer from hot to cold. Obviously the temperature around connecting
tubing and measuring spiral is about 65 °F. As the actuating medium in
the bulb reaches this temperature, the fluid outside the bulb starts to
liquefy. The transition from the vapor to the liquid phase, and vice
versa, takes time as expressed in the illustrated time delay. It will be
noticed that this time delay is much longer when the temperature is
lowered than when it is raised. The explanation is that with the rising
temperature only the interface between liquid and vapor has to be
brought to the new temperature. With a decreasing temperature all the
vapor in the bulb must first be lowered to the new temperature.
This illustrates that control under such conditions would be unreliable.
The ordinary Class II (vapor-filled) thermometers should never be used
under cross-ambient conditions. There are, however, dual-filled Class II
systems available which circumvent this shortcoming. In dual-filled
bulbs, the volatile liquid and its vapor cannot leave the bulb. They only

* “Response Speeds of Pressure Type Thermometers,” Bulletin No. 60-1, published by The
Brown Instrument Company, Philadelphia, Pa., (1942).
170 AUTOMATIC CONTROL

partially fill the bulb space. The remaining space, as well as connecting
tubing and measuring spiral, are filled with a non-volatile liquid. The
fluids do not intermix. These bulbs are well suited for any temperatures
within their range.
Hornfeck* made a number of measurements on resistance thermom-
eters and thermocouples. Using them in air at approximately 1000
ft/min. gives the following time constants:
Resistance thermometer 0.07 min.
Iron-Constantan thermocouple
No. 14 wire 0.13 min.
No. 19 wire 0.05 min.
These time constants are sufficiently close to consider the response of
either resistance themometers or thermocouples as equally fast. It is in-
teresting to see that the thinner wire, i.e., less mass, hence lower heat
capacity, results in considerably faster response.
The effect of dilferent capacities is even more pronounced in a com-
parison between
(a) a resistance thermometer in a stainless steel well of 0.600 outside
and 0.563 inside diameter
with (b) a thermocouple of No. 12 wire in a stainless steel well of 0.563
outside and 0.438 inside diameter.
The time constants in min. are approximately as follows:

Thermocouple Resistance Thermometer

Air at 1000 ft/min. 2.27 0.67


Air at 5300 ft/min. 1.20 0.33
Water at 60 ft/min. 0.95 0.12

The mass of the thermocouple well is about three times that of the
resistance thermometer well. Besides this, the comparatively heavy
thermocouple wire adds even more to the larger capacitance. This illus-
trates that a comparison between two systems can be made only if they
are operated under identical conditions. The contribution of the capaci-
tance can also be seen from the above figures because the difference in
time constant is largest where the capacitance/resistance relation is also
largest, i.e., in water.
MEASURING ELEMENTS 171

This is even more noticeable by increasing the air velocity from 1000
to 5300 ft/min. Since the resistance decreases with the higher velocity,
i.e., the capacitance/resistance ratio increases, the thermocouple time
constant increases from about 3.40 to 3.65 times the time constant of
the resistance thermometer.
The decrease of thermal resistance due to increased speed of convec-
tion currents in gases has been discussed before. This is not limited to con-
vection currents. Thus, when measuring air temperature, the thermal re-
sistance and hence the time constant depends on the velocity of air that
passes over the measuring element. This is illustrated in Figure 9-11. For

Figure 9-1 1. Change of time constant with air velocity.

example, the relative time constant for 600 ft/min. on this diagram is
about 1.1, and for 5300 ft/min. about 0.5. Hence, if an actual time con-
stant of 3 minutes has been determined for 600 ft/min., then the cor-
responding time constant for 5300 ft/min. would be (0.5/1.1) X 3 = 1.36
min.
An attempt has been made to establish a table for the time constants
of thermocouples and resistance thermometers. In view of the various
factors that have been shown to control the time constant, the limits
through which each time constant may vary must be consequently large.
Correlation of available data gives the following approximations:
172 AUTOMATIC CONTROL

Time Constant in Min.

with metal with ceramic


Controlled medium bare protection protection
tube tube

Water and dilute solution


with forced convection 0.0005-0.05 0.01-1.2 0.2-3.0
Saturated high-pressure
steam, molten metals,
water and dilute solutions
with natural convection 0.01-0.1 0.1-5.0 0.3-5.0
Oil and saturated steam at
atmospheric pressure 0.02-0.2 0.2-7.0 0.7-10.0
Air at normal duct
velocities 0.03-0.3 0.3-10.0 2.0-12.0
Heavy oils, syrups, etc. 0.08-0.8 1.0-20.0 4.0-20.0
Air with natural
convection 0.15-1.5 2.0-60.0 8.0-60.0

Comparing this table with the values given previously for filled-system
thermometers, it will be noticed that for thermocouples and resistance
thermometers in metal protection tubes, time constants are both shorter
and longer than with filled-system thermometers. The explanation is
that a thermocouple or resistance thermometer hanging free in a well is
slower in response than the usual filled-system thermometer, but that
high-speed elements have been developed that have shortened the time
constants considerably. Nevertheless, filled-system thermometers have
been made faster in response too.
Three basic approaches are available to design a temperature-measur-
ing primary element with shorter time constants;
1. increase of outside surface area
2. decrease of mass
3. improvement of thermal contact within the element.
A few examples will now be given where one or the other of these
approaches has been used, first in filled-system thermometers, then in
thermocouples, and last in resistance thermometers.
Figure 9-12 shows a capillary coil bulb. The extended length of such
a bulb is about 30 ft, which obviously results in a large increase of area
over an equivalent 6 in. standard bulb. The disadvantage is that it
cannot be used with a well without losing its advantage of large outside
MEASURING ELEMENTS 173

Figure 9-12. Capillary coil bulb.

surface area. Henee, it lacks mechanical protection and the ease of


removal of a standard element. However, in air duets, dryers, etc. it may
be used to great advantage.
The Taylor Transaire temperature transmitter is illustrated in Figure
9-13. This design permits using a considerably smaller primary element

ZERO
RELAY ENT SCREW

20 PSI
AIR S

MAIN SPRING
RECEIVER TEMPERATURE AND BAROMETRIC
PRESSURE COMPENSATING BELLOWS

ZERO ADJUSTING SPRING

STOP PLATE
SPE ED-ACT
CALIBRATION LOCK SCREWS
ADJUSTING
WHEEL

NOZZLE

NOZZLE LOCK NUT

BAFFLE
UB> BASE
FORCE BALANCE B

-TEMPERATURE
SENSITIVE BULB

Figure 9-1 3. Transaire temperature transmitter.


(Courtesy of Taylor Instrument Cos.)

than others. The actuating medium in a filled-system thermometer is


not only inside the bulb, but also outside in the connecting tubing and
measuring spiral. This means that not only the bulb but other parts as
well are temperature sensitive. This can be minimized by making the
ratio of actuating medium inside bulb to actuating medium in conneet-
ing tubing and measuring spiral as large as possible. This results in a
certain minimum bulb volume. By placing, however, a transmitter right
at the primary element and eonverting the element’s response into an
air pressure signal, the volume outside the bulb becomes much reduced
174 AUTOMATIC CONTROL

and the bulb can be made correspondingly small. The bulb sizes are as
small as % by 3 in.
Furthermore, this transmitter is built to provide rate action. Due to
the relatively long time constants, rate action in the controller is gener-
ally desirable. In this case, it is built into the transmitter and the primary
feedback signal has the corresponding correction. Rate action in the
controller is not required.
Figure 9-14 shows a pencil type thermocouple which makes the iron
protection tube part of an iron-constantan thermocouple. The constan-

CONSTANTAN WIRE

Figure 9-14. Pencil-type thermocouple.

tan element, which mechanically is considerably weaker than iron, is


fully enclosed by the iron. This gives minimum mass with relatively good
mechanical protection without well. However, easy exchangeability is
again sacrificed. Response speed, however, is considerably increased.
A high-speed thermocouple developed by Leeds & Northrup for steam
temperature control is illustrated in Figure 9-15. To improve thermal
contact between thermocouple and well, spring pressure is applied on
the sleeve which forces the silver plug—containing the measuring junc-
tion—against the base of the well. The purpose of the silver plug is to
assure a large area of contact by a material of low thermal resistance.
A similar approach has been used in Foxboro’s “Dynatherm” for a
resistance thermometer, as illustrated in Figure 9-16. The temperature
-sensitive resistance element is wound on a silver core. Metal foil in
the base of the well assists the heat transfer to the solid silver tip and
core. Thus the effect of the air space between resistance temperature
detector and inside wall of the well is minimized and heat flow is rapidly
and evenly distributed through the aluminum foil and the silver parts.
Nothing has been said so far about radiation-type primary elements.
Since they do not depend on convection and conduction, the corre-
sponding time-delaying factors do not apply. Only the mass of the
radiation-sensitive element (usually a thermocouple or thermopile) must
be considered. Time constants are 0.2 to 0.5 sec.
MEASURING ELEMENTS 175

Figure 9-15. High-speed thermocouple. (Courtesy of Leeds & Northrup)

Where time constants are relatively long, as in the majority of tem-


perature measurements, the speed of the controller becomes a negligible
factor in the control system, and any improvement there is without
influence on the whole. In choosing a controller for a temperature-
controlled process, it is advantageous to know: (1) the process gain in
units change of controlled variable per inch displacement of final con-
trol element; (2) the dead time; (3) the load changes that can be expected,
in terms of temperature changes that would take place if no control
were to be applied; (4) the speed with which these load changes occur;
(5) the allowable maximum temperature deviation.
Dead times of various components in the control loop are added.
176 AUTOMATIC CONTROL

GLASS
EXTENSION
FIBER-
TUBE
WRAPPED
LEADS

SILICONE
SEAL -WELL
CERAMIC
INSULATOR

AIR SPACE RESISTANCE


WINDING

TIP-
SENSITIVE
CAPSULE

SOLID SILVER
TIP & CORE ALUMINUM
FOIL

Figure 9-1 6. ‘Dynatherm” resistance temperature detector.


(Courtesy of Foxboro Co.)

Time constants can be negiected except one, if tliis one is more tlian twice
tTie otliers. It is tlien possible to malce tlie following decisions:

1. If the load changes are slow, no rate action is required;


2. If the floating rate to be calculated according to page 42 results in
a satisfactory answer, this action should be chosen.
3. If proportional-position action as calculated in accordance with
the method on page 47 gives a satisfactory answer, this should be
chosen; otherwise proportional plus reset plus rate action should be
chosen.

This shows that the immediate recourse, as it is frequently talcen, to


proportional plus reset plus rate action for temperature control, is work-
able but is certainly not the most economical.
10. CONTROLLERS

The controller links the measuring means and the final control element.
The purpose of the controller is to regulate the control point at which
the controlled variable is to be maintained, to adjust the proportional
speed or band of the signal to the final control element, and to provide
responses which may depend not only on the magnitude of a deviation
from the controlled variable, but also on the rate of change and the
duration of this deviation. How these functions are achieved in different
types of controllers is described in the following pages.

Pneumatic Controllers with Position Balance


Pneumatic controllers in which the input is a position differ basically
from those where the input is force. As previously pointed out, the posi-
tion system is primarily used with recorder controllers or indicator
controllers.
Set Point Mechanisms. The first function of any controller is to
compare the primary feedback from the measuring element with the
reference input from the set point mechanism. This is essentially a
mathematical operation:

(Reference input) — (Primary feedback) = Actuating signal

For the purpose of this mathematical operation, mechanical linkages


are used like those shown in Figure 10-1. Diagram (a) is self-explana-
tory, the flapper position is the result of the set point adjustment as well
as of the input. In diagram (b) an eccentric disc is mounted on a shaft.
The input signal is rotation of the shaft. To change the set point, the
relative position of shaft and eccentric disc is adjusted manually by
rotating the disc on the shaft. In diagram (c) flapper and nozzle are
independently positioned, one by the input, the other by the set point
adjustment.
Figure 10-2 illustrates an actual summarizing (or subtracting) linkage
as used by the Mason-Neilan Regulator Co. This linkage is carried by
the pen mechanism. The input signal from the measuring element is
transmitted by the element link to the pen arm. The control pin on the
pen connector engages a slot in the control lever and transmits this
177
178 AUTOMATIC CONTROL

Figure 10-1. Set point mechanisms.

Figure 10-2. Mason-Neilan set point mechanism.

motion to the so-called gimbal link which provides the actuating signal
for the flapper.
The index-connecting link transmits the position of the set point
adjustment knob to the summarizing linkage. Turning the knob shifts
CONTROLLERS 179

the position of the gimbal hnk with respect to the pen. The arrangement
is basically that of Figure 10-la.
Controller Mechanisms. From the set point mechanism the actuat-
ing signal is applied to the controller mechanism, where different con-
troller actions are produced. Two-position action is not possible with
pneumatic control, unless special relay devices are used. This is because
the flapper motion can only produce a gradual change of the nozzle
back pressure and not an on-off action. However, since a flapper motion
in the order of 0.001 in. usually suffices to change the nozzle back pres-
sure through its full range, arrangements without feedback such as
shown in Figure 10-1 can be adjusted to provide proportional bands of
one per cent or smaller and may then be considered two-position con-
trollers for most practical purposes.
Flapper-nozzles without feedback are available for proportional bands
up to approximately 10 per cent. Beyond this, feedback becomes a
requirement. It must be realized that with a total motion in the order
of 0.001 in., the flapper is accurately positioned within 0.00001 in. when
a change of one per cent in the controller output signal is desired. As
a matter of fact, a high degree of accuracy for considerably less motion
is actually obtained.
Feedback mechanisms may be divided in those repositioning the
flapper and in those repositioning the nozzle. An example of each is
described here.
Figure 10-3 illustrates the operation of a proportional-position con-
troller as developed by the Minneapolis-Honeywell Regulator Co. The
set point mechanism follows the same principle as the one previously
described. The flapper position is determined by the displacement bal-
ance between actuating signal and feedback signal. The latter is obtained
from the controller output signal through the bellows. The function of this
mechanism is similar to the set point mechanism, i.e., a subtraction of two
signals. In this case, the equation reads

(Actuating signal) — (Controller output signal) = Flapper position

The difference with the set point mechanism, however, is that in the
equation of the proportional-position mechanism, a closed loop relation
exists: the flapper position determines the nozzle back pressure which
determines the controller output signal which in turn determines the
flapper position. The relationship can be expressed in form of a diagram,
as in Figure 10-4. The feedback mechanism is expressed by the summing
point in which the controller output signal is subtracted from the actu-
ating signal. The pilot valve amplifies the result and the symbol used is
that of an amplifier.
180 AUTOMATIC CONTROL

Figure 10-3. Proportional-position controller with flapper feedback.

Going back to Figure 10-3, the proportional band adjustment is


indicated by an adjustable fulcrum between two links. This amplifies to
a certain extent the feedback motion of the bellows. It is shown as an
amplifier in the feedback loop of Figure 10-4; an arrow indicates that
the gain of this amplifier, i.e., the proportional band, is adjustable.
PILOT
VALVE

Figure 10-4. Diagram of proportional-position controller.

One other feature is essential in this construction. As seen in Figure


10-3, the flapper is kept in contact with a pin on the whiffletree by
means of the spring. This prevents the flapper from being forced against
the nozzle with subsequent mechanical damage. When the whiffletree
swings out beyond the point at which the flapper contacts the nozzle.
CONTROLLERS 181

contact between whiffletree and flapper is simply lost and no damage


is done.
Figure 10-5 illustrates a somewhat different method of feedback. It
follows a principle used by the Taylor Instrument Cos. The feedback

Figure 10-5. Proportional-position controller with nozzle feedback.

operates on the nozzle instead of the baffle as in the previous arrange-


ment. The proportional band is adjusted by means of a parallelogram
system of levers which converts the horizontal movement of the bellows
into a smaller or larger vertical movement of the nozzle.
The action of the parallelogram is further illustrated in Figure 10-6. The
arrangement has two fixed pivots, A and B. Of these pivots, B can be
manually shifted and constitutes the proportional band adjustment
which, as mentioned before, is equivalent to changing the gain in the
feedback. Figure 10-6 illustrates two adjustment positions. The two
upper diagrams apply for one, the two lower for the other. By displac-
ing the push rod through a horizontal distance s, the nozzle tip moves
vertically, increasing distance d-^ to t/g. In changing the pivot B from B^
to ^2, and displacing the push rod again through the same distance s,
the resulting increase of distance to d^ is considerably more than d^
to d^. A diagram for this arrangement would be the same as that in
Figure 10-4 which can be applied for either the flapper or the nozzle
feedback.
Mechanisms for reset and rate action differ from those for proportional-
position action because their effect includes a time function. The stand-
ard method for obtaining time functions in pneumatic elements consists
in combining a resistance and a capacitance, as shown in Figure 10-7.
A change of air signal will displace the bellows. However, the response
of the bellows is delayed because of the resistance and the capacitance in the
182 AUTOMATIC CONTROL

— J k-S

Figure 10-6. Adjustment of proportional band by parallelogram.

ruwv

n
AIR SIGNAL

RESISTANCE

Figure 10-7. Pneumatic RC network.

circuit. For example, an increase in pressure of the air signal results in


flow across the orifice resistance. This flow continues until the pressures on
both sides of the resistance are equalized. The air volume required to
equalize the pressures depends on the magnitude of the capacitance.
Hence it follows that the time required to transmit the change in air
pressure, i.e., the air signal to the bellows, depends on the magnitudes
of resistance and capacitance.
The combination of a resistance R and a capacitance C to produce
the described action is called SLYI RC circuit. Frequently, the capacitance
CONTROLLERS 183

in the bellows and connecting lines in combination with the orifice is


large enough to provide the effects of an RC circuit. In this case the
additional capacitance of the volume chamber in Figure 10-7 is not
needed.
Air flow through the orifice is roughly determined by the flow equa-
tion Q = kA Xhis means that the flow rate diminishes as
the pressure difference decreases. The result is that the increase of the
bellows pressure and hence its movement follows a curve like the one
in Figure 10-8. Since the pressure differential becomes increasingly

Figure 10-8. Bellows motion in RC circuit.

smaller, the flow rate and with it the bellows motion becomes slower.
Mathematically, a finite position will never be reached and the curve
approaches some finite value continuously without ever reaching it.
Under practical conditions, a finite position is obtained. Nevertheless, it
would be difficult to give an exact figure for the time it would take the
bellows to reach a new position under certain defined conditions. This
limitation has led to the usage of the time constant, which has been dis-
cussed before. The time constant is the time the bellows needs to change
through 63.2 per cent of the maximum motion that results from a signal
change.
The method by which proportional-position and reset actions are
combined is illustrated in Figure 10-9. Upon a change, e.g., an increase
of controller output signal, the proportional position bellows will act as
feedback without delay. This will minimize the flapper motion. The
effect of the reset bellows will be gradual because of the RC network.
Its action opposes that of the proportional position bellows, moving the
flapper away from the nozzle, thus increasing the controller output
signal beyond the first impulse which was controlled by the proportional-
position feedback. This means that positive feedback is added to nega-
tive feedback. Figure 10-10 is the resulting diagram. The RC network
is indicated by a separate box. Since the reset feedback that passes
through it is subtracted in the measuring point, and the resulting signal
184 AUTOMATIC CONTROL

Figure 10-9. Proportional plus reset feedback.

Figure 10-10. Schematic drawing of proportional plus reset controller.

becomes subtracted a second time at the summing point where it is


added to the actuating signal, the net effect is that the reset feedback is
positive.
The proportional band adjustment affects both feedbacks, propor-
tional and reset. This is necessary because reset action is to repeat the
effect of the proportional-position action by a certain number of times
per minute, which is expressed in the reset rate. Hence, the effect of the
reset action must be proportional to that of proportional action which
is assured by a common gain adjustment.
The reset rate as such can best be adjusted by making the resistance
variable. This is accomplished either through a needle valve which pro-
vides the necessary adjustable restriction or by a capillary, the length of
which can be changed.
CONTROLLERS 185

Rate action implies that the controller output signal be changed with
the rate of change of the actuating signal. Rate action is provided by
adding another RC arrangement as illustrated in Figure 10-11 (cf. Figure
10-10). The dashed line indicates an alternate connection of essentially

Figure 10-11. Schematic drawing of proportional plus reset plus rate controller.

the same effect. This second RC network delays the feedback action,
thereby allowing the actuating signal to pass undiminished through the
first summing point. This is further illustrated in Figure 10-12. The
bellows opposing the reset bellows is delayed in its action by an addi-

tional restriction and capacity tank. The consequence is that the pro-
portional-position feedback lags as desired. The magnitude of the lag
depends on the magnitude of the controller output signal as well as on
the speed with which this change occurs. This results in a larger con-
troller output signal than would be obtained with undelayed propor-
186 AUTOMATIC CONTROL

tional-position signal; and this delay will be proportional to the speed


with which the actuating signal changes, which is the purpose of rate
action.
From the foregoing, it can be concluded that the lag elfect of a pneu-
matic RC network produces reset action when used in positive feedback,
and rate action in negative feedback.
The effect of reset action is actually the same as if the set point were
to be changed. When the controlled variable is below the actual set
point, the reset action pushes the effective set point and with it the effec-
tive proportional band upward, thereby increasing the corrective action
of the controller.
This may result in a difficulty with three-term (proportional plus reset
plus rate) controllers when the set point is changed or when a process
is started up and the controlled variable is outside the proportional band
but rapidly approaching it. As the controlled variable enters the pro-
portional band, the rate action results on a controller output signal which
is proportional to the rate of change of the controlled variable. This
decelerates the controlled variable and reduces overshooting. Since,
however, the effective proportional band has changed because of reset
action, this effect takes place only when the controlled variable is far
within the actual proportional band. The result is considerably more
overshoot than would be obtained if the rate action could be made
independent of the reset action as in other pneumatic controllers
described further below.

Pneumatic Controllers with Force Balance


The force balance controller is used where set point and feedback
signal are available as calibrated pressure rather than linked with indi-
cator or recording mechanisms. This makes them essentially blind con-
trollers which may be used with or without recorder or indicator. Figure
10-13 shows the essential parts of a Foxboro Consotrol proportional
plus reset controller. The outstanding functional part of the unit is the
floating disc which acts both as the force-balance detector and as the
flapper of a conventional flapper-nozzle system.
Each of the four bellows exerts an upward force on the disc. The net
effect of all forces acting simultaneously is the establishment of the
horizontal position of the disc, its nearness to the nozzle, and hence the
controller output signal.
In operation, any change in pressure in either the primary feedback
or set point bellows slightly raises or lowers the corresponding side of
the floating disc and causes a change in nozzle back pressure which
operates the pilot valve to increase or decrease the controller output sig-
CONTROLLERS 187

A.

Figure 10-13. Principle of Foxboro Consotrol proportional plus reset controller.

nal. The feedback of this signal acting on the proportional bellows


re-establishes a balance of the forces on the disc. The change of con-
troller output signal is thus proportional to primary feedback or set
point pressure change.
If there is a sustained differential between set point and primary feed-
back, a sustained difference between proportional and reset bellows
pressures will result. Air will flow from one to the other, causing the
disc to alter the nozzle opening and to continuously change the con-
troller output signal to maintain this difference. This is the reset action.
The force balance floating disc rests against the proportional band
adjusting lever which provides an adjustable fulcrum about which the
floating disc moves. If the position of the proportional band adjusting lever-
fulcrum is changed, the moment arm of each bellows about the fulcrum
becomes also changed and a different set of forces is required to balance
the disc. For example, when the lever is set so that the distance of the
proportional bellows from the fulcrum axis is twice that of the primary
feedback bellows, a two-psi primary feedback change is balanced by a
one-psi output change. A proportional band of 200 per cent is the result.
188 AUTOMATIC CONTROL

When the distance of the proportional bellows from the axis is half that
of the primary feedback bellows, a one-psi primary feedback change is
balanced by a two-psi output change, which is equivalent to a 50 per cent
proportional band.
In 1947 the first stack-type controller was introduced by the Moore
Products Co. It was a radical departure from existing pneumatic con-
trollers and resulted in an entirely new design concept. Figure 10-14

Figure 10--14. Principle of Moore proportional-position controller.

shows the principle of the Moore proportional controller together with


the pilot valve which becomes an inherent part of the control circuit.
The upper diaphragm of the pilot valve is of smaller area than the
lower. Therefore, in order to balance the forces across the two-diaphragm
assembly, the pressure on the upper diaphragm must exceed that under
the lower diaphragm by a certain amount. These pressures correspond
to nozzle back pressure and controller output signal, respectively. Sup-
pose an excess of 2 psi nozzle back pressure suffices to balance the con-
troller output signal. This means that the balanced pilot valve main-
tains a 2 psi pressure difference between controller output signal and
nozzle back pressure. The nozzle bleeds directly into the controller
output. The purpose of this arrangement—to approach null-balance
conditions—has been described in a previous chapter.
The flapper is part of a rigid stem connected to three diaphragms.
CONTROLLERS 189

When any of these diaphragms deflect, the whole stack including the
flapper is bound to move.
The diaphragm between chambers B and C has half the area of either
of the other two diaphragms in the stack. The pressure in chamber B,
for example, exerts force in two directions: one through the left dia-
phragm and one through the right. Since the area is twice as much on
its left, the force in this direction is also twice that of the force to the
right. By considering forces to the right positive and those to the left
negative, the following equation for balanced conditions can be applied;
A - B + 0.5B - 0.5C + C - D = 0
or simply
A - 0.5B + 0.5C - Z) = 0
Since the areas of A, B, C, and D arc equal, they cancel out, and it is
only necessary to insert the pressures that exist in the respective chambers.
Figure 10-14 shows certain air pressure values which are assumed to
represent initial conditions. Suppose the primary feedback signal changes
and increases the pressure in chamber .S to 10 psi. The increased air
pressure exerts a force toward the left and the flapper moves toward
the nozzle. This raises the nozzle back pressure and as a consequence
the pilot valve acts to increase the pressure of the controller output signal.
The increased output pressure is fed back into chamber A. If the
needle valve of the proportional band adjustment is closed, the pressure
in D will remain constant.
In order to balance for the new condition it is necessary that
A = 0.55 - 0.5C + D
and inserting values,
A = 5 - 3 + 6 = ^
This means that the controller output signal increases by two psi for
four psi change in primary feedback signal, which corresponds with a pro-
portional band of 200 per cent.
To increase the gain of the controller and obtain, e.g., twice the
response in the controller output signal for a given change in primary
feedback, i.e., a proportional band of 50 per cent, the needle valve of
the proportional band adjustment is slightly opened.
This changes the behavior of the controller because the feedback,
which is directly applied to the A chamber, is now also active in the D
chamber, although less because the needle valve reduces the air pressure.
As air flows through the needle valve, increasing the pressure in the D
chamber, air also flows from D to C through the restriction shown. This
will not increase the pressure in C since the set point pressure is con-
190 AUTOMATIC CONTROL

trolled by a regulator which, once it is set, will keep the pressure constant.
The result is that the pressure in D is maintained between the controller
output signal and the set point pressure. Once these two pressures are
given, pressure in D depends on the amount of opening of the needle
valve.
If the needle valve is positioned so that the increase from 6 to 10 psi
in controller output signal raises the pressure in chamber D to 9 psi,
then a 4:3 relation exists between the two pressure increments. For a 4
psi change in primary feedback as in the above example, it is then
necessary to change the controller output signal to 14 psi in order to
obtain balance. This will raise the pressure in chamber Z) to 12 psi, and
inserting the values in the above equation will confirm the balance.
Figure 10-15 adds reset action to the mechanism illustrated in Figure
10-14. The connection between chambers C and D is taken out. To main-

Figure 10-15. Principle of Moore proportional plus reset controller.

tain the feedback pressure in Z) a second nozzle is added. Chamber D con-


nects through a restriction to nozzle B.
The back pressure in the B nozzle under otherwise static condi-
tions is controlled by a flapper rigidly connected to a diaphragm. The
back chamber of the diaphragm constitutes the capacitance of the reset
action. Air which flows through the needle valve of the reset adjustment
would gradually flex the diaphragm, moving the flapper toward the B
nozzle and thereby increasing the back pressure. Closing the reset
CONTROLLERS 191

adjustment when the nozzle is wide open maintains a minimum back


pressure.
The air pressure in the D chamber assumes a value between the con-
troller output signal, i.e., the A chamber pressure, and the B nozzle
back pressure. The wider open the proportional band adjustment is, the
closer will be this value to the controller output signal and, hence, the nar-
rower will be the proportional band.
At this point, opening the reset adjustment by a certain amount
results in a gradual build-up of pressure behind the diaphragm, moving
the flapper toward the B nozzle. This increases the pressure also in
chamber D which in turn closes nozzle A, and consequently raises
the controller output signal. A gradual increase of controller output
signal is thus produced as long as an unbalance exists. The rate of this
increase is determined by the amount of opening of the reset adjustment.
Figure 10-16 shows how rate action is added to the schema of Figure
10-15. The unit is inserted between the primary feedback and the propor-

RATE ADJUSTMENT

Figure 10-16. Rate action assembly for Figure 10-15.

tional-plus-reset assembly. The pressure of chamber E is applied to cham-


ber B of Figure 10-15. Air is admitted to chamber E through a restric-
tion and flows out to atmosphere through a nozzle. As the outlet area
through the nozzle is reduced by the approaching flapper, the pressure
in E increases. Assuming that the diaphragm between E and E is about
0.2 times the area of the diaphragm between E and G, then
Q2E - 0.2F + E = G
or 02E + 0.8F = G
In balanced conditions E = E and therefore E = E — G. When the pri-
192 AUTOMATIC CONTROL

mary feedback is 6 psi initially and increases to 10 psi, the pressure in


E will also rise because the flapper moves toward the nozzle. The pres-
sure rise in F is delayed because of the restriction and capacitance in
the chamber. Initially, it is 6 psi. The pressure in E would have to rise
to 26 psi in order to effect balance under these conditions, and would
therefore amplify the primary feedback signal considerably. Gradually,
however, the pressure in F also rises and the pressure in E is reduced
accordingly. The amplification in chamber E depends on the rate ad-
justment and also how fast the primary feedback signal changes.
Figure 10-17 is the diagram of this arrangement. Comparing this with
the three-term controller of Figure 10-11 shows a significant difference;

Figure 1 0-1 7. Schematic drawing of Moore proportional plus


reset plus rate controller.

the rate action is now completely separated from the proportional plus
reset mechanism. The result is that the shifting of the effective propor-
tional band due to reset action is without effect on rate action. When
the controlled variable is outside the actual proportional band, rate
action will dampen the process as soon as the controlled variable enters
the actual proportional band of the controller.
The Moore three-term controller is not built in separate units as may
be implied from the preceding description. Actually all these units
including the pilot valve are combined in one cylindrical assembly
measuring 3.375 in. in diameter and about 11 in. in length.

Hydraulic Controllers
Pilot valves and flapper-nozzles as used in pneumatic controls are
rarely used in hydraulic controls; four-way valves and jet pipes are gen-
erally preferred. Since the output from the jet pipe has considerable signal
range both in capacity and pressure it can be frequently used without a
four-way valve. Figure 10-18 shows however direct coupling of the two
CONTROLLERS 193

Figure 1 0-1 8. Hydraulic floating controller.

components in order to amplify the jet pipe signal. This combination of jet
pipe and four-way valve is called a booster.
The primary feedback is applied directly to the jet pipe by means of a
diaphragm. It is counter-balanced by a compression spring. Thus the
actuating signal—the deflection of the jet pipe—is the result of two
opposing forces: the primary feedback and the set point spring. The set
point adjustment simply changes the compression of the spring.
When the primary feedback increases, the jet pipe swings counter-
clockwise. This directs oil to the oriflce which opens to the left back
chamber of the booster spool. The resulting spool motion, moving toward
the right, tends to re-center the two orifices under the jet pipe. When
this is achieved, pressure in the left and right back chambers are equal
and the spool motion stops. In other words, the spool faithfully follows
the motion of the jet pipe. For the rest, the booster action is that of any
four-way valve. When the spool is displaced to the right, oil flows into
the right chamber of the actuating cylinder. The resulting motion of the
actuating piston continues as long as the jet pipe is displaced from its
mid-position. The larger the deflection, the wider open are the ports of
the four-way valve, and the faster is the actuating piston motion. The
action is therefore proportional-speed floating.
Hydraulic controllers frequently use proportional-speed floating
194 AUTOMATIC CONTROL

action, but by means of feedback they are readily converted into pro-
portional-position controllers. This is illustrated in Figure 10-19. The
primary feedback is applied to a bellows. The bellows deflection is bal-
anced by the spring characteristic of the bellows itself and a tension
spring. The set point is adjusted by changing the tension.

Figure 1 0-1 9. Hydraulic proportional-position controller.

The crossarm pivots about B when the bellows moves, and about A
when the actuating piston moves. A change in signal displaces the bel-
lows, rotating the crossarm about B and positioning the spool of the
four-way valve. This admits oil to one side of the actuating piston. The
resulting motion is fed back to the crossarm and, by rotation about A,
to the spool which is thereby recentered, stopping further motion of the
spool. Thus a position is obtained which is proportional to the primary
feedback signal. The proportion of position to signal, i.e., the propor-
tional band, is determined by the position of pivot B.
Although the hydraulic controller is preferred for proportional-float-
ing action and is readily adaptable for proportional-position action, the
combination of the two into proportional plus reset control is somewhat
more difficult. The main cause is the incompressibility of the fluid. The
resistance-capacitance network of the pneumatic circuit is not readily
applicable under these conditions. Pressure change by compression does
not apply for liquids. The method used for a hydraulic proportional plus
reset controller is shown in Figure 10-20.
CONTROLLERS 195

RESET
ADJUSTMENT

•, ,v ' S <S

-^777^
o 1
rr

/I
STABILIZER / %
■ PIVOT /f
iS OILINLET

PROPORTIONAL^
BAND ADJUSTMENT/

1-AAA/^"^

ACTUATING
CYLINDER

Figure 10-20. Hydraulic proportional plus reset controller.

Control actions are obtained by the hydraulic stabilizer. Suppose the


primary feedback signal rises. This moves the jet pipe counterclockwise.
The resulting increase of pressure in the corresponding receiving orifice
will drive the piston in the stabilizer to the right. The small bypass flow
through the reset adjustment may be disregarded here. The oil volume
which is displaced from the right side of the stabilizer piston flows into
the actuating cylinder. The actuating piston thus assumes a position
which is proportional to the stabilizer position.
The movement of the stabilizer piston feeds back to the jet pipe by
means of a lever—rotating about pivot A—and a spring, rebalancing the
jet pipe. Thus proportional-position action is obtained. The proportional
band is adjusted by shifting pivot The piston of the actuating cylinder
is positioned in proportion to the primary feedback signal.
Displacement of the stabilizer from its middle position compresses the
spring in the stabilizer toward the left or right, according to the direc-
tion of the displacement. In any case the spring force will tend to return
the stabilizer piston to its center position. The volume which is displaced
196 AUTOMATIC CONTROL

in this motion must flow through the reset adjustment, which is a needle
valve. Hence the amount of opening of the needle valve, i.e., the adjust-
ment of the reset rate, determines the speed with which the stabilizer
piston returns to its center position.
The result of the return motion of the stabihzer piston is that, through
the feedback linkage, the jet pipe continues by reset action the motion of
the actuating piston which was initiated by the proportional-position
action.

Electric Controllers
The term electric controllers, as used here, includes electronic con-
trollers. The simpler forms of electric control are most widely used in air
conditioning and heating control systems. However, they are not neces-
sarily restricted to this use and their inherent simplicity and proven re-
liability should entitle them to many additional control tasks.
A two-position control is illustrated in Figure 10-21. The primary con-
tacts may be operated by various means such as the bimetalhc element

PRIMARY
CONTACT

of a thermostat or the electrodes of a level controller, as was shown in


Figure 4-3. In the position shown, primary contact 5 just closed and the
held winding of the motor is energized through contacts C and B. As soon
as the motor starts rotating, the cam mounted on its shaft will close con-
tact D. The held winding remains energized through contacts C and D
CONTROLLERS 197

independent of the action of the primary contacts until the motor has
rotated through 180 degrees. At this point, the cam opens contact C and
the motor stops. It now requires closure of the A contact to again en-
ergize the motor and drive it through another 180 degrees. The motor
action is used to position a final control element.
A typical electric single-speed floating controller was described before
(Figure 4-8); the same type of shaded-pole reversible single phase
motor may be used for proportional control action as illustrated in Fig-
ure 10-22. This is the principle of the Honeywell Series 90 circuit. In this

Figure 10-22. Electric proportional controller.

arrangement, the primary feedback is applied to a bellows or equivalent.


The bellows positions the sliding contact of slidewire C. A relay
is built into the motor unit. It consists of relay coils E and F, and
contacts A and B. If more current flows through relay coil E than
through F, contact A will close. Conversely, if more current flows through
F than through E, contact B will close. The motor will move in one direc-
tion or the other, depending upon which contact is closed. The sliding con-
tact of a second potentiometer, D, is mounted on the shaft of the motor.
In the condition shown, both potentiometers, C and D, are in such posi-
tions that exactly the same amount of current flows through relay coils
198 AUTOMATIC CONTROL

E and F. When the primary feedback signal increases the sliding contact
of C moves clockwise. This increases the current through E and decreases
the current through F. Consequently, contact A closes and the motor
starts rotating in clockwise direction. This moves the sliding contact of
D until currents through E and F are again the same and the relay con-
tacts are broken. Thus the amount of motor movement is proportional
to the primary feedback signal.
A somewhat different arrangement, using the same motor, is illustrated
in Figure 10-23. This has the advantage over the previous one in provid-

Figure 10-23. Electric proportional controller with null balance.

ing a null balance at the measuring element. The primary feedback is


applied to a diaphragm. The resulting force is balanced by a moving
coil. Any unbalance results in a deflection, closing either contacts A or
B. This energizes the positioning motor which in rotating positions the
sliding contact mounted on its shaft. The resulting change in voltage ap-
plied to the moving coil rebalances the forces and opens the circuit.
In order to avoid contacts and slidewires, a number of other arrange-
ments are available. Generally, they belong in a higher price class.
Typical for these are synchros (see Figure 7-13) which operate through
amplifiers—electronic or magnetic—and servos to obtain proportional-
position action.
Where reset and rate actions are required, combinations of capacitances
and resistances as illustrated in Figure 10-24 are generally used. The lag
network operates by the same characteristics as its pneumatic counter-
part (Figure 10-7). Under balanced conditions, the voltage drops across
R and the load resistance add up to the d.c. supply voltage E. Suppose
CONTROLLERS 199

E =: LOAD

\
(a) LAG NETWORK

II

E
/ R’ '
LOAD

\
(b) LEAD NETWORK

Figure 10-24. Electric RC networks.

the supply voltage (Figure 10-24a) decreases. Without the capacitor, the
voltages across the resistances would drop correspondingly. With the
capacitor, however, an additional discharge current will flow through
the load resistance, which diminishes gradually until the capacitor
charge corresponds with the new value of E. Thus the capacitor has a
delaying effect in establishing balanced conditions after a change. It is
therefore called a lag network. The situation is the same whether the
supply voltage increases or decreases. A curve showing the change of
the current through the load after a change in supply voltage would
have the same contour as the curve of the bellows motion in a pneumatic
RC circuit shown in Figure 10-8. The magnitude of the time constant is
proportional to the product of capacitance and resistance. A condition of
this arrangement is that d.c. voltage is required.
The lead network shown in Figure 10-24b is another method of
combining electrical resistance and capacitance. Again under balanced
conditions, the circuit behaves the same with or without the capacitor.
When the supply voltage E changes, however, the capacitor becomes in
effect a conductor, shunting out resistor R. As the capacitor readjusts
its charge, its effect gradually diminishes and finally current will again
flow only through R. The result is that a change in voltage produces in-
itially a much more pronounced change of voltage across the load re-
sistance, which gradually subsides and finally assumes the proportion
that is given by the resistance values. The voltage across the load resist-
ance actually leads to the supply voltage change and is proportional to
its rate of change. The lead network provides rate action in the forward
path. If used for negative feedback, it provides reset action. This is the
reverse of the pneumatic RC circuit and the lag network shown in Figure
200 AUTOMATIC CONTROL

10-24a, which provide reset action in positive feedback and rate action
in negative feedback.
Figure 10-25 illustrates the use of RC networks in a proportional plus
reset plus rate controller as built by Leeds & Northrup. Two Wheatstone

O.C.
SUPPLY

Figure 10-25. Principle of Leeds & Northrup proportional plus


reset plus rate controller.

bridges—a control bridge and a feedback bridge—are combined in this


arrangement. The sliding contact of slidewire A is mechanically linked
with the-pen arm in the recorder. The set point is adjusted mechanically.
Deviation from the set point moves the sliding contact at A, unbalanc-
ing the control bridge. This results in a voltage due to the current flow
through the measuring leg of the bridge. Since, however, resistors R^, R^,
and and capacitor Cj constitute a lead network, rate action is provided.
Rate time is adjusted by changing the variable resistance R^.
A chopper, essentially a vibrating reed, rapidly switches from to ,62

applying to the amphfier either voltage in rapid pulses. As long as a dif-


ference between and persists, the amplifier puts out a signal to the
62

final control element which is coupled with potentiometer B.


Motion of the final control element and hence of the sliding contact
at B unbalances the feedback bridge. This produces a voltage across
capacitor C and resistors R and R^. As long as the controlled variable
2 2

does not return to the set point, an voltage other than zero is main-
tained. To provide an equal voltage, a current must flow through re-
62
CONTROLLERS 201

sistor charging capacitor Cg. As charges, current in the feedback


bridge tends to diminish. To keep charging current flowing, the final
control element must continue to move, producing a corresponding
change in position of the sliding contact at B, until is again at zero
with the controlled variable at the set point. Reset action is thus pro-
vided and reset rate can be adjusted by means of variable resistor R^.
The proportional band of the control action can be adjusted at G and H.
Figure 10-26 shows the arrangement of the two lead networks, one in
the forward path for rate action, the other as negative feedback for re-

Figure 10-26. Schematic drawing of Leeds & Northrup controller.

set action. The two variable resistors express the alternate means for ad-
justment of proportional band. In this, as in practically all electric and
electronic controllers, rate action is located before reset action. This re-
duces overshooting in those cases where the controlled variable is
initially outside the proportional band as has been previously discussed.
Electric and electronic controllers, as described so far, use a position
feedback obtained from the final control element. This is different from
the conventional pneumatic controller which feeds back its own output
signal. This output signal is then transmitted to the final control element.
The electronic controllers described in the following are in this and
various other respects a close equivalent of the pneumatic controller.
Figure 10-27 shows the principle of the Manning, Maxwell and Moore
controller. The system is based entirely on d.c. signals. This refers to the
output of the measuring element as well as to the output of the controller.
The input to the controller is obtained from the primary feedback trans-
mitted from the measuring means and a set point mechanism, the
principle of which is illustrated in Figure 10-28. The primary feedback

Figure 10-27. Schematic drawing of Manning, Maxwell and Moore controller.


202 AUTOMATIC CONTROL

ACTUATING
SIGNAL

Figure 10-28. Principle of set point arrangement.

is applied to a moving coil. Suppose this signal increases, resulting in


repelling the coil. This moves the beam, to which the coil is fastened, in
a clockwise direction, and the vane will enter between the inductance
coils of an oscillator amplifier. The d.c. output of the amphfier, which is
the actuating signal, is thereby increased. The actuating signal is fed back
to a second coil winding, which is wound so that an increase of current
produces increased attraction and tends to move the beam counter-
clockwise. The balance of the beam is determined by
(Actuating signal) + (Set point spring) = Primary feedback.
Hence, by changing the spring tension, the level of the actuating signal
is raised or lowered, and the set point is thereby adjusted.
The resulting output is applied to the controller, illustrated in Figure
10-27. The first amplifier provides rate action, which is obtained by a
lag network in negative feedback. If rate action is not required, this part
of the circuit can be removed. A second amplifier provides the necessary
voltage and power amplification. Proportional band is adjusted by a
variable resistor in the negative feedback. A lead network provides the
desired reset action.
The Swartwout controller requires a different approach, because the
output signal of the measuring means is an a.c. signal derived from a
differential transformer. This signal is matched with a second a.c. signal
CONTROLLERS 203

which is manually adjustable and represents the set point. The resulting
actuating signal is applied to the controller. The controller is illustrated
in Figure 10-29. The first amplifier provides proportional band adjust-
ment in its feedback. From there the signal must be converted into a d.c.

Figure 10-29. Schematic drawing of Swartwout controller.

signal in order to make RC lead and lag networks applicable. This is


done in a phase-sensitive rectifier.
The phase-sensitive rectifier converts the signal which originated from
a differential transformer. In doing so, it must be able to distinguish in
which direction from the center the core of the differential transformer
is displaced. The output in the midposition is zero and increases by equal
increments when displaced from this position. The phase relationship to
the primary is such that, assuming it is zero in one direction, it will be
180 degrees in the other direction. The phase sensitive rectifier responds
to the phase shift by a shift of polarity in the direct current. It is then
possible to raise the reference level by a summation circuit. For example,
adding a constant voltage of 3 volts to a signal that changes between
— 2 and +2 volts, results in a signal changing from 1 to 5 volts.
Reset and rate action are provided by networks which in their prin-
ciple are similar to those described before. The controller output is a
d.c. signal.
Interaction
Practically all controllers—pneumatic, electric or electronic—exhibit
some interaction between the adjustments of proportional band, reset
rate, and rate time.
This means that a difference will generally exist between nominal set-
tings of proportional band, reset rate and rate time and their effective
values, depending on their relative adjustments. The change in effective
proportional band* may be large enough to make the control system
unstable when changing the reset rate or the rate time. The product of
reset rate, in repeats per minute, and rate time, in minutes, should al-
ways be 0.25 or smaller. Otherwise instability may result.

* Young, A. J., “An Introduction to Process Control System Design,” Instruments Publishing
Co., Pittsburgh, Pa., (1955).
n. FINAL CONTROL ELEMENTS

The most common final control element is the sliding stem valve with
pneumatic actuator as illustrated in Figure 11-1. The term actuator is
used here in preference to “motor operator” which is recommended by
the Automatic Control Terminology of the American Society of
Mechanical Engineers.

Figure 11-1. Diagram of sliding stem valve with pneumatic actuator.

The output signal of a pneumatic controller is applied to the top of


the diaphragm. The range of the signal in standard applications is 3 to
15 psi. As the air pressure exceeds 3 psi the valve tends to move down-
ward, compressing the spring. The force required to compress a spring
is, in the ideal case, directly proportional to its deflection. In practice,
however, a valve spring may deviate up to 5 per cent from linearity.
Changes in temperature increase the nonlinear behavior even more.
204
FINAL CONTROL ELEMENTS 205

Another source of nonlinearities is the change of effective area in the


diaphragm described previously.
The relatively minor effect of nonlinearities in the final control ele-
ment has been stressed before. It will be shown in the following pages
that the action of the valve in the control loop includes a number
of other nonlinear responses, all of which have nevertheless little if any
effect.
At this point, it may only be pointed out that the nonlinearity, e.g., of
the controller output signal, can go so far that it nullifies even the char-
acterization of the control valve, and that, furthermore, particularly in
flow and temperature control systems, the primary feedback signal is
frequently far from linear.

Forces in a Spring-opposed Diaphragm Valve


The force exerted by the air pressure on the diaphragm compresses
the spring until the force stored in the spring is equal to the sum of the
forces from the air pressure, the friction and the thrust.
The friction is caused by the bearings for the valve stem, while the
process fluid and its pressure drop across the valve port result in forces
on the valve plug that act largely as vertical thrusts. The spring force
must be large in order to minimize the effect of friction. The area of the
diaphragm must be correspondingly large to be able to compress the
spring.
Suppose the diaphragm area is 140 sq in. The minimum air pressure
of 3 psi is to produce a force of 420 lb and the maximum pressure of 15
psi results in a force of 2100 lb. In the case of a 2-in. stroke, a spring
constant of (2100 — 420)/2 = 840 Ib/in. is required. The spring has to be
precompressed by Vi in. in order to provide the initial force of 420 lb.
If the thrust is 280 lb, the pressure must increase to 5 psi before the
valve begins to move. Similarly, if the valve is at midstroke and the
thrust changes by 100 lb, the valve position will change by 0.12 in., or 6
per cent of its full stroke. This illustrates that a spring-opposed diaphragm
valve should only be used where comparatively small thrusts are involved.
Friction acts in a similar manner. For example, a 20 lb breakloose
force needs an increase of 0.143 lb in controller output signal, i.e., more
than 1 per cent, before the control valve begins to move. This is equiv-
alent to a dead zone of 2 per cent which may quite effectively contribute
to the instability of a control system.
Stroke Time of the Spring-opposed Diaphragm Valve
The time needed to stroke a valve from fully open to fully closed de-
pends largely on the effective area of the diaphragm. This makes the
206 AUTOMATIC CONTROL

smaller valves—which have smaller diaphragms—the faster ones. For a


pressure change from 3 to 15 psig, typical stroke times in round figures
are about as follows:

stroke valves—2 sec/in.


Wi” stroke valves—3 sec/in.
1" stroke valves—5 sec/in.

These values, as any others based on a fuU-range pressure change, furnish


only comparative data about the response in an actual control loop. In
the first place, signal pressures change rarely over the complete range
from 3 to 15 psi. In the second place, the effect of the valve action on
the process decreases the signal from the controller and correspondingly
slows down the speed. The speed of a given valve depends essentially on
the port area of the pilot valve in the controller—disregarding time lags
in the transmitting circuit.

Valve Positioners
The valve positioner is a device which compares the actual position of
the valve with the controller output signal. As such it eliminates most of
the shortcomings that are due to the above described method of using a
spring force as an expression of the valve position. The valve positioner
is generally an auxiliary device attached to the conventional spring-op-
posed diaphragm valves. In itself it may represent either a force-
balanced or position-balanced mechanism.
Figure 11-2 is a diagram of a position-balanced valve positioner. The
controller output signal is applied to a bellows which is spring opposed
and assumes a position proportional to the signal. It moves a lever using
point yl as a fulcrum, which operates a pilot valve that admits secondary
air to the valve diaphragm. The valve position is fed back by a linkage
to the lever, turning it about point B. This repositions the pilot to a point
where motion of the valve ceases. The valve has assumed a new position
in proportion to the controller output signal applied to the bellows.
Lever C can be adjusted to change the proportion which A moves with
respect to valve motion. This alters the feedback and determines the
valve stroke per unit change of controller output signal.
Figure 11-3 illustrates a force-balanced valve positioner. The lever is
displaced against the force of the feedback spring. For example, the con-
troller output signal increases, moving the bellows against the force of
the feedback spring (neglecting the spring rate of the bellows). The
pilot opens, admitting secondary air to the top of the diaphragm, mov-
ing the valve downward. This stretches the feedback spring, increasing
its force and positioning the pilot and returning it to its initial condition.
FINAL CONTROL ELEMENTS 207

CONTROLLER
I OUTPUT
I SIGNAL

Figure 1 1-3. Force-balanced valve positioner.


208 AUTOMATIC CONTROL

at which point the valve settles in its new position. The stroke adjust-
ment follows the principle of the position-balanced valve positioner. It
is obtained by ehanging the feedback ratio by means of the lever ratio.
The valve positioner reduces the air volume involved in the controller
output signal system, since the air space above the diaphragm is filled
with secondary air. Within the limitations described below, this results
in faster response of the valve.
A valve positioner will, in general, not increase the response speed of
a valve if a change of the controller output signal from 3 to 15 psi is con-
sidered, unless the secondary air supply is more than 15 psi. This is be-
cause the pilot port area in the controller as well as in the positioner are
of the same order of magnitude. However, for minor changes of con-
troller output signal, the amplification of the positioner takes effect and
produces faster response than can otherwise be obtained.
The positioner amplifies the controller output signal by a certain gain
factor. This gain, however, does not affect the control loop as such.
The control valve with the valve positioner represents a minor loop
within the major control loop. This minor loop is composed of the con-
troller output signal, the pilot valve, the valve stem motion, and the
feedback to the controller output signal.
The gain of the positioner is limited by the pressure of the secondary
air supply. A typical valve positioner may require a change of 0.07
psi in the controller output signal to move the pilot valve from fully
closed to fully open. If the control valve is closed at 3 psi and the
secondary air supply is 15 psi, then an inerease of the controller output
signal from 3 to 3.07 psi or more will produce an initial pressure drop of
12 psi across the pilot. This results in correspondingly high rate of flow
to the diaphragm top, which decreases gradually as the new position is
reached. On the other hand, suppose the valve carries at a given position
14 psi on the diaphragm. A change in controller output signal of 0.07
psi in controller output pressure wiU now result in an initial pressure
drop of only 1 psi across the pilot and the valve motion will be corre-
spondingly slower.
The gain and the resulting time constant of a control valve with valve
positioner depends therefore on the position of the valve. This is particu-
larly noticeable when the secondary supply pressure is not higher than
15 psi. This behavior of the valve positioner can explain instability of a
control system at certain controller output pressures.
The main advantage of the valve positioner is that in general it has a
comparatively high gain, and much larger forces can therefore be pro-
duced to overcome sticking and friction of the control valve than is pos-
sible without positioner.
FINAL CONTROL ELEMENTS 209

Due to the high gain of the valve positioner, it reduces considerably


the tendency of the valve to change its position with variations in thrust.
An increase in thrust, for example, will generally not be able to displace
the position by more than one per cent without increasing the diaphragm
pressure to the maximum.
Fast changes in thrust may, however, change the valve position be-
cause of the time lag in the circuit. In general, these position changes are
not harmful, or at least less harmful than the pressure changes in the
process fluid which produce the variations in thrust. Accurate control
can hardly be obtained where sudden pressure changes in the process
fluid occur, and the changing valve position may become the visible ex-
pression of a more basic trouble in the control system. If the thrust
changes are slow they may cause some response in the valve position
but without affecting control in a closed system.

Valve Characteristics
Nonlinearities may exceed acceptable limits. This is particularly the
case in the relation between valve port area and flow through the valve
as well as in the process response itself. To reduce these nonlinearities,
valve plugs of different shapes, as shown in Figure 11-4, are available.
These plugs have nonlinear characteristics in their relation between valve
lift and valve port area. This relation is called the valve characteristic.

■—^ —
AA

LOJU

EQUAL PERCENTAGE

Figure 11-4. Typical valve plugs.


210 AUTOMATIC CONTROL

Figure 11-5 illustrates the open loop of a temperature control system.


The input to the controller is the measurement signal without conver-
sion, i.e., the input changes 1°F per 1°F change in the controlled
variable.

Figure 11-5. Temperature control system.

The controller output signal changes 2 psi/°F change in input. This


signal is converted by the valve actuator into a valve stroke with a gain
of 0.1 in./psi. The valve port area changes 0.5 sq in. for each inch
in valve travel, assuming a linear valve characteristic. The flow through
the pipe line varies at the rate of 200 gpm/sq in. port area, and the tem-
perature responds to flow changes with 0.5°F/gpm. The total loop gain
is expressed by °F/°F X 2psi/°F X 0.1 in./psi X 0.5 in.Vin. X 200
gpm/in.^ X 0.5°F/gpm = 10
This gain should be maintained—ideally—for any set point, for any
pressure drop, and for any load conditions. The fact is that the gain in
the piping, for example, is not constant. Where this occurs, compensa-
tion may be possible by characterization of the valve plug. For example,
if the gain in the pipe line changes to 100 gpm/in.^ while simultaneously
the gain in the plug increases to 1 in.Vin., the overall gain would remain
the same and the dynamic performance of the system would not be af-
fected. The ideal valve characteristic maintains a constant gain in the
control loop.
Figure 11-6 shows valve characteristics corresponding to the plugs il-
lustrated in Figure 11-4. Valve characteristics are not tailor-made. Im-
provements can be expected, although perfectly constant gain through-
out the system and operation is an unattainable ideal.
The flow rate through a valve is expressed by the equation
Q = k,A (11-1)
FINAL CONTROL ELEMENTS 211

Figure 1 1-6. Valve characteristics.

where Q is the flow rate, a constant, A the port area, and p is


the pressure drop across the valve port. As long as p remains constant,
the relation between flow rate and port area is linear. There are, how-
ever, two additional pressure drops changing with the flow rate and hence
affectingp. These are the pressure drop in the valve body and in the pipe-
line. Generally, the drop in the valve body can be neglected, which will be
done in this discussion.
Considering a total available system pressure P and a pressure loss in
the pipeline of p^ the pressure drop across the valve is then given by
p = p-p,
and hence equation (11-1) may be written
Q = k^A yJP -p. (11-2)
212 AUTOMATIC CONTROL

Obviously, the smaller can be made in proportion to P, the more


linear will be the operation.
Pipeline loss is generally calculated by means of Fanning’s equation
which, in one practical version, reads

p, = 0.0066
where
is the pressure loss in psi,
is the specific gravity,
L is the pipe length in feet,
D is the pipe diameter in inch,
/ is the friction factor, and
B is the flow rate in barrels per hour.
The friction factor is obtained from a suitable table, such as given in the
Mechanical Engineers’ Handbook by L. S. Marks.
The rate of flow for various port areas can be calculated as in the fol-
lowing example. The fluid is crude oil of 0.9 specific gravity which
is pumped through 300 ft of 2-in. wrought iron pipe. Normal maximum
pumping rate is 400 barrels/hr. Total pressure available is 250 psi.
The first step consists in determining the pipeline pressure drop at
various pumping rates. This is done by means of Fanning’s formula and
the results are shown in the following table:

TABLE 1.

Bbl/hr

40 1.6
120 12.2
200 32.5
300 68.0
400 118.0

Using conventional methods, a valve size is now chosen, which in this


case shall be 1.5 in. This corresponds to an approximate port area
of 1.75 sq in. It may be assumed that the normal maximum flow of 400
barrels per hour corresponds to a port area of 80 per cent of maximum.
Inserting these values in equation (11-2) and solving for gives
400
= 25
(1.75 X 0.8) V250 - 118
This value of k^ can now be used to solve by means of equation (11-2)
FINAL CONTROL ELEMENTS 213

for the valve port area at the different values of flow and pipe line pres-
sure drop as given in Table 1. The results are listed in Table 2.
TABLE 2.

%of % of
Bbl/hr A
max. area max. flow

40 1.6 0.1 6 9
120 12.2 0.3 17 27
200 32.5 0.5 29 45
300 68.0 0.9 52 67
400 118.0 1.4 80 90
445 145.0 1.75 100 100

The last row of values has been obtained by inserting a few trial-and-
error values in the calculations for Table 1 to obtain the flow which cor-
responds to the maximum port area of the valve.
The results of Table 2 are graphically represented in Figure 11-7. The
gain of change in flow rate vs change in port area at any point can be
easily determined as shown in the illustration. At about 15 per cent of
maximum port area the gain is 1.64. At about 68 per cent the gain has
decreased to 0.7. In Figure 11-6, the gain for these same port openings
has been shown on the linear-contoured and the equal-percentage
characteristics. Thus two graphs are available: one for the pipeline char-
acteristics (Figure 11-7), the other for the valve characteristic (Figure
11-6). It remains to correlate these data.
The gain of flow rate over valve lift is the product of
Port Area Flow _ b a
Lift Area c b
The symbols a, b and c refer to Figures 11-6 and 11-7 and may be re-
placed by a', b', c', b'^, c\, or Z?2, c^, whichever of the small triangles is
referred to in the graphs. Table 3 shows the various gains:
TABLE 3.

Port
A B c Combined Gain
Linear Equal
Opening
V-Port Percentage Pipeline A X C B X c

15% .92 .7 1.64 1.51 1.14


68% 1.2 2.45 0.7 0.84 1.71

In this example, it appears that the equal-percentage characteristic


would give somewhat better results since it changes its gain by 1.71:1.14 =
214 AUTOMATIC CONTROL

PER CENT OF MAXIMUM AREA

Figure 11-7. Flow-area characteristic of pipeline.

1.5, while the linear V-port characteristie ehanges the combined gain by
1.51:0.84 = 1.8. However, in the first place, the difference is smaller than
the uncertainties which are inherent in the calculations; in the second
place, comparison would be necessary between more than two points.
In fact, these points must be carefully chosen to be in line with the most
probable operating ranges. Evaluation of gain changes in the process will
further influence the choice of the valve characteristic. This systematic
approach, however, is only a further application of the method outhned
here. Important are the relatively high nonlinearities which have to be
accepted.

Rangeability
Characterized valves are not able to operate over the full range. As
the valve approaches its closed position, the valve characteristic cannot
even approximately be maintained, and a valve operating in this range
FINAL CONTROL ELEMENTS 215

would give erratic results in a control system. This limitation is expressed


by the rangeability factor. If the valve can be used between 10 and 100
per cent of its stem travel, the rangeability is = 10. If it is limited
to a range between 20 and 100, it is = 5. A good practical figure
is an average rangeability of 8, approaching 15 on larger valves, but de-
creasing to as much as 5 on small valves.
In determining the rangeability required for an actual case, it is neces-
sary to take into account pressure drop across the valve as well as flow.
Calling the maximum flow Q^, and the corresponding valve pressure drop
/?i, and calling the minimum flow and the corresponding valve pres-
sure drop P2, the required rangeability R is given by

For example, if the maximum flow were 360 gpm at 12 psi and the
minimum flow 60 gpm at 27 psi, the required rangeability would be

Process Gain
It is generally assumed that the process gain is constant under alter-
ing load conditions. Closer analysis of a process may show that this is
not so and a valve characteristic may be chosen for partial compensa-
tion of these gain variations.
Changes of set point will affect the gain of almost any process. A
tuning of the controller for one set point may cause instability or exces-
sively slow corrective action at another set point. It becomes difficult, if
not impossible, to consider all the various effects and combinations in
selecting a valve characteristic. However, it is never lost time to consider
certain possible combinations of (a) valve characteristic, (b) flow through
the valve, (c) load, and (d) set point, and to determine the gain under
these conditions. Conditions may be determined beforehand which have
the highest gain and are most likely to produce instability under actual
operation.

Conclusions on Valve Action in Control Loop


The question may well be raised how it is possible that satisfactory
control can be obtained at all under conditions where disturbing non-
linearities may be reduced but not eliminated. To answer this, static and
dynamic conditions must be considered separately. The nonlinearities
may produce a static inaccuracy, and hence a deviation of the controlled
216 AUTOMATIC CONTROL

variable from the set point. In floating controllers and those that include
reset action, these deviations are automatically corrected by the con-
troller action. In proportional controllers, the offset resulting from load
changes is generally greater than similar consequences of the nonlinear-
ities. Therefore, in either case, considerable nonlinearity can be tolerated
without having effect on the control system.
Dynamically, the nonlinearities require a tuning of the controller that
assures stability under the conditions of highest gain. In practical opera-
tion, controllers are usually adjusted with enough stability margin to
assure stable operation even when conditions are occasionally much
closer to instability than originally expected. However, it is here where
nonlinearities may occasionally cause trouble, since the controller is
adjusted for specific load conditions or a certain set point, and changes
in either may produce instability and require changes in the controller
adjustments.

Other Control Valves and Actuators


So far only the spring-opposed diaphragm valve has been considered.
Most of the features are similar to other types like butterfly or Saunders
Patent valves, and to other actuators or combinations like the electro-
pneumatic relay or the electrohydraulic valve actuator. From the view-
point of control systems, some particulars of the butterfly valves and
the final control elements in electronic control systems are of special
interest.
Butterfly Control Valves. Butterfly valves are particularly suited for
controlHng large flows, especially at low pressures. They are also used to
advantage in lines carrying considerable amounts of suspended matter,
which would cause excess clogging of plug-and-seal types of valves. They
usually operate through about 60 degrees of their rotary movement.
These valves produce practically no pressure drop when they are fully
open. A butterfly valve is shown in Figure 11-8.
The unbalanced forces across the discs of butterfly valves, which the
actuator has to overcome, are comparatively high. The tendency of
these forces is to close the valve. Their magnitude is proportional to the
pressure drop across the valve and the cube of the disc diameter, and
depends on the disc position.*
The unbalanced forces push the disc against its bearings and friction
results. This friction must be overcome by the actuator. In closing the

* Dally, Charles A., “Butterfly Control Valves,” Instruments (Dec. 1952).


FINAL CONTROL ELEMENTS 217

valve, the actuator is assisted by the unbalanced forces, but in opening


it is opposed by the friction plus the unbalanced forces themselves. The
torque required to position a large butterfly valve, e.g., 24 in., may
amount to 6000 foot-pounds and more under normal operating
conditions.

Figure 11-8. Butterfly valve.

Such large and changing torques will have an influence on the control
system to be chosen. Pneumatic systems are frequently no longer ade-
quate and hydraulic systems will be preferred.

Electropneumatic Relays. The electropneumatic relay converts the


output signal of an electronic controller into a pneumatic signal which in
turn is applied to the conventional spring-opposed diaphragm valve
with or without positioner. Figure 11-9 illustrates the principle as used in
the Swartwout Power Relay. The controller output signal, i.e., the input
218 AUTOMATIC CONTROL

signal into the relay, is applied to a moving coil which is mounted on a


lever. On the right the lever is fastened to a feedback diaphragm. When
the electrical input signal increases, the resulting increase of force in the
moving coil tends to move the lever away from the magnet. This depresses

ELECTRICAL MOVING VENT TO


INPUT COIL ATMOSPHERE

^'FEEDBACK DIAPHRAGM

SUPPLY AIR

LOADING PRESSURE

VENT TO
ATMOSPHERE

BOOSTER RELAY DIAPHRAGM

Figure 1 1-9. Electropneumatic relay—Swartwout principle.

the feedback diaphragm. Its downward motion displaces the pilot valve
stem, opening the lower port of this valve. The result is that the air pres-
sure under the feedback diaphragm as well as under the booster relay
diaphragm increases. The pilot valve remains open to the air supply
until the pressure under the feedback diaphragm has increased sufficiently
to balance the force due to the electrical input signal.
The increased air pressure under the booster relay diaphragm moves
the booster valve stem upward and opens the upper port of this valve.
Supply air then flows through the port, increasing the loading pressure
and the pressure on the upper side of the booster relay diaphragm. The
upper port of the booster valve closes as soon as the pressures on both
sides of the diaphragm are equal. This reestablishes balanced conditions,
but at a loading pressure increased in proportion with the increased
electrical signal.
The action is similar on decreasing electrical signals. In either case the
loading pressure follows the changes in input signal in linear relation.

Electrohydraulic Actuators. A diagram of an electrohydraulic valve


actuator is shown in Figure 11-10. The moving coil is fastened to a jet
pipe. When the signal current increases, the coil is pulled into the magnetic
field and the jet pipe rotates in counterclockwise motion. The result is that
the piston moves downward together with the plug of the control valve
to which it is linked. The feedback lever is connected to the piston stem
FINAL CONTROL ELEMENTS 219

MAGNET ASSEMBLY

Figure 11-10. Schematic drawing of Askania electrohydraulic valve actuator.

by means of a roller and a spring and follows its motion. This results in
a stretching of the feedback spring as the piston moves toward a new
position. As the force of the feedback increases and approaches the
opposing force of the moving coil, the jet pipe returns to its neutral posi-
tion, and once this position is attained, the piston stops. The position
of the piston is proportional to the magnitude of the signal current.
12. CONTROL SYSTEMS

It is one thing to consider a controller, e.g., a temperature controller,


as an isolated phenomenon that is connected into a closed loop with
some imaginary process unit, and another to be aware of the numerous
details that influence the action—and hence the selection—of a control
system within the totality of an industrial production arrangement. The
difference is one between specifying control components and engineer-
ing a control system.
In illustrating control systems it is convenient to use a system of sym-
bols and identifications which allows representation in the form of a
flow plan. Such a system has been developed by the Instrument Society
of America and is contained in their “Recommended Practices.”

InstfutDSOt pfoccss Jsipmg All lines


and- hydraulic lines
to be
Instrument air lines TT
fine in
relation
Instrument electricol leads to process
piping
Instrument capillary tubing —X———X——X——X——Xr

Locally Board
Mounted Mounted
Locally Board Locally Board
Mounted Mounted Mounted Mounted

0
Basic symbols for Basic symbols for Basic symbols for
instrument with combination instrument transmitter
single service and or device with two
function services or functions

Basic symbol Basic Symbol for Basic symbol for 3-way Basic symbol
for electrically actuated piston actuatiecl body for for safety
spring-opposed valve (solenoid or valve (hydraulic any valve (relief) valve
diaphragm valve motor) or pneumatic)

220
CONTROL SYSTEMS 221

Basic symbol for Basic symbol Sasic symbol showing Basic symbol
sclfaduafed for manually opera- pneumatic transmission showing pneumatic
(infcjraljregulafin^ ted control taitre instrument (electric connection from
yoive transmission same instrum ent to spring -
except for type of opposed diaphragm
connection) voive

Figure 12-1. Basic instrumentation symbols.

Basic instrumentation symbols are shown in Figure 12-1. In order to


identify control components and instruments in a flow plan the follow-
ing letters and combinations are used.

LETTERS OF IDENTIFICATION
Definition and Permissible Positions in Any Combination

FIRST LETTER SECOND LETTER THIRD LETTER


LETTER Process Variable Type Reading or Additional
or Actuation Other Function Function

A Alarm Alarm
C Conductivity Control Control
D Density — —
E — Element —
(Primary)
F Flow — —
G — Glass —
(No Measurement)
H Hand — —
(Actuated)
I — Indicating —
L Level — —
M Moisture — —
P Pressure — —
R — Recording —
(Recorder)
S Speed Safety —
T Temperature — —
V Viscosity — Valve
w Weight Well —
222 AUTOMATIC CONTROL

The following may be used optionally as a first letter for other process
variables:
1. “A” may be used to cover all types of analyzing instruments.
2. Readily recognized, self-defining chemical symbols such as CO , 2

O , etc., may be used for these specific analysis instruments.


2

3. The self-defining symbol “pH” may be used for hydrogen ion con-
centration.
Furthermore, it is permissible to insert a lower case “r” after “F” to
designate Flow ratio. Likewise, lower case “d” may be inserted after
“T” or “P” to designate Temperature difference or Pressure difference.
If the flow contains several units of equal letter identification, it be-
comes necessary to supplement the general identification by a numerical
system, to establish its specific identity. Thus, the identification TRC-1,
refers to a temperature recording controller and to item No. 1.
Figure 12-2 applies this system for the typical instrumentation of a
fractionating column. The feed to the fractionating column is maintained
constant at a certain rate which is established by the set point of the
flow recording controller (FRC-2). This set point is automatically ad-
justed by the level recording controller (LRC-1). The level control
maintains not a fixed but an average level. This arrangement combines
average and cascade control as will be described further below. Its pur-
pose is to reduce fluctuations in the feed flow rate as much as possible.

Figure 12-2. Typical instrumentation of fractionating column.


CONTROL SYSTEMS 223

The situation is somewhat similar in the control of the flow rate of


the heating medium. The flow rate has to be kept constant even though
upstream or downstream pressures may change. On the other hand,
temperature in the fractionating column must be kept constant. Hence
the purpose of the cascade control in this case is to change the set point
of the flow recording controller (FRC-3) by means of the temperature
recording controller (TRC-4). The remaining control loops in this case
are single-loop controls and need no further description as such.

Averaging Control
In the feed tank of Figure 12-2, it is quite permissible for the level to
fluctuate between certain hmits, but it is of greatest importance that the
flow to the fractionating column does not change abruptly. A control
system of this kind, purposely permitting variations of the controlled
variable which are larger than required by the system dynamics, is
called an averaging control system.
Suppose the maximum permissible rate of change in flow is 3.75
gpm/min. For a control valve with linear characteristic and a flow
capacity of 37.5 gpm at the prevailing pressure, the minimum time for
full stroke should be 10 min. This can be effected with hydraulic con-
trollers by selecting the corresponding flow rate of the actuating medium
and the piston area. In pneumatic controllers, conditions are such that
the timing of the final control element is not practical. The approach
to the problem is then different and will be discussed further below.
Before the speed of the final control element is adjusted, it is neces-
sary to ascertain that the rate at which the flow into the tank changes
is always considerably below 3.75 gpm/min., or that it exceeds this rate
only intermittently. The time during which the flow into the tank can
exceed the maximum rate of change depends on the capacitance of the
tank, the rate of flow, and the tolerance between maximum and minimum
level. With feed tanks, the tolerance is generally so high that consider-
able amounts can be absorbed. Furthermore, limit controls may be pro-
vided to take safety measures in case these limits are exceeded.
Boiler drums are more critical in this respect. In their case the flow
into the drum can be changed only gradually to avoid shocks. The level
is also allowed to fluctuate but ±2 in. is usually considered the limit.
Furthermore, the flow rates may be considerably larger than those for
the feed drum. The load may change rapidly, but careful study should
be given to the possibility that the load might change faster than the
feedwater flow, draining the boiler drum beyond the minimum level
before the control valve in the feed line can follow.
It was stated that in pneumatic control a simple speed adjustment of
224 AUTOMATIC CONTROL

the final control element cannot be used. Hence, while for the hydraulic
controller a proportional-speed floating controller could be used, pro-
portional-position or a proportional plus reset action would be the choice
if the final control element is pneumatically actuated. The point is then
to select the widest possible proportional band. A method for determin-
ing the proportional band is described in the following paragraphs.
For a tank of 8 ft in diameter, i.e., an area of about 7200 sq in., and
a permissible level variation of ±20 in., the volume contained between
minimum and maximum level is approximately 1250 gallons. Figure
12-3 illustrates this case. For a maximum flow of 250 gpm into the tank,
FLOW IN

Figure 12-3. Averaging control system.

it is necessary that the valve reaches a position at maximum level at


which 250 gpm are allowed to flow out of the tank.
A minimum flow of 50 gpm into the tank may exist, and hence at
minimum level the valve should still be sufficiently open to allow an
outflow of 50 gpm. If the valve is rated at maximum capacity of 300
gpm, this means that the controller output signal should be 5 psi at
minimum level and 13 psi at maximum level. Since these levels are 40
in. apart, the proportional band of the controller is a 60 in. change of
the controlled level, i.e., the flow rate changes by 5 gpm for every inch
of change in tank level.
The proportional band is thus given by

P = (12-1)
where ^
Pis the proportional band expressed in inches of controlled tank level,
V is the maximum flow capacity of the valve, in gpm,
L is the tolerance between maximum and minimum level, in inches,
and
Q is the maximum sudden change of flow into the tank, in gpm.
CONTROL SYSTEMS 225

The valve capacity depends on the level in the tank. The maximum
level is the most critical. It is hardly conceivable that the maximum
level coincides with the maximum flow change, but this condition may
be assumed in order to be on the safe side. It is also assumed that the
valve has linear characteristics. Such a valve could be employed in this
system. If valve characteristics are changed, the fastest change in rate
for a given displacement should be established.
The greatest and fastest change of flow into the tank would be a
sudden increase of 200 gpm, from 50 to 250 gpm. Since the volume
between maximum and minimum level is 1250 gallons, the level rises
200 gpm X 40 in.
6.4 in./min.
1250 gal.

Consequently, the valve would initially open at the corresponding rate


of 5 gpm/in. rise in level which is 32 gpm/min.
This procedure can be expressed in the equation.

q/mm. = (12-2)
eP
where
q/min. is the rate of change of flow obtained by valve action, in
gpm/min.
Q is the maximum sudden change of flow into the tank, in gpm
e is the tank capacity between maximum and minimum level, in gal.
The other factors correspond with those of equation (12-1). Obviously,
in either case any other consistent units may also be used.
This equation can be used in various ways. For example, the fastest
permissible rate of change of flow is given as 16 gpm/min., while all
other conditions are the same as before. In this case, q/min. = 16 and
insertion of values into equation (12-2) allows solving for the propor-
tional band P. The result is a proportional band of 120 in.
Since the equation contains the tank capacity between maximum and
minimum level, it permits selection of the correct tank for a given set
of conditions.
It has been assumed in the preceding that Q, the maximum sudden
change of flow into the tank, is equivalent to a change from minimum
to maximum flow into the tank. Actually such a sudden change may
never happen, or if it does happen, it is permissible that limit switches
open dump valves or take similar precautionary measures. Under such
conditions, the value of Q may be chosen correspondingly smaller.
Certain modifications are possible in the assumptions contained in
the foregoing. For example, the flow into the tank may reach its maxi-
226 AUTOMATIC CONTROL

mum only intermittently. In this case the Q in equation (12-1) is sub-


stituted by the factor M, which expresses the difference between the
maximum and minimum flows that can be expected. Hence

P = ^L (12-3)
M
The proportional band P, becomes now correspondingly larger and its
insertion in equation (12-2) results in either a smaller tank or a slower
maximum rate of change of flow from the tank. It should be noted that
the factor Q in equation (12-2) remains unaltered, since it is assumed
that fluctuations through the maximum range may occur intermittently.
If this is not the case, further corrections are possible.
When using equation (12-1), a proportional-position controller is to
be used. The case of equation (12-3), however, requires proportional
plus reset control. The reason is that offset decreases the absorption po-
tential of the tank. Using a proportional-position controller without
reset means considerable offset in view of the large proportional band
used in averaging control. The control point may therefore drift con-
siderably from its set point and be near one of the level limits when a
sudden load change occurs. As long as the proportional band is adjusted
according to equation (12-1), the level limits can hardly be exceeded.
No advantage from reset action can be expected.
If the sensitivity of the controller is further decreased, in accordance
with equation (12-3), then reset action is required to assure that the
control point is always driven back to the set point, and that sufficient
absorption capacity in the tank is available for sudden load changes.
It is advisable that limit switches are provided in such arrangements to
provide dumping or similar precautionary measures in case load changes
are occasionally larger or more persistent than expected.
Cascade Control
Cascade control has two purposes. One is to control for two or more
different kinds of load changes. The other is to improve the dynamics
of a system by effectively reducing the dead time of the system.
Figure 12-4 shows the level control system of a feed tank, similar to
that in Figure 12-2. Again the main purpose is to avoid too rapid
changes in the feed flow. Averaging liquid level control is used. However,
this protects only against sudden changes of flow into the feed tank. In
this case another variable becomes important,'namely, upstream and
downstream pressure changes. The flow must be kept constant, no
matter what the pressure conditions are. It still must change with level.
In other words, two kinds of load changes must be controlled: pressure
CONTROL SYSTEMS 227

and tank level. The solution is illustrated in both the flow diagram and
the block diagram of Figure 12-4. A second control system is connected
into the level control loop. The level controller “cascades” into the flow
controller. The method is basically simple. All that is required is a flow

a — FLOW DIAGRAM

b - BLOCK DIAGRAM

Figure 12-4. Cascade control system.

control loop with a reference input obtained from the level controller.
Thus the flow is kept automatically constant at a given set point. Changes
of upstream or downstream pressure cannot affect it. As the level
changes, however, the reference input signal changes also. This readjusts
the set point and with it the flow rate controlled by the flow controller.
Ii is interesting to observe from the block diagram that this is equiva-
lent to breaking the process into two parts—the “flow” process and the
“level” process—and tapping a point between these two parts to obtain
an additional feedback loop within the control loop.
Figure 12-5 represents a polymerizer with a cascaded temperature
control system. This particular system was designed by the Bristol Co.
The polymerizer kettle is heated with hot water which passes through
a jacket around the kettle. The temperature inside the kettle must be
kept constant. This temperature depends not only on the load condi-
tions of the kettle but also on changes in steam pressure and quality,
as well as on changes in cold water pressures and temperature.
The arrangement chosen is to use one temperature recording con-
troller, TRC-1, to measure the temperature of the product inside the
228 AUTOMATIC CONTROL

Figure 12-5. Polymerizer with cascade control system.

polymerizer kettle. A second temperature recording controller, TRC-2,


measures the temperature in the hot water jacket. This controller is to
keep the jacket temperature constant. It accomplishes this by position-
ing three control valves. As long as the valves in the drain and cold water
hnes are closed, the hot water circulates through the bypass. These valves
are always in identical position and as they open, more and more cold
water is admitted, reducing the temperature in the jacket.
The drain and cold water valves are fully open at a signal pressure of
3 psi from controller TRC-2. As the signal increases they close gradually.
At 9 psi they are closed completely and remain closed with further in-
crease of pressure. On the other hand, the steam valve remains closed
up to 9 psi and begins to open as the pressure rises further, being fully
open at 15 psi. Thus at 9 psi all three valves are closed and water cir-
culates but no heat or cold water is added. If the temperature drops
below this point, the steam valve begins to open. If the temperature
rises, cold water is added.
The TRC-1 controller in this case is a proportional plus rate controller.
No reset action is added to avoid overshooting on start up since this is a
frequently started batch process. The TRC-2 controller has proportional-
position action only.
This process consists of two sharply separated capacitances. One is
the capacitance of the jacket, the other is the inside of the kettle. Both
capacitances are separated by the inner kettle wall. It has been pointed
out previously that the response curve of a multi-capacitance process
can be split into a time constant and an equivalent dead time. The dis-
advantages of dead time, actual or equivalent, have been discussed at
length. In the polymerizer control under consideration, the cascade
system has actually eliminated this handicap by splitting the capacitances
and providing an intermediate feedback signal. Figure 12-6 is the cor-
CONTROL SYSTEMS 229

responding block diagram. In its arrangement, this diagram does not


differ from Figure 12-46. The feedback from between the two capaci-
tances, however, illustrates an additional characteristic of feedback,
which is an improvement of the dynamic characteristics. Without it, the

Figure 1 2-6. Schematic drawing of polymerizer with cascade control system.

detection of temperature changes would be delayed since the heat flow


comprises two capacitances with a resistance connecting them. The
introduction of a second temperature controller in the jacket signals
changes in the process temperature before they actually take place.
The cascade controller requires two or more controllers, each of which
contains a complete measuring system and a complete controlling unit.
One of these controllers is the master controller. It positions the set
point in one or more other controllers, such as secondary, pneumatic
set, or autoset controllers. In the block diagram of Figure 12-6 control-
ler TRC-1 is the master controller. The other controller, TRC-2, is the
secondary controller.
While the master controller is a conventional one-, two-, or three-term
controller, the secondary controller requires modifications. The signal
from the master to the secondary controller covers a certain range. In
simple cascade controllers, this signal range is equivalent to a change of
the set point index in the secondary controller moving from one extreme
to the other. This results in certain limitations in adjusting the control
actions of both controllers. Manual adjustments are therefore frequently
provided to change range and zero of motion of the set point index in
the secondary controller.
The secondary controller is generally designed to also permit reversi-
bility of action. There are two interrelated controlled variables in a cas-
cade control system: the master or primary variable and the secondary
variable. If an increase of the magnitude of the master variable raises
the set point of the secondary variable, the action is called direct. If it
lowers the set point, it is reverse action.
Figure 12-7 illustrates the operation of the Taylor pneumatic set
controller. The set point is adjusted pneumatically by the input air sig-
230 AUTOMATIC CONTROL

Figure 12-7. Schematic drawing of pneumatic set controller.


(Courtesy of Taylor Instrument Cos.)

nal from the master instrument. The pneumatic adjustment mechanism


is indicated by solid lines to differentiate it from the controller mecha-
nism. With an increase in input air signal entering the pneumatic ad-
justment unit through line (1) at (6), the set point will be raised. This
means that this is direct action. If reverse action is required, line (1)
attaches at (5).
With the direct action as shown, an increase in signal moves pin (11)
toward the right. Parallelogram (12) transmits this movement to arm
(15) which is pivoted at (13), moving it upward. The resulting upward
motion of pivot (17) is transmitted through link (18) to bell crank (2)
which pivots at (25). Link (23) which attaches to the floating gear (19),
to which the set point index is attached, is drawn downward. This moves
the set point index to the left.
CONTROL SYSTEMS 231

The amount of travel of the set point index per psi change in input
air signal, is regulated by the span adjustment knob (4). Travel increases
as the knob is turned to higher dial readings. The dial is calibrated in
inches of index travel per psi change in input air signal. It covers a range
of 0 to 0.4 in. per psi.
The span adjustment for a required set point index travel per unit
change in input air signal can be determined by the use of the equation
A/B = C, where A is the linear distance of index travel in in., measured
on a radius from the center of the chart, B is the total change in input
air signal, and C is the setting of the span adjustment knob in in./psi.
For example, if the set point index is to move 3 in. for a variation in
input signal from 3 to 15 psi, then the dial is set for
3
— = 0.25 in. per psi

A span adjustment of 0 in. per psi reduces the pneumatic set con-
troller to a conventional pneumatic controller. It no longer responds to
changes in input air signal and controls at the set point which is deter-
mined by the adjustment of knob (21).
Various methods are used for determining the upper and lower limit
through which the set point index of the secondary controller moves.
One method is to make the span adjustment zero. The set point index
position under this condition defines the index position at any other
span adjustment when the input air signal is 9 psi, i.e., at the midpoint
between 3 and 15 psi. Another method is a zero adjustment which sets
the lower point of the range through which the set point index moves.
In either case the range of set point index motion is shifted either upscale
or downscale.
Adjustable control-point limit stops are also frequently built into the
secondary controller. If the secondary variable has to be maintained
within certain limits, the stops can be so positioned that the control point
does not exceed these limits.

Ratio Control
The purpose of ratio control is to maintain the relative magnitude be-
tween two variables without being concerned about their absolute mag-
nitude. The simplest form is the open-loop arrangement in Figure 12-8.
Flow through line A is measured, and supplies the actuating signal for
the flow controller. A valve in line B is positioned by the flow controller.
This arrangement contains all the uncertainties of an open loop. The
pressures in line B are assumed to be constant and the control valve to
be linear. This is obviously an idealized condition which cannot be
232 AUTOMATIC CONTROL

obtained. However, in some cases it may be sufficiently approached to be


acceptable when only an approximate ratio is required. Under such condi-
tions the example in Figure 12-8 provides a simple and inexpensive
arrangement.
LINE A

FLOW
LINE A CONTROLLER LINE B

Figure 12-8. Open-loop ratio control system.

Considerable improvement is obtained by the closed-loop method il-


lustrated in Figure 12-9. Flow in line A, which is called the primary
flow, is now compared with flow in line B, called the secondary flow.
The controller maintains the secondary flow at a rate which can only be
varied when the primary flow changes.

LINE A

LINE B

Figure 12-9. Closed loop ratio control system.

The measurement signal from the primary flow r goes into a ratio
mechanism. This multiplies the input r by a factor k which is given by
the manually adjustable ratio setter. The summation point receives thus
a reference input signal equal to kr and combines it with the measure-
ment signal b from the secondary flow. The difference between kr and b
is the actuating signal for the controller. The output signal from the
controller regulates the secondary flow to make the difference between
kr and b equal to zero. Hence
kr — b = 0 or r/b = k (12-4)
which means that the ratio between r and b, i.e., between primary and
secondary flow, is equal to k. In other words, the control system main-
tains a ratio which is given by the adjustment of the ratio setter.
CONTROL SYSTEMS 233

Figure 12-10 shows a hydraulic ratio controller. Dilferential pressure is


measured across an orifice plate through which the primary flow passes.
This differential pressure is transmitted to diaphragm (1). The force on
the diaphragm is applied to counterlever (2) and over the ratio slider (3)
to the jet pipe.

TO FINAL CONTROL

Figure 12-10. Hydraulic ratio control system.

The secondary flow is measured by similar means. The resulting force


from diaphragm (4) is directly transmitted to the jet pipe. The position
of the jet pipe is a function of the two forces which are due to the pri-
mary and secondary flow respectively. These forces balance around the
pivot point of the ratio slider, and the equation
d rp
d-y~ ^d +
is obtained from the lever dimensions d andy as shown in Figure 12-10,
using F’l for the force from the primary flow and F for the force from
2

the secondary flow. Since this shows that


F,^d-y
F 2 d+y
it illustrates that the ratio F^/F^ can be changed. The dimension d
is fixed but dimension y' is adjustable by means of the ratio setter.
When either of the two flows changes, the balance between the two
forces is disturbed. The jet pipe deflects, thereby providing a controller
234 AUTOMATIC CONTROL

output signal for the final control element. The resulting adjustment of
the secondary flow reestablishes the ratio balance.
Frequently, the ratio is automatically adjusted in response to a third
process variable. This case is illustrated in Figure 12-11 where the feed

PRODUCT

TEMPERATURE
SET CONTROLLER

COLUMN

Figure 12-11. Cascade ratio control system.

flow and the reflux flow of a distillation column are maintained at con-
stant ratio. As the temperature rises the ratio of feed to reflux is sup-
posed to increase. As shown, the temperature is measured and the
output signal of the temperature controller is utilized to adjust the ratio
between feed and reflux. Actually there are two controlled variables, the
temperature and the ratio. There are also two complete controllers.
Flence this is a cascade system. However, the temperature controller cas-
cades into a ratio system. The block diagram, by showing that one proc-
ess variable—the feed—is not part of the closed loop, clearly expresses
that a non-cascade system is also involved.

2- and 3-Element Control


Equation (12-4) in its first version, i.Q.,kr — b = 0, shows that the ratio
controller may be considered a subtracting control system just as weU as
a ratio system. In fact, the difference is in the viewpoint rather than in
CONTROL SYSTEMS 235

the action. The change in ratio can be considered as altering the set point
by subtracting from it some other value. An illustrative example is found
in maintaining the level of boiler drums. When the steam demand in-
creases, a momentary “swelling” of the water in the drum occurs. Simi-
larly, when the demand drops, momentary “shrinking” takes place. Re-
sponse of the level controller to these phenomena produces undesirable
upsets of a balanced feedwater flow.
In the two-element system shown in Figure 12-12, the output signal
of a flow controller is used to adjust the set point of a level-recording

STEAM

Figure 12-12. Two-element control system.

controller. The level controller regulates the valve position in the feed-
water line. As long as the steam flow is constant, the system responds
only to changes in level. If the steam flow changes, the set point of the
level controller is automatically readjusted to provide for a correspond-
ing change in feedwater flow.
The result is that in case of increased steam flow, the level controller
will call for less feedwater flow because of the swelling, while the steam
flow controller calls for increase of feedwater flow. The two signals
cancel out. The feedwater flow remains constant until the momentary
swelling subsides. At that point, the increased demand of the steam flow
results in increase of feedwater flow. By this method load changes result
in a minimum of upset in the boiler steaming condition.
When a permanent change of set point, and hence of level, cannot be
tolerated, a feedback signal from the final control element may balance
the flow signal, as well as the level signal. An arrangement of this sort is
illustrated in Figure 12-13. Change in steam flow changes the differential
pressure across the orifice. The differential pressure is applied across a
diaphragm and acts as a force on the beam which positions a four-way
236 AUTOMATIC CONTROL

valve. The resulting displacement of the actuating piston operates the


control valve. The primary feedback from the level operates in similar
fashion. For example, when the level rises, the float motion relaxes the
spring that connects it to the lever. This results in a clockwise motion of

Figure 12-13. Two-element control system with feedback.

the lever. The four-way valve admits oil to the top of the piston, tending
to close the valve. This downward motion feeds back to the level by means
of the spring restoring the balance at a new position of the control valve.
In case of increased steam demand and momentary swelling, the level
tries to move the lever clockwise, while the increased pressure differential
opposes this motion. The result is that the control valve does not move
until the swelling subsides. Normally, a manual set point adjustment is
added, which, however, is not shown in the diagram.
Figure 12-14 illustrates both methods of two-element control. In
either case the flow signal is used as reference input for the level control
system. The valve-position feedback in the second case, however, sub-
tracts from the flow reference input, and under balanced conditions, the
input into the second summation point becomes zero, as far as the flow
signal is concerned. The actuating signal into the level controller is then
merely the difference between manual set point and primary feedback.
The three-element feedwater control system combines a typical cas-
cade level control system with the steam flow correction. It is illustrated
in Figure 12-15 and is particularly suitable where load disturbances in
the feedwater supply side are to be expected. The three-element system
permits, furthermore, the use of floating control. Since the boiler drum
has no self-regulation, floating control cannot be used with it unless a
three-element system is used. The latter converts the arrangements into
CONTROL SYSTEMS 237

a - WITHOUT VALVE-POSITION FEEDBACK

b - WITH VALVE-POSITION FEEDBACK

Figure 12-14. Schematic drawings of two-element systems.

Figure 12-15. Three-element control system.

a flow control system with self-regulation into which the output of the
level controller is cascaded.

Totalizing
In oil refineries, steel mills, etc., by-product fuels are frequently burnt
either separately or in combination with coal, oil, or natural gas in order
to obtain an economical source of thermal energy for steam boilers.
Multiple-fuel combustion presents some interesting control problems.
One is that the total Btu supply in fuel must be maintained. Generally,
238 AUTOMATIC CONTROL

the by-product is used as base fuel. However, the by-product supply is


rarely constant. One, or sometimes two, make-up fuels are therefore re-
quired in order to maintain a constant Btu supply for the furnace.
Another problem is the ratioing of air in proportion to the fuel. Fre-
quently, some indirect methods are used, such as maintaining a con-
stant ratio between steam and air flows in a steam boiler. Flue gas
analysis may also be employed in order to proportion air to the oxygen
content of the flue gas.
The most direct method, however, is the direct ratioing of air to fuel,
either separately for each fuel or for the sum total of the fuels. As com-
pared with gas analysis methods, this provides faster control, since cor-
rection can be made at the moment when the fuel flow rate changes, and
not after the results of the change become measurable in the flue gases.
A disadvantage, however, of proportioning air to fuel flow is a re-
sultant volumetric ratio. The correct ratio between air and fuel is based
on weight. If either the air or a gaseous fuel change in pressure or tem-
perature, the ratio will lose accuracy, unless provisions are made which
correct for these changes.
Figure 12-16 illustrates the principle of the Hays fuel totalizing con-
trol system. Total Btu input is the controlled variable of this system. The
desired magnitude of total Btu is adjusted by means of the Btu control
potentiometer either manually or automatically by a steam pressure con-
troller or temperature controller.
In the arrangement shown, either fuel 1 or fuel 2 may be the base fuel,
and the remaining fuel, the make-up fuel. The flow rate of the base fuel
may vary, yet the total Btu input is maintained at the desired value by
automatic adjustment of the make-up fuel. Which one of the two fuels is
to be the base fuel and which the make-up fuel can be determined by
a manual selector switch.
Flow meters measure the flow rates of the two fuels. These meters are
provided with mechanisms for square root extraction, so that the output
is proportional to flow rate. A potentiometer in each flow meter is
positioned in proportion to the rate of the respective flow.
The output of the two potentiometers is averaged by use of a center-
tapped autotransformer, and this average voltage is applied to the coils
of the relay. The relay contacts operate a valve actuator which adjusts
the control valve of the make-up fuel. When the relay closes one con-
tact, the actuator opens the valve; when it closes the other contact, the
valve closes. The valve maintains its position when none of the contacts
is closed.
With the Btu control potentiometer in a given position representing a
definite Btu input demand, and the base fuel being burned in a quantity
CONTROL SYSTEMS 239

Figure 12-16. Principle of Hays fuel totalizing control system.

less than sufficient to meet the demand, the electrical relay closes the
corresponding contact. This opens the valve of the base fuel until the
flow as measured by the flow meter reaches the desired level, and the
relay contact is again opened.
Figure 12-17 shows the principle of an Askania multiple fuel control
system. In this case, two fuels are burnt in a furnace and air is supplied
to match the fuels. The arrangement is as follows:
A small blower or compressed air source provides a pilot air flow
through two lines, each about one inch in diameter. A ratio controller
A measures on its primary diaphragm the differential pressure across an
orifice plate in the fuel line I. By means of a hydraulic flow through a
jet pipe, this controller positions the actuating piston of the control
valve in C of the pilot air flow arrangement. An orifice plate in the C
branch provides a differential pressure that is imposed on the secondary
diaphragm of the controller A. A rate of pilot air flow is thus obtained in
C, which is directly proportional to the rate of flow of fuel I.
Ratio controller 5 in a like manner measures the flow of fuel by means
of an orifice plate in fuel line II, and in turn provides a means of con-
240 AUTOMATIC CONTROL

PILOT AIR FLOW


FUEL @

Figure 12-17. Askania totalizer and air ratio control system.

trolling an air pilot flow in proportion to this fuel, and thus to its air re-
quirement.
Therefore, controllers A and B regulate the rate of air flows C and D
in such a way that the air flow in each line is always proportional to the
flow of each fuel. These air flows are then brought together and the total
flow, measured by means of oriflce E, provides the primary signal for
ratio controller F. The combustion air is regulated by this last controller,
which maintains it in fixed proportion to the common pilot air flow which
represents the sums of fuels I and II.
The ratio sliders A and B may be set for different air ratios cor-
responding to the requirements of each fuel. They can also be auto-
matically set as a function of temperature and pressure of gaseous fuels,
and thus correct for changes in density. The same holds true of the com-
bustion air ratio controller.

Time-Schedule Control
Time-schedule controllers regulate batch processes according to some
predetermined time schedule. Various methods are described here and
illustrated in Figure 12-18; these are available in the line of time-
schedule controllers made by the Foxboro Co.
The automatic shutdown controller starts its cycle when a push button
is depressed. This opens the valve and leaves it open until the controlled
CONTROL SYSTEMS 241

SHUTDOWN AUTOMATIC
POINT SHUTDOWN
CONTROLLER

ELAPSED
SET POINT
TIME
CONTROLLER

SET POINT ELAPSED


TIME
CONTROLLER
WITH
DEFERRED
ACTION

CAM-SET
CONTROLLER

CYCLELOG
CONTROLLER

\
\

Figure 12-18. Action from various types of Foxboro time-schedule controllers.

variable reaches a predetermined value, the shutdown point. The valve


then shuts down automatically.
The elapsed time controller starts a preselected time period. Within this
time, it brings the controlled variable to its set point and then controls
it at that point. At the end of this period the valve is automatically
closed.
The elapsed time controller with deferred action differs in that the
elapsed time is counted from the time where the controlled variable
reaches the set point.
The cam-set controller regulates a schedule according to the contour of
a rotating cam. Where the full 360 degree cycle of a rotating cam is not
242 AUTOMATIC CONTROL

required, the cam is returned to the starting point by a manual


adjustment.
The Cyclelog controller is used to pre-heat a batch to any selected base
temperature. It then holds the controlled variable at that temperature
for a preset period of time. After this time the temperature is allowed to
rise at a preselected rate to the top set point. It holds the batch at the
top set point for a certain period of time, after which it automatically
shuts down the process. Two steps may be added as indicated by
the dashed line in Figure 12-18. They consist of (1) a predetermined
prolongation of the time which the valve stays open by pressing a push
button, and (2) the provision of a controlled rate of cooling.
These are only a few of a large number of variations which are used
in control systems where the control of a process becomes a time
function.
13. INDUSTRIAL APPLICATIONS

Only experience can teach the proper application of the principles of


automatic controls. New combinations of automatic controllers and
their accessories come up continuously. Some practical examples, taken
from a number of industries, are described in the following pages.

Steam Generation
Figure 13-1 shows the automatic control system of a steam-generating
apparatus. It follows closely an installation by the Hagan Corporation

Figure 13-1. Automatic control system of steam generator.

243
244 AUTOMATIC CONTROL

at the Hazelwood By-Product Plant of Jones & Laughlin Steel Corpora-


tion.* Coke-oven gas is used as base fuel and tar is used to make up for
deficiencies in the supply of coke-oven gas. The tar is preheated to a
temperature of 250°F which makes its consistency comparable to that of
fuel oil. Atomizing steam is added to the tar for easy combustion.
The pressure in the steam header is a direct function of the demand
on the steam generator. This pressure controller is the master controller
of the entire combustion system. The sequence by which it operates is
as follows.
Suppose the steam header pressure drops. This is sensed by pressure
controller PRC-1 and results in an increase of auxiliary steam to the
turbine which drives the forced draft fan. The increased rate of flow of
combustion air is measured and changes the set point of flow ratio con-
troller FRC-4. A ratio computer converts the air flow signal into a pro-
portional signal which is applied to the ratio controller. This raises the
rate of flow of coke-oven gas. The increased input of fuel in proportion
to combustion air results in an increase of steam generation, returning
the steam pressure to its set point.
The pressure of the coke-oven gas is maintained by controller PRC-3.
This has the advantage of constant upstream pressure for the flow con-
troller, which is thus independent of varying pressure drops in the gas
lines resulting from changes in flow rate due to position changes in the
flow control valve.
The coke-oven gas flow is maintained constant by the FRC-4 flow
controller, but it changes in proportion to the combustion air flow. A
summarizer provides a signal to flow ratio controller FRC-3. This signal
is equal to the fuel demand from the ratio computer minus the fuel
supply as measured in the coke-oven gas line. As long as the coke-oven
gas copes with the fuel demand, the control valves in the tar and atomiz-
ing steam lines are closed. When a deficiency is signaled by the sum-
marizer, the set point of the FRC-3 controller is automatically raised
and hot tar and atomizing steam start to flow simultaneously.
Furnace draft is controlled to limit the amount of cold air filtering
from the outside through the brickwork of the furnace and to protect
from the danger of back firing in case of over-pressure in the combus-
tion chamber. The furnace draft controller PRC-2 is a ratio controller.
Steam pressure is the primary variable. The final control element is a
valve in the auxiliary steam element which controls the steam to the
turbine which drives the induced draft fan.

* Peth, H. W., Kotsch, J. A., and Gilmer, H., “Highly Automatic Steam Generation,” Instru-
ments and Automation, (Dec. 1955).
INDUSTRIAL APPLICATIONS 245

In addition to the control system illustrated in Figure 13-1, the flame


in the combustion chamber is continuously monitored. This prevents the
flow of coke-oven gas or tar into the furnace without immediate com-
bustion. The combustion detector is connected into an electric relay sys-
tem which shuts down air, gas, and tar flow in case of flame failure. This
is done by additional valves not shown in the diagram. The same electric
relay system is utilized to provide shutdown in case of fan failure, either
in the induced draft or the forced draft system, of feedwater failure, or
of power failure.

Soaking Pit
Before steel ingots pass to the rolling mill where they are pressed to
their ultimate shape—sheet metal, bars, etc.—they are heated in a soak-
ing pit to a high, uniform temperature. The most desirable condition is
a uniform plasticity at the highest possible temperature without melting
any of its constituent elements. Figure 13-2 shows a control system used

by Askania Regulator Co. for premixed gas-fired soaking pits. Blast


furnace gas and coke-oven gas are used as fuel. They are combined in a
mixer and pass from there to a booster where they are brought to the
pressure required by the burners. This pressure is measured by pressure
controller PC-3, and is controlled by regulating a butterfly valve in the
blast furnace gas line supplying the mixer. For every change in gas load
there will be a corresponding change in the blast furnace gas supply in
order to maintain the mixed gas pressure at the desired set point.
The primary elements which measure the rates of flow of blast furnace
gas and coke-oven gas are butterfly valves which substitute the conven-
tional orifice plate. The differential across the butterfly valve is measured
as a function of the flow rate. Flow controller FC-1 changes the opening
of the butterfly valves with changes in flow. Hence the equivalent of a
variable area meter is obtained. This provides greater accuracy through-
246 AUTOMATIC CONTROL

out the control range and avoids the limitations in rangeability that are
characteristic for fixed orifice plates.
Proportioning of the blast furnace gas and coke-oven gas is provided
by flow ratio controller FC-2. The primary variable is the blast furnace gas.
When the rate of flow of the blast furnace gas changes, the set point of
FC-2 is automatically readjusted and the control valve in the coke-oven
gas line is repositioned to maintain the constant ratio between the two
gases. Another flow ratio controller, FC-4, proportions fuel gas to air
flow.
PC-5 is the soaking pit pressure controller. The pressure is measured
in one of the side walls and transmitted to the controller which positions
a damper in the stack through which the waste gases escape. Positioning
the damper controls the pressure in the soaking pit. This prevents suc-
tion in the pit which may otherwise result at low fuel rates and produce
excessive air infiltration. Not only is heat loss the result of air infiltra-
tion, but the flow of gases in the pit itself may be disturbed, upsetting
the uniformity of heating in the various sections. On the other hand, if
excessive pressure builds up, heat is wasted at openings, cracks, and
through the soaking pit walls, resulting in serious economic losses.

Water Treatment
The automatic control system of a water treatment plant is illustrated
in Figure 13-3. The diagram is based on the Floneywell control system*
at McAlester, Oklahoma. Water is pumped from a large lake reservoir

Figure 13-3. Automatic control system of water treatment plant.

to a second lake for storage. From there it flows by gravity toward the
mixing basin, shown in the illustration.

* Collins, R., “McAlester Modernizes,” Instrumentation, Vol. 10, No. 2, (1957).


INDUSTRIAL APPLICATIONS 247

The flow into the mixing basin is controlled by the flow ratio controller
FIC-2. When the level in the clarifier drops, the level controller LIC-3 ad-
justs automatically the set point in FIC-2 to admit more flow to the
treatment plant.
The measured flow is signalled to the chlorinator which adds chlorine
to the water in proportion to rate of flow. Furthermore, flow meter FR-1
records and integrates the raw water flow. A flow accumulator which
parallels the flow meter integrator starts the feeder timer after each 2315
gallons of raw water. The feeder timer actuates both alum and lime
feeders and adds these chemicals to the raw water for 30 seconds. This
process reduces the turbidity of the water to less than 10 parts per
million.
The water then flows by gravity into a clarifier, then into sand filters
and on to a million-gallon clear well. No measurable concentration of
suspended solids are contained in the clear well water. The chlorine
residual is about 0.4 parts per million.
Flow to the clear well is regulated by ratio flow controller FIC-5 with
level controller LIC-6 for automatic set point adjustment.
The pressure drop across the filter is indicated by PIA-7. As the filters
get clogged, the pressure drop increases. It is allowed to rise a certain
amount, but when this amount is exceeded, an alarm sounds. This sig-
nals that the filter needs cleaning. The operator starts the backwashing
cycle, which removes the filter from the continuous water supply cycle.
Since, however, a number of filters (not shown in the diagram) are
parallel, the load is taken over by them and the process is not interrupted.
When the alarm sounds, the operator presses two buttons; one to
silence and reset the alarm, the other to start a timer which rotates eight
cams on a common shaft, making and breaking certain electrical con-
tacts to produce the backwashing cycle as follows:
Solenoid valve SV-1 is energized, shutting off the controller output
signal from LIC-4 and releasing the air to the control valve. This closes
the filter influent valve.
After ten minutes, the timer energizes solenoid valve SV-2, admitting
air to the drain valve and opening it. At the same time SV-3 is also en-
ergized, which closes the filter effluent valve. Now the solenoid valve
SV-4 is energized and opens the valve in the surface wash line for one
minute. This done, solenoid valve SV-5 admits air of 6 psi pressure
through the diverting relay to the valve in the wash water line. The di-
verting relay has the characteristic of always passing the air which has
the highest pressure. After 2.5 min. solenoid valve SV-6 energizes and 13
psi air pressure now flows to the control valve. This sequence of a lower
and then a higher pressure produces first a low baskwash rate followed
248 AUTOMATIC CONTROL

by a faster backwash rate. After another five minutes SV-6 is deenergized


and the system returns to a low backwash rate for 2.5 min. Remaining
steps return the system to normal operation by closing the wash and
drain valves and opening the influent and effluent valves. The effluent
valve is opened only after the filter has been refilled to proper height.

Air Conditioning
The control of humidity and temperature in an air conditioning sys-
tem is illustrated in Figure 13-4. The air which flows into the humidifler
becomes saturated after being sprayed with water. This water is main-

Figure 1 3-4. Air conditioning control system.

tained at constant temperature by means of split range valves for water,


and is steam controlled from a cascade temperature controller. The
humidifier represents a process capacitance, and since the response of
TRC-2 is delayed more than is suitable for good control, it is aided by
controller TRC-1 to improve the dynamic response.
The saturated air then passes through a heater where, by means of
steam coils, the temperature is brought to that maintained by temper-
ature controller TRC-3.

Solvent Recovery
Figure 13-5 illustrates the control system of a solvent recovery unit.
The diagram is based on an installation* by Taylor Instrument Companies
at Lederle Laboratories, Pearl River, New York. This unit recovers or-
ganic solvents which are needed in large quantities for obtaining aureo-
mycin from fermentation mash. The unit consists of four distillation
columns. The incoming feed contains four component solvents which
are designated a, b, c, and d. In the first column, a liquid mixture of a,

* Englund, S. W., and Giesse, R. C., “Solvent Recovery at Lederle Labs,” Taylor Technology,
Vol. 5, No. 2, (1952).
INDUSTRIAL APPLICATIONS 249

A — Feed I with solvents a, b, c and d


B — Steam
C — Water
1 — First Column VII — Reflux drum
D — Product with solvents a ond b
II —Second Column VIII — Calandria
E — Feed II with solvents c, d and HjO
III — Third Column IX — Decanter
F — To waste
IV — Fourth Column X — Reflux drum
G — Solvent b
V —Condenser XI — Pump tank
H — Waste layer for recovery
VI — Feed Drum
J — Solvent c
K — Solvent d

Figure 13-5. Automatic control system of solvent recovery unit.

b and water is distilled off. The remaining solution passes on to eolumn


II and is combined with a second feed flow containing c, d, and water.
Solvent b is recovered from column III, and solvents c and d are
recovered from the overhead and bottom product, respectively, of column
IV. These recovered solvents are in essentially pure form and are returned
to the production process. The bottom layer in the decanter is passed
on as waste to another column which is not shown in the diagram.
Feed flow and steam are admitted to column I in controlled quantities.
The distillate leaves at the top and passes to a condenser where it
is liquefied. From there it flows to the reflux drum and is then either re-
turned to the column for another distillation cycle or drained as product
containing a and b components. The amount of reflux is controlled by
the recording temperature controller TRC-4. The condensate temper-
ature is also measured and regulates the rate of flow to the condenser.
The flow of bottom product from column I to II is combined with
another feed flow. The total flow is maintained constant by FRC-7 ad-
mitting as much feed flow as required for maintaining the constant flow
rate. Water is admitted to column II by means of a manually controlled
valve with a position adjuster at the control panel. The development of
suitable analysis control instruments would probably allow automatic
250 AUTOMATIC CONTROL

control in this case. As it is, bottom and overhead analyses are made at
periodic intervals. The bottom product of column II goes to waste.
Column III sends its overhead through reflux and drains periodically
by another manually controlled valve to recover solvent b. The bottom
product passes through a feed drum with level control into column IV.
Heat is supplied by means of a calandria through which part of the
liquid circulates. The steam flow is kept constant by ratio controller
FRC-14. If the column temperature drops, the rate of steam flow is in-
creased by readjusting the set point of FRC-14 by means of temperature
recording controller TRC-13. The bottom product of column IV passes
into a pump tank. It eonsists of solvent d. The overhead is condensed
and flows from the condenser to a decanter. The bottom layer flows out
to another column. The top solvent layer goes to the reflux drum, and
is then split into reflux and solvent c. The reflux is maintained constant
by ratio controller FRC-18. Any exeess of this flow is drained as solvent
c. An inverse ratio between reflux and waste layer flow from the deeanter
is maintained by readjusting the set point of FRC-18 by means of
FRC-19. This is done to return sufficient solvent to the column, though
some is removed to the additional column.

Gasoline Production
A control diagram for a Perco motor fuel alkylation process is
shown in Figure 13-6. The arrangement is based upon a control system
proposed by Berger and Peters,* using a number of analytical instru-
ments for control purposes. This process takes olefins, i.e., propylene,
butylenes, and amylenes, and alkylates them with isobutane, using
liquid hydrofluoric acid as catalyst. Non-gasoline-range hydrocarbons
are thus converted into high-octane blending stocks.
It is not intended to analyze the process in detail but rather give an
over-all concept and then show how some of the variables are controlled.
An electronic control system with electrohydraulic valve actuators is
assumed.
The liquid olefins enter the process at A, while isobutane for alkyla-
tion is added through B. Both feeds, after passing through their respec-
tive driers, are combined with additional recycled isobutane, and enter
the reactor II. Hydrofluoric acid, the catalyst, enters from the bottom.
After the alkylation reaction is completed in the reactor, the mixture
flows into the settler III, where the hydrofluoric acid separates from the
hydrocarbons. Most of the acid is recycled to the reactor. A portion.

* Berger, D. E., and Peters, W. D., “Converting Non-Gasoline-Range Hydrocarbons into


High-Octane Blending Stocks,” Control Engineering, (Sept., 1956).
251

_g
.y C O X
12 u _x
*u Q)
r= O w. < C c “D
o
1 S’ ^ C
D o
^ I O D
c o -Q
O E
^ o
(U ^ X i C<1) •“X
V)

a
LW

<D o a, ^
‘x 0-C~D 0"D 3-2
o o
;u o ^’n.— Cc O Q-“D,_ • —n ^ E
-O ^ “D
-C -C
D O
o 0) *u a)Jl^<i>oa)Oa)>si *u ^ ’u £ 0 ^
J) (D 0)
CO (/) <^ QS^ LL
Ds:ii-^QUa:UQX<>o O < to 1 Z
I I I I I' 'I I I I I I I I I I I I I I I
_==>>> < £0 U □ UJ u. O I
i/i
> > i X X X 10

Figure 1 3-6. Automatic control system for Perco motor fuel alkylation proce
252 AUTOMATIC CONTROL

however, is deviated through the acid rerun unit where water and acid
oils are separated from the hydrofluoric acid by distillation. To replenish
the acid in the system, a fresh acid storage tank, VI, is provided, which
receives its charge directly from the tank car.
The hydrocarbons from the settler flow into the deisobutanizer, VII,
where the isobutane and lighter components are separated. The bottom
product of this column is normal butane and alkylate which is drained
off at H. The overhead product provides recycle isobutane for the re-
actor. A small portion of this isobutane, however, is sent for fractiona-
tion to the depropanizer, XI, to obtain upgraded recycle isobutane, and
to the hydrofluoric acid stripper, XII, for the production of propane,
which is removed at G. The hydrofluoric acid is recycled to the reactor.
The diagram of this process is equipped with a number of analytical
control instruments. Thus, the feed into the reactor is analyzed for
olefins by ARC-3 and for isobutane by ARC-4. A ratio computer per-
mits setting the ratio of these two components. The signal from this
computer to the flow ratio controller FRC-2 is the difference between
the desired and the measured ratios, and readjusts the set point of
FRC-2. Since FRC-2 controls the isobutane, it will change this flow until
the desired ratio is reestabhshed.
ARC-7 analyzes the hydrofluoric acid for water and acid-soluble oil
contents. It controls the set point of FRC-6 which controls the feed flow
to the acid rerun unit. This feed flow is heated by a heater not shown in
the diagram. A second flow controlled by FRC-5 by-passes the heater.
Thus the ratio of these two flows is determined by the purity of the re-
cycle hydrofluoric acid.
The reflux into the deisobutanizer is controlled by flow ratio controller
FRC-14. The set point of this controller is automatically adjusted by
ARC-15 which analyzes for N-butane concentration in the column.
Another analytical instrument, ARC-16, determines the isobutane con-
centration and controls the steam input into the reboiler by readjusting
the set point of FRC-17. A propane concentration analyzer, ARC-20,
controls the amount of deisobutanizer overhead product that is passed
into the propanizer.
Another interesting arrangement is the weight controller WRC-10. By
means of a strain gauge, the weight of the contents in the settler is de-
termined. When it drops under a set minimum, the two-position con-
troller starts the motor pump to get additional hydrofluoric acid from
the storage tank.
INDEX

Accuracy, 8 Cahbration errors, 10


specifications of, 9 angular, 10
Actions, see Controller actions zero, 10
Actuating signal, 1, 53, 177, 179 Capacitances, 4
Actuators, 204 electrical, 5, 115
electrohydraulic, 218, 219, 250 thermal, 5
Air-conditioning systems, 36, 37, 38, 248 volume, 5
American Society of Mechanical Engineers, weight, 5
ASME, 1, 23, 65, 204 Capacitors, 115, 116, 199
Amplifiers Capillaries, 99
as null detectors, 128 flow equation, 99
electronic, 126, 127, 128 Capillary coil bulb, 172
magnetic, 128, 129, 130 Cascade control, 223, 226-231
full wave, 130 pneumatic set, 229-231
pneumatic, see Pilot valves span adjustment of, 231
Amphtude, see Sinusoidal curve, amplitude polymerizer, 227-229
Askania Regulator Co., 239, 245 ratio, 234
Averaging control, 223, 224, 225 Cascade controllers, 229
Choppers, 128, 200
Backlash, 11, 93, 134 Colhns, R., 246
Balances Consotrol, 186
force Control agent, 2
with deflection system, 136, 137, 138, 144 Control elements, final, 43, 204-219
with null system, 138, 139, 140, 144 Control loop
square-root extraction, 158 closed, 1, 25, 232
position, 136, 137, 144, 145, 146 gain, 210
hydrauhc, 145 open, 25, 231
pneumatic, 145 Control point, 3
“Bellofram,” 93, 150 Control systems
Bellofram Corporation, 93 cascade level, 226-227
Bellows, 94 distillation column, 234
spring rate, 94 flow plan, symbols & identifications, 220,
Berger, D. E., 250 221
Bode diagrams, 65, 67, 73 fractionating column, 222
Booster, 193, 218 fuel totalizing, 238
Bourdon tube, accuracy of, 9 multiple fuel, 239, 240
Break frequencies, 79, 82 three-element, 236, 237
Bristol Co., 99, 121, 227 two-element, 235, 236

253
254 INDEX

Control valves, 15, 204-219 Dead zone, 103, see also Dead band
butterfly, 216, 217 Deflection systems, 136-139
characteristics, 209-214, 215 Deviation reduction factor, 47
diaphragm, 205, 217 Diaphragms, 92-94
flow rate, 210 effective area, 92, 93, 205
four-way, 105-110 effective diameter, 92, 93, see also effective
flow forces of, 108, 109 area
hole-and-plug, 109, 110 linearity, 93, 94
overlap and underlap, 107 metal, 94
linearities, 204, 214-216 slack, 92
pressure drop, 211 Differential gap, 28, 31
rangeabiUty factor, 214, 215 Differential transformers, 117, 118, 202, 203
speed, see Stroke time computing with, 152
stroke time, 206 linearity, 118
Controlled medium, 2 Drift, 128
Controlled variable, 1, 2 Dynatherm, 174
Controller actions, integrating, see Reset
action
Englund, S. W., 248
Controller switch, multi-position, 31
Controllers
discontinuous, 7 Fanning’s equation, 212
electric, 196-203 Feedback
electronic, 196, 201, 217, see also electric internal, 88, 229
force balance, 186 primary, 1, 2, 177
hydraulic, 38, 192-196, 223 Flapper, see Flapper-nozzle
interaction, 203 Flapper-nozzle, 44, 45, 95-98
master, 244 flow equation, 95, 96
mechanisms, 179 hydraulic operation, 98
pneumatic, 223, .224, 231 hnearity, 96
with force balance, 186-192 nozzle back pressure, 96, 98
with position balance, 177-186 rate of flow, 95, 96
stabihty of, 203 size, 97
stack-type, 188 special forms, 98
Corrective action, speed of, 6 free-vane, 99
Critical setting push-pull, 98, 150
hmitations for, 43 Floating rate, 39, 40, 50
of proportional band, 50, 81 adjustments, 40-42
Cylinders, 111 method of determining, 42
double-acting, 111 Floating speed, 39, 40
single-acting. 111 critical setting, 42
Flow rate, 154, 210
Dally, Charles A., 216 Foxboro Company, 174, 186, 240
Damping Frequency response, 61-86
critically damped response, 41 block diagram, 71, 72
overdamped response, 40 conclusions, 86
Vi-decay ratio, 41 data
14-decay ratio, 41 combination of, 71
Dead band, 11,14 graphic representation of, 65, 73
Dead time, 5, 24, 35, 165-172, 175, 226 technique of taking, 63-65
INDEX 255

Frequency response (coni.) Instability, see Stability


margins Instrument Society of America, ISA, 220
gain, 69 Integrating action, see Reset action
phase, 69
safety, 69 Jet pipes, 103-105
of proportional plus reset controller, 79-82 hnearity, 104
of proportional plus reset plus rate con- sizes, 105
troller, 82-86
of proportional-position controller, 77-79 Kirk, D. B., 142
of proportional-speed floating controller, Kotsch, J. A., 244
74-77
stability conditions, 69, 70, 74, 76, 78 Lag network, 199, 202
steady-state conditions, 79 Lead network, 199-202
Friction factor, 212 Ledoux bells, 154, 155
hmitations of, 155
Gain, see Process gain Leeds & Northrup, 174, 200
Galvanometers, 119, 120, 137, 138 Linahan, T. C., 167
magnetic field of, 119 Linearity, 11, 154
Gasoline production, 250-252 absolute, 12
Generators, pneumatic sine-wave, 64 in control systems, 14, 205
Giesse, R. C., 248 in final control elements, 204, 209
Gilmer, H., 244 independent, 11, 113
normal, 11, 113
Hagan Corporation, 243 zero-based, 11
Hays Corporation, 238 Linkages
Heat cam and counterweight, 155
capacity, 162 disadvantages, 157
conductance, 162 lever system, 156, 157
conductor, 162 mechanical, 155-157
convection, 163, 164 Liquid-level control system, 27
film resistance, 163 Load changes, 3, 226, 235
propagation, 162-165 Loop, see Control loop
quantity, 162 Lost motion, 8, 10
radiation, 162, 174
Hooke’s law, 88, 136 Magnetic fields, 119, 123, 129
Hornfeck, A. J., 164, 170 Magnitude ratio, 63, 65, 67, 73, 84
Hysteresis, 12 Manipulated variable, 2
in control systems, 14 Manning, Maxwell and Moore, 201
magnetic, 13 Marks, L. S., 212
mechanical, 13 Mason-Neilan Regulator Co., 177, 178
with bellows, 94 Massachusetts Institute of Technology, MIT,
with torque motors, 135 109
Measuring means, 87
Impedance, 116 Minneapolis-Honeywell Regulator Co., 179,
Inductance, 116, 117, 129 197, 246
mutual, 117 Moore, C. B., 141
self-inductance, 117 Moore Products Company, 142, 188, 190, 192
Inductance coils, 116, 119 Motion, simple harmonic, 62, 64
Induction motors, see Servomotors Motor coasting, 35, 36
256 INDEX

Motor operator, 204 Potentiometers, 112-115, 139


Moving coils, 130-132 conclusions, 114
with electromagnet, 132, 218 linearity, 113
Multiple-fuel combustion, 237 resolution, 114
resolution sensitivity, 113, 114
Pressure duplicators, 138, 140, 142
Needle valves, 100 Primary elements, 2
flow equation, 100 radiation-type, 174
Neutral zone, 33 response time, 154
Nichols, N. B., 23 Process gain, 20, 22, 215
Noise, 114, 128 factor of, 208
Nonhnearities, see Linearity Processes
Nozzle, see Flapper-nozzle multi-capacitance temperature-controlled,
NuU systems, 138, 139, 140, 141 56, 228
linearity of, 140 temperature-controlled, 42
Nyquist criterion, 66 with dead time, 23
Nyquist diagrams, 65, 66 with multi-capacitances, 22, 41, 66
with single capacitance, 22, 41
Offset, 2, 46, 79 Proportional band
Oppelt, W., Dr., 47 adjustment of, 48, 59, 181, 182, 184
Orifices, 95, 99, 104 critical setting of, 50, 84
Oscillators, 120, 121 method of calculation of, 49, 50, 224
amplifier, 202 Proportional plus rate control, 228
sensitivity, 121 Proportional plus reset control, 51, 52, 53,
Overshoot, 6, 30 183, 184, 190, 194, 224
methods to reduce, 31 adjustments of, 55
Proportional plus reset plus rate control, 55-
59, 185, 192, 200
Period, 62, see also Frequency response adjustments of, 60
Peters, W. D., 250 Proportional-position control, 23, 43, 50, 53,
Peth, H. W., 244 66, 179, 180, 183, 188, 194, 197, 198,
Phase angle, 63, 65, 67, 69, 73, 75, 77, 84 224
Phase lag, see Phase angle adjustment of proportional band, 49, 50,
Phase reversal, 68 see also Proportional band
Phase shift, 69, 118 Proportional-speed floating control, 38-42,
Phototubes, 121-122 50, 53, 193, 224
gas-filled, 121-122 adjustment of floating rate, 41,42
photomultiplier, 121-122 Protecting tubes, 164
vacuum-filled, 121-122
Pilot relay, 142, 143, see also Pilot valves
Pilot valves, 45, 100-103, 142-143 Ramp function responses, 57
amphfication factor of, 98 Rate action, 56, 58, 82, 174, 185, 191, 199,
continuous-bleed, 100 201, 202
hnearity, 101, 102 Rate time, 58, 59, 82, 85, 203
non-bleed, 102 adjustments of, 59
Pipehne Ratio control, 231
characteristics, 213, 214 hydrauUc, 233
pressure drop in, 212 totalizer and air ratio, 240
Pneumatic relay, see Pilot relay Ratio slider, 233
INDEX 257

RC circuit, 182, 183, 186, 199, 200 Soaking pit, 245, 246
lag network, 199, 202 Solvent recovery, 248, 249
lead network, 199-202 Sorteberg bridge, 150-152
Recorder, 64 square root extraction, 150, 152
Recorder-controller Specific heat, 162, see also Heat capacity
electric, 144 Spring force, 89, 90, 92, 205
pneumatic, 144 Spring rate, 88, 89, 90
Rectifier, phase-sensitive, 203 Springs
Reference input, 1, 177 compression, 89, 205
Relays, electropneumatic, 217, 218 helical, 88, 95
Remanence, 14 precompressed, 90, 102
Reset action, 52, 53, 79, 183, 187, 190, 199, tension, 89
201, 202 Square root extractions, 150, 154, 158
Reset rate, 53, 54, 55, 79, 82, 85, 203 Stability, 7, 66, 70, 74
adjustments, 52, 54, 59, 184 Stabilizer, hydrauhc, 195
Resistances, 4 Static characteristics, 8
Resistors, 114, 148, 149, a/50 Rheostats Static friction, 8, 10, 11, 114
Resolution sensitivity, 10 Steam generation, 243
of differential transformer, 118 Step change, see Step input
of potentiometer, 113 Step function responses, 17, 18, 25, 26, 52, 57
of rheostat, 10 nonhnearity of, 17
Response rate, 19, 20, 22 of proportional plus reset controller, 53
Response time, 165, see also Time constant of proportional-position controller, 46
Rheostats, 10, 114, 115 Step input, 18, 20, 52, 57, 161
Stiction, see Static friction
Saturation, 128, 129 Summing point, 1, 2
Scales Swartwout Co., 202, 217, 218
linear, 73 Synchro control transformer, 125
logarithmic, 67, 73 Synchro differential transmitter, 125
Scientific Apparatus Makers Association, Synchro receiver, 124
SAM A, 165 Synchro transmitter, 123, 124, 126
Self-regulation Synchros, 122-126, 198
process with, 19, 27
process without, 18, 75
Servoamphfiers, 133 Taylor Instrument Cos., 173, 181, 229, 230,
Servomotors, 126, 132-134, 147, 148, 198 248
hmitations of, 134 Temperature control, 154, 210
spring-restrained, see Torque motors choosing a controller for, 175, 176
Set point, 2, 3, 32 dynamic errors, 159, 166
mechanisms, 177, 178, 201 static errors, 159
Shearer, J. L., 107 Temperature controllers, 29, 175
Single-speed floating control Temperature gradient, 163
conclusions, 38 Temperature reproduction, 159
with self-regulation, 36 precautions, 160
without self-regulation, 33-36, 197, 236 Thermal lags, 161-165
Sinusoidal curves, 62 Thermocouples, 170-172, 174
amphtude of, 62, 64 Thermometers
Sinusoidal input, 61 filled-systems, 169, 172
frequency of, 63 resistance, 170-172
258 INDEX

Three-term control, see Proportional plus Two-position control, 196


reset plus rate control conclusions, 32, 179
Thyratrons, 126 with dead time, 29
Time constants, 21, 165-172, 183 without dead time, 27
approach to shorter, 172
of filled-system thermometers, 169
Vacuum tubes, 126, 127
of resistance thermometers, 170-172
Valye actuators, see Actuators
of some controlled media, 168
Valve plugs, 209
of thermocouples, 170-172
Valve positioners
Time functions, in pneumatic elements, 181
force-balanced, 206
Time lags, 3
gain of, 208
Time-schedule control, 240-242
position-balanced, 206
automatic shutdown, 240
Valves, see Control valves
cam-set, 241
cyclelog, 242
Water treatment plant, 246, 247
elapsed time, 241
WeUs, see Protecting tubes
elapsed time with deferred action, 241
Wheatstone bridge, 146-150, 200
Torque motors, 134, 135
hnearity of, 147
Totalizing control, 237-240
square root extraction, 150
totahzer and air ratio, 240
Transaire temperature transmitter, 173
Transistors, 126, 127 Young, A. J., 203
junction, 126
point-contact, 126 Ziegler, J. G., 23
s.
•ts ■

*. ^

k.’

You might also like