136.

00-2

Instrumentation & Control - Course 136

CONTROL THEORY REVIEW

Consider the level control of an open tank wi th a vari-

able demand outflow and variable inflow regulation by control
valve throttling.

Clearly, the level will only remain constant when the

inflow (gi' just equals the outflow (go). If the demand
increases, the level will begin to drop off until the inflow
has increased sUfficiently to provide a mass balance and
stablize the level.
A transmitter is required for measuring the tank level
and developing a signal that can be displayed remotely for
level indication, alarm monitoring or be input to a computer
or an analog controller for level control. An electronic
transmitter produces a 4-20 rna (now accepted as the industry
standard) or a 10-50 rna signal representative of process var-
iation. If the transmitter was suitably calibrated the
signal will vary linearly from 4-20 rna as the tank level
changes from 0-100%. The transmitter is just the control
loop da ta link wi th the process. The current process sta te
will be continually advised by the transmitter in an analog
fashion.

November 19B1 - 1 -
 136.00-2

~----~

LT: Level Transmitter

The transmitter signal can be monitored by an electronic

controller which is a voltage input device. A dropping
resistor is placed in the current loop to develop the requir-
ed voltage signal. For example a 250 ,II dropping resistor
will develop 1-5 VDC from a 4-20 rna signal. The transmitter
signal or measurement signal can be compared to the set point
(desired operating point) to determine the process error sign
and magnitude. '

ERROR = SET POINT - MEASUREMENT

E = SP M

This error can now be sui tably amplified by the con-

troller to produce a corrective control signal of 4-20 rna.

SP

I E:::.-ror

M -#AMP >----:l>

This can be represented in symbol form as follows:

Mea~~~nt__f \ __ .£.ontrol:..
Signal \V Signal

LIe: Level Indicating Controller

- 2 -
 136.00-2

The control (outPut) signal is applied to a field mount-

ed transducer (liP) which converts the 4-20 rna electronic
signal to an equivalent 20-100 kpag pneumatic signal. The
pneumatic signal provides the driving force to position the
spring opposed diaphragm actuated control valve.

Control
----- liP
Signal

Flow

The complete loop with all the equipment combined would

result in an operative, closed loop, negative feedback con-
trol system.
qi----i¥l-----......,

e
,

IL .. _

L..._--:l>qo

Assume the control valve is an air to open style1 in-

creasing the pneumatic signal to the valve will drive it more
open. The controller must, now respond to process deviations
in a particular fashion defined as the controller action. A
controller with direct action will respond to a process
deviation above the set point by increasing the control
signal. That is to say, an increasing rnea~urement results in
an increasing control signal. Direct action is shown as
(increasing, increasing) or ( t t ) . F9llowing this same rea-
soning, reverse action will be shown as (increasing, decreas-
ing) or (H I •

The controller action required for the level control

system developed must be reverse. As the level rises above
the set point, the control signal will decrease, allowing the
valve to go more closed. Control actions are changed in
electronic controllers by simply positioning a switch so that
a controller is easily adapted to the requirements of the
system.
 136.00-2

The level control loop can also be sketched in block

form rather than as a physical system.

sp - + -
Error,
AMP
I lip VALVE
TANK
LEVEL

XMTR

For general representation, this sketch c'an be simpli-

fied by considering the control amplifier, lIP, valve, and
process blocks to be combined into one block designated as
the forward gain block G.

E C
-l> G
-
B

R = Reference input or set point

E = Error, deviation from set point
G = Forward loop gain operator for controller, valve, etc
H = Feedback gain operator representing the transmitter
B = Feedback signal
C = Controlled variable (level in this example)

Ideally, the controlled variable should be maintained at

the set point by the control system. The ratio of controlled
variable to set point (C/R) can be calculated by referring to
the previous closed loop sketch. From the sketch:
 136.00-2

E =R - B, EG = C, CH = B

E = R-B
C
G =R - CH

C = GR - CHG
C + CHG = GR

C(l + HG) = GR
C G
R = I + HG

C
This ratio of R is called the control ratio. Later in
the course this ratio will be referred to as the closed loop
transfer function and will describe the ratio of controlled
variable to input excitation as a function of the disturbance
frequency. The control ratio in the frequency domain would
appear as:

cqw) = G(jw)
R(Jw)

Consider the system developed so far being subjected to

a disturbance such as a step increase in demand with the
level controller only able to provide on/off control (like a
house thermostat). The valve will be driven fUlly open or
closed depending on the process position wrt the set point.
As soon as the process reaches the set point, the valve will
drive to the other extreme. It should be relatively easy to
visualize the resulting level control with on/off control.
The process will settle out in a continuous cycle about the
set point.

Notice that if the level is too high, the inflow valve

is driven more closed and vice versa.

The control system must act in opposition t,o the devia-

tion in an attempt to restore the process to the set point.
The following sketch represents the process cycling about the
set point and the corresponding control system response.

- ,-
 136.00-2

Peak Deviation (Input J

/ f
Conslont
Amplitud~

__t
Controller Output

Although the control system reponds to the error, it

would be convenient to consider the subsequent process
response to the corrective action. Consider the dotted sine
waveform representing valve opening and the solid sine wave-
form representing level variation about the set point.

In the inflow valve is drlven open to provide maximum

input to the tank, the level shoulo rise. Notice that the
level does not peak until 1/2 cycle later. (See the two
shaped peaks.) The level response is delayed or is said to
lag behind the valve motion hy 1/2 cycle or 180°. The cUf-
ference in phase angle between a sinusoidal output and input
is called the phase lag. Consider the instant when the valve
opening is peaking negatively and the process curve is peak-
ing positively. The sine angle corresponding to a negative
peak will be 270° (sin(270) = -1) while that corresponding to
a positive peak will be 90° (sin(90) = 1).

Phase lag = output - input

¢=900-270°

¢ = -180'

The system phase lag will always be negative so that the

negative sign is usually deleted, lag being implied.

The system gain can also be considered for this cycling

condition. Gain in control theory will be the ratio of per-
cent change in output to percent change in input for some
block or dev ice.

%11 output
GAIN =
%11 input
 136.00-2

Since the amplitude of the cycling is constant, the loop

gain must be one. Otherwise the amplitude would be attenuat-
ed (Gain <1) or the process would drive to some limiting
value (Gain 1). Loop gain is simply the combined effect of
all gain components in the system, such as controller gain
(k e ), valve gain (k v ) and process gain (k p ).

Loop gain = kckvk p

The critical conditions for continuqus process oscilla-
tions are a loop gain of one and a phase lag of 180°. Such
sustained, constant amplitude cycling is defined as marginal
stability. An operational system should avoid the marginal
stability state by ensuring that the loop gain is less than
one when the phase lag approaches 180 0 • The phase lag in a
stable system will usually be less than 140 0 preventing sus-
tained oscillations.

In order to produce a full cycle of oscillation, a total

rotational angle of 360 0 must be present. ,The controlle;r
always introduces a phase angle of 180 0 by acting in opposi-
tion to the process deviacion. This constant 180 0 lag by the
controller is ignored in phase considerations. The remaining
180 0 (to prov ide 360 0 ) lag is ,-supplied by such things as
valve reponse, process capacitance, and dead time.

The limitation of on/off control is that the valve wil:).

be stroked to its ex treme pas i tion wi th an error of 1 % or
10%. Even though only a small correction is required, full
valve travel (over correction) will resulc. If the control
system could respond proportional to the error, a more stable
system should result. A variable gain controller is required
so that if the process is close to the set point, the valve
will be throttled slightly.

Proportional Control

The optimum response of a proportional control system

would approximate the quarter decay recovery curve. The
quarter decay curve is just underoamped response to a distur-
bance where each successive positive peak is 1/4 of the pre-
vious one.
 136.00-2

Consider the level control system under proportional

control being sUbjected to a demand increase. The level
fluctuations will be much more stable than was the case under
the on/off control.

% Level

Time
Level Response Following An Upset

A proportional controller can have setting units of

either gain or proportional band. Proportional 'Band (%PB) is
defined as the percentage change by the controlled variable
about the set point (symmetrical) which will produce a 100%
change in control signal. '

%6input
%PO = x 100
%l'loutput

Gain and proportional band are reciprocally related.

Gain = 100
%PB

Assume that the level varies 10% above and below the set
point before the control signal drives through 100% change.
The controller must have a proport~onal band of 20% or a gain
of 5.

The output signal of a proportional controller will

respond proportional to the error. The manipulated variable
is dependent upon the control signal, so allow that the mani-
pulated variable will be varied proportional to the error.

mae

Including a proportionality constant results in:

m = ke
rn = manipulated variable magnitude,
k = controller gain,
e = process error, deviation from the set point.

- 8 -
 136.00-2

In this example the manipulated variable will be adjust-

ed in proportion tc;> the error in an attempt to maintain
stable control of the process. If the controller wa$
successful in holding the measurement at tqe set point, then
the error would be reduced to zero. Obviously the manipulat-
ed variable cannot be allowed to drop to zero in a dynamic
system, so ~he basic control equation must be modified
slightly. A bias term can be included to provide a control
signal value if the error is zero.
m = ke + b
b = bias term

The bias term is supplied so that the control signal

will drop to zero when the process is at the set point. In
general the bias value is considered to be 50% of the control
signal for an ideally aligned proportional controller. That
is the contr.ol signal output of a flE!rfectly aligned propor-
tional controller will be 50% when the measurement equals the
set point regardless of the controller gain.

The for~ula representing proportional control is achiev-

ed by sUbstit~ting 100 in place of k.
%PB .

100
m =
%PB e + b

The block diagram commonly used to represent a propor-

tional contrdller is shown below:

R _~~_-'=.e_ __k ]_~m",---".

I- c

R = reference' input or set point,

k = controller gain,
m = manipulate..d variable,
c = feedback signal from the controlled variable,
e = error. ~
The proportional band of the controller would be adjust-
ed to provide optimum control of a given process. If a
transient disturbance occurs the controller will stroke the
valve in proportion to the error to provide corrective
action.

_ a _
 136.00-2

Offset

Proportional only control will result in a process off-

set if the process supply or demand conditions of the system
have been varied. Offset is defined as the stable deviation
of a process away from the set point after a proportional
only control system has responded to a process supPly or
demand change. Imagine a system under proportional control
in which the error has been red uced to zero wi th m = b =
50%. If the %PB = 50, and a load change requires the mani-
pulated variable to rise to 60% to achieve equilibrium; then
determine the error required to provide this value of m.

100
m = %PB e +
b

100
60 = 50
e + 50

10 = 2e

e = 5%

If the proportional band was 200%, calculate the offset

which would result.

100
m = %PB
e + b

100
60 = 200 e
+ 50

10 = .5e

e = 20%

The offset is a function of the proportional band and

can be a substantial error depending upon the proportional
band and the magnitude of the load disturbance. Clearly
this type of steady state error would not be acceptable in
many control sys terns. Note tha t the problem of offset can
not be corrected by narrowing the proportional band since the
system might then be driven to instability by excessive loop
gain.

A practical example of offset can be shown using a level

system with a variable demand represented by two pumps.

- 10 -
 136.00-2

---
-._-j

P,
., ~q
0

P,
~
qo

System Data

Controller Scale = 0 - 100%

Set point = 60%
% PB = 50%
Valve = 0 - 200 l/min, linear flow characteristics
Pl = 100 l/min
P2 = 40 l/min

Initially, only PI is running, the level is at the set

point and stable. Since the measurement is at the set point,
the control signal will be 50% (bias value). Inflow will be
50% of 200 l/min or 100 l/rnin which just matches the outflow
due to pump Pl'
Pump P2 is now switched on simulating a demand increase,
outflow exceeds inflow and the level begins to drop. The
level controller responds proportional to the error and
drives the valve more open until the inflow matches the out-
flow. The level must drop enough so that the error modified
by the controller gain will stroke the valve sufficiently to
restore equilibrium.

Calculation of Offset

Initial outflow: 100 + 40 = 140 l/min

Initial inflow : 100 l/min
Required change in inflow: +40 l/min
40
Required %~valve: 200 x 100 = 20%

%6output = Gain x % error

20 = 2 x % error
%error = 10 (this is the offset)
 136.00-2

The level must drop 10% before the inflow again matches
the outflow and the level stops dropping. An offset of 10%
would exist in this particular case.

Straight line approximations can be used to sketch such

a control response. Initially the level will drop at a maxi-
mum rate (line 1) since there is the greatest mismatch
between inflow and outflow. As the level drops, the control-
ler will drive the control valve more open to decrease the
mismatch so that the level will not drop so rapidly (line 2).
The rate of level drop will decrease until the level is
eventually held steady at the new offset position (line 4).

% Level

-4
----- 3

Time

A second control mode, reset or integral mode is requir-

ed to eliminate offset.

Reset Mode

Reset or integral mode is selected to eliminate offset

in a control system. The offset problem in the previous
example can be corrected by opening the inflow valve more
than the equilibrium proportional response requires. Reset
mode will integrate the error, changing the control signal
until the measurement is brought back to the set point.

Consider the proportional response to a given process

change. If the error is not zero, reset mode will begin to
contribute to the control signal. The time required for
reset to duplicate the original proportional response can be
noted in minutes. This time is called the reset time and is
the number of minutes required to repeat the original
proportional response. Reset time is stated in Minutes Per
Bepeat (of the proportional response) and is designated as
MPR.

- 12 -
 136.00-2

Some manufacturers mark the integral mode setting in

reset rate units of Repeats (of proportional response) Per
Minute (RPM). Reset- time and reset rate are reciprocally
related.

1
MPR = RPM
eg, 5 MPH is equivalent to 0.2 RPM.
Pure integral mode output can be represented as:
1
m "" R ft edt
a
m = manipulated variable
e = error
R = reset time in MPR

Assume that the process to be controlled has a sinu-

soidal error developed about the set point.
--p--"I

L---':L~ ..... I--__-l'-__ Set Point

This error can described by the general function:

e = A sin (2 TI t)
P
e = error
A = amplitude
t = time
P = period,
The integral mode output will be:
. (2,t)
1 it A Sln
m = R - - dt
P
a
AP Cos(2TIt)
m = - 2'lTR P It
a
AP cos(2TIt) +~
m = - 2'lfR P 2TIR

- 13 -
 136.00-2

AP
Allow that rn o = 21TR '
the manipulated variable value
at time zero.

m = - ~~R Cos (2;tl = roc

The effect of reset mode on the loop gain and phase

angle should be determined. The gain of the integrator can
be found from the ratio of the integrator output amplitude to
input amplitude.
GR = AP'2 nR = P
2nR

For any given cycle period, the gain of the integrator

depends upon the reset time selected. Notice that reset mode
gain will decrease as the period of cycling decreases. Reset
can be considered as sup·plying high steady state gain for the
elimination of offset, but will have a limited gain contribu-
tion for a transient disturbance.
In order to find the phase angle contriJ:jution of rest
mode, the output and input angles must be compared. The
input to the integrator unit was the function Sin (2nt) while
. ---p
the output is -Cos

Graphically a cosine function leads a sine function by

90°.

+1 Cosine

o
/
/Sine

-1
,J

A negative cosine function will be inverted so that the

output will lag the input by 90°.

+1
Cosine

o Sine

-1

- 14 -
 136.00-2

An ideal integrator exhibits a 90° phase lag regardless

of the period of the input. Since phase lag is an undesir-
able quality, the addi ticn of reset to a control system is a
trade off between the elimination of offset and decreasing
the stability of the system.

Proportional Plus Reset

This control combination allows the response speed and
stability of proportional control while eliminating offset
from the system. The general formula representing this two
mode combination is:
100% 1 t
m = %PB (e + R fo edt)

The following block diagram is commonly used to repre-

sent proportional plus reset control.

'-------- c

Ti = integral time or reset time (MPR)

Since both gain and phase considerations are involved,

the combination of proportional plus reset control can be
represented by vector methOdS.

100
l:'ll"

100 P
PB (2rrR)

- 15 -
 136.00-2

(G pR )
, = (100)' + ((100)(..E-..) )'
PB PB 211'R

(..E-..)' 1/'
GpR = 100 + (1 + 2.R )
PB

100 P
PB (2.R) P
Tan(~) = =
100 2'R
PB
P
~PR = tan-! ( 2.R) .
(lag)

It has been found that optimum control will be obtained

with a proportional plus reset controller if reset contri-
butes appoximately 11° lag at the marginal stability condi-
tion. Reset would be specified on a system where offset can
not be tolerated and the system could be subjected to process
supply or demand changes. Examples of proportional plus
reset control would be steam drum level contro1, or feedwater
flow control.
If reset time is too low '"for the process application,
then reset mode will change the signal faster than the pro-
cess can respond. The signal will be integrated past the
normal limits (above 20 rna in a 4-20 rna loop) by reset. The
valve can only be physically fully opened or completely clos-
ed even though the signal exceeds its normal range limits.
When the process finally does respond to the valve being at
the extreme position, the process must cross the set point
(change the sign of the error) before the reset term begins
to integrate back down. The net effect is a cycling process
more extreme than on/off control since the valve does not
begin to change status until after the set point is crossed.
This condition where the control signal exceeds normal limits
due to reset action is referred to as reset wind-up.

The following simplified control program demonstrates

the potential problem of reset wind-up.

PROGRAll

Sum = 0 •••••' •• -••••••••••••••••••• initialize summing location

10'''' ReadT ....••....••.....••.... determine temperature
El = Set-T ••••••••••••••••••••••• calculate error
Sum=Sum + El ••. ~ ..•••..•••....••. increment sum
SIG = k* (El + Sum) ...•••...••....• calculate signal
R
Output SIG ...••....•••....••....• apply signal to DAC
service other loops
Go to Hilt

- 16 -
 136.00-2

Assume that the proportional plus reset temperature con-

trol loop is executed once each second and the error (El),
sum value (Sum) and % signal (BIG) are recalculated each
second. If a particular error, say 5%, exists and the pro-
cess is slow to respond, then every second the sum will be
incremented by 5%. Imagine the consequences of this constant
5% error being summed every second for three minutes if the
reset time is very low. The reset term becomes very large
wrt the proport ianal term so that even though the process is
eventually restored to the set point, the signal does not
change significantly. The process must now rise above the
set point so that the sign of the error will change and then
the reset mode can beg in to integrate the excessive signal
back down (effectively Sum = Sum - EI).

A properly adjusted proportional plus reset controller

should restore the process to the set point in a stable
fashion following a process supply or demand change.

% Level
-_ . Set Point

Time

Correct Reset Time

If excessive reset rate is applied, then the resulting

over correction will result in the process cycling about the
set point. Reset cycling is caused by too many RPM or too
low a setting MPR.

- 17 -
 136.00-2

Set Point
% Level

Time

Reset Rate Too Fast

Derivative Mode

Derivative mode or rate mode responds to the rate of

change of the deviation of the process from the set point.
Derivative mode can be achieved by taking the first deriva-
tive of the error.

m ::: D de
dt

m ::: manipulated variable

D = derivative time in minutes
e = error
Derivative time is the time advance of the control
signal with proportional plus derivative control over the
response to the same disturbance with only proportional con-
trol. Derivative response is easily visualized for a system
which is subjected to an error which changes linearly wrt
time. (Say 2%/min.) The proportional response will be a
ramp keyed by the changing measurement and dependent upon the
controller gain.
 136.00-2

% control
Signal

o 1 2 3 4 Minutes

Proportional Response to a Ramp Disturbance

Derivative mode will only respond while the error is

changing. Assume a derivative time of 2 minutes, then deriv-
ative will cause the control signal to be now', that signal
value that proportional would cause two minutes from now;
while the error is changing at the same rate.

% control
Signal

o 1 2 3 4 Minutes

Proportional Plus Derivative Response

Practical derivative response will be somewhat rounded

as the error just begins to change or just stops changing.
(Similar to charging or discharging in an RC circuit.) The
net effect of the larger magnitude control response will be a
smaller process deviation away from the set point and improv-
ed stability since derivative is acting to oppose a changing
error. (Braking effect.)

Consider the response of a pure derivative device being

sUbjected to a sinusoidal error.

- 19 -
 136.00-2

Asin(2 n t)
m = Dde = D d( p)
dt
dt

m =
DA2 n
p
Cos (2~t)

The input function is sin(2nt/P) while the output func-

tion is Cos (2nt). The output must lead the input by 90° due
p
to the relationship of sine and cosine functions. An ideal
derivative mode exhibits a phase lead of 90° regardless of
the period of the input.

Gain for the pure rate mode is found from the ratio of
output amplitude to the input amplitUde.

DA2n
= -p- = 2n D
p
A

Notice that if the natural period is short, then the

gain of the derivative mode will be large. There will be
large derivative gain from a high frequency disturbance.

Noise in a high frequency process will be amplified con-

siderably by an ideal derivative unit. This is why deriva-
tive mode is not specified for fast processes such as flow.
The resulting over correction and prompt process response
will soon result in uncontrolled cycling. A high limit is
always placed on derivative gain to prevent high frequency
instability. This derivative gain limit is usually about ten
and is called the rate amplitUde. The magnitude of the rate
amplitude is set during the instrument manufacture.

There is one best setting for derivative time for any

given process. Too much derivative time will stroke the
valve excessively resulting in process cycling with a short
period.

% Level Set Point

%PB Response to a Step Change

- 20 -
 136.00-2

% Level I \.,.. -----.~ - - Set point

'PB + Correct Rate Response

% Level
--_._------- Set Point

Too Much Rate Action

The correct adjustment of derivative mode should mlnl-

mize the process deviation and max:imize the stability of the
application.

Proportional Plus Derivative Control

This two mode automatic ,control combination can be
represented in general by:
m = 100 (e + D de ) + b
%PB dt

Notice that the bias term is required for ~his combina-

tion since control will revert to straight proportional in
the steady state. If the error stops changing then de ~ 0
dt
and the control equation becomes:
100
M = rn (e) + b

A control system with proportional plus derivative con-

trol will still experience offset When subjected to a process
supply or demand change. Reset must be included if offset is
a problem in the control system.
The following block diagram represents a proportional
plus derivative controller.

R m
---"'---;~

Td ~ Derivative time in minutes.

- 21 -
 136.00-2

The resulting gain and phase angle for a proportional

plus derivative controller can be determined by the same
vector methods used with the proportional plus reset model.

GPD '1
100 (2nD
PB P

2
(G pD )2 = (100)2 + « 100) (2nD))
PB PB P

, 1/,
GpD = 100 (1 = (2nD) )
PB P

100 (2nD)
PB P 2nD
Tan(¢) = 100 =
P
PB

¢pD = tan- 1 (2 n D) (lead)

P

Empirical results from operative loops have shown that a

maximum lead of 40 0 would be the ,practical phase contribution
limit for a proportional plus derivative controller. The
derivative mode would be adjusted to produce this lead at the
marginal stability condition. This lead contribution effec-
tively shifts the frequency at which the 180 0 lag occurs to a
higher value improving the stability of the system.

Theoretical Three Mode Controllers

The three modes discussed can be combined to form a pro-

portional plus reset plus derivative controller (PIO). This
will result in a controller which can respond proportional to
the deviation and the rate of deviation while eliminating the
steady state error. This combined function can be represent-
ed mathematically as:

- 22 -
 136.00-2

m = 100
PB
(e + 1
R f
t edt + D de)
dt
o

Notice that with the addition of reset, the bias term is

not required since the integration of the error provides a
floating bias value.
The block diagram representing a three term controller
is shown in the following sketch.

R
~€+ Itris Tat--_m=-~ +
1- c

The resultant gain and phase angle is dependent on the

mode settings selected and the frequency of the Rrocess dis-
turbance. The derivative time will always be set to provide
a larger lead contribution than the lag from reset, so that a
net phase lead will result to aid stabiliza~ion. The result-
ant three mode vector representation will be:

A G pID
100 2rrD 'V'I
PB --p- I
~PID
I
100 P 100
PB 2rrR PB
'if

( GpID) P ")
rR)

P 2 liz
Gp1D = 100 (1 + 2rrR) )
PB
 136.00-2

100 ( 2 TID - P )
PB P 2TIR 2".D _ P
Tan(¢) = =
100 P 2JTR
PB

¢PID = tan- 1 ( 2 PTID -

TiTR
p) (lead)

Since the lag of reset is offset by the lead of rate, a

greater amount of reset rate can be used in a three term con-
troller than is possible with only proportional plus reset.
A net lead of 25° is the expected phase contribution from an
optimumly tuned three mone controller.

Cascade Control

qi ---j;:,Q------,

riP

Consider this level system being subjected to a demand

disturbance followed by a supply disturbance. If a demand
increase occurs the tank level will begin to drop. The level
controller will respond proportionally, stroking the valve to
a new position in an attempt to stop the level from dropping
further. Should a supply decrease occur at this time, the
flow through the valve will be less than expected for the
given valve position. Tank level will now drop further from
the set point until equilibirium can again be restored. Min-
imizing this sort of upset would be desirable in many appli-
cations.

The problem can be simplified by realizing that the

supply disturbance is the only addition to this system com-
pared to previous examples. If the flow fluctuations can be
steadied, then the level system should perform adequately. A
flow control loop can be placed on the inflow line and a com-
mand for a particular flow can be developed by the level
controller.

- 24 -
 136.00-2

qi I' A/Q

r
FT
l/
IMG
I
L_ FI __ J I

i Ii
SP
L_Q- LT

qo

This control combination is called a cascade system.

The control signal output from the level controller is now
the set point for the flow controller. Both controllers in
this system must be reverse action. The fl,ow controller
action is dictated by the valve in the flow system. The
level controller is reverse action since an increase in level
requires a decrease in flow.

If a demand increase occurs, the tank level begins to

drop and the output of the level controller will increase.
This signal will raise the set point of the flow controller,
effectively calling for increased inflow. Should a supply
decrease occur, the drop in flow is sensed by the flow trans-
mitter and compared to the requested set point of the flow
controller. The flow valve is throttled as necessary to hold
the flow at the requested rate. Flow disturbances are elim-
inated before they can disturb the large capacitance, slower
responding level system.

The general format for a cascade system is for the major

lag controller to develop the set point for the minor lag
controller. (Major lag sets Minor lag.)

r---~~ Major Lag r---~ Minor Lag

Flow
Level

- 25 -
 136.00-2

The major lag controller is usually referred. to as the

primary controller. The output of the primary controller is
then the set point for the minor lag or secondary con-
troller. After finishing 136.00-3 (Control Mode Settings),
consider how you would start up and tune a cascade system
such as this level control example.

- 26 -
 136.00-2

i\SSIGNMENT

L Sketch an electronic control loop used to maintain the

temperature of 020 leaving a heat exchanger.. The
effluent temperature is sensed by a surface mounted
RTD. The valve used on the cooling service water line
must fail open in the event of lost instrument air.
Discuss one cycle of operation and determine the con-
troller action.

2. Sketch a simplified block diagram representing the cas-

cade control of the system in Question 1. Identify the
major and minor lags.

3. Sketch the control loop equipment necessary to provide

the cascade control of the HX system so that both ser-
vice water supply variations and bleed flow changes can
be corrected. Sta te controller act ionl3 ~ I

4. Explain clearly how offset can occur in a proportional

controlled system which is subjected to a process supply
or demand change.

5. Discuss the net result of using derivative mode on a

flow application which can experience a surge flow
transient.

6. From a general stability consideration, what is the

advantage of derivative and disadvantage of reset mace
in a given control system?

7. What is the effect of reset wind-Up on a control system?

8. What phase lead will a controller produce if the deriva-

tive time is zero minutes? Show this as a vector
c::kpt-_ch.
 136.00-2

9. A proportional plus derivative controller is utilized on

an application where the process cycles 6 times in 2
hours. (Marginal stability.) Control settings of 75%
PB and 2 minutes derivative are selected.
(a) Determine the theoretical resultant gain of this
controller.
(b) Determine the phase angle contribution for this
control combination. ( lead or lag)

(c) State if you feel this derivative time is cot'rect

based on the empirical model or if the derivative
time shoUld be increased or decreased.
(d) Can you pred ict what the approximate deri vati ve
time shoUld be in minutes? What general informa-
tion is required7 (See 136.00-3, to check your
answer. 1

10. Assume the application in question 9 is to be controlled

with a proportional plus reset controller. Control
settings of 75% pa and 4 minutes per repeat for reset
are selected.

(a) Determine the theoretical resultant gain of this

controller.

( b) Determine the phase angle contribution for this

combination (lead or lag) •

(e) State if you feel this reset time is correct based

on the empirical model or if the MPR shoUld be
increaseo or decreased.

(d) Can you predict what the approximate reset time

should be in MPR'? What general information is
required? How would you obtain this data?

M.J. MacBeth
D.E. Tennant

_ On