ARCHIVED Corporate Engineering Standard

Maintenance and Construction Procedure: PR38

PR38
Controller Tuning

Table of Contents

1. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Introduction—controller tuning . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Controller tuning precautions . . . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Trial settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

5. Controller modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
6. Controller and valve actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
6.1 Types of controller action . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
6.2 Determining controller action . . . . . . . . . . . . . . . . . . . . . . . . . 7
6.3 Controller action—cascade loops . . . . . . . . . . . . . . . . . . . . . 8
6.3.1 Select correct controller action for primary controller . . . 8
6.4 Valve action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
6.5 Split-range valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
7. Controller tuning procedures—all modes . . . . . . . . . . . . . . . . 9
7.3 Upset generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
7.5 Gain adjustment procedure . . . . . . . . . . . . . . . . . . . . . . . . . 10
7.6 Reset adjustment procedure . . . . . . . . . . . . . . . . . . . . . . . . 11
7.7 Derivative procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
8. Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
8.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
8.2 Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
8.3 Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
8.4 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
8.5 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
8.6 Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
8.7 Floating control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
8.8 Batch switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Document reaffirmed February 1991 / Document archived March 2004

Contact ENGG::STANDARDS on E-mail for more information.
This document may be used and reproduced for DuPont business only. Page 1 of 23
PR38
ARCHIVED
Controller Tuning

9. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
10. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

List of Figures
Figure 1. Trial settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Figure 2. Product cooler temperature control . . . . . . . . . . . . . . . 7
Figure 3. Gain too low, proportional band too wide . . . . . . . . . 12
Figure 4. Optimum gain, proportional band . . . . . . . . . . . . . . . 12
Figure 5. Gain too high, proportional band too low . . . . . . . . . 12
Figure 6. Reactor temperature cascade control . . . . . . . . . . . . 13
Figure 7. One-quarter ratio decay cycle . . . . . . . . . . . . . . . . . . 14
Figure 8. Constant gain on process signal . . . . . . . . . . . . . . . . 18
Figure 9. Variable gain on process signal . . . . . . . . . . . . . . . . . 18
Figure 10. Adaptive gain used to stabilize control . . . . . . . . . . . 19
Figure 11. Characterizer to linearize divider . . . . . . . . . . . . . . . . 19

Document reaffirmed February 1991 / Document archived March 2004 2

PR38
ARCHIVED
Controller Tuning

1. Scope
1.1 This procedure provides information and guidance to aid the user
in the adjustment of the controller portion of a process control loop. The
procedures may be used with Pneumatic, Electronic (analog), Single
Loop Digital Controllers (SLDC), Distributed Control Systems (DCS),
and Direct Digital Control (DDC).
1.2 This procedure is designed to aid plant maintenance personnel.
Much more information and formal theory and techniques are available
in the literature shown in section 10. A number of experts are available
in the Company for special problems.

2. Introduction—controller tuning
2.1 The quality of control achieved with an automatic control system
depends upon several factors:
a. Characteristics of the process; such as dead time, time constants,
and nonlinearities.
b. Quality of the measurement and control devices and their current
condition. For example, a sticking control valve will introduce nonlin-
earity and erratic behavior and adversely affect control. A thermowell
with an insulating coating will slow down the temperature measure-
ment and thus cause slow response to changes. A long pipeline
bringing brine to a cooler may be filled with warm brine if there has
been no demand for some time.
c. Extent and type of process upsets.
d. Controller tuning of the loop affected, or possibly, of other interacting
control loops.
2.2 To be successful, the process of tuning (the controller adjustments)
must be systematic and logical. This procedure presents proven meth-
ods and principles.

3. Controller tuning precautions

3.1 Assessment of the effects of controller tuning changes will require a
test of the control loop by observation of system response either during
normal operation or in response to a deliberately caused upset. Be cer-
tain that those responsible for safe and proper operation of the process
(Operations) agree to both the magnitude and direction of any proposed
changes, or have them make such changes. The problem is to introduce
an upset large enough to be observed and to provide usable information,
but small enough to keep from causing an unacceptable situation, either
to the operation or the equipment. The upset might be caused by moving
the set point or by some other action to change the operating conditions.
Many modern controllers respond to a change in set point in a different
manner than a change in the measured variable. For these the set point
change is not always a good test of system response.

Document reaffirmed February 1991 / Document archived March 2004 3

PR38
ARCHIVED
Controller Tuning

3.2 Before making any changes, note and record the control valve load-
ing (output signal) required to hold the process at an acceptable value,
and the “as-found” controller settings. This will permit an orderly recovery
to manual operation if required and to return to the original tuning set-
tings if operation is not improved. Records of the reasons for change,
the changes made, and the results of the changes can be very valuable
in understanding the operation of the loop. It is very possible that the
tuning settings which are successful for one set of operating conditions
will be unacceptable at another. A compromise set of settings may be
needed, or it may be necessary to retune when operating conditions
change. Use of adaptive gain, see 5.2, may make it possible to better
satisfy all operating conditions with a single set of tuning parameters.
3.3 Note that other system upsets that might occur while tuning is in
progress can mask the effects of adjustments. Do not change controller
tuning without making an effort to prove satisfactory operation at the
new values, and to be certain that the control loop is stable. It is best
to observe operation under actual load change conditions for a long
enough time to create confidence in the results.

4. Trial settings
4.1 A reasonable estimate of controller settings may often be made
from a limited knowledge of the process. A number of rules of thumb
have been developed as a guide in tuning. These are usable only for
continuous processes and are for “regulatory action,” where the goal is
to keep a constant process value (PV) in the face of process upsets.
4.2 “Servo-action” type control is required if the PV is to closely follow a
changing set point (SP). This may be required in some batch processes
with programmed SP changes. Here it is necessary to have a system
which ideally is faster in response than the set point change. There is
less guidance in the process control literature on this problem than for
regulatory-action control.
4.3 Figure 1 shows controller settings that can be expected to give satis-
factory regulatory control in most loops, based upon the type of variable
being controlled. These settings may not result in the optimum response.
However, if the process is stable, they will be adequate for noncritical
loops and further tuning will not be needed. Consider these as starting
point for loops without known characteristics. These values are based
upon theory and experience and have been proven by years of use.

Document reaffirmed February 1991 / Document archived March 2004 4

 PR38
ARCHIVED
Controller Tuning

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
Figure 1. Trial settings

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
Reset

ÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁ ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁ ÁÁÁÁÁÁ
(Repeats/ Derivative
Process type Gain PB Min/rpt minute) (time)

ÁÁÁÁÁÁÁÁ
Flow
ÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
0.4 250 0.1 10 –

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁ ÁÁÁÁÁ
ÁÁÁÁÁÁ
Pressure—close
coupled(1) ÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
0.5 200 0.2 5 –

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
Pressure—
ÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
2 50 2 0.5 –

ÁÁÁÁÁÁÁÁ ÁÁÁÁÁ
ÁÁÁÁÁÁ ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
accumulating(2)

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁ
Level–tight(3)
ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
1.3 80 5 0.2 –

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁ ÁÁÁÁÁ
ÁÁÁÁÁÁ ÁÁÁÁÁÁ
Level—averaging(4)

ÁÁÁÁÁ
ÁÁÁÁÁÁ
Temperature—fast
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
(5)

ÁÁÁÁÁÁ
1.3
(6)
80
–
2
–
0.5
–
0.2

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
measurement(5)

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁ
Temperature—slow
ÁÁÁÁÁÁ
ÁÁÁÁÁÁ
1 100 2 0.5 2

ÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
ÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁ
measurement(5)

Notes:
(1) Such as in header pressure control
(2) Such as in vessel pressure control
(3) Level must be held close to set point
(4) Level may deviate from set point
(5) The value for derivative should be made equal to the time constant of the measurement if that can be determined

5. Controller modes
5.1 There are many types of controllers. Assuming a constant SP, each
type of controller responds to a change in PV in a unique way. This type
of response is the controller “mode.” The actual response (change in out-
put value) may be affected by the magnitude of the process change, the
duration of the change, the rate of change, or some combination of
these. Some modes are described below:
On-off. The output is 100 percent or 0 percent. A thermostat is an exam-
ple of this. Used for certain simple applications, usually where the mea-
sured variable changes slowly.
Gap control. Has a deadband, usually centered on the SP. Response
outside of the deadband may be any of several modes.
Proportional (P). The output changes proportionally to the error, i.e.,
to the difference between process and set point values.
Proportional-integral (Pl). The output changes proportionally to the
error plus the sum (integral) of error times the duration of the error.
Proportional-integral-derivative (PID). The output changes proportion-
ally to the error plus the sum (integral) of error times the duration of the
error plus the rate of change of the measured variable.
Proportional-derivative (PD). Output changes proportionally to the error
plus the rate of change of the measured variable.

Document reaffirmed February 1991 / Document archived March 2004 5

PR38
ARCHIVED
Controller Tuning

Integral (I), or “integral only” or “floating.” The output changes pro-

portionally to the magnitude and duration of the error, without any initial
proportional response.
5.2 Other modes such as nonlinear gain are available. Some digital con-
trollers provide configurable selection of controller type. One, for exam-
ple, allows selection, by configuration, of four types of control algorithms.
These types are:
Proportional-Integral-Derivative (PID),
Proportional-Derivative (PD),
Integral-Derivative (ID), and
Proportional-Integral-Derivative with Adaptive Gain (PIDAG).
5.3 This choice of modes makes it possible, at times, to optimize con-
troller tuning and operation, thus achieving a quality of process control
difficult or impossible to obtain with the simpler standard analog PID
controller.

6. Controller and valve actions

6.1 Types of controller action
The controller “action” defines the direction of change in controller output
in response to a change in the process input. This action must be deter-
mined by taking into consideration the effect of all other elements in the
loop, such that the final control element (usually a control valve) changes
in the proper direction to counteract the process change.
6.1.1 There are several different ways to express controller action.
Some manufacturers define “Direct” and “Reverse” controller actions in
a different manner than other manufacturers. By convention today, con-
troller action is usually (and best) expressed as:
A process change that causes the transmitter output to increase,
vs,
The required controller output change, Increase or Decrease, that prop-
erly counteracts the transmitter output Increase.
Thus: Increase–Increase or Increase–Decrease
6.1.2 Another definition, sometimes used, is most commonly expressed
as where:
Direct: Increased process change and transmitter output causes an In-
crease in the controller output.
Reverse: Increased process change and transmitter output causes a
Decrease in controller output.
6.1.3 Many digital controllers display a “+” to indicate direct action and
“–” to indicate reverse action.

Document reaffirmed February 1991 / Document archived March 2004 6

PR38
ARCHIVED
Controller Tuning

6.2 Determining controller action

6.2.1 Figure 2 shows a typical temperature control loop which throttles
cooling water to a product cooler. The required action for the controller
can be determined as follows:
• If the incoming hot product temperature increases, the output signal
from 3TT will increase.
• To counteract the increased incoming temperature: since the valve is
air-to-close, and it must now open to provide more cooling, its signal
must decrease.
• Since the increase in temperature requires the valve signal to
decrease, the controller action must be “increase-decrease.”
Figure 2. Product cooler temperature control

6.2.2 A similar analysis can be made for any loop to determine the re-
quired controller action. Note, however, there occasional variations in
design that reverse the logic, and these must be taken into account
when determining controller action. An increase in the process variable
may result in a decrease in the transmitter output. Reverse-acting trans-
mitters, and intentional reversal of transmitter process connections are
examples. Refer to DR130S, Fail Safe Design of Process Control Loops,
for a discussion of this. See also 6.4.1.
6.2.3 An inverter function may be intentionally configured within the
controller, on either the process or output signal, to reverse the signal.

Document reaffirmed February 1991 / Document archived March 2004 7

PR38
ARCHIVED
Controller Tuning

6.3 Controller action—cascade loops

6.3.1 Select correct controller action for primary controller
While in cascade control, the output signal from the primary controller is
the set point for the secondary controller. In a manner very similar to that
in 6.2.1, the primary controller action is selected, such that its output
will change with a change in the process signal, in a direction which
will counteract the process change.
6.3.2 Referring to Figure 6, assume something causes the vessel tem-
perature to increase, thus causing an increase in the signal from the
temperature transmitter. Since the secondary controller is controlling the
steam flow (heating) to the jacket, an increase in its set point should
cause an increase in the temperature in the vessel. Thus it follows
that an increase in the process signal requires a decrease in the primary
controller output signal, thus decreasing the steam flow. The required
controller action is increase-decrease.

6.4 Valve action

6.4.1 A control valve may control heating or cooling fluids. A valve may
add fluid to a vessel or, take it from a vessel. Water may be used for
heating, if its temperature is higher than the process material. The effect
of throttling the valve, on the process, must be known to properly select
controller action.
6.4.2 Valve action is normally determined by the perceived safe action
when the air supply fails. Typically, a spring in the actuator is expected
to cause the valve to move open (air-to-close or fail-open) or to close
(air-to-open, or fail-closed).
6.4.3 A reversing relay may be used in the signal to the valve, or the
valve may have a reverse-acting positioner.
6.4.4 If any one of the above signal-reversing features were present
in the simple loop shown on Figure 2, the analysis of controller action
would result in the controller action being increase-increase. But note
that if the logic is inverted twice, the action would then be increase-
decrease again.

6.5 Split-range valves

6.5.1 Analysis of split-range, multiple valve systems follows the same
logic as above, to determine controller action. These valves traditionally
have had full-stroke spans of 3 to 9 or 9 to 15 psig.

Document reaffirmed February 1991 / Document archived March 2004 8

PR38
ARCHIVED
Controller Tuning

6.5.2 The “split-range” systems above require use of nonstandard

positioners; valves with ranges different from the standard, nominal
3 to 15 psig (or 4 to 20 madc). With a digital controller, all of the “two-
valve, split-range” functions can be implemented in the controller, thus
allowing use of standard, full range, calibrations for the valve positioners.
This reduces valve cost, and makes maintenance much easier. The con-
troller output can be split internally into two output signals, with each
signal scaled and/or inverted as required, to be functionally identical to
systems with nonstandard valve positioner ranges. This scheme pro-
vides more accurate valve positioning than the split-range positioner
spans, and should be used where practical.
6.5.3 Pneumatic or electronic signal relays are often used to get split
range action while using the full range calibration for the positioners.
6.5.4 See 8.5 for typical split-range applications.

7. Controller tuning procedures—all modes

7.1 Most control loops are tuned by experimental trial and error during
operation. This is the most certain method to use because the process is
the “perfect” model of itself. In order to avoid waste of time and effort it is
necessary to be methodical and logical. See precautions in 3.1 for safety
considerations.
7.2 For cascade control loops, that is, where a secondary or “slave” con-
troller receives a set point from a primary or “master” controller it is nec-
essary to tune the secondary controller first for good response, then tune
the primary controller.

7.3 Upset generation

See precautions in 3.1. Some processes are continuously subject to up-
sets. For these, there may be no need to generate an upset. For others,
or to determine response at different operating conditions, it may be nec-
essary to cause an upset. For the older pneumatic controllers simply
blocking an air vent, for a moment, with a finger caused the controller
output to increase quickly. With later controllers a set point change is
an upset. However, with many modern controllers a set point change is
treated differently than a measured variable change: the output will ramp
smoothly at the reset rate without any gain response. (A similar situation
occurs with the integral-only controller, but for these, the output always
changes at the reset rate.) The controller can be switched to manual, the
output changed, then switched back to automatic. For controllers with
automatic tracking of set point in manual operation the set point may
have to be adjusted back to the original value. In every case it will be
necessary to understand the equipment in use in order to properly,
and safely, tune the controls.

Document reaffirmed February 1991 / Document archived March 2004 9

PR38
ARCHIVED
Controller Tuning

7.4 These procedures apply to the most commonly used controller algo-
rithms, including various combinations of gain (proportional), reset, and
derivative functions. It does not include information on “integral only”
(floating) controller adjustments. Digital controllers may include adjust-
able parameters beyond these three functions.
7.4.1 Adjustments should be made in the sequence given below.
7.4.2 Some digital controllers include an adjustable, low-pass filter in the
input signal block. A typical adjustment range is from 0.001 to 10 Hz.
The filter is effective in removing noise or spikes occurring on otherwise
normal signals. Most loops will operate properly at a value of 10 Hz. Fil-
ter settings of too low a frequency may conceal the true process perfor-
mance and result in slow response to upsets.
7.4.3 Some digital controllers also permit adjustment of derivative gain
(a value which exists, but is not obvious, and is fixed in amplitude, in
analog controllers). A typical range is from 1 to 30 (gain). Operation at a
value of 10 will provide derivative response, at a given derivative time
setting, quite similar to earlier analog controllers. (Derivative gain values
in analog controllers typically vary from 6 to 10.)

7.5 Gain adjustment procedure

a. Switch to manual control, if available. Let the process stabilize and
settle out at an acceptable value. Remove the effect of reset and de-
rivative. If there is no direct way to switch these off, then set reset to
minimum response, that is, a large number of minutes/repeat or a
small number of repeats/minute. Set derivative to a small number of
minutes.
b. Set the gain at a reasonable value based on Figure 1 “Trial Settings,”
or a knowledge of the system. Cycling is less likely with a relatively
low gain or high proportional band. However see also the discussion
of cycling at too low a gain under reset adjustment, see 7.6.1
c. Switch to Automatic Control.
d. Observe operation. If the process remains stable, then make a set
point change (occasionally, some other variable can be changed to
test the system). Another scheme is to make a relatively large set
point change (20–50 percent) for just long enough for the valve to
change position and then return the set point to the original value.
This creates an upset but retains the original set point. Some trial
and error for the amount and time of this upset may be needed.

Document reaffirmed February 1991 / Document archived March 2004 10

PR38
ARCHIVED
Controller Tuning

e. Observe the controller output change and the response of the pro-
cess. See Figures 3, 4, and 5. If the process cycles, reduce the gain
to one-half of the original value, and observe the result. If the process
goes out of control, switch to manual, set the valve at the appropriate
value as determined above and wait for the system to settle out. Then
reduce the gain and try again. If the process is unresponsive and
shows little change, then increase the gain by a factor of two. Keep
increasing gain until the process cycles, then reduce gain to one-half.
Note the time, in minutes, between successive cycles of the process.
f. There is normally little value in making small changes in controller
tuning settings. A typical control loop will have substantial changes
in process gain and the control valve gain over the range of opera-
tion. There is also little value in attempting to determine the exact
best gain setting before the reset and derivative are nearly at the final
settings. See Figures 3, 4, and 5 for examples of idealized response
at different gain settings.

7.6 Reset adjustment procedure

a. The initial reset setting, in minutes per repeat, should be equal to the
“natural period” of the process. This is the time, in minutes, between
successive cycles of the process, as determined in “e” above. Set the
reset to this value. If the period is not determined, then set reset to
typical values shown on Figure 1, for the type of application, or near-
est similar.
b. Observe the result, or make a set point change as above. Observe
if the change in reset action results in improved control or degraded
control. If the loop starts to cycle then the reset action is too great and
the reset time should be doubled or the repeats/minute cut in half.
Cycling caused by excessive reset will be at a much lower frequency
than cycling caused by excessive gain. Gain has the control effect of
moving the control valve immediately in response to process upsets.
Reset has the effect of more slowly changing the control valve in re-
sponse to a sustained difference between set point and measured
variable. Continuing reset action will gradually eliminate the offset. If
a very small error is more important than fast response to upset, then
it may be desirable to increase the reset effect and decrease the gain
to get both a strong reset action and retain stability.
7.6.1 It is also necessary to realize that if reset is present, and in many
controllers there is no way to completely eliminate it, and if the process
has capacitance, that is, is a level control or pressure control then two
“integrating blocks” are present and the system can be unstable if the
gain is too low as well as if gain is too high. This must be considered a
possibility if a particular loop seems to act poorly. See a control theory
book for more explanation.

Document reaffirmed February 1991 / Document archived March 2004 11

 PR38
ARCHIVED
Controller Tuning

7.6.2 In a very few situations, a poor valve positioner, combined with a

spring and diaphragm actuator will have an “integrator” effect with a time
constant near that of another integrator in the loop and can reduce loop
stability. In this case the positioner should either be replaced with a bet-
ter one (faster, higher gain), or the positioner removed and a volume
booster relay installed to speed up the valve. This situation is rare.

Figure 3. Gain too low, proportional Figure 4. Optimum gain, Figure 5. Gain too high, proportional
band too wide proportional band band too low

Document reaffirmed February 1991 / Document archived March 2004 12

 PR38
ARCHIVED
Controller Tuning

Figure 6. Reactor temperature cascade control

7.7 Derivative procedure

7.7.1 The most common use of derivative with regulator-action control
is to compensate for the effect of a time lag in the measurement. This
means that the derivative time should be set to the measurement time
constant. Another use of derivative is to reduce overshoot after a set
point change, for servo-action control. Derivative provides a stabilizing
effect on a control loop. However, it should not be used where the mea-
sured variable is noisy, that is, where there is a considerable variation in
the process signal at a frequency faster than the output device and con-
trol loop can react. Derivative, unless preceded by a good low pass filter
will amplify the noise without improving control. A longer derivative time
setting (larger number) is an increase in the derivative effect. Start with
an estimated or predicted value and observe the result. If the derivative
time is set too long, then after a set point change, as the measured vari-
able approaches the set point, the controller will act too soon and the
process variable will stop moving directly towards the set point, and
will move only at the reset rate. Repeat the set point changes, tuning
changes, and observations until satisfactory operation is seen.
7.7.2 Some controllers have adjustable derivative gain, see 7.4.4.
7.7.3 An “Inverse Derivative Relay” is an analog device which acts like
a low pass filter. This is used to reduce wear in control valves. A similar
function is available in the digital devices. The smoothing should be only
enough to take the noise out of the output signals. Excessive filtering will
cause the loop action to be sluggish in the face of upsets.

Document reaffirmed February 1991 / Document archived March 2004 13

PR38
ARCHIVED
Controller Tuning

8. Applications
8.1 General
8.1.1 The operating goal should be known for each control loop. For ex-
ample, when tuning a loop to 1/4 ratio decay, see Figure 7, the expected
result usually is tight control for upsets due to changes in the process.
The process is held as close to the set point as possible. This is called
“regulator-action’’ type control. The valve or other manipulated variable
may make wide excursions to accomplish this, and in many cases this
is the desired performance.
8.1.2 In some cases this tight control is not the desired action. It may
be more desirable to achieve a smoothing or averaging effect on fluid
flow to, or from the vessel, see 8.3.1.
Figure 7. One-quarter ratio decay cycle

8.1.3 A different criteria is used for good performance in the face of set
point changes, as in batch processes. This has been called the “servo-
action’’ type control. Also, batch control loops often will benefit from
use of a batch switch, see 8.8.

8.2 Flow
See Figure 1 for trial settings. Generally flowmeters have considerable
noise on the signal and low gain, fast reset is suitable.

Document reaffirmed February 1991 / Document archived March 2004 14

PR38
ARCHIVED
Controller Tuning

8.3 Level
8.3.1 For receiver level control, it is often desirable to allow the level to
vary over a wide range, to minimize flow changes to upstream or down-
stream process equipment.
8.3.2 For example, a particular level might be allowed to vary between
20 percent and 80 percent, with no harm to the receiver operation. For
this case, a proportional-only controller (no reset) is a good choice, and
this tuning can be calculated. The allowable excursion is 60 percent of
scale, 80 minus 20. The gain that will fully stroke the valve over this level
range is 1/0.6, equal to 1.67 gain units. With set point at 50 percent, the
valve will be just open at one allowable limit, and just closed at the other.
Note: typical of the sort of required understanding of the details of the
equipment is that to implement this scheme in the Moore 352 requires
(1) both the set point and manual reset blocks not track in Manual, and
(2) the manual reset value be set to 50 percent, so the controller output
is 50 percent when set point and process signals are equal.
8.3.3 Also available are various types of nonlinear control. The level
could be allowed to move freely with some reset effect but no propor-
tional effect until certain limits are reached, then, at the limits, the pro-
portional gain takes effect. In controllers where reset is developed as a
function of gain, reset effects will be very small when gain is low because
limits have not been exceeded. For the case of gap control without reset,
experience with this sort of control is that the process tends to “hang”
near one limit or the other, and that the expected smoothing does not
occur. Flow remains constant until the level reaches one of the limits
and then changes in response to the controller gain.
8.3.4 Another variation on the idea is to use a conventional Pl controller,
tuned for low gain to get good smoothing, but also install another con-
troller with proportional-only action and relatively high gain to enforce
greater valve action as the level approaches the limit of operation. Typi-
cally this other controller is a high-gain-plus-bias relay. Signal selecting
relays then determine which signal goes to the valve.
8.3.5 Other schemes have been used. For example, the “error squared”
algorithm has very low gain action near the set point but increases rap-
idly as the error increases.
8.3.6 A “gap” action controller has low or no gain for small values of
error and higher gain when the error exceeds the preset limits.
8.3.7 Note also, that the above discussion on averaging control could
be applied to some pressure control loops.

8.4 Temperature
Temperature is the one application which most often uses derivative
action. It is most typically used to offset the time lag in the temperature
measurement.

Document reaffirmed February 1991 / Document archived March 2004 15

PR38
ARCHIVED
Controller Tuning

8.5 Pressure
8.5.1 In vessel pressure control the control system typically either vents
to relieve excess pressure, or adds gas (nitrogen, air, etc) to keep the
pressure up to the set point. These systems use two control valves in
some variation of a split-range design. See 6.5.2. Use of digital control-
lers can allow use of standard, rather than split-range valves, and that
approach is recommended where practical.
8.5.2 Probably the most common variation of split-range systems is, for
fail-safe design, to have the vent valve fail open on loss of instrument air,
and the gas valve fail close. In this case, the vent valve is air-to-close,
with a split-range of 3 to 9 psig. The gas valve is air-to-open, ranged 9 to
15 psig. The controller is increase-decrease; with its output at 9 psig
both valves are closed. Thus an increase in the vessel pressure causes
the transmitter output to increase, and the controller output to decrease.
This in turn causes the gas valve to further close (if operating conditions
are such that it is adding gas to the vessel), or, the vent valve to further
open (if operating conditions are such that the vessel is venting at the
time).
8.5.3 A less common variation on split-range systems, is a fail-safe
requirement that the vessel be “bottled-up” on instrument air failure; vent
valve closed, full pressurizing gas supply pressure on the vessel. A con-
troller action analysis, as in 5.2.1, indicates the controller is increase-
increase; the vent valve is 9–15 A.O.; the gas valve is 3–9 A.C. Another
situation would call for the two valves supplying heating and cooling to a
vessel to act differently to controller output signals and for positioner air
supply. It is possible, with a reverse acting positioner (increasing input
signal/decreasing valve opening) to close on increasing signal, but also
close on loss of supply air. Thus, switching off the positioner supply air
will shut both valves.

8.6 Ratio
8.6.1 Ratio control has commonly been done by taking a fraction of the
primary flow signal, say 70 percent, and applying that signal as set point
to the secondary controller, rather than dividing one flow signal by
the other to get the actual ratio value, and using the value as PV. See
Figure 8. The reason for this is twofold:
• While it is very desirable to use the actual ratio as the controlled and
displayed variable, see Figure 9, the pneumatic analog divider relays
once used are notoriously inaccurate at low signal inputs, and drift
with time. (However, various special instruments have been used
successfully, and newer electronic and digital computing relays are
far better.)

Document reaffirmed February 1991 / Document archived March 2004 16

PR38
ARCHIVED
Controller Tuning

• The ratio value, calculated by dividing one flow by the other, is non-
linear. When the denominator is small, then a small change in the
(denominator) flow rate makes a large change in the calculated ratio.
When the denominator is large, the opposite is true. Restated, the
“process gain” varies significantly with different operating conditions.
Thus, controller tuning parameters properly adjusted for one set of
operating conditions may be unsuitable for another.
8.6.2 The PID adaptive gain digital controller provides a solution for both
problems listed above. See Figure 10.
• The ratio value can be calculated in the controller, digitally. This
makes an accurate ratio value available for both control and display,
and eliminates the “analog drift.”
• The adaptive gain (AG) signal to controller changes the controller
gain value in response to the designated input. When configured so
the AG block has as its input the “denominator” flow signal, the con-
troller gain is very low at low flow values, and increases as the flow
increases. Thus the system is essentially linearized, and the tuning
parameters need not be changed as operating conditions change.
8.6.3 It may also be possible to rearrange the signal path and accom-
plish ratio control with a multiplying block. This avoids the problems
involved in any dividing device. (This is the equivalent of 8.6.1 above;
the actual ratio value is not present in this scheme.)
8.6.4 Other means are available in many digital controllers to linearize
control responses. For example, the signal exit the divider block could be
routed through a characterizer block, and thence to the controller block.
See Figure 11.

8.7 Floating control (ID)

8.7.1 When a controller using the ID (integral-derivative) control algo-
rithm experiences a step input in its process signal, its output does not
immediately change: there is no “proportional response” to the change
in input. Rather, its output starts changing, usually slowly, in a direction
which will cause the process signal to return to set point. The rate of the
output change is governed by the reset time setting.

Document reaffirmed February 1991 / Document archived March 2004 17

PR38
ARCHIVED
Controller Tuning

Figure 8. Constant gain on process signal

Figure 9. Variable gain on process signal

Document reaffirmed February 1991 / Document archived March 2004 18

PR38
ARCHIVED
Controller Tuning

Figure 10. Adaptive gain used to stabilize control

Figure 11. Characterizer to linearize divider

Document reaffirmed February 1991 / Document archived March 2004 19

PR38
ARCHIVED
Controller Tuning

8.7.2 The ID control algorithm is not appropriate for and is not normally
used for ordinary transmitter valve-controller control loops. Further, con-
troller response as shown on Figure 7 is not normally obtained with ID
control. It is essential that the required operation of the process be com-
pletely understood before attempting tuning. For example, similar to the
receiver level control described in 8.3.1, it is possible that the quickest
return of process to set point is the worst strategy to meet the process
requirements.

8.8 Batch switch

8.8.1 If a controller with reset is unable to return the process to set
point, the controller output will eventually be driven, by reset action, to
either its minimum or maximum value. For an ordinary flow loop, the con-
troller action most likely will be increase-decrease, see 6.6.1, and the
valve air-to-open. For this situation, if the process stays below set point,
the controller output will end up at more than 100 percent. This situation
is called “reset windup.” It most frequently occurs in batch operation,
where the control valve is forced closed by the batch sequencer, when
some process parameter is satisfied, and stays closed until the next
batch is run.
8.8.2 The problem with reset windup and batch operation surfaces when
the valve is allowed to reopen. At this instant, with a flow loop, the pro-
cess value will be zero, the set point at whatever value the batch pro-
grammer generates, and with standard 2 or 3 mode controllers the
controller output will be greater than 100 percent. Thus the A.O. control
valve will go wide open, and the process will start to rapidly increase.
The controller output will not change, and the valve will stay wide open,
until the process reaches and passes the set point. Thus a large over-
shoot occurs, and the process may cycle excessively until it eventually
stabilizes at the set point.
8.8.3 The reset windup problem may be eliminated by use of a “batch
switch” function in the controller (available in many analog pneumatic
or electronic and digital controllers). This switch modifies the controller
feedback signal. Its effect is that proportional action occurs, and the
valve loading starts to decrease, long before the process reaches set
point. In fact, the batch switch can cause sluggish start-up, and this
problem then can be eliminated by use of available bias or preload that
modifies the batch switch action. Note that batch switch action is entirely
different from either proportional or integral output limit functions.

9. Definitions
9.1 Automatic control system (control loop)—A control system which
measures a variable and operates to correct or control the deviation from
a selected reference.

Document reaffirmed February 1991 / Document archived March 2004 20

PR38
ARCHIVED
Controller Tuning

9.2 Controller—The device which compares the set point (SP), the de-
sired value, to the process variable (PV), and develops an output signal
(CO) to go to the final control element. The controller may be a physical
device or could be all or part of the program code in a computer. (The
Moore 352 controller is a computer; so are the “multi-loop controllers”
which simultaneously run many control algorithms and output to many
control valves.)
9.3 Control loop—The various parts used to control one variable, con-
sisting of a measuring device, a decision device, and a final element.
9.4 Controller mode—The manner in which a controller reacts to a
change in the measured variable, assuming a constant set point.
9.5 Controller tuning—The adjustment of the various controller modes
to produce acceptable control.
9.6 Cycling—This refers to a sustained oscillation in the measured vari-
able caused either by the control loop or some external influence.
9.7 Critical damped—A system in which the measured variable returns
to set point, after upset or load change, in the shortest practical time
without overshoot or undershoot.
9.8 Damping—Damping is an aspect of control loop dynamics which
tends to reduce cycling. It is usually accomplished by reducing the gain
of the system.
9.9 Direct action—Or Direct acting, an increase in PV causes an in-
crease in controller output. However, it has also been defined as exactly
the opposite. The only safe term to use is Increase(PV)/lncrease(CO). In
some digital systems the sense of action is expressed as positive or neg-
ative gain.
9.10 Derivative—Also called rate action, is the response of the control-
ler to the rate of change in the process variable. In some physical imple-
mentations this effect is also seen for changes in set point.
9.11 Derivative time—The time associated with derivative or rate ac-
tion; the time that the controller acts “before” it would without derivative.
9.12 Deviation—The difference between the set point and the mea-
sured variable.
9.13 Final control element—The device which changes the process in
response to the controller output signal. This is often a control valve.

Document reaffirmed February 1991 / Document archived March 2004 21

PR38
ARCHIVED
Controller Tuning

9.14 Gain—The ratio of the change in output due to proportional action

to the change in input signal for an amplifying, or attenuating device.
See also Proportional Band. Every device in a control loop has gain. The
transmitter converts the process variable (flow, temperature, etc) to
the standardized signal. Thus 5°C change could be expressed as 2 psi
change. A transmitter with a range twice as wide would have a 1 psi
change. The control valve, responding to a 1 psi change in controller
output signal might change a temperature 5 degrees. A larger valve
would have changed the temperature 15 degrees for 1 psi change in sig-
nal. Thus we can define the gain of a control valve in process terms or in
flow terms. The controller is (usually) the most convenient place to adjust
loop gain. The second most convenient place is often the transmitter
range. The control valve size is usually set by the need to pass a certain
flow at certain conditions and cannot easily be changed, unless the
valve is oversized. The change in controller output caused by the gain
or proportional term is GAIN times the difference between set point
and process variable.
9.15 Offset—A sustained difference between SP and PV.
9.16 Overdamped—Describes a system which responds to a change in
set point without overshoot, but also slower than optimum.
9.17 Overshoot—In response to a load change or upset, the controller
will move process value back towards set point. If the process value
goes past set point, it is called overshoot.
9.18 Proportional band—This phrase was used in earlier literature and
will be found as a calibration scale on older instruments. Proportional
Band (PB) = 100 percent/GAIN, or GAIN = 100 percent/PB.
9.19 Reverse action—See Direct Action.
9.20 Reset—Also called Integral. Where the change in controller output
is proportional to the duration and magnitude of the offset between set
point and process value. Adjustment scale units are in Repeats/Minute.
This is the time for the effect of initial proportional action to be duplicated
again (repeated). The reset adjustment dial on some controllers are
marked with the inverse, minutes per repeat.
9.21 Transmitter—The device which converts the signal from a sensor
to a standard process control signal, e.g., 3–15 psig, 4–20 MADC, or
other. The sensor may be included in the transmitter, or may be external,
for example, the transmitter could be a resistance to current converter
and the sensor a RTD. Another device, called a “primary element” may
also be involved, for example, an orifice plate.
9.22 Valve loading—The output signal from a controller, or the input
signal to the control valve.
9.23 Underdamped—Describes a system which responds to a change
by overshooting.

Document reaffirmed February 1991 / Document archived March 2004 22

PR38
ARCHIVED
Controller Tuning

10. References
1. Shunta, Joseph P., “The Calculation of Controller Tuning Parame-
ters,” Accession Report No. 16853 (May 1984).
2. St. Clair, D. W., “Controller Tuning and Control Loop Performance,”
Accession Report No. 16699 (April 1983).
3. Buckley, P. S., “Techniques of Process Control,” John Wiley & Sons
(1964).

Document reaffirmed February 1991 / Document archived March 2004 23