Complete PID Manual PDF

Section 1.
Basic control
considerations
Regulatory Control and PID loop
tuning
© 2019 Crabtree Controls Ltd. All rights reserved 1
Learning Objectives
 Identify some of the components that contribute to

process lag
 Review the basics of process reaction
 Describe first and second order process response
 Recall the problems associated with digital sampling
 Explain the difference between inherent and installed
valve characteristics
 Discuss the sources of the dead band and hysteresis
 Review the sources of electrical noise and its
mitigation
1
Basic elements
 One of the major problems in regard to the field of PID

control is a complete lack of standardisation.
 There are none.
 None in regard to the configuration of the controller
 None regarding the actual working principle
 None regarding the terminology used by different
manufacturers.
 This is not going to change!
Basic elements
The four basic elements of any control system comprise:

 process;
 transducer (sensing element plus transmitter);
 final control element; and
 controller.
2
Basic elements
Controller
Process
Transducer (sensor + transmitter) Final control element
Process lag
Opening the valve …

… will cause the level to start to rise immediately.
3
Process lag
 However, new value only reached some time later

 Time is dependent upon the size and shape of the tank
and the flow through the valve.
 This process lag is primarily the result of three process
characteristics:
 capacitance,
 resistance, and
 dead time.
Capacitance
 Capacitance of a process is a measure of its ability to

hold energy
 Likened to inertia.
 Should not be confused with capacity
 Two processes with the same capacity might have very
different capacitances.
4
Capacity and capacitance
 Consider two tanks – both having the same liquid

volume capacity of 32 k (32 m3).
 However, if both tanks are emptied at the same
constant flow rate, then the level of Tank B will fall at
twice the rate ofA tank A.
Tank Tank B
B 4m
A 2m
Capacity = 32 m3 Capacity = 32 m3
 In other words, Tank A has twice the liquid volume

capacitance with respect to the liquid level, as in B, i.e.
16 k/m versus 8 k/m.
Tank A Tank B
B 4m
A 2m
Capacity = 32 m3 Capacity = 32 m3
5
 Paraffin in Tank A requires about 4kJ of heat energy to raise

the temperature 1ºC
 Water in tank B requires nearly double the heat energy (7.7
kJ) to raise the temperature 1ºC
 Tank B has nearly twice the thermal capacitance of Tank A.
Tank A Tank B
(32 m3) (32 m3)
Paraffin Water
Heat energy required = 4 kJ Heat energy required = 7.7 kJ
Resistance
 Heat transfer cannot be instantaneous and heat energy

must be applied for a given amount of time before there is a
given change in temperature through the wall of the piping.
 This is termed the ‘resistance’ of the process
Fluid in Steam in
Steam out Fluid out
6
Resistance
 Resistance may be likened to the ‘opposition to flow’

 Determined by a number of factors – including the
thickness of the piping and the thermal resistance of the
process medium.
Fluid in Steam in
Steam out Fluid out
Dead time
 The dead time, or transport lag, is the time taken for the
material, discharged from the loading hopper to be detected
by the belt weigher located some distance away.
Loading hopper
Hopper feed
actuator
Material travel
Conveyor belt
Belt weigher
Transport lag = 10 min
7
Dead time
 Deadtime is often given the symbol ()

 Deadtime introduces more difficulties in automatic
control than any of the other lags
 Every effort should be made to keep it to a minimum.
 In many applications it is possible to measure the
process closer to the actuation point.
 Thermocouples are often located on the basis of access
convenience rather than process to optimisation and can
be moved closer.
Definition of terms
 Controlled variable is called the Process Demand

(PD) or Manipulated Variable (MV) or simply the OP.
 The process variable is also sometimes referred to as
the Measured Variable (MV).
Process Process
Variable (PV) Demand (PD)
Process
Transducer (sensor + transmitter) Final control element
8
Process reaction
 If PD subject to a step change…

 …. by how much will the PV change?
? PV PD
Process
Process reaction
 Process Gain given by:
Δ PV %
KP 
Δ PD%
Process
9
Process reaction
 Assume a step change in the PD of 10%

 If PV also changes by 10% then the Process Gain (KP) is 1
 And if the PV only changes by 5% then the Process Gain
(KP) is 0.5
 And if the PV changes by 20% then the Process Gain (KP)
is 2
PV PD
Process
Process reaction
 Generally, the process gain for a self-regulating process

should lie between the 0.5 and 2.0.
 Why?
 In practice KP is determined by two factors:
» the valve size and
» the transmitter span.
 Ideally we’d like to use the full span of both devices.
 The valve, being a mechanical device, has a very limited
rangeability and may suffer from problems such as
hysteresis and backlash.
10
Process reaction
 Therefore if the valve were adjusted from 0-100% of the full

open position, and the measurement also responded from
0 - 100%, then the process gain would be unity, which is
ideal.
 As a general rule of thumb, if the process gain is greater
than unity the valve is oversized, and …
 …if it is less than unity, it is usually due to the transmitter
span being too wide.
 One disadvantage of using an oversized valve is that in
order to achieve a desired response, the proportional gain
in the controller has to be correspondingly reduced.
 This may lead to problems on controllers having limited
low-end gains.
Setting span too high
● Setting the transmitter span too wide is common practice

and, initially, results from uncertainty regarding the process.
● If the temperature needs to be controlled at 550ºC, and
normally ranges over 450 to 650ºC under manual control,
then the transmitter span need only cover that range.
● Setting the span to cover say 300 to 700 ºC means that the
process gain is halved.
(PV) (PD)
Process
11
Process reaction
 One disadvantage of using an oversized valve is that in

order to achieve a desired response, the proportional
gain in the controller has to be correspondingly reduced.
 This may lead to problems on controllers having limited
low-end gains.
 Typically, many commercial controllers are limited to a
minimum gain of 0.1 (or 1000% proportional band).
 If the valve is oversized it may not be possible to set a
low enough gain in the controller for use on fast
processes (such as flow) that are tuned with low gain
and fast integral. the valve is oversized.
Process reaction
 If a transmitter is spanned too wide, resulting in a process

gain of less than unity, problems can occur with the quality
of the measurement.
 For example if the process gain were 0.5, it would mean
that only the first 50% of the transmitter span was being
used.
 If this was on, say, an orifice flow it would severely limit the
useful flow range and measurement accuracy.
 Secondly a low process gain also necessitates having to
increase the controller proportional gain to achieve a
particular control response.
 This makes the output of the controller more ‘jumpy’, and
will also further amplify signal noise – resulting in increased
valve wear and possibly increased control variance.
12
Process reaction
 Again it should be noticed that the process gain will never

be completely linear.
 Nonlinearity is a function of the final control element, the
process, and the transducer.
 Generally, however, most nonlinearity will be due to the
valve itself.
 That is, the process gain will, typically, vary as a function of
the valve operation.
Process reaction
 If PD subject to 30% step change…

 ...how will the PV react?
? PV PD
Process
13
First order lag
As the liquid level (h) increases,

the outflow Q2 also increases
until the outflow equals the inflow
Q1 – at which point the level will
have reached a steady state level
(H).
First order lag
The rate of change of level is proportional to the difference

between the inflow and outflow divided by the cross-
sectional area of the tank:
dh Q1  Q 2

dt A
dh
A  Q1  Q2
dt
Making the simplified assumption that the outflow (Q2) is
proportional to the level (h) then:
Q2  k  h
where k is a constant determined by the resistance R of the
valve (V2).
14
First order lag
Substituting for Q2:

dh
A  Q1  k  h
dt
dh
A  k  h  Q1
dt
A dh Q
 h  1
k dt k
h  Q1 1 ε kt 
replacing k by 1/ where () is time constant:
  
t
h  Q1 1 ε  
 
First order lag
• Ball-cock valve system

• (a) the tank is empty and the
H valve is fully open
• (b) as the water level rises
(a) towards the maximum level
(H), the ball float rises and
starts to close the valve.
H
h
(b)
15
1st-order response
Assuming a time constant () of 1, application of this

equation results in:
1.0
0.8
 h
0.6 1 0.632
h 2 0.865
0.4
3 0.950
0.2 4 0.982
5 0.993
0
0 1 2 3 4 5

1st-order response
 Classic first-order lag response in which level (h) is plotted

against time (T).
 If the tank had continued to rise at its initial rate it would fill
the tank in the time defined by time-constant (1 minute).
1.0
0.8
0.6
h
0.4
0.2
0
0 1 2 3 4 5
t
16
1st-order response
 The time-constant () is defined by the time required for

the variable to reach 63.2% of its final value.
 The time required to reach 99% of the final value is
usually given as 5 x  (e.g. in this example 5 minutes).
1.0
0.8
0.6
h
0.4
0.2
0
0 1 2 3 4 5
t
1st-order response
 In practice, an idealised first order differential lag is rarely

encountered
 Typically, a number of such lags are combined in series.
 A number of tanks (each representing a first order lag)
arranged in series in which the size of each tank (and
therefore the time constant ) is progressively decreased.
/3
/2
/3

/2 /3
2 3
1
17
1st-order response
 As the number of lags in series is increased, the response

to a step-input change moves away from the familiar first
order lag and becomes more and more ‘S’ shaped.
1.0
0.8
0.6
1
h
0.4
2
3
0.2
0 1 2 3 4 5
0
Time constants
1st-order response
This ‘S’ curve response, when combined with deadtime,

gives rise to this typical response.
18
Self-regulating flow control
 Self regulating: step change in manual results in change of PV

that steadies out at new steady-state
 Flow loops are very fast and have very little capacitance
 PV can be very noisy unless flow meter is installed properly
 Flow meter should be placed upstream of valve
PV MV
FT
Process reaction curve
 Here, the process exhibits a very rapid increase from zero

to a new steady-state condition with virtually no deadtime
 Typical of a flow control loop
PD = 30%
PV
19
Self-regulating small volume pressure control
 Again – very fast and has very little capacitance

 PV can again be very noisy
MV PV
PT
Self-regulating temperature control
MV PV
Steam
TT
Process medium
Condensate
20
Batch end-point control
MV PV
Steam
TT
Process medium
 Integrating process
Condensate
Integrating level control
Flow in
Output Input
LT
Process medium
 Integrating process
21
Boiler tank response
 First problem relates to level measurement.

 As shown, would the level indicator work?
Steam out
Fire stack
Level
indicator
Steam
Fire box V1
Water in
 When the water in the boiler is cold the gauge shows the
same level as the internal level.
 However with the presence of bubbles the internal level will
be much higher.
Steam Steam
Water Difference
level in levels Water with
Water bubbles
22
 Assume system is in equilibrium.

 What happens to the level if steam demand is increased?
Steam out
Fire stack
Steam
Level
indicator
Fire box V1
Water in
Second Order Response
 A second order response is analogous to the control

element containing a spring or damper
 So, instead of the equation being of this order:
A dh + h = H
K dt
 It looks more like this:
d2x + 2b dx + 2 = a
dt2 dt
 Where:
a = step input
b = damping factor
 = natural frequency of resonance
23
Second Order Response
Curve’s characteristic shape determined by:

 natural frequency () at which the system will oscillate
 damping factor (r) – how quickly the oscillations die out.
1.2
1.0
0.8
0.6
h H
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10
t
Under-damped
 If the damping factor (r) is less than 1 the system is under-

damped and shows a rapid change to the new value
 But overshoots the steady-state value and produces a
number of oscillations.
1.2
1.0
0.8
0.6
h H
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10
t
24
Critically damped
 If the damping factor is equal to 1 the system is critically

damped
 Characterized by a fast rise without overshoot.
1.2
1.0
0.8
0.6
h H
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10
t
Over-damped
 If damping factor is greater than 1 the system is over-

damped
 Characterized by a slow rise without overshoot.
1.2
1.0
0.8
0.6
h H
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10
t
25
Dead band plus under-damped 2nd order response
 Dead time is a condition of no response for a given time interval

 It shows up on the reaction curve as a straight line – delaying
whatever other lags are present.
1.2
1.0
0.8
0.6
h H
0.4
0.2
Deadtime
0
0 1 2 3 4 5 6 7 8 9 10
t
Process lag
 Comparison of the effect of various values of damping

factors
1.2 r = 0.5
r = 0.7
r = 0.4
1.0
0.8
r = 1.5
r=1
0.6
h H
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10
t
26
Digital sampling
● Assume a sinusoidal PV having a periodicity of 5 s

● And assume grab samples are taken twice each second…
● …i.e. the sampling rate is 10 times the frequency of the analog
input
● Here, the reconstructed signal is a reasonable representation of
original.
5s Analog input
Reconstructed signal
Digital sampling
● If the analog signal is sampled only once each second… …

● …i.e. the sampling rate is now ony 5 times the frequency of the
sampled signal…
● … the digitised version is still representative of the input
frequency.
5s Analog input
27
Digital sampling
● However, if the analog signal is under-sampled e.g only once 4

seconds…
● …the reconstructed signal can result in a waveform whose
frequency bears and no relationship to the original analog signal.
● This is called aliasing
5s Analog input
Digital sampling
● And if the PV is sampled at the same rate as the analog input

signal…
● …the reconstructed signal would appear to be a constant d.c.
value.
5s Analog input
28
Aliasing
● There are several steps we can take to avoid aliasing.

● The most obvious is to ensure that the analog input is sampled
at a rate of at least twice the highest frequency present in the
signal.
● This is called the Nyquist limit.
● In practice it is customary to use a sampling rate that is at least
five times the highest frequency component present in the
signal.
● Because the sampling rate is usually fixed at some maximum
value …
● …the solution may lie in ensuring that no high frequency
components are present in the signal that could be aliased.
Anti-aliasing filter
● A low-pass ‘anti-aliasing’ filter is often located within the

transmitter to remove any frequency components above the
Nyquist frequency (half the sampling frequency).
Ideal response
1.0
Signal attenuation
Practical response
-3 dB point
Frequency
29
Process reaction masking
● Here, the PV filter receives the PV signal, reduces the high frequency
noise and passes it on to the controller
Process reaction masking
● Because the filter takes some

time to react, its output
responds much less to high
frequency signals than to low
frequency signals.
● In other words, the high
frequencies (like noise) are
filtered out while the low
frequencies (like process
variable changes) are passed
through
30
Dangers of filtering
● This is the response of a noisy process variable…

● … as a result of a 10% change in setpoint…
● …followed by a 10% step-change in load.
10% setpoint change
10% step load change

● The response to the setpoint change is an overshoot of

250%...
● …whilst the response to the 10% load change indicates
that the controller has brought the process back to the
setpoint with little overshoot.
250% overshoot
on SP change
PV responds well on
10% setpoint change load change
10% step load change

31
● In order to reduce the noise level, filtering has been applied to the
PV.
● The filtered PV seen by the operator shows no sign of overshoot
and the load change reaction shows only a small offset.
Filtered PV
response
Correct placement of sensors
● Use bare-wire thermocouples where possible to reduce

thermal capacity.
● Insulated thermocouples and thermowells can
considerably increase the thermal capacity of the system.
● When using thermowells make use of a thermo-paste to
reduce thermal lag.
● Position the sensor as a close to the actuation point as
possible.
32
Correct placement of sensors
● In flow measurement, position the flow sensor upstream

of the final controller element.
● Most flow measuring instruments have a better tolerance
to obstructions downstream of the sensor than upstream.
● For example, a typical magnetic flowmeter can tolerate an
upstream obstruction of five pipe diameters but only
requires two pipe diameters downstream.
Measuring the wrong parameter
● Would this system work?

● Measuring the wrong process variable results in feed-
forward control and drift.
Controller
33
Control Valve Characterisation
Inherent valve characteristics
 The characteristics of the valve are determined by the

physical internal shaping
 Commonly called the inherent valve characteristics
 Sometimes termed manufactured characteristics
 Determined with a constant pressure drop across the valve
34
100
Quick Characteristic curves
opening
80 ● linear,
Percentage flow
Linear
● equal percentage,
60
● quick opening
40
20
Equal
percentage
20 40 60 80 100
Percentage stroke
Linear characteristic
● Simply a straight line
100
relationship between flow
80 and stroke at constant
pressure drop.
Percentage flow
Linear
60
40
20
20 40 60 80 100
Percentage stroke
35
Equal percentage
● equal increments of rated travel
100
will give equal percentage
80 changes of existing flow.
Percentage flow
● e.g. every 10 % change in

60
stroke the flow might produce a
40
change 100 %..
● at a low stroke of less than 50
20
Equal %, the flow is low relative to the
percentage total available flow.
20 40 60 80 100
Percentage stroke ● This would provide good
throttling capability due to small
gain of flow relative to stroke
Quick opening
● The quick opening
100
Quick characteristic produces rapid
80
opening increases in flow with small
stroke increments.
Percentage flow
60
● very close to maximum
40 available flow is achieved at
approximately 30 to 40 % of
20 stroke.
● This form of characteristic is
20 40 60 80 100 most often applied where
Percentage stroke
throttling is not the primary
concern.
36
Characteristic profiling
Plug outlines used to obtain:

● (a) equal percentage
● (b) linear and
● (c) quick-opening flow characteristics.
(a) (b) (c)
Characteristic profiling
For cage-guided trim, flow characterisation is

determined by the shape of windows in the
cylindrical cage
● (a) equal percentage
● (b) linear and
● (c) quick-opening flow characteristics.
(a) (b) (c)

37
Installed characteristics
 The inherent flow characteristics are determined at a

constant pressure drop.
 In practice as the flow rate increases other portions of
the process piping take up some of the available system
pressure drop
Pressure drop Pressure drop
across piping across valve
 Hence, as the flow rate increases, the pressure drop across

the valve decreases
 Characterised by the Pressure Drop Ratio (PDR)

 Given by:
P valve
PR 
P system
 A PDR of 1.00 indicates that the valve takes all of the
available pressure drop, at all positions, and there is no
drop across the system
 This indicates a relatively small valve in a very large line.
Pressure drop Pressure drop

across piping across valve
38
 As the PDR becomes smaller there is an increasingly

larger pressure drop across the system.
 Usual practice is to size valve and other components
such that the valve takes 25 to 50% of the available
system pressure drop when it is wide open
 This corresponds to a PDR of 0.25 and 0.5, respectively
Installed linear flow characteristic
100
80
Percentage flow
60
40
P valve
PR 
P system
20
20 40 60 80 100
Percentage stroke
39
Installed equal percentage flow characteristic
100
P valve
80 PR 
P system
Percentage flow
60
40
20
20 40 60 80 100
Percentage stroke
A good rule of thumb is:

 use Linear when the pressure drop across the valve is a
large proportion of the total pressure drop.
 use Equal Percent if in doubt.
 use Linear for level control.
 use Equal Percent for pressure control.
40
Deadband
 Deadband is an area of a signal

range or band where no action
C
occurs.
 Gear teeth with backlash exhibit
deadband.
Output
 There is no drive from the input

to the output shaft from A to B …
 …and only after the backlash
A B has been taken up will there be
a drive from the input to the
Input
output (B to C).
Deadband
 On reversal there will be no drive

from the input to the output shaft
D C from C to D …
 …and only after the backlash has
been taken up will there be a drive
Output
Deadband from the input to the output (D to

A).
 Dead band can usually be
attributed to valve friction (packing,
A B
actuator seal rings, and valve seal
Input rings) and pneumatic dead zones
in the instrumentation.
41
Dead band
 A well-engineered valve should respond to signals of 1% or less

to provide effective reduction in process variability.
 However, it is not uncommon for some valves to exhibit dead
bands as great as 5% or more.
 In a recent plant audit, 30% of the valves had dead bands in
excess of 4%.
 Over 65% of the loops audited had dead bands greater than
2%.
Hysterisis
 As the mass of the weight is

increased, then, according
to Hooke’s law of elasticity,
the extension of the spring
(x) would be directly
proportion to the applied
load (M) or, in simple terms,
strain is directly
proportional to stress.
42
Hysteresis
 In reality, the spring is unlikely to

obey Hooke’s law perfectly and
Extension (x)
instead the relationship would be

non-linear – with the extension
increasing as the load is
Load increased.
increasing
Load (M)
Hysteresis
 However, as the spring is unloaded

each weight that produced a
specific extension as the load
increased will now produce a slightly
Load
Extension (x)
longer length.
decreasing
 Hysteresis is thus a path dependent
characteristic that can be attributed
Load to materials not being able to return
increasing to their original shape and size after
being stretched or deformed.
 In most pure electronic systems,
Load (M)
hysterisis can be effectively ignored
and thus the deadband and
hysterisis are one and the same.
43
Hysteretic error
 Hysteresis is often displayed as

the hysteretic error.
Output
Hysteretic error
Input
Deadband + Hysteresis
 For control valves, dead band

is the dominant nonlinearity
whereas hysteresis is a
secondary effect.
 Allowable dead band plus
Output
Hysteretic error hysteresis is typically specified

to be ≤ 1% for the entire valve
assembly ̶ comprising the
valve, actuator, and positioner.
Deadband
Input
44
Testing for hysteresis/deadband
Equal input
step changes
Valve stem
position indication % offset in valve
position
Testing for hysteresis/deadband
FINAL CONTROL ELEMENT MAX. ACCEPTABLE

Spring and diaphragm 3%
Spring and diaphragm with positioner 1%
Piston with positioner 1%
Variable speed drive 1%
45
Sources of hysteresis/deadband
● Undersized actuator
● Seal friction in rotary valves
● Friction in plug valves
● Inadequate air supply
● Loose or worn linkages in actuator connector
● Defective or improper calibration of I/P converter
● Loose or worn linkages in positioner connector
● Defective or improper calibration of positioner
Undersized actuator
● On reversal valve is effectively stuck
● Valve positioner starts to force excess air pressure into
the actuator
● Once it moves and the static friction is overcome,
overshoot occurs
● Result is likely to be continuous cycling
46
Undersized actuator
Equal input
step changes
Valve stem
position indication Overshoot indicates
negative hysterisis
Stick-Slip without positioner
Slip Hysteresis
Stick
47
Testing for non-linearity
50.0
37.5
PV
25.0
12.5
PD
0.0
Testing for non-linearity
140 kPa 2 140 kPa 4

20 -100 kPa
1 3
Positioner
PD I/P 5 Mechanical
linkage
4 – 20 mA converter 20 -100 kPa
Valve stem
4 position
4 - 20 mA transmitter
48
Exercise G
WHAT YOU SHOULD HAVE OBSERVED IN THIS EXERCISE:
● Without a positioner, stem friction can cause valve sticking –
limiting the resolution of valve travel
● This can be detrimental in any type of feedback loop.
● With a positioner, the effect of stem friction is largely overcome
in the steady state.
● However, a dynamic effect is produced, approximating that of a
relatively high frequency, damped second order system.
● If the valve is in a slow loop (e.g., temperature), this is probably
of no consequence.
● However, if this is in a fast loop, such as a flow loop, this in itself
may have a detrimental effect on loop behavior.
Signature analysis
• The spikes at each end of the curve verify that a solid stop
has been reached at both ends of travel.
• The top stop is often used as the reference point when
establishing benchset.
160
Top stop
140
Actuator pressure (kPa)
120 2 x Friction Slope

100
Closed
80
60
40
20
0 Bottom stop
10 0 10 20 30
Travel (mm)
49
Signature analysis
• The separation of these lines is the result of the friction

band.
• The higher the friction the wider the separation.
160
Top stop
140
120 2 x Friction Slope

100
Closed
80
60
40
20
0 Bottom stop
10 0 10 20 30
Travel (mm)
Signature analysis
• The huge increase in friction as the valve travels towards

the open position is probably due to some form of galling..
160
140
120
100
Closed
80
60
40
20
0
10 0 10 20 30
Travel (mm)
50
Footprint torque characteristic
Galled
seat
Torque demand
Stem requires
lubrication
Packing
too tight
Footprint
torque profile
Closed Valve Position Open
Definition of Noise
● Defined as any undesirable electrical signal which

distorts or interferes with an original or desired electrical
signal.
● One Man’s Signal is another man’s Noise...
51
Electrical Coupling of Noise
● Impedance (or Conductance) Coupling

● Electrostatic (or Capacitive) Coupling
● Magnetic (or Inductive) Coupling
Impedance Coupling
Unbalanced lines
A A
B B
C C
Finite impedance develops voltage drop

that is coupled into the other inputs
52
Impedance Coupling
Balanced Differential lines
A Balanced differential lines A
B Balanced differential lines B
C Balanced differential lines C
Capacitive and Inductive coupling
 Capacitive coupling is a result of high Voltage

interference
 Inductive coupling is the result of high Current
interference
 Both high voltages and high currents are usually present
at the same time but …
…. their treatment must be considered separately
53
Capacitive coupling
 Capacitive coupling – the result of high Voltage interference
High voltage source (dV/dt)
Capacitive coupling
to the signal lines
C1 C2
R1 R3
I1
Transducer RL
I2
R2 R4
Capacitive coupling
 Shield provides a lower impedance path to ground
High voltage source (dV/dt)
i
C
R1 C3 R3
i
I1
Transducer RL
I2
R2 C4 R4
The Shield must be grounded to

provide a return current path
ONLY ONE GROUND
54
Shielding performance
Capacitive Coupling
 Copper braid (85% coverage) 100:1 (20 dB)

 Low resistance multi-layered screen 3200:1 (35 dB)
 Aluminium mylar tape 6300:1 (38 dB)
Inductive (magnetic) coupling
● Inductive (Magnetic) Coupling is the most frequent cause

of interference in industrial signal circuits
● Also ... the most difficult form of interference to deal with
● Main problems are at Power Frequency (or harmonics)
 Coupled into the control and communication circuits
 Mainly from power cables, conduits, cable trays, etc
running near to or parallel to the communication cables
55
Inductive coupling
● The current in conductor A

produces a magnetic field
 A portion links with the adjacent
A conductor B
 This induces a voltage in
conductor B
B
 The induced voltage is directly
proportional to the rate of change
of the magnetic flux
Inductive coupling
 Inductive coupling – the result of high Current interference
High current source (di/dt) produces: magnetic flux
Transducer RL
Induced voltages
56
Equivalent circuit for inductive coupling
● The induced voltages depend on :

 Rate of change of current (di/dt) in conductor current
source
» "Area" of signal loop
» Distance between signal loop and current source (D)
High current source (di/dt)
D Vinduced
D
Transducer RL
Vinduced
Inductive coupling
 In reality flux linkages will be set up along the length of

the signal loop
Transducer RL
57
Use of twisted cable
 The Voltage induced in the adjacent twists tends to be

equal and opposite … and sums to zero
Transducer RL
Performance ratios
 10 twists/m (Category 3) 14:1
 40 twists/m (Category 5e) 140:1
Low permeability shielding
● Diverts the magnetic field using a magnetic material
Steel Steel Wire

conduit Armour
Area of low Area of low

magnetic field magnetic field
Signal cable Signal cable
Magnetic field Magnetic field
58
Magnetic shielding
● The high current source produces a flux

 This induces eddy currents in the shield
 With both ends grounded there will be a circulating current
Transducer RL
Magnetic shielding
● According to Lenz's law the induced current will have a

direction that opposes the changes in the flux linkage
Transducer RL
59
Inductive shielding
● In practice, good magnetic shielding is achieved by :
 using a balanced connection with twisted pairs
 running signal lines in steel conduit or SWA
 conduit or SWA grounded at both ends
 inductive shield should NOT be part of the signal path
Conclusions
● If source of disturbance (e.g. power cable) is itself

enclosed in SWA grounded at both ends, induced current
flow in SWA will reduce the overall magnetic field outside
the power cable
● Power cables on steel cable tray or ladder, grounded at
both ends, will have a low overall magnetic field
● If a signal cable is run in a non-metallic (e.g. plastic)
conduit, a separate shield wire grounded at both ends,
should be run with the signal cable
60
Practical recommendations
Remove the causes of interference

» Segregation of cables
» Power and comms cables on completely different routes
» Divert large surges to ground
Shield the electronic circuits from interference
» Ground capacitive shields at one end
» Ground magnetic shields at both ends
» Multi-channel cables should have separately shielded pairs
with overall shield
» Exposed sections of twisted pair as short as possible
» Shield continuity should be maintained
Reduce the effect of coupled noise and interference
» Use balanced differential connections for communications
» Use twisted pairs wherever possible for signal cables
Cable spacing
IEEE 518-1982
 Level 1. Analog signal less than 50 V
 Level 2. Analog signal greater than 50 V
 Level 3. Switching signal greater than 50 V and up to 20 A.
 Level 4. Switching signal up to 1000 V and up to 800 A.
61
Covers
● Trays manufactured of
metal
● Earthed with full continuity
for full length
● Fully covered
Cable spacing
Both in separate tray

 Level 1/4 670 mm.
One in tray and one in conduit
 Level 1/4 460 mm.
Both in separate conduit

 Level 1/4 310 mm.
62
Do’s and don’ts
● Do use a high-quality twisted pair screened cable

● Do ground the screen at one point only.
● Do ensure that you use a dedicated cable tray system for
instrumentation cables.
● Do use galvanized metal trays.
● Do replace covers.
● Do ensure that the cable tray is grounded at regular intervals
(yes, that’s right – more than once).
● Do look at a minimum separation distances between high
current/voltage bearing cables.
Do’s and don’ts
● Do run high current/ voltage bearing cables in a separate

conduit.
● Do use galvanized steel conduit in preference to aluminium.
● Do ground the conduit at regular intervals.
● Do run high current bearing cables at right angles to your
signal cables.
● Do use fibre-optic cable in high noise environments.
63
Do’s and don’ts
● Do not run a high current/ voltage carrying cable in the same

tray as your instrument cables – even if there is a divider
between them.
● Do not rely on filters to sort your problems out.
● The use of a low-pass filter to smooth out a noisy process
variable signal, and thus reduce noise, has been used for many
years.
● Reliance on such a solution, however, gives rise to two major
problems:
 Aliasing; and
 Reaction masking.
64
Section 2. PID control
tuning
Open Loop Control

 Control action is not a function of PV
 No self correction when PV drifts
 Control based on measured disturbances
 Example:
Speed = f (A)
A = accelerator Speed = Process
position Variable
 Control Action = Position of Accelerator

 Control result = Speed of Car
 BUT: Speed could vary with changing load
1
PID control
 The term PID refers to the three adjustments that are to

be found on the controller:
 Proportional,
 Integral
 Derivative.
 PID control has been around since about 1940 and has
its roots embedded in pneumatics – with several of the terms
dating back to this period.
 The PID controller remains the most widely used method
of controlling a closed-loop industrial process – accounting
for more than 95% of all systems.
PID control
 PID control has survived longer than many expected.

 Many expected advances like:
 fuzzy logic,
 dynamic matrix theory,
 model predictive theory and,
 neural networks
 …to render this relic of the 1930s and ’40s obsolete.
 However PID did not go away...
 …and the ability to tune PID loops is a core competency.
 However, universities and engineering schools have tended
to concentrate on the next generation of modern theory-based
control technology (which has never really arrived)…
 … dramatically reducing the time spent on the real world of
PID loops and tuning.
2
ON/OFF control
Steam in
V1
SP
PV ERR On/off Output Actuator

A  controller
V2
Product in
Thermocouple Steam
out
Product
out
ON/OFF control
Open Open Open
Valve
Closed Closed Closed
160ºC
+10ºC
Temperature (ºC)
Off Off Off

150ºC Setpoint
On On
-10ºC
140ºC
20ºC
Time
3
ON/OFF control
● The controller Output can have only two values:

● Output (Max) and Output (Min) – dependant on whether
the error (ERR) is positive or negative.
Output (Max)
Output (Min)
_ 0 +
Error (ERR)
ON/OFF control with deadband
● Deadband — “the range through which an input can

be varied without initiating a response”.
Open Open Open
Valve
Closed Closed Closed
160ºC
Temperature (ºC)
+10ºC
Off Off Off
150ºC Setpoint
6ºC Dead-band
144ºC
On On
-10ºC
134ºC
20ºC
Time
4
ON/OFF control
 In On/Off control there are only two possible controller

outputs – no matter the value of the error signal
 Thus process is forced to oscillate.
 In order to avoid forced oscillations it is necessary to
reduce the gain for small values of ERR.
 This is achieved through the use of proportional control
 The procedure is simple:
 measure the process variable (PV);
 compare the PV with the target value or setpoint (SP)
 use the difference (the error signal, usually
abbreviated to ERR) to determine the output (OP) of
the controller
Proportional control
SP
Summing block
ERR = SP - PV Controller Reverse action
 KC
OP
PV
PD
Thermocouple
Steam in
Process fluid in Process fluid out
Steam out
5
 The error signal acts on the proportional controller

having an amplifier with gain KC.
 The OP which determines the process demand (PD) is
thus the error times the gain.
 OP = ERR * KC
Output
(Max)
Output
(Min)
_ +
Error (ERR)
 When the temperature is

Setpoint
150ºC low the large error signal
ensures that the valve is
Temperature
fully open
 As the temperature
increases, the controller
20ºC
reduces the deviation signal,
Time which progressively closes
100% the control valve.
 This action continues
Error signal
until, ideally, the process

temperature slowly and
smoothly reaches the
setpoint, whereupon the
0%
Time deviation signal is zero and
the control valve is fully
closed.
6
Proportional offset
+SP 150
25 Controller OP = 45%
- KC
PV 125
Proportional offset
+SP 150 OP = 48%
25 Controller OP = 45%
- KC
135 15
PV 125
7
Proportional offset
● The process temperature will not reach the setpoint

temperature.
● Instead it will assume a lower value of 125 ºC in order to
produce a deviation signal of sufficient magnitude to
maintain the valve in a position…
● …that keeps the temperature at 125ºC
● This difference, due to the effect of the load, is called
proportional offset.
Proportional offset
● How big an error signal is required will depend largely on

the gain (KC) of the controller.
● One solution lies in increasing the controller gain KC.
● This means that the error of 25ºC will now produce a
larger OP.
● This, in turn will increase the PV and thus reduce the
error signal.
8
Proportional offset
Thus, the fundamental concept behind a proportional

controller is that the output of the controller is proportional
to the difference between the set point and the control point.
Reducing proportional
offset
Setpoint
150°
K=8
C
K=4
Process variable
K=1
20°
C
Time
Proportional offset
New set
180ºC
point
Proportional
offset
Set point
150ºC 150ºC
Proportional
offset
125ºC
Time Time
(a) (b)
9
Proportional offset
● Can we live with this problem?

● In some cases yes.
● Proportional control can be used by itself with stable 1st
order systems if small constant steady state errors can
be tolerated.
● Typically this might arise where the sensor inaccuracy is
large or even unimportant as might happen in liquid level
control where only an approximate level value is needed
for correct operation.
● In addition, an error can often be tolerated in the inner
loop of a cascaded control system.
SP
Manual bias
ERR = SP - PV Controller
 KC
control
OP
PV
PD
Error
Thermocouple
Steam in
Steam out
10
Proportional offset
● The manual bias or manual reset control acts on the final

control element in the form of a d.c. bias or offset voltage.
● In this manner the reset control is adjusted to offset the
output to a point where the error signal is zero i.e. until the
valve is 50 % open.
● Under stable conditions, this is the setting required to
maintain the temperature at the setpoint of 150C.
● The amount of offset required does not have to be actually
calculated.
● It only requires that the operator adjust the manual reset
control slowly, allowing the process loop to adapt to the
changing conditions, until the deviation signal, as indicated
on the meter, is zero.
Manual reset
● What is the OUPUT?

 So…the OUPUT = KC* (SP-PV) + MR
SP
ERR = SP-PV ERR* KC MR
 KC
PV
OUTPUT
Process
11
Proportional band
● Proportional Band or PB
● Change in transmitter signal (expressed as percentage of
its range) that is required to produce a 100% change in the
controller output
● Proportional Band is given by:
PB = 100/Kc
Proportional band
● Proportional Band is the range through which the PV must travel..

● ….to cause the controller output to change by 100%.
100
80
PV span (%)
60
40
20
0
0 20 40 60 80 100
Controller output (%)
12
Proportional band
● So, if PV travels 100%...

● … to give a controller output change of 100%...
● …the Proportional Band is 100%
100
80
PV span (%)
PB = 100%
60
40
20
0
0 20 40 60 80 100
Proportional band
● But, if PV only travels 60%...

100
80
PV span (%)
60 PB = 60%
40
20
0
0 20 40 60 80 100
13
Proportional band
● And, if PV only travels 20%...

100
80
PV span (%)
60 PB = 20%
40
20
0
0 20 40 60 80 100
Proportional band
KC =0.2
PB = 500%
100
KC =1
PB = 100%
80
KC =5
PV span (%)
60 PB = 20%
KC = 
40 PB = 0%
20
0
0 20 40 60 80 100
Valve opening (%)
14
Automatic reset
 An automatic system should carry out the same operation

as the manual system
 While an ERR continues to exist the system should
increment (or decrement) the PD until the ERR signal is
reduced to zero.
 At this point the SP and the PV are the same.
 Automatic reset is implemented using integral action
Input Integral block Integral
Integral action
8 16
7 14
6 12
5 10
Input
4 8 Integral
(number of
squares
3 6 per unit
time)
Input
2 4
1 2
0
0 1 2 3 4 5 6 7 Time8
15
Integral action
8 16
7 14
6 12
5 10
Input Integral
4 8
(number of
squares per
3 6 unit time)
2 4
Input
1 2
0 Time
0 1 2 3 4 5 6 7 8
Integral action
6 12
5 10
4 8
3 6
Input Integral
2 4 (number of
squares per
2 unit time)
1
0 0
-2 -2
-4 -4 Time
0 1 2 3 4 5 6 7 8
16
Integral
SP I
ERR Controller
 KC
OP
PV
PD
Error
Thermocouple
Steam in
Steam out
Integral action
● Here, Integral action is expressed in time.
0.67 min
1 min
2 min
10% integral
change
Valve position
Long
10%
proportional
change
Time
17
Integral action
● Here, Integral action is expressed as how far it has increased

(or decreased) in 1 minute
0.67 min
1½ repeats/min
1 min
2 min
1 repeat/min
10% integral
change ½ repeat/min
Valve position
10% 1 minute
proportional
change
Time
Integral action
Proportional output
Step error input

K
 Resultant output
I
Integral output
18
Integral action
● The terms Integral time and Repeat time can be confusing.

● Integral time and Repeat time is expressed in similar
fashion:
 integral time: minutes/repeat or seconds/repeat
 repeat time: repeats/minute or repeats/second
● Rarely will the controller indicate which parameter is being
used and the user, therefore, has to refer to the handbook.
● Foundation Fieldbus has decided to settle on integral time
expressed in seconds per repeat.
● PI control is used for a wide range of process applications
but is generally confined to fast self-regulating systems
exhibiting a first-order characteristic with little deadtime.
● Notably these would be typified in the control of flow and
low-capacity pressure systems.
Manual Reset
ERR = SP – PV
OP = KC* ERR + MR
SP
ERR MR
 KC 
PV
OP
19
Integral windup
● If an error signal (ERR) persists for some time, the

Integral term (and therefore the Output) will continue to
ramp up (or down) for as long as the error exists.
● In most cases feedback action ensures that the error
signal is, in fact, reduced.
● However, if a continuous error exists which cannot be
eliminated, then the output would continue to increase or
decrease until the final control element reaches its
physical limit of fully opened or fully closed.
Integral windup
● In a pneumatic or electronic analog controller, the
integrator output will continue to ramp until a physical limit
such as air supply pressure or a supply voltage, was
reached.
● At this the integrator is 'saturated' or fully 'wound up'.
● However, in a digital computer where there is,
theoretically, virtually no upper or lower limit, the
integrator could carry on until the output reaches a value
close to infinity.
20
Integral windup
SP
 P + I controller
PV
Isolating valve Control valve
● If the isolating valve closes the PV will fall to zero.

● This will produce a continuous error signal that acts on the
integral term to ramp it up towards infinity
● The output itself saturates at its upper limit – typically 20 mA.
Integral windup
 PV falling to zero
results in continuous  Proportional output
error (ERR)
K
 Output saturates at
SP   high limit (e.g.20 mA)
I Integral ramps
up towards
infinity
 PV falls to zero
when isolating valve
closes
21
Integral windup
● So, we’ve now arrived at a situation where the valve is

fully opened and the output is saturated.
● But, never forget that the integral term is still ramping up.
● What happens when the isolating valve is opened?
Integral windup
 PV rising to maximum
results in continuous  Proportional output
negative error (ERR)
K
 Output remains
SP   saturated at its high
limit of, e.g.20 mA
I Integral ramps down

from its previously high
Remember Integral value
 When isolating valve has been ramping up
opens PV rises to its all the time that the
maximum – determined by isolating valve was
the control valve closed.
22
Integral windup
Main control
valve ON
OP
SP
PV
Integral windup
Main control
valve OFF
OP
SP
Anti-windup
PV
23
Integral windup
Manufacturers also make use of a number of proprietary

solutions for dealing with windup. These include:
● Conditional Integration.
If the controller output is saturated, and input and output are of the
same sign, then set the integrator input to zero.
● Limited Integration.
If the integrator is saturated, and controller input is of the same sign,
then set the integrator input to zero.
● Tracking Anti-Windup.
If the controller output is saturated, reduce the integrator input by some
constant, times the difference between unsaturated and saturated
controller output.
Integral windup
● Generally these measures work with only varying

degrees of success and seldom eliminate the
problem.
● Consequently it is better to try and ensure that windup
does not occur in the first place.
● One of the most common forms of implementation is
when the controller output hits a limit, the proportional
band is forced in the other direction.
● The objective is that, on the next start-up or control
recovery, there will be a significant reduction in the
overshoot.
24
Controller Gain = Kc
SP
ERR Controller
 KC
OP
PV
Process Gain = Kp PD
Steam in
Steam out
Stability
SP +
- ERR
 Controller
Kc = 0.5
PV Loop gain = 1 OP
Kp = 2
25
Stability
SP +
- ERR
 Controller
Kc =10
PV Loop gain = 1 OP
Kp = ?
Stability
SP Disturbance ERR Disturbance

+2 +2
+
-
 Controller
Kc = 0.5 OP Disturbance
-1
Kp = 2
PV Disturbance
-2
26
Stability
SP Disturbance ERR Disturbance

+2 +2
+
-
 Controller
Kc = 0.5 OP Disturbance
-1
180° phase
shift
Kp = 2
PV Disturbance
+2
Stability
SP +
- ERR OP
 Kc
PV
Kp
ERR = SP - PV ERR = SP - Kp * OP
ERR = SP - Kp * Kc * ERR ERR + Kp * Kc * ERR = SP
SP
ERR (1 + Kp * Kc) = SP ERR =
(1 + Kp * Kc)
27
Considerations for Critical frequency
Total process loop phase shift of 180°
 Gain = 1 ............ continuous oscillation

 Gain < 1 ............ signal decays
 Gain > 1 ............ increasing oscillation
Demonstration. (15 min)
 Demonstration of instability using Proportional only

 (Set Reset Off)
28
Testing frequency response
Input Output
Process
/X/ /Y/ 
/Y/
/X/ Input
Output

Bode plot
8
6
Since 0 db corresponds to a gain
magnitude of 1, the shaded area lying
4
above 0dB on the upper graph is
2 potentially unstable.
Gain (dB)
Frequency (Hz)
0
0.02 0.04 0.1 0.2 0.4 1 2 4
-2
-4
-6
-8
90
120
Phase lag ()
150
Frequency (Hz)
180
0.02 0.04 0.1 0.2 0.4 1 2 4
210
240
270
29
Bode plot
8
6
4
2
Gain (dB)
Frequency (Hz)
0
0.02 0.04 0.1 0.2 0.4 1 2 4
-2
-4
-6
-8
Here, the shaded area lying at less
90 than 180° is potentially stable.
120
Phase lag ()
150
Frequency (Hz)
180
0.02 0.04 0.1 0.2 0.4 1 2 4
210
240
270
Bode plot
8
The locations at which stability and instability occur,
6 respectively, are termed:
4
2
Gain (dB)
Frequency (Hz)
0
0.02 0.04 0.1 0.2 0.4 1 2 4
-2
-4 Gain crossover
-6
-8
90
120 Phase crossover
Phase lag ()
150
Frequency (Hz)
180
0.02 0.04 0.1 0.2 0.4 1 2 4
210
240
270
30
Bode plot
8
When the gain falls to unity the phase lag is:
6
4
2
Gain (dB)
Frequency (Hz)
0
0.02 0.04 0.1 0.2 0.4 1 2 4
-2
-4
-6
-8
90
120
Phase lag ()
150
Frequency (Hz)
180
0.02 0.04 0.1 0.2 0.4 1 2 4
210
240
270
Bode plot
8
And when the phase lag is 180º the gain is only:
6
4
2
Gain (dB)
Frequency (Hz)
0
0.02 0.04 0.1 0.2 0.4 1 2 4
-2
-4
-6
-8
90
120
Phase lag ()
150
Frequency (Hz)
180
0.02 0.04 0.1 0.2 0.4 1 2 4
210
240
270
31
Bode plot
8 Phase
6 margin 90
Phase
4 crossover 120
Phase lag ()

2 150
Gain (dB)
Frequency (Hz)
0 180
0.02 0.04 0.1 0.2 0.4 1 2 4
-2 210
Gain
-4 crossover Gain 240
-6 margin 270
-8
Exercise K. Bode plot (45 min)
 Carry out Exercise 1: Section 5.0 ‘Frequency response’
Period Freq Cont PV Ampli Time Phase

(Min) (cyc/min) Output Ampli Ratio Lag (Deg
60 0.0167 10.0 12.7% 1.27 8.0 48

30 0.0333 10.0 8.65% 0.865 7.5 90
15 0.0667 20.0 7.3% 0.365 5.5 132
7.5 0.1333 20.0 2.44% 0.122 3.75 180
3.75 0.2667 20.0 0.54% 0.027 2.25 216
32
Polar plot
Frequency Magnitude Phase angle

(Hz) (absolute) (°)
0.06 1.57 -19
0.18 1.42 -40
0.36 1.23 -61
0.61 1.06 -80
0.97 0.89 -100
1.1 0.76 -121
1.2 0.66 -140
1.3 0.57 -160
1.8 0.50 -180
2.5 0.44 -200
3.4 0.39 -220
3.6 0.35 -242
Polar plot
1.4
1.2
Imaginary axis (90)
1.0
0.8
0.6
0.4
0.2 Real axis (0)

180 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 190

-0.2
1.57
-0.4
-0.6 0.06 Hz
-0.8
-1.0
(a)
-1.2
-1.4 270
33
Polar plot
1.4
1.2
1.0
0.8
0.6
0.4

180 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2

-0.2
-0.4 400
-0.6
1.42
-0.8
-1.0 0.18 Hz
-1.2
-1.4 270
Polar plot
1.4
1.2
1.0
0.8
0.6
0.4

180 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2

-0.2
-0.4 610
-0.6
-0.8 1.23
-1.0
-1.2 0.36 Hz
-1.4 270
34
Polar plot
1.4
1.2
1.0
0.8
0.6
0.4

180 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2

-0.2
-0.4
800
-0.6
-0.8
1.06
-1.0
-1.2 0.61 Hz
-1.4 270
Nyquist diagram
1.4
1.2
1.0
0.8
0.6
3.6 Hz 0.4
3.4 Hz
2.5 Hz 0.2 Real axis (0)
180 1.8 Hz 0.2 0.4 0.6 0.8 1.0 1.2 1.4
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2

-0.2
1.3 Hz
1.1 Hz -0.4
0.06 Hz
-0.6 Increasing
1.1 Hz
-0.8 frequency
0.97 Hz
0.18 Hz
-1.0
0.61 Hz 0.36 Hz
-1.2
-1.4 270
35
Nyquist diagram
Circle of unity radius

1.0
Gain margin
at phase
crossover Phase
crossover
1.0
-1.0
Phase margin Gain crossover

at gain
crossover
-1.0
Nyquist diagram
1.0
Circle of
unity radius
1.0
-1.0
Unstable plot
-1.0
Stable plot
36
Effect of integral action
Proportional
output (in-phase)
Input
P

Resultant output
I
Integral output
(lags by 90°)
Proportional output
Integral output
37
Integral
output
Process
lag
Proportional output
Derivative action
Proportional
output (in-phase)
Input
P

Resultant output
D
Derivative output
(leads by 90°)
38
Derivative action
Derivative
output
Proportional output
Derivative action
Derivative
output
Proportional output
Integral output
Resultant
39
Derivative action
Integral
output
Process
lag
Derivative
output
Proportional output
Instability in Closed Loop
Process
lag
Integral
output Proportional output
40
Integral
SP I
ERR Controller
 KC
OP
PV
Derivative
D PD
Thermocouple
Steam in
Steam out
Derivative action
(B)
Error
signal
(A)
Derivative Needle
output pulse
(ideal)
Derivative Stretched
output pulse
(real)
41
PID control
Derivative output
D
Proportional output
Step error input
P 
Resultant
output
I
Integral output
Derivative action and noise
Here is a statement taken straight from a so-called authoritative

source:
● “Derivative action can be added if desired. If the process value
signal is not noisy and the derivative is on the measured PV and
not on the error, derivative action should stabilize the controller.
In general, if derivative action is added, the controller gain (KC)
can be increased and the integral term (Ti ) can be decreased.
BEWARE: derivative action often behaves poorly on
industrial processes.”
● Strictly, what they’ve said is true.
● But, it’s very misleading since it would appear to advocate that
we shouldn’t use it.
● Indeed several organizations have taken such statement so
much to heart that they actually prohibit the use of derivative
action.
42
● Earlier, it was noted that if the process lag is small it is

impossible to produce a total loop phase lag that equals
180° and become unstable.
● In general, flow control loops have very small process lags
and would, therefore, never require the application of
derivative action.

● It is important to recognize that that the derivative acts, not on the
amplitude but on the rate of change of the input – in this case the
error signal.
● The result is that derivative action amplifies noise
● Because of excessive noise on the PV signal, derivative action
act poorly on many industrial processes.
● The solution lies, not with the arbitrary prohibition of derivative
action but in applying good instrumentation practice to minimize
the noise in the first place.
● Alternatively, it is customary to incorporate a low pass adaptive
filter within the controller in order to get rid of any high frequency
noise within the system.
● Where possible, do not rely on such filtering
● Derivative is normal expressed in time setting of minutes although
the Fieldbus Foundation has settled on time expressed in
seconds.
43
Digital control
 Invariably, the PV is an analog signal.

 The conversion into a digital signal (D/A converter) requires the
analog signal is sampled at a regular intervals in order to build up
a representation of the original that is as accurate as possible.
 During each sample, separate calculations are made to
determine the proportional, integral and derivative values.
 These independent calculations are then used to increment or
decrement the existing output value.
 In other words, the calculations do not determine the absolute
value of the output but only changes that have occurred during
the sample period.
 This is called an Incremental Algorithm.
Digital control
 At time T1 the difference

OP2 between the PV and the SP will
give rise to an error ERR1.
 This ERR1 acting on the
OP1 controller KC will thus give rise
PV to an output:
Scan period KC x ERR1 = OP1.
 And at T2, the difference
ERR2 between the PV and the SP will
ERR1
give rise to an error ERR2 that
SP will act on the controller KC to
T1 T2 give an output:
KC x ERR2 = OP2.
44
Digital control
OP2  In this case, OP2 is a totally

independent calculation and is
not dependent on the previous
OP1 value OP1.
PV  In practice the application of the
Scan period Incremental Algorithm is largely
transparent to the user.
ERR2
 However, there are
consequences that the user
ERR1
SP
needs to be aware of.
T1 T2
Parallel algorithm
 This configuration is
most often used as a
P teaching model
 Often referred to as:
» ‘ideal parallel’
Error input Output – Process
(ERR) Demand (PD)
» ‘non-interacting’
I 
» ‘independent’
» ‘gain independent’
45
Parallel algorithm
From this we can deduce the following:
d (ERR)
KC ×ERR
P ERR
PD  ERR  KC K I  ERR dt  K D  dt

(ERR) Demand (PD)
 ERR dt
K I × 
ERR = Error signal
PD = Process Demand (output)
dd((ERR ) KC = Proportional gain
KDD×
dt
dt KI = Integral gain
KD = Derivative gain
Parallel algorithm
The equation is now:
 KI 
PD  ERR K C   s TD  K D  
KC
P  s TI 

(ERR) KI Demand (PD)

s TI  This algorithm might appear
ideal …
 ….except, maybe, from the
fact that each function block
KD ×
sTD has its own amplifier – adding
to the overall cost.
 In reality, the independence of
each function block leads to a
rather strange problem.
46
Parallel algorithm
 Consider the output of a parallel PI controller …
Proportional output
Step error input

P
 Resultant output
I
Integral output
Parallel algorithm
 If the contribution from the proportional block had a magnitude 10%,
then the integral time is given by the time it takes to raise the output a
further 10%.
 In other words, the contribution of the integral action has 'repeated' the
proportional contribution in T minutes.
 Here, the controller gain KC = 1 and the integral time is set such that it
will repeat the proportional action in T minutes.
Output
10% integral
step
10% proportional
step
T minutes
47
Parallel algorithm
 If the gain is halved (KC = 0.5) then for a 10% step change in error,
the proportional action would produce an output step of 5%.
 However, in a parallel algorithm the integral action would continue
ramping up at the same rate such that the proportional action is
repeated in half the time (T/2 minutes).
 In effect, therefore, when the gain is reduced, because the integral
action remains the same, it gets faster in relation to the proportional
action and the loop tends towards instability.
Output
5% integral step
5% proportional step
T/2
minutes
Parallel algorithm
The phase lag of the resultant
Integral of proportional and integral
action
action is still short of the
conditions for instability:
» a phase lag of 180 and
» a gain of equal to or
Resultant greater than 1.
Proportional output Process lag
48
Parallel algorithm
Integral
action Increasing the integral action
may cause the resultant to meet
the classical conditions for
instability.
Resultant
Parallel algorithm
Integral However, decreasing the

action proportional action may also
cause the resultant to also meet
the classical conditions for
instability.
Resultant
49
Parallel algorithm
Integral Decreasing the proportional

action
action may cause the resultant to
also meet the classical conditions
for instability.
Resultant
Parallel algorithm
Typical PV response of PI controller with a parallel algorithm to a

10% step change in the Setpoint with the proportional gain (KC = 1)
and the integral time set to 0.17 minutes.
65.0
10% Setpoint change
60.0
(engineering units)
Process Variable
PV response
55.0
KC = 1
50.0
I = 0. 17
45.0
0.0 30.0 60.0 90.0 120.0 150.0
Time (seconds)
50
Parallel algorithm
Here, the integral time remains the same at 0.17 minutes but the
proportional gain has been reduced (KC = 0.25).
65.0
60.0
(engineering units)
Process Variable
PV response
55.0
10% Setpoint change
KC = 0.25
50.0
I = 0. 17
45.0
0.0 30.0 60.0 90.0 120.0 150.0
Time (seconds)
Ideal algorithm
Also described as:
 ‘ideal parallel’,
 ‘non-interacting’,
 ‘ISA’,
 and at least one company
 ‘interacting’.
P
SP
OP
 K I 
PV
D
51
Ideal algorithm
 The P, I and D control blocks are (more or less) independent of each
other.
 The term ‘ideal non-interactive’ might better be thought of as
identifying the algorithm rather than describing it.
 The term ‘ideal’ does not imply that it is an ideal system but rather
that it is only suitable under ideal conditions where an ideal, noise-
free PV exists – not the typical input from a field sensor.
 A classical teaching model, frequently used in textbooks, this
algorithm is also used for computer optimisation and process
modelling.
Ideal algorithm
 The equation for this algorithm is now:
 1 

PD  KC  ERR 1   s TD 
 s T 
I
P
SP
OP
 K I 
PV
D
52
Ideal algorithm
 This particular arrangement has the advantage over the parallel

algorithm in that only one amplifier is used and that changes in the
gain (K) actually affect both the integral and derivative actions.
 Surprisingly this makes it increasingly non-interactive as far as
adjustments are concerned.
 1 
PD  ERR K   s TD 
 C sT
I

Ideal algorithm
Integral
action
Resultant
In the Parallel algorithm

Proportional output Process lag we saw how decreasing
the proportional action
may cause the resultant
to also meet the classical
conditions for instability.
53
Ideal algorithm
Integral
action  As the proportional gain is
varied, in this case reduced,
the integral gain is also
reduced.
Resultant
Process lag  Thus, although the magnitude

of the resultant changes, the
Proportional
output phase angle remains the same.
Series algorithm
 The origin of the series algorithm has its roots in pneumatic control
 Designers found it difficult to build parallel structures using pneumatic
components.
 With the introduction of both analog and digital systems, many
manufacturers introduced designs that were based on the series
algorithm.
KI =1 KD =1
SP
ERR OP
 KC I  D 
PV
54
Series algorithm
 The equation for this algorithm is thus:
 1  
PD  K  ERR 1    1  sTD 
 sTI  
KI =1 KD =1
SP
ERR OP
 KC I  D 
 1 
PV 1   1  sTD 
 sTI 
Series algorithm
 The theoretical output of a derivative block is a sharp needle pulse

 In older pneumatic and analog electronic controllers the response
was limited by natural lags inherent in the system.
 In practice the pulse is stretched out.
(B)
Error
signal
(A)
Derivative Needle
output pulse
(ideal)
Derivative Stretched
output pulse
(real)
55
Series algorithm
 In a digital system the maximum output value is only be limited by the

resolution of the D/A converter used within the system.
 Moreover, since a digital system is likely to make use of an
incremental algorithm, then even the smallest change from the
previous scan would be seen as a step input
 The controller would thus increment or decrement the existing output
value.
 The use of a straight derivative block would be unusable, since the
output would be continually jumping up and down between limits.
Series algorithm
 Common practice to include a low pass filter in front of the

derivative block
 This limits the excursion and stretches the pulse
KI =1 KD =1
SP
ERR Low pass OP

 KC I  filter
D 
PV
56
Series algorithm
 This becomes:
 1   1  
PD  K  ERR 1       1  sTD 
 sTI   1  sTLPF  
 1  1  sTD  
PD  K  ERR 1    
 s T I  1  sTLPF  
KI =1 KD =1
SP
ERR Low pass OP

 KC I  filter
D 
PV    
1 
1
 
1
 1  sTD 
 sTI   1  sTLPF 
Series algorithm
 It is important to appreciate that this filter is not intended to be used as
a noise filter
 If excessive noise exists on the PV then use should be made of the
proper noise filter within the system that filters the incoming PV signal.
 A serious drawback of this arrangement is that each time the
Derivative value (TD) is changed, it would be necessary to change the
filter time constant.
 Consequently, in practice, the low pass filter response TLPF is
configured so that it is directly geared to the value of the derivative
time, TD.
TLPF = TD
where:  is a constant and is usually fixed by the manufacturer at
some value between ¼ and 1/8.
 With  fixed at 1/8 , for example, the derivative component would be 8
times more powerful than the Low Pass filter.
57
Series algorithm
 1  1  sTD  
PD  K  ERR 1    
 sTI  1  sαTD  
 
 1  1  sTD  
PD  K  ERR 1    
 s T I  1  s TD  
  8  
Series algorithm
 Users should be aware that the value of  is sometimes user
adjustable and arbitrarily set to a nominal default by the manufacturer.
 This can lead to ridiculous situations.
 Michael Brown, a well known practitioner in the field of loop
optimisation, reported on a plant where the default value of  had
been set at unity.
 The derivative chain thus becomes:
 1  sTD  
 
 1  sTD  
 The filter directly cancels out the derivative term. Totally pointless!
 In another incident the manufacturer’s default had been set to 2.5.
 Now, the denominator is larger than the numerator and the
complete block acts as large lag.
58
Alternative algorithms – PID
 Here, control is based on the ERROR signal – the difference
between the SP and the PV.
 The ERROR signal itself cannot thus differentiate between a
disturbance of the PV and a change in the SP. In most processes,
changes in the PV are normally smooth.
 However the operator can make a step change to the SP resulting
in a step change in the ERROR signal – with possible major
damage
P
SP
OP
 K I 
PV
D
Alternative algorithms – PID
 This algorithm is usually referred to as PID – inferring that

the ERROR signal acts on all three components: the
Proportional, the Integral and the Derivative.
 This is also sometimes referred to as a ‘Type A equation’.
P
SP
OP
 K I 
PV
D
59
Alternative algorithms – I-PD
 In this algorithm, normal (relatively slow) fluctuations of the PV are

applied directly to the Proportional and Derivative components
 The ERROR signal acts only on the Integral.
 This implies that step changes made to the set point SP are also only
going to be applied through integral action.
SP
ERR OP
 I 
PV
K P
D
Alternative algorithms – I-PD

 This is also sometimes referred to as a ‘Type C equation’.
 The result is that even if the operator should make excessive step
changes, the OP will only ramp up or down under the action of Integral
action.
SP
ERR OP
 I 
PV
K P
D
60
Alternative algorithms – PI-D
 In any control loop where Derivative action is applied, large set point
changes can result in excessively large changes to the OP.
 This is simply the result of the Derivative action acting on the rate of
change of the ERROR step input.
 In the ‘Type B equation’, only the relatively slow PV changes are
applied to the Derivative control action.
P
SP
ERR OP
 K I 
PV
D

 In the normal PID algorithm where derivative action is base on the
ERROR signal, the system provides a faster initial take-off and will
suppress overshoot for a set point change.
 However, in the PI-D algorithm, where derivative action is based on
variations of the PV, it works against a setpoint change, since it is
unaware that the PV should be changing initially and that the brakes
should only be applied to the process as it approaches setpoint.
P
SP
ERR OP
 K I 
PV
D
61
 Again, many manufacturers rather make use of some form of setpoint

velocity limiting
 Largely unknown to users, this configuration is used as a default for
the vast majority of controllers on the market.
 In some cases manufacturers only provide this one configuration.
P
SP
ERR OP
 K I 
PV
D
Fail-to-open action
● For a reverse acting controller, as the PV drops from 100% to

0% …
● …the output changes from 0 to 100%.
● So, when PV falls to 0% the OP is 100%
100
80
PV span (%)
60
40
20
0
0 20 40 60 80 100
62
Fail-to-open action
 In this example the valve is fail-open: in the event of signal

failure to the valve, the valve opens
 Thus a 0% signal to the valve causes it to be fully open and a
100% signal causes it to be fully closed
 But, remember, with reverse action when PV falls to 0% the
OP is 100%
 Consequently, to achieve fail safe we must have direct action.
 Subsequently, in order to achieve a condition in which:
ERR= SP – PV
 We must reverse the output of the analog output function block
Fail-to-open action
SP Direct or reverse acting?

 ERR = SP-PV PID
PV
Reverse output Yes or No? AO

No: 0 – 100% = 4 – 20 mA
Yes: 0 – 100% = 20 – 4 mA
AI
I/P
PT
Process
63
Normal action
SP ERR = SP-PV Reverse acting
 100%
PID
0%
100%
PV
0%
Direct output AO
0% 4 mA
AI 100% 20 mA
I/P
4 – 20 mA 3 psi
PT 15 psi
Process
Reverse action
Direct acting 100%
SP ERR = SP-PV
 100%
PID
0%
PV
0%
Reverse output AO
100% 20 mA
AI 0% 4 mA
I/P
4 – 20 mA 15 psi
PT 3 psi
Process
64
Section 3. Loop tuning
tuning
What are we trying to achieve?
 Ideally all loops should perform optimally to capture all

economic benefits.
 However this is not economically feasible.
 Therefore, efforts should be focused on the key control loops
that directly affect the product-measured qualities.
 These include,
» yield
» purity
» cost to produce and
» up time (reliability).
 Field equipment deficiencies on key control loops should be
eliminated prior to tuning.
1
“No amount of advanced control which relies on the
use of poor field instrumentation can be expected to
yield worthwhile benefits.
Thinking of control as a hierarchy, everything must

work well at the lower levels for the higher levels above
to work.”
M.J. Oglesby (ICI Engineering Technology)
What are we trying to achieve?
Generally, the minimum performance requirements laid

down by industry are:
 Process gain > 0.5 and < 2
 Hysteresis/deadband < 2% (1% with positioner)
 Stiction < 0.5%
 Noise < 2%
 Linearity (G max/G min) < 2
If you are not meeting these requirements, any attempts to

optimise your loop by ‘twiddling’ with the knobs on your
controller are likely to be a complete waste of time.
2
Loop tuning – methods
 How many tuning methods are there?

 3?
 5?
 10?
 In fact there are more than 75 different methods.
 Why so many methods?
 Much depends on what you are trying to do!
 Are you trying to optimise on process disturbance
changes?
 Are you trying to optimise on set-point changes?
 To start with there are vast differences in the
requirements for a typical flow loop and those on a
surge level tank
Loop tuning – methods

Type of loop Characteristics Required control
Flow and liquid Fast response with no time PI controller with intermediate controller
pressure control delay gain
Usually with small high-
frequency noise
Liquid level Noisy due to splashing and High gain PI controller for integrating
control turbulence process.
Conservative setting for averaging
control when used for damping the
fluctuation of the inlet stream.
Gas pressure Usually fast and self regulating PI controller with small integral action
control (large reset time).
Temperature Wide variety of the process Use PID controller to stabilise the
control nature response.
– Usually slow response with
time delay
Composition Similar to temperature control Effectiveness of derivative action is
control usually with larger noise and limited.
more time delay
3
Possible responses
Setpoint change
Unstable:
Increasing
oscillation
Continuous
oscillation (r = 0)
Under damped
(r > 0 < 1)
Critically damped
(r = 1)
Over damped
(r > 1)
Settling time
 Theoretically the time taken for the system to settle

completely after a change in setpoint is infinite.
 However, in practice a ‘tolerance limit’ is defined which
can be expressed as a percentage of the setpoint change.
 A typical value of the tolerance limit is 5% and once the
plant is within this band it is considered to have achieved
the setpoint.
4
Quarter amplitude damping
 One of the most commonly used performance criteria is termed

‘Quarter Amplitude Damping’ (QAD) …
 …in which each cycle peak is one quarter of the amplitude of
the previous peak
A
1/4A
 Although, for many applications, this is an adequate and easily

achievable response, it is not necessarily either a desirable
response or even an optimal response.
 What’s wrong with Quarter Amplitude Damping?
 It is not safe enough – it does not allow enough gain margin for
process dynamic changes that could result instability.
 A quarter wave response consists of approximately 4 cycles.
5
 This means that the valve has to reverse eight times every time
there is a control error change.
 Is this really desirable?
 Do you really want to work your valves so hard?

 What happens if there is excessive deadband + hysterisis?
 On each reversal of the valve the controller output has to ramp
through the hysteresis band under the command of the
controller's integrator.
 This slows down the control, and increases variance.
 On slow processes such as temperature it can literally take
hours before the process finally settles out.
 This means that on many loops, quarter amplitude damped
tuning results in much worse control variance than even critical
damping, where no valve reversals at all are made on a control
error change.
6
Minimizing the area
 The basic aim is to keep the error signal to a minimum.
 Thus the shaded area is the integral of the error and could be
used as an index of performance: the smaller the area, the better
the response.
Process
Setpoint variable
+
+ + +
− − − −
Object: minimal area

Time
Minimizing the overshoot

 However, merely reducing the integral is not necessarily the
solution.
 For example, here, the integral (the shaded area) is smaller.
 Unfortunately the peaks are also considerably larger.
Process
variable
7
Minimizing the overshoot
 This may be overcome by taking the integral of the square of
the error – known as the Integral of the Square Error (ISE).
 Where:
 2
ISE   e (t ) dt
0
Process
variable
Minimizing the area
Process
Setpoint variable
+
+ + +
− − − −
Object: minimal area

Time
(a)
Integral (area under
curve)
Total area of
the error signal
Time
(b)
8
Minimizing the area
Process
Set point variable
Proportional
offset
Time
(a)
Integral (area under
curve)
Increasing
area of the
error signal
(b) Time
Before we start
 You should have some idea of what you’re trying to do and

what you can expect to do.
 You should have some idea of how fast and how far the process
is going to react to changes in the manipulated variable
 You should know and agree on how much change you will allow
in the process and in the manipulated variable.
 You should know and agree on who is going to make the actual
adjustments to the controller settings
» You should also be confident that they know how to do it
physically
» You should also be confident that they are not likely to
become confused if something has to be done in a hurry
9
Before we start
 You should know and agree on who is going to switch between

manual and automatic
 You should know and agree on who is going to make the
setpoint changes
 You should have knowledge of the safety interlocks and any
other safety concerns for the process
 You should pick a time to tune such that you will be available
for a reasonable time afterwards should problems arise
 You should record the existing settings and the controller
output in the event you want to return to them either in a hurry or
simply to the the system as you found it.
 You should make sure the changes are communicated to all
operating person.
 If a log book is kept for this purpose, use it.
Open loop response
 In many regards the open loop reaction method is the most

obvious solution.
 We introduce a step change to the Process Demand (PD) and
observe the Process Variable (PV).
 By observing the process reaction we can immediately obtain
information regarding the dynamics of the process itself.
 This makes this method very useful for troubleshooting since we
usually have some idea as to what the response should look like.
 Consequently, if it doesn’t match up to what we had expected, it
gives us some clues as to the nature of the problem.
10
Open loop response
 Critically, we can judge straight away if the process is self-

regulating.
 Further, we can also determine the process gain (KP) and the
two critical time constants involved:
 the deadtime () and
 the response time ().
 Armed with information regarding the dynamics of the process,
we are now in a position to adjust the controller parameters
(Proportional, Integral and Derivative) to produce a ‘best-fit’ tuning
solution.
Open loop response
Two things to note:

 Firstly this is a classical first order response plus dead-time (FOLPDT).
65.0
60.0
55.0
50.0
45.0
0.0 20.0 40.0 60.0 80.0 100.0
Time (seconds)
11
Open loop response
 Secondly, it is a self-regulating process in which the PV has
evened out to a new steady state value as a result of a step change
applied to the process demand (PD).
65.0
60.0
55.0
50.0
45.0
0.0 20.0 40.0 60.0 80.0 100.0
Time (seconds)
Open loop response

PD = 60 – 50 = 10% PV = 62 – 46. = 16%
KP = PV/ PD = 16/ 10 = 1.6
65.0
60.0
55.0 PD
PV
50.0
45.0
0.0 20.0 40.0 60.0 80.0 100.0
Time (seconds)
12
Open loop response
 = 10 seconds
65.0
60.0
PD
PV
50.0
45.0
0.0 20.0 40.0 60.0 80.0 100.0
 Time (seconds)
Open loop response

 = 20 seconds
65.0
60.0
PD
PV
63.2%
50.0
45.0
0.0 20.0 40.0 60.0 80.0 100.0
 
Time (seconds)
13
Open loop response
 KP = 1.6
  = 10.0 seconds
  = 20.0 seconds.
Summary of Open Loop Tuning formulae
Kp = Process gain  = Dead time Response time

P - Control Kc =
Kp 
0.9 
PI - Control Kc =
Kp 
TINT = 3.33 
PID Control Kc = 1.1 

Kp 
TINT = 2.0 
TDER = 0.5 
14
Summary of Open Loop Tuning formulae
 KP = 1.6
  = 10.0 seconds
  = 20.0 seconds.
PID Control
1.2  1.2 * 20
Kc = Kc = = 1.5
Kp  1.6 * 10
TINT = 2.0  TINT = 2.0 * 10 = 20
TDER = 0.5  TDER = 0.5 * 10 = 5
Quarter amplitude damped response

 KC = 4.9
 TINT = 9.8 seconds
65.0
60.0
(engineering units)
Process Variable
55.0
50.0
45.0
0.0 40.0 80.0 120.0 160.0 200.0
Time (seconds)
15
Lambda response
 KC = 0.8
65.0
60.0
(engineering units)
Process Variable
55.0
50.0
45.0
0.0 40.0 80.0 120.0 160.0 200.0
Time (seconds)
Cohen-Coen response
 KC = 6.1
65.0
60.0
(engineering units)
Process Variable
55.0
50.0
45.0
0.0 40.0 80.0 120.0 160.0 200.0
Time (seconds)
16
Typical analyser response
 KC = 4.1
65.0
60.0
(engineering units)
Process Variable
55.0
50.0
45.0
0.0 40.0 80.0 120.0 160.0 200.0
Time (seconds)
Approximate FOPDT
 In reality, few processes will exhibit the ideal FOPDT

response
 Rather, because most processes comprise a series of
lags, they will take on the S-curve response in which there is
no clear demarcation between where the deadtime () ends
and the response time () starts.
17
Approximate FOPDT
100%
of PV
PD
 
Indeterminate area
Approximate FOPDT
0.632
 PV
PD
 
Deadtime Response time
18
Approximate FOPDT
 This problem may be overcome by drawing a tangent to

the steepest part of the curve.
 What is now, strictly speaking, the ‘apparent’ deadtime
() is calculated from the point at which a step was applied
to the PD, and the point at which the tangent crosses the
horizontal line indicating the previous steady state value of
the PV.
 This point also marks the starting point from which the
response time () may be calculated (63.2% of the PV).
Tangent to point
of steepest rise
100% of PV
63.2
% of
PV
 
PD
19
Approximate FOPDT
 It is not always easy to precisely determine the steepest
part of the curve and hence the point at which the tangent
crosses the line may not be exact.
 In such cases, where some uncertainty exists, is it better
to round the figures are up or down?
 Here it is useful to examine the Ziegler Nicholls formulae
again:
1.1 
 Gain (KC)
KP 
 Integral time (TI) 2.0 θ
 Derivative time (TD) 0.5 θ
Approximate FOPDT
1.1  TINT  2.0  TDER  0.5 

KC 
KP 
 Clearly, if you under-estimate the deadtime () you are
estimating that the process is easier to control than it really
is.
 This results in a higher calculated value for the gain (KC)
and a smaller calculated value for the integral time (TI) –
both tending to produce oscillation and subsequent instability
in the control loop.
 It follows that over-estimating the deadtime will move the
calculated values towards a lower gain and longer integral
time – both tending towards better stability.
20
Approximate FOPDT
1.1  TINT  2.0  TDER  0.5 

KC 
KP 
 If you estimate the response time () too long, you are
estimating that the process is easier to control than it really
is.
 This over-estimation results in a higher calculated value
for the gain (KC) – again tending towards instability
 Consequently, if there is any uncertainty in estimating the
process parameters, the basic rules are:
Use a longer estimate for the deadtime ().

Use a shorter estimate for the response time ().
Lambda Closed loop response
 While the Ziegler-Nichols aims QAD, Lambda tuning aims

at a first-order lag plus dead time response to a set point
change.
 Lambda tuning offers the following advantages:
 The PV will not overshoot its set point after a disturbance or
set point change.
 Much less sensitive to errors made in determining process
dead time…
 …in lag-dominated processes too easy to under- or over-
estimate the relatively short process dead time.
 Ziegler-Nichols tuning rules can give really bad results when
the dead time is measured incorrectly.
 Tuning is very robust – consequently the control loop will
remain stable even if the process characteristics change
dramatically from the ones used for tuning.
21
 Lambda-tuned loops absorb disturbances better – passing

less of it on to the rest of the process.
 This is very attractive in highly interactive processes.
 Consequently, commonly used in paper industry to prevent
the entire machine from oscillating due to process
interactions and feedback control.
 The user can specify the desired response time (actually the
closed loop time constant) for the control loop.
 This provides one tuning factor that can be used to speed up
and slow down the loop response.
 However … Lambda tuning has a drawback:

 Lambda sets the integral time (TINT) equal to the process
time constant ().
 Consequently, if the process has a very long time
constant, the controller will have a very long integral time –
making recovery from disturbances very slow.
22
Lambda tuning procedure
 Carry out openly loop test to determine the process model

parameters KP,  and .
 We now need to choose a closed loop time constant () for
the control loop
 A large value for () will result in a slow control loop, and a
small () value will result in a faster control loop.
 Typically the value for () should be set between 1  to 4 .
Where there is a high degree of confidence use:
=
 This will provide a relatively fast response to set point
changes.
 If  <  the controller is likely to be overly aggressive.
Where there is a low degree of confidence use:

 = 2 to 4.
 The tuning relations may be calculated as follows:
K
=
τ
K
+
C
λ θ
P
TI
θ2
=τ+
λ + θ
TD
=
E
R
23
Percentage
SP PV
63.2%
  Time (seconds)
Integrating processes
 In some cases the process may not reach a new steady-
state value.
 Such integrating processes are typical of level-control
loops.
 A pseudo-integrating process can also exist in, for
example, a temperature loop in which the new equilibrium
value is well above the normal operational region.
 In either case, it is no longer possible to determine the
gain graphically.
 However, there is nonetheless, sufficient information to
allow us to determine the controller tuning values.
 Draw a tangent to the point of steepest rise and extend it
below the initial equilibrium value to intersect the vertical line
representing the step function change in the PD.
24
Integrating process
C = vertical distance
Tangent to point of
steepest rise
KP C

τ θ  PD
C
1 .1  τ
KC 
K P θ

PD
Continuous Cycling Method
Why?
1. Whilst the deadtime () and the response time () can
often only be approximated, the period can be measured
very precisely.
2. This means that the quality of the tuning parameters is
much higher.
3. The loop stays in Auto – thus making it safer for difficult
and very fast loops.
4. The test includes the effects of valve hysterisis and dead-
band as well as system non-linearities that are dependent
on direction and rate of change.
5. The test also includes the dynamics and special features
of different controller algorithms.
25
Why not?
Despite these advantages there are several reasons why it is
difficult to implement the closed loop test:
1. In many applications, a sustained oscillation may not be
acceptable to supervisory and operations personnel.
2. It is sometimes difficult to control the amplitude of the
oscillation – with a small change in set point yielding a
larger than expected oscillation amplitude.
3. Several tests may be required over a relatively long period
of time – during which the production output is out of spec.
4. This again, may not be acceptable to the production
personnel.
1. Put controller in P control only – do not use I or D control. This

would entail setting TI to a maximum (or turning integral action off
2. Set TD to zero.
3. If necessary set control to ERR = (SP - PV)
4. Set the controller to Automatic
5. Make a small step-change to the SP
6. Increase the value of K until continuous equal-amplitude cycling.
Note: A large amplitude is not required, and, in fact, need only be
large enough to distinguish the oscillation from the measurement
noises band.
7. Record the value of gain – referred to as the ultimate gain (KU)
8. Measure the period of the cycling (e.g. from one peak to the next).
This is called the ultimate period (PU)
26
Continuous constant
amplitude oscillation
A
SP
PU
Quarter
amplitude
damping
A
SP 1/4A
27
Quarter
Continuous constant amplitude
amplitude oscillation damping
1/4A
A
SP
PU
P - Control Kc = 0.5 x Ku
PI - Control Kc = 0.45 x Ku
TINT = Pu /1.2
PID - Control Kc = 0.6 x Ku

TINT = Pu /2
TDER = Pu /8
28
Default settings
Process type K I D
Flow 0.8 0.1 -
Liquid level 5 10 -
Reactor temperature 5 5 I/4
In-line temperature 0.5 1 I/4
Column temperature 0.5 5 I/4
Liquid pressure 0.3 0.1 -
Gas pressure 5 5 0.05
Reactor pH 1 2 0.5
Neutraliser pH 0.1 5 1.2
Inline pH 0.2 0.5 -
Improving ‘as found’ tuning
 This closed-loop method , proposed by Dr. Howard L.

Wade, is based on making a series of step changes to the
setpoint (SP), whilst the loop is in Automatic) and
observing the period (P) and the amplitudes A and B.
 From this we can calculate:
1. The ratio P/TINT
2. The decay ratio A/B
 Using the Period/Integral time ratio we adjust the integral
time
 Using the decay ratio (the ratio of the amplitude (A) of the
first positive peak to the amplitude (B) of the second
positive peak as shown below) we adjust the gain.
29
The ratio P/TINT – adjust integral time

The decay ratio A/B – adjust the decay ratio ratio
Period (P)
A
Amplitude
Time

 When a process is exhibiting Quarter Amplitude Damping
(QAD), then the oscillation period (P) will lie between 1.5
to 2 times the integral time (TINT):
1.5 TINT  P  2TINT
 Thus, the ratio P/TINT should lie between 1.5 and 2.
 And if it isn’t (it’s either lower or higher) we adjust the

integral time to:
TINT  0.6 P
30
 For QAD, the decay ratio should be around 0.25.
 If not we make an adjustment to the gain (KC) according
to the gain adjustment chart:
1.6
1.4
Multiply present gain by:
1.2
1.0
0.8
0.6
0.4
0.2
0
0 0.25 0. 5 0.75 1.0
Decay ratio
● Let’s start off with a common situation in which a step

change to the set point produces this oscillation:
● The oscillation period (P) is 16.2 minutes and the ratio of

the amplitudes A and B is 0.66.
● Further, we find the tuning settings as:
● KP = 1.5
● TINT = 2.5
31
● What’s the P/TINT ratio?
● According to the data it’s:
P/TINT ratio = 16.2/2.5
= 6.48
● Since this is greater than 2, the rule is to choose a new
integral time according to:
TINT  0.6 • P
● So, in this example what would we would set TINT to?
0.6*16.2 = 9 min.

● The next step setpoint change shows the following:
● A sluggish response with an increase in the oscillation

period (P) to 17.5 min and a decay ratio that is practically
zero.
32
● What’s the P/TINT ratio?
● According to the data it’s:
P/TINT ratio = 17.5/9
= 1.9
● Since this is less than 2, we leave the integral alone and
turn our attention to the gain (KC).

● Since the decay ratio is nearly 0 then, according to the
gain adjustment chart how much do we increase the gain
by?
1.6 According to the gain
adjustment chart it is
1.4
necessary to increase the
1.2 gain by a factor of 1.4 –

1.0 from 1.5 to 2.1.
0.8
0.6
0.4
0.2
0
0 0.25 0. 5 0.75 1.0
Decay ratio
33
● With the gain (KC) set to 2.1 the integral time (TINT) is
now 9 minutes.
● Here’s the result of the next step setpoint change:
● Here the oscillation period (P) has reduced to 14 min to give:

P/TINT ratio = 14/9
= 1.6
● and the decay ratio has increased to 0.12.
● So what’s next?

● Since the P/TINT ratio lies between 1.5 and 2.
● We again leave the integral alone and turn our attention to
the gain (KC).
1.6 ● Since the decay ratio is
1.4 0.12 then, according to
the gain adjustment chart
1.2
how much do we
1.0 increase the gain by?
0.8 ● According to the chart the
0.6 gain is now set to 1.2 *
2.1 = 2.52
0.4
0.2
0
0 0.25 0. 5 0.75 1.0
Decay ratio
34
● So, the gain (KC) is now 2.52
● And the integral time (TINT) is still 9 minutes.
● Here the oscillation period (P) has reduced to 13.4 min to

give:
= 1.49
● What’s next?

● Since this is again outside the limits of the P/TINT ratio
between 1.5 and 2.
● We need to choose a new integral time according to:
TINT  0.6• P
● So, in this example what would we would set TINT to?
0.6 * 13.4 = 8 min.
35
● So, the integral time (TINT) is now 8 minutes
● And the gain (KC) is still 2.52
● Now the oscillation period (P) is still 13.4 min to give:

= 1.67
● Is this within the prescribed limits?

● Yes it is – so what’s next?
● Well, the decay ratio now 0.18. By how much must the
gain be increased?
1.6 According to the chart KC
must increase by a factor
1.4
of about 1.05
1.2
So what’s the new level?
1.0 Well it was to 2.52 so
0.8 the new gain should be:
0.6
0.4 KC =2.65
0.2
0
0 0.25 0. 5 0.75 1.0
Decay ratio
36
● And here’s the result of the next step setpoint

change:
● And here’s the result of the next step setpoint

change:
● Well, that’s just about it.

● We’ve got the period down to 13.2 minutes and the
decay ratio down to 0.2
● Not quite QAD but a vast improvement on the original.
37
Tuning according to Pessen 1
● Tuning for no overshoot on start-up

● Variation of continuous cycling method
K C = 0.2 * KU
PU
TINT =
3
PU
TDER = 2
Tuning according to Pessen 2
● Tuning for small overshoot on start-up

● Variation of continuous cycling method
K C = 0.33 * KU
PU
TINT =
2
PU
TDER = 3
38
Comparison of methods
Closed ZN Pessen 1 Pessen 2

0.6 x Ku 0.2 x Ku 0.33 x Ku
KC
TINT Pu /2 Pu /3 Pu /2
Pu /8 Pu /2 Pu /3
TDER
Level control
 The vast majority of level loops only require control

between two limits:
» UL% = Allowed upper limit (%)
» LL% = Allowed lower limit (%)
» SP% = Set-point (%)
LC
UL%
SP%
LT
LL%
FOUT
39
Averaging level tank control
 Also need to determine:

» V = Volume between 0 and 100% (ft3 or m3)
» Q = Maximum flow rate (ft3/min or m3/min) when
valve is fully open
1 %
0 -
0
% S
 Then calculate:
K
=
U
L
P
%
C
1

1 %
0   
0 -
% L
K
=
S
P
L
%
C
2
   
 KC is then set as larger of KC1 and KC2
 Finally calculate:
4 Q
V K
TI
=

N
T

C
Averaging level tank control
 Assume the following: » UL% = Allowed upper limit (%) = 70

» LL% = Allowed lower limit (%) = 40
» SP% = Setpoint (%) = 60
» V = Volume = 100 m3
» Q = Maximum flow rate = 5 m3/min
UL%  Then calculate:
1 %1
0 -0 -
0
% S% L
1 01 6
0 -0 0
0 60 -
K
=
=
1
0
SP%
U
L
P
%
8m
C
1
   
0
LL%
K
=
=
5
S
P
%
L
%
4
0
C
2
   
 Since KC1is the larger of the two
values we set:
KC = 10
4 Q
V KC
4 5
1 1
0 0
0
4m
TI
=
 
 
N
T
 
40
Simplified level tank control
 A simplified liquid level control system
FIN
LC
LT
L
FOUT
Idealised level tank control
 An idealised liquid level control system
FIN
LC
LT
L
FC
FT
FOUT
41
Idealised level tank control
Attributes Implications
 The tank has a constant cross-  The control loop constitutes a
sectional area. linear system
 The level controller is cascaded  There is no effect from the
to a flow controller. upstream or downstream
pressure, or line loss.
 A valve positioner is installed  The system is not affected by
on the flow control valve. the size of the valve.
 All inflow goes to outflow – the  There is no dead time in the
tank is merely a buffer storage. loop.
 The size of the tank is  The dynamics of the flow loop
substantial. are significantly faster than the
dynamics of the vessel.
 The level controller operates at  Response to set point need not
a constant set point. be considered
Tight tank level control
 Whereas most self-regulating processes require three

parameters (KC, , ) to approximate the process
dynamics…
 …most liquid level loops only require one (occasionally two)
parameter(s).
 In place of process gain (KC) use is made of the tank
residence time (TR).
 This value may be calculated or it may be determined from
process tests.
42
Tight tank level control
 Calculation of the tank residence time (TR) is given by:
Q
u
a
n
t
i
t
y
o
f
f
l
u
i
d
b
e
t
w
e
e
n
u
p
p
e
r
a
n
d
l
o
w
e
r
l
e
v
e
l
t
a
p
s
TR
M
a
x
i
m
u
m
f
l
o
w
r
a
t
e
t
h
r
o
u
g
h
v
e
s
s
e
l

π 4
d
h
2 F
TR
=
 

m
 where:
 D = diameter of vessel (ft. or m)
 H = distance between upper and lower level taps (ft. or m)
 Fm = maximum controllable throughput through vessel (ft3/min
or m3/min)
Tight level tank control
 Determination of TR from process tests is carried out as

follows:
 Put the controller in manual.
 Change the output by a small amount (V).
 After a period of time (T)…
 … restore the output to its original value.
Valve
T
Flow
V
F
 Note that the outflow will change by an amount(F)

 During this time the level (L) will begin changing and only stop
when the output is restored to its original value.
Level
L
43
Tight level tank control 1
 Determination of TR from process tests is carried out as

follows:
Put the controller in manual
Flow
L
T
Level
F T
Valve
V T
 V = amount you changed the controller output (%)

 L = change in level (%)
 F = change in the flow (%)
 T = duration of test (min)
Tight level tank control 1

• V = Volume = 100 m3
• Q = Maximum flow rate = 5 m3/min
• TL = Hold-up time = 100/5 = 20 minutes
• Also assume worst case input disturbance of
10% …and level not to deviate by more than
5%
8m • FIN = 10%
100 m3
• LMAX = 5%
• Settle out quickly (0.05 decay ratio)
0
.
5
0 5
1
0
0
.
5 Δ
0 L
Δ
FI X
K
=
=
1
K
=
 
N
C
C
M
A
0
.
7 1
4 .
2
0
0
.
7 K
4 C
TL
TI
=
=
1
4
.
8
m
i
n
4m 
TI
=

0
44
Section 4. Advanced control
tuning
Cascade control
1
Single Loop Feed Heater Control
Inlet Outlet
SP
T
TC
PV
Condensate OP
out
Steam
in
 Problem
 OP drives valve position independently of fuel flow
pressure
 Fuel flow pressure changes disturbance on outlet
temperature
 Solution
 Add another control loop to keep fuel flow constant
2
Cascade Control
Inlet Outlet
SP
T
OP
TC
PV SP
Condensate
out PV
FC
OP
Steam
in F
Advantages of Cascade Control
 Any disturbances that affect the slave variable are

detected and compensated for by the slave controller
before the have time to affect the primary control variable
 Controllability of outside loop is improved because inner
loop speeds up the response of the process’s dynamic
elements between the control valve and the slave variable
 Non-linearity of the process in the inner loop is handled by
that loop and removed from the more important outer loop.
 However…
 … considerable more investment
 Therefore important that considerable improvement
results
3
Principle of Cascade Control
 Add FC loop to monitor and control fuel flow

 TC loop drives SP of this controller
(drives true flow and not valve position)
Cascade Control Terminology
 Downstream Controller (Slave)

 Also secondary controller
 SP driven by another controller’s OP
 Upstream Controller (Master)

 Also primary controller
 Controller whose OP drives SP of secondary
controller
4
Cascade Control
Inlet Outlet
SP
T
OP
TC
PV SP
Condensate
out PV
FC
OP
Steam
in F
Cascade control
Inlet Outlet
SPtemp
T
OPtemp
TC
PVtemp SPflow
Condensate
out PVflow
FC
OPflow
Steam
in F
5
PV tracking
 Controller in manual mode

 Choice available
 Operator selects optimum OP manually
 Optimum OP Value  no correction needed
 No ERR permitted
 SP = PV (PV Tracking in Manual)
 Mode  Auto  SP remains
Cascade control
Inlet Outlet
SPtemp
T
OPtemp
TC
PVtemp SPflow
Condensate
out PVflow
FC
OPflow
Steam
in F
6
Initialisation
 Kind of manual mode

 Secondary SP drives OP of primary
» PV tracking takes place in primary
» FC in manual mode
» OP driven by operator to target temperature T2
» PV of FC has correct value to achieve T2
» PV - Tracking of FC  SP of FC has correct value
» TC now obtains OP = SP (FC)
» Secondary into cascade
» Primary automatically switches to AUTO mode.
» Smooth BUMPLESS transfer from MAN to AUTO
Cascade control
Inlet Outlet
SP
T2
OP
TC
PV SP
Condensate
out PV
FC
OP
Steam
in F
7
Feedforward control
Would this work?
Even if goal is constant level?
Control
action
Controller
Disturbance: flow
changes
8
Would this work?
 Goal is to keep level constant

 Manipulated variable = inlet valve position
 Control action : Inlet flow = Outlet flow
 This is an example of Feedforward Control

 If Feedforward control is only type of control - PV will
drift over time because of:
 Natural Evaporation
 Delay in valve positioning etc.
Feedforward control
 Open loop control

 Value to be controlled (PV) is not used to calculate
control action
 Principle is: manipulate a variable of process so that
it compensates for impact of process disturbances
9
Inlet Outlet
SP
T2
TC
PV
Condensate
out
OP
Steam in
Feedforward control
Inlet Outlet
F1 T1 SP
OP OP
FC
FC
Condensate
out PV
F
Steam in
F1 C
F T 2  T 1 Lead/Lag
h
C h T2
10
Feedforward control
 Advantages
 Operates much faster
 Cannot cause instability
 Can reduce load upsets by an order of magnitude
 Corrects disturbance before it upsets the measured
variable
 Tuning is generally less critical
 Even poorly tuned feedforward control is better than
none on processes with slow response to load
response
Feedforward control
 Disadvantages
 Cannot correct for unmeasured load disturbances
 Does not take into account the control error
 Therefore PV will drift over time
 Solution:
 Combined Feedforward/Feedback control
11
Inlet Outlet
T2
T2
TC
F1 T1 
Condensate SP
OP OP
out FC
FC
PV
F
F
F1 C
T 2  T1 Steam in Lead/Lag
h
C h T2
Combined Feedback/Feedforward control
 Feedback applied as ‘nudging’ control to correct for

drift
 Therefore very non-aggressive
12
Tuning approach
 Strong feedback control has tendency towards

instability
 Therefore emphasise is on Feedforward control
 Use Feedback to correct PV drift
» Tune secondary controller
» Tune Feedforward control
» Tune feedback control using Ziegler-Nichols
» Assess speed of drift of PV
» If drift of PV insignificant reduce K of primary
Ratio control
13
Ratio control
 Ratio Control is used when the control of one flow must
be in a fixed relation to another.
 Two or more inputs to a process must be varied to
satisfy process demand.
 The values of the variables must be kept in a
prescribed ratio to satisfy the composition or condition
of an end product.
 Basically, there are two groups of control
configurations:
» Series
» parallel
Series configuration
Master
controller
RSP
‘LEADER’ Control Flow variable A
system
A
RSP
‘FOLLOWER’ Control Flow variable B
system
B
14
Series configuration
 The series approach is a ‘lead and follow’ system

 To satisfy the demand of the process, a master
controller will determine the value of one of its input
variables (Ratio Set Point).
 This variable will be measured, and the quantity of the
second input variable will be controlled as a function of
the first.
 This means that as the first input variable rises, so will
the value of the second.
 If the leading variable falls to zero, so must its follower.
 This interlock between the variables can become a
valuable safety feature.
Parallel configuration
Master
controller
RSP
‘LEADER’ Control Flow variable A
system
A
RSP
‘FOLLOWER’ Control Flow variable B
system
B
15
 The parallel approach allows the master controller to set

the value of both variables.
 Instead of one control system leading and the second one
following, the parallel concept ties both systems to the
direct command of the master.
 As process demand changes, the master controller adjusts
the values of both control systems simultaneously.
 Main advantage is that because there is no interlock,

both systems are independent of one another and thus
any noise occurring in one variable will not be reflected
in the second.
 Also, simultaneous updating of both systems could
result in a smaller deviation from the desired ratio
during changes in process demand.
 The obvious disadvantage of parallel ratio control is
that there is no interlock safety and thus if one variable
drops to zero, the second will not.
16
Single series ratio loop with fixed ratio
A
P
PV I/P
x Controller
A OUT
RSP
SP determined by consistency
of final mixed product
Master
controller
x I/P
P
B
 The simplest method of accomplishing ratio control.

 The SP of the master controller is determined by measuring
the consistency of the final product blend.
 This establishes the value of flow ‘B’.
Single series ratio loop with fixed ratio

A
P
PV I/P
x Controller
A OUT
RSP
Master
controller
x I/P
P
B
 The flow is measured and its percentage equivalent becomes

the set point of the flow ‘A’ controller.
 If flow ‘B’ is 20%, then 20% becomes the set point for flow ‘A’.
 In this example, the pipe for flow ‘A’ is much larger than ‘B’.
 This allows a ‘preset’ ratio to be established.
17
Single series ratio loop with adjustable ratio
P
PV I/P
x Controller
A OUT
RSP
Master
Ratio controller
x I/P
relay
P
B
 A ‘ratio relay’ allows the ratio of the flows to be easily adjusted.

 The system remains serial because although the value of flow
‘B’ is multiplied by the ratio it sets the value of Flow ‘A’ and
thus any fluctuations in flow ‘B’ will be reflected in ‘A’.
Series ratio loop using remote set-point flow controller
P
PV I/P
x Controller
A OUT
RSP
RSP
Controller
Ratio PV
B OUT
x I/P
relay
P
B
Master
controller
 Instead of the master controller assigning a valve position for

the ‘B’ system, now it is determined by the controller.
 A better control system for flow ‘B’ will result in a more even
control of Flow ‘A’.
18
Series ratio loop using remote set-point flow controller
A
P
PV I/P
x Controller
A OUT
RSP
RSP
Controller
Ratio PV B OUT
x I/P
relay
P
B
Master
controller
 The system has the same interlock feature as the previous

examples had.
 How well the systems stay together on a change in process
demand will depend on their response characteristics.
Deadtime compensation
19
Dead time
Loading hopper
Hopper feed
actuator
Material
travel
Conveyor
Belt
belt
weigher
Transport lag = 10
min
 The dead time, or transport lag, is the time taken for the
material, discharged from the loading hopper to be detected
by the belt weigher located some distance away.
 During this period any changes produced by the actuator
will not be detected.
Dead time
Loading hopper
Hopper feed
actuator
Material
travel
Belt
weigher
 Of course, this problem could be fairly easily overcome by a

moving the belt weigher to a point immediately after the
discharge hopper.
20
Dead time in paper forming and drying section
Calendering
rollers
Basic weight
Headbox measurement
Screen Take-up reel
Dryer section
 Thick stock is discharged onto a moving belt screen where it

becomes a mat.
 As the mat moves always from the headbox, the water drains
away and eventually produces a mat that is dry and stable
enough to be drawn into the drying section for further drying.
Calendering
rollers
Basic weight
Headbox measurement
Screen Take-up reel
Dryer section
 At this point of the ‘basic weight’ is measured

 This is used to control the head box discharge in order to
maintain the ‘basic weight’ to a preset value.
 Clearly, the length of time it takes for the mat to travel from the
headbox to the ‘basic weight’ sensor represents a pure dead
time.
21
Calendering
rollers
Basic weight
Headbox measurement
Screen Take-up reel
Dryer section
 Here, the controller measures the ‘basic weight’, compares this

with the SP and generates an output.
 Now it must wait for the deadtime period before seeing a result
of this change in the value of the controller output.
Calendering
rollers
Basic weight
Headbox measurement
Screen Take-up reel
Dryer section
 If the controller expects a result before the deadtime has
elapsed, and none occurs, it will assume that its last change had
no effect and will continue to increase its output until the PV
senses a change has occurred.
 By this time it will be too late and the controller will have
supplied too much or too little material from the headbox.
22
Solutions
 The easiest solution is to ‘de-tune’ the controller to a slower

response rate.
 The controller will then not overcompensate unless the dead
time is excessively long.
 The Integral mode of the controller is very sensitive to deadtime
since during this period of inactivity of the PV an ERR term is
present and the integrator is busy ramping the output value.
 Ziegler and Nichols determined that the best way to ‘de-tune’ a
controller to handle a deadtime of D minutes is to reduce the
Integral time constant TI by a factor of D2 and the Proportional
constant by a factor of D.
 The derivative time constant TD is unaffected by dead time since
it only occurs after the PV starts to move.
Solutions
 An alternative method of overcoming the problem is if we could

inform the controller of the deadtime period and give it the
patience to wait and be content until the deadtime has passed.
 Now, detuning and making the whole process very sluggish,
would not be required.
23
Smith Predictor
 In 1957 O.J.M. Smith, of the University of California at Berkeley

proposed the predictor control strategy
SP
Controller Industrial process

Real PV +

Dynamic Dead-time
-
Predicted simulation simulation
PV
+
Predicted PV
 +
Smith Predictor
SP Industrial
process
Controller Dynamic
Dead-time response
+
Simulator 
Dynamic
-
Dead-time response
Error = 0
24
Smith Predictor
SP Industrial
process
Controller Dynamic
Dead-time response
Dead-time Dynamic +
simulation simulation 
Dynamic
-
Dead-time response
Error = 0
Smith Predictor
SP Industrial
process
Controller Dynamic
Dead-time response
Dynamic Dead-time +
simulation simulation 
Dynamic
-
response Dead-time
Error = 0
25
Smith Predictor
SP Industrial
process
Controller Dynamic
Dead-time response
Dead-time Dynamic +
simulation 
simulation
Dynamic
-
response Dead-time
Error = 0
26
Section 5. System integration
tuning
Understand the process

● Why does this process exist?
● What are the Safety, process, and quality constraints?
● How does this process relate to the main product being
produced?
● What would happen if it were bypassed?
● What are the cost impacts of this process?
● What are the quality impacts of this process?
● How does it impact on other loops?
● Does the SP remain constant?
● How fast are disturbances?
1
● Review P&IDs
● Consult with:
 process engineers
 instrument technicians
 operators
● Look for the fundamental reasons for this part of the process.
Does it exist to:
 reduce variation?
 blend?
 change the product?
● You may find that there are conflicting descriptions of

the process objectives.
● This is paydirt!
● Get everyone together and have a discussion about
the real process objectives.
● Gaining alignment on the purpose of the process will
quite often resolve all kinds of other operating
problems.
2
Top 20 Mistakes
Greg McMillan and Stan Weiner, two of the world’s great

authorities in the field of process control, list the top 20
mistakes made in the field of process instrumentation.
Mistake No. 1
Control valve pressure drops that are not engineered.

● Most of the valves in your plant end up being oversized.
● This is accentuated by the tendency for the mechanical,
process, and instrument engineer to each put in some
excess capacity and then choose the next larger size.
3
Mistake No. 2
The use of on-off valves for control valves.

● The primary culprits of the supposed transition from “on-off” to
throttling are rotary valves.
● Taking a ball or butterfly valve, slapping a rack and pinion or
scotch yoke piston actuator and a spool type positioner and
calling it a control valve will create lots of repetitive cycles and
questions like “what is wrong”.
● Also beware of names that mean the opposite. For example, the
“high performance” isolation valve is really the “lousy
performance” throttling valve because the high sealing friction
and shaft windup cause excessive stick-slip.
● Remember that “tight shutoff” means “high friction.”
Mistake No. 3
The placement of differential head or insertion type (e.g.

vortex, annubar, thermal mass) flow meters downstream
of the control valve.
● It screws up the velocity profile and the next performance
review of you and your meter.
● Yet websites, newsletters, and books have this arrangement
proudly displayed as a logo for process control.
4
Mistake No. 4
Recalibration as the solution to instrument problems.

● This gives the appearance of doing something.
● The real problem is usually associated with the operating
conditions, application, or the installation.
● For example, the orifice could be worn, the steam tracing could
have boiled out the liquid in the sensing lines, connections
could be plugged, condensate could have collected in sensing
lines, and the composition and any of the physical properties
(density and viscosity) could have changed.
Mistake No. 5
Tight control of level in a surge tank.

● The purpose of this volume was to absorb changes in flow and
for the level to roll with the punches.
● Keeping the level within a few per cent of set point passes
them on and defeats the purpose of the tank.
● Most oscillations in a process can often be traced back to an
over zealous person tuning a level controller and an
excessively tight set of low and high level alarm limits.
5
Mistake No. 6
Hydro test and flush of control valves and instruments.

● This is a good idea if you want to see how tight a person with a
wrench can make the valve packing; whether the vendor really
meant the pressure over range limit, or whether your valve trim,
insertion meter, or electrode can withstand welding rods and
pipe wrenches travelling at 3 m/s.
Mistake No. 7
Equalization lines without a purge or tracing.

● It is not a case whether you view the line as half full or half
empty.
● You need to choose full or empty and then install a purge or
heat tracing to guarantee it.
● Trusting technicians to periodically empty or fill lines is an
exciting exercise in wishful thinking.
● Vapours will condense and fill up an empty line, or changes in
pressure and temperature will empty a full line at the least
opportune times.
6
Mistake No. 8
The over use of differential head meters to measure

flow.
● Orifices stink.
● To use them for mass flow, borders on the absurd.
● Apparently, engineers are oblivious to measurement noise,
poor rangeability from the square root relationship, the shift in
meter coefficient with wear and Reynolds number, the change
in density with composition, and the seven leak points.
● There are too many better choices with less noise, more
rangeability, and better long term accuracy and reliability, such
as Coriolis, magnetic flow, and vortex meters.
● Insertion versions are available for large pipelines.
Mistake No. 9
The installation of temperature sensors in baffles.

● This is a good idea if you want a smooth and gradual response
in temperature while your reaction runs away.
● The insertion of sensors in baffles coated with glass or full of
cigarette butts and sand makes trends so slow you will think
you are in the Caribbean.
7
Mistake No. 10
The over-use of thermocouples for control.

● If you like to adjust set points to compensate for drift and
increase trend display scales to flatten out noise and scatter,
don’t want to use derivative action, then save the $200 and go
with the thermocouple (TC).
● Just tell the process engineer the TC is faster and stronger
than the RTD, and don’t mention that it is inside a thermowell
whose design dominates the time response and reliability of the
installation.
● Be sure to save an extra $800 by omitting the smart transmitter
and take the TC directly into a DCS.
● Just remember to set the trend compression high to make the
A/D noise disappear and turn off the rate action.
Mistake No. 11
The use of anything other than the most simple, reliable,

and accurate measurements.
● If you pride yourself on your intelligence and creativity and are
into extreme sports while on the job, go for the most exotic and
complex instruments and analysers you can dream up.
● Forget about Coriolis density meters.
● They are just too accurate and reliable to be anything but
boring after they are installed.
● Instead pick an analyzer that likes a well heated, ventilated,
and air conditioned environment.
8
Mistake No. 12
Installation of pressure gauges and electrodes before start-

up.
● This is a great idea if you have stock in these companies.
● Just think of all the gauges and electrodes that get busted
during construction and water batching.
Mistake No. 13
Steam tracing left on during the summer.

● This is great for conducting chemistry experiments, such as
how hot does the process fluid need to get before its gets
sticky, solidify, or polymerize.
● Also, it provides a good check on sensor temperature ratings
and safety programs for technicians to wear gloves.
9
Mistake No. 14
The use of field mounted switches.

● This is lots of fun if you like to guess the trip settings,
especially for field pressure and temperature switches.
● Limit switches on valves may be unavoidable but make
sure they are encapsulated and have no mechanical
components.
● The higher amp rating of other types doesn’t do you much
good if they get hung up or corroded.
● Just remember that limit switches, which are designed to
tell you if a valve has failed, fail more often than the valve.
Mistake No. 15
The use of home-made instruments, analysers, and

algorithms.
● The use of special fabricated or designed stuff instead of off the
shelf standard solutions is exciting for the innovator but not for
everyone else who has to support or use it.
● The testing is prejudiced and the documentation is nonexistent,
but don’t let this stop you. After all that next promotion or
graduate degree may just be one special instrument or
algorithm away.
10
Mistake No. 16
The placement of all your eggs in one basket.

● Not a good idea unless you are big or stupid enough to
absorb the loss.
● The placement of interlocks or control system that can
disable an entire plant in one computer or controller or on
one power supply or circuit is risky business.
Mistake No. 17
The use of too much integral action (too small integral

time) in reactor and column temperature controllers.
● Integral has no sense of direction and will only make a
correction after it crosses the line – much like a 90 year old
driver.
11
Mistake No. 18
The use of not enough derivative action in temperature

and pH controllers.
● Turn off the rate in these applications and experience first hand
the exhilaration from acceleration.
● Just make sure the trend recording shows the new highs
reached by exothermic reactors or steep titration curves.
Mistake No. 19
The omission of positioners on fast loops.

● The real life problems introduced by bench settings, high
packing friction, shaft windup, high seating friction, and the
capabilities and diagnostics of digital smart positioners
make their omission a crime that should be punishable by a
person forced to guess what the valve position is for every
valve without a positioner.
12
Mistake No. 20
The use of steam traps in large or critical applications.

● Why not make the performance of millions of dollars worth of
process equipment and instrumentation hinge on a $100 stream
trap?
● One reason is that the use of a small vessel, level sensor, and
controller to hold-up condensate might cost a few thousand
dollars.
● Who cares if the stream trap stays shut and liquid condensate
backs up into the process equipment and reduces the heat
transfer area?
● Who will notice if the steam trap stays open, and steam blows
through to the condensate system?
● Engineers and operators have gotten sick to their stomach
looking at the trend recordings of evaporators, reactors, and
columns with cheap traps.
13
Section 6. Laplace transforms
tuning
Laplace Transformations
 So why do we need to know anything about Laplace

Transforms?
 Well, firstly, it’s not essential.
 I’d estimate that more than 99.9% of the population have
never heard of them – never mind use them.
 But what about in control engineering?
 Well, that’s a different matter.
 Pick up virtually any book on the subject and you’re like to
be faced with at least one transfer block making use of
Laplace Transforms.
 Why?
 Well, the fact is that they make life easy for us.
1
Transfer Functions
 As we’ve seen, we make use of block diagrams to

represent the composition and interconnection of a system.
 When used together with Transfer Functions we can
represent the cause-and-effect relationships throughout the
system.
 A Transfer Function is defined simply as the relationship
between the input and output signal of a device.
Rf
Ri Rf
Vi – Vo   Vi
Vo Ri
+
Transfer Functions
 In block diagram representation the gain of the amplifier

is G and the output Vo (ignoring the inversion) is:
Vi G Vo Vo  G Vi
2
Transfer Functions
 Supposing we had two amplifiers in series having

gains G1 and G2 respectively.
Vi G1 Vo G2 Vo
 What would be the total gain of the system?
Vo  G1  G2 Vi
 Very simple!
 Why?
 There is no time dependency
Transfer Functions
 In the integrator circuit the output is time dependent.

Cf
Ri 1
Ri  C f 
Vi – Vo   Vi dt  A
Vo
+
 We no longer have a straightforward relationship

between the input and the output
 The output (Vo) is now described in a linear Ordinary
Differential Equation (ODE).
 The most important point to recognize is that the output
is now dependent on time and it is said to be dynamic and
operating in the time domain.
3
Transfer Functions
 In the integrator circuit the output is time dependent.

Cf
Ri 1
Ri  C f 
Vi – Vo   Vi dt  A
Vo
+
 We no longer have a straightforward relationship

between the input and the output
 The output (Vo) is now described in a linear Ordinary
Differential Equation (ODE).
 The most important point to recognize is that the output
is now dependent on time and it is said to be dynamic and
operating in the time domain.
Laplace Transforms
 The Laplace Transform is a powerful tool used to solve

a wide variety of problems by transforming the difficult
differential equations into simple algebraic problems
where solutions can be easily obtained.
 Applying Laplace Transforms is analogous to using
logarithms to simplify certain types of mathematical
operations.
 By taking logarithms, numbers are transformed into
powers of 10 (or e – natural logarithms) to allow
multiplication and division to be replaced by addition and
subtraction respectively.
 Similarly, the application of Laplace Transforms to the
analysis of systems, which can be described by linear
ODEs, enable us to solve ODEs using algebra instead of
calculus.
4
Laplace Transforms
 They also provide a straightforward method for handling

the mathematical time shift associated with dead time
equations.
 Thus, complicated analysis can be performed in a
straightforward manner.
 Once the result from a transformation has been obtained
we can then apply the Inverse Laplace Transform to
retrieve solutions to the original problems.
Laplace Transforms
ODE problem
Difficult
ODE solution
5
Laplace Transforms
Algebraic
ODE problem Laplace Transform
problem
Very Easy
Algebraic
ODE solution Inverse Laplace Transform
solution
Laplace Transforms
 The definition of a Laplace Transform is:


L[ f (t)]   f (t)estdt  F(s)

where:
O
L [ f (t )]
= the symbol for the Laplace Transformation in the
brackets; .
s = complex variable (s = a +jb) introduced by the
transformation.
 Ho-hum….you thought that this was supposed to make
things simple?
6
Laplace Transforms
 Probably the easiest way to look at this ‘s’ operator is

that it represents the derivative relative to time:
d
s
dt
Laplace Transforms
 So, a derivative transformation block having a derivate

time Td :
sTd
 Continuing along this route, the reciprocal of derivative,

1/s is the integral.
 Thus the integral transformation block is:
1
sTi
where Ti is the integral time:
7
Transformation blocks
1
G =1 1  sT
Gain block First order lag
1
sTd 1  sT1   1  sT2 
Derivative block Second order lag
1 sT
sTi 1  sT
Integral block Lead block
Laplace Transforms
 The table below shows a number of standard Transforms
for different types of inputs
Block type Transform
Unit impulse 1
1
Unit step
s
1
Unit ramp
s2
1
e  at sa

sin  t
s2  2
 Let’s look at a unit step

8
Laplace Transforms
 If we applied a unit step to a Gain Block what would we

expect out?
 So, a unit step in….
1 1
G =1
s s
Step input Step output
 ….results in a unit step out
Laplace Transforms
 So let’s look at an integral block

 And let’s apply a step input
 The output is given by:
O
u
t
p
u
t
1 1
 
s s Ti
O
u
t
p
u
t
1

s 2 Ti
1 1
1
s sTi s 2 Ti
Step input
9
Laplace Transforms
 If we look to see what this is:

 We see it’s a unit ramp
Unit impulse 1
1
Unit step
s
1
Unit ramp
s2
1
e  at sa

sin  t
s  2
2
Laplace Transforms
 So, for a step input …

 ….we get a ramp output (with time TI)
 Which is what we would expect from an integral
block
1 1 1
2
s sTi s T
i
Step input Ramp output
10
Laplace Transforms
 We can apply the same rationale to a Derivative block

 Let’s apply a step input
 The output is given by:
O
u
t
p
u
t
1
  sTd
s
O
u
t
p
u
t
 1  Td
1
s
sTd 1.Td
Step input
Laplace Transforms
 If we look to see what this is:

 We see it’s a unit impulse
Unit impulse 1
1
Unit step
s
1
Unit ramp
s2
1
e  at sa

sin  t
s2  2
11
Laplace Transforms
 So, for a step input

 ….we get a impulse output
 Which is what we would expect from an derivative
block
sTd
Step input Impulse output
First order lag
 The general form of a First Order Lag:
K K
1 s
1 s

Response time
12
Deadtime
 The general form of deadtime:
e  θs e  θs

Deadtime
FOLPDT
 When combined the form is:
K e - s K e - s
1  s 1  s
 
Deadtime Response time
13
 The process of applying Laplace transforms to analyze a
linear system involves the following basic steps:
1. Convert the system transfer function, or differential
equation, to the s-domain by replacing ‘d/dt’ with ‘s’.
2. Convert the input function(s) to the s-domain using
the transform tables.
3. Algebraically combine the input and transfer function
to find an output function.
4. Use partial fractions to reduce the output function to
simpler components.
5. Convert the output equation back to the time-domain
using the tables.
14
LABORATORY EXERCISE 1
PROCESS DYNAMIC CHARACTERISTICS
OBJECTIVE: To become familiar with various forms of process dynamic characteristics, and to learn a
method of constructing a simple process model from step test data.
Optional: To become familiar with obtaining data from frequency response tests.
PREREQUISITE: Completion of PC-ControLAB tutorial (under Help | Tutorial ) or an equivalent

amount of familiarity with the program operation.
BACKGROUND: All process have both steady state and dynamic characteristics. From a process
control standpoint, the most important characteristic is the process gain. That is, how much does the
process variable (PV) change for a change in controller output. If both the PV and the controller output
are expressed as normalized variables (i.e., 0 - 100%), then the process gain is a dimensionless number.
The two most important dynamic characteristics of a process are the amount of dead time in the process
and its time constant. Real processes rarely exhibit a response of a pure first order lag (time constant)
and dead time, but can often be approximated as a first order lag and dead time.
This exercise tests for the process gain, dead time and time constant for both a “pure” process (can be
exactly represented as first order lag plus dead time) and for a more realistic process which can only be
approximated as a first order lag plus dead time.
1. RUNNING THE PROGRAM

Start Windows.
Run PC-ControLAB.
2. FIRST ORDER LAG PLUS DEAD TIME PROCESS

Click on View | Display Size | Bigger
Click on Process | Select Model.
Highlight “Folpdt.mdl" (First Order Lag Plus Dead Time) and press Open.
Press Zoom and change the PV scale range to 50-75. (Note that the PV scale has already been
converted to 0 – 100% of measurement span.)
With the controller in MANUAL, press Out. Note the initial values:
Present PV: ________
Present Controller Output: ________

Regulatory Control and PID loop tuning Exercise 1. Process dynamic characteristics
Key in a new output value of 45.0.
(Hint: After the numerical value is keyed in, wait until a vertical green line on the grid is just crossing the
grid boundary before pressing ENTER. This will make it easier to estimate subsequent times.)
After the PV has stabilized at a new value, press PAUSE.
Final value of PV: ________
What type of process response does this appear to be? ________
How much did the PV change? ________
How much did you change the controller output: ________
Process Gain:
∆ PV (%)
KP =
∆ OP(%)
________
How long after the controller output change before the PV started changing?
Dead Time (θ) ________
Calculate 63.2% of PV change: ________
Actual value of PV at 63.2% of change: ________
How long after the PV started changing (i.e, at the end of the dead time) before the PV crossed the
63.2% point?
Time Constant (τ) ________
Select Process | Change Parameters. Observe the values listed for Dead Time and
Time Constant. Do these parameter values agree with what you observed?
________
Select Process | Initialize.
Select Process | Change Parameters.
Select ‘Dead Time’ and change its value to 2.0 (minutes).
Select ‘Process Gain’ and change its value to 1.0.
Select ‘Time Constant’ and change its value to 3.0 (minutes).
Press CLEAR.
Page 7.1.2
Change the controller output from 35.0 to 45.0. Observe the response. Is this what you would expect?
________
You have just observed the response of a pure first order lag and dead time process. Very few, if any,
processes are this “clean.” We will now look at a process with unknown dynamics, but we will attempt to
approximate it with a first order lag plus dead time model.
3. UNKNOWN PROCESS
Click on Process | Select Model. Highlight “Generic” and press Open.
Notice that the PV scale is now in Engineering Units, rather than in percent.
(If not, then select View | Display Range | Engineering Units)
Upper end of scale (corresponds to high end of transmitter range) ________
Low end of scale (corresponds to low end of transmitter range) ________
Span of PV ________
Press ‘OUT’
Initial value of PV (in engineering units) ________
Initial value of PV (in percent of span) ________
Controller Output: ________
Change the controller output from 35.0 to 45.0.
When the PV reaches (apparent) equilibrium, press Pause. Does this look like a true
first order lag plus dead time process? ________
Does it look ‘approximately’ like a first order lag plus dead time process? ________
What is the final value of the PV (to the nearest whole number)? ________
How much did the PV change, in engineering units? ________
How much did the PV change, in percent of span? ________
Estimate the process gain. Kp ________
To estimate the dead time, draw (or visualize) a tangent to the PV curve, drawn at the point of steepest
rise. From the time of controller output change to the intersection of this tangent with the initial steady
state value is the apparent dead time.
Apparent dead time: (θ) ________
Different observers might estimate anywhere between 1½ to 2 minutes. For the purpose of calculating
controller tuning parameters, it is better to take the longer value where there is any uncertainty, since that
will produce more conservative controller tuning.
Page 7.1.3
The apparent time constant is the time from the end of the dead time to 63.2% of the process rise.
Apparent time constant: (τ) ________
NOTE: One of the uses that can be made of the estimates of process gian, dead time and time constant
is to calculate controller tuning parameters. (See Laboratory Exercise 9, PID Tuning from Open Loop
Tests.) Since dead time is more difficult to control than a first order lag, then if you estimate dead time too
short, you are estimating that the process is easier to control than it really is. This will result in controller
tuning parameters that cause the loop to be overly aggressive. Similarly, if you estimate the time constant
too long, you are estimating that the process is easier to control than it really is, and again the resulting
controller tuning parameters will cause the loop to be overly aggressive. On the premise that is one is to
make an error, it is better to err in the conservative direction than in the aggressive direction, then the
following pragmatic guidelines can be given:
If there is any uncertainty in estimating the process parameters, estimate the dead time on the
long side, and estimate the time constant on the short side.
Page 7.1.4
4.0 OTHER FORMS OF STEP RESPONSE

This section will explore other forms of step response.
4.1 Negative Process Gain
Select Process | Select Model. Highlight “generic3.mdl" and press Open.
Record the initial values:
PV: __________ °F __________%
Controller output: __________%
Increase the controller output by 10%. When the PV reaches equilibrium, record the following:
Final value of PV: __________°F __________%
Final value of controller output: __________%
Process Gain:
∆ PV (%)
KP =
∆ OP(%)
∆ PVfinal (%) - ∆P𝑉initial (%)

KP =
∆ O𝑃final (%) - ∆ O𝑃initial (%)
= ________
4.2 Integrating Process
An integrating process is one which will not achieve a natural equilibrium. The control of liquid level,
where either the vessel inflow or vessel outflow is fixed by an external set point, is a common example of
an integrating process.
Select Process | Select Model. Highlight “level2.mdl” and press Open.
This process model simulates a level control application in which the tank outflow is an independent
process disturbance (load), and the controller controls a valve in the inflow line. This process model has
some elements of "realism" that are not needed in this exercise; These features will be used in
subsequent laboratory exercises. For the present, we will eliminate these "realism" elements.
Highlight the parameter labelled ‘Valve Pos: 0 = No; 1 =Yes.’ (Block 9) Change this parameter value to
1.0.
Select the parameter labelled ‘Levl Noise: 0=No; I =Yes’. (Block 64) (Use the scroll bar at the left hand
side of the parameter list, if this parameter is not visible initially.)
Change this parameter value to 0.0.
Page 7.1.5
Press Clear.
Change the controller output from 35.0 to 38.0. This simulates an increase in inflow to the
tank.
What does the PV (tank level) do? ___________________________________
Does it appear that the PV is going to reach an equilibrium? ________
Before the PV reaches a limit (0 or 100), press the button StepIncr above the controller faceplate. This
simulates an increase in outflow from the tank. Note the increase in the trace labeled 'LOAD'.
If the PV is now falling, then the rate of outflow must exceed the inflow. Slightly increase the controller
output (to increase the tank inflow) to attempt to stabilize the level. If the PV is rising, decrease the
controller output.
Can you eventually (approximately) stabilize the level by manual control? ________
4.3 Inverse Process Response
Occasionally the process response to a step change in controller output, or to a load change, is initially in
the opposite direction to that expected from "first principles.' This is normally due to some underlying,
second order effect. Once the second order effect disappears, then the process responds in the expected
manner. A response such as this is called an "inverse response.” An example of inverse response is the
“shrink-and-swell” effect of steam boiler drum level. If the steam draw-off and steam rates are in
equilibrium, then the drum level will be constant. On an increase in steam draw without a corresponding
increase in feedwater rate, a decrease in drum level would be expected. Initially, however, due to the
reduction in pressure and consequent flashing of water into steam, the level rises; if the excess steam
draw is maintained, the level eventually begins to drop. This is qualitatively demonstrated in this exercise.
Select Process | Select Model and recall the Level2 process.

Select the parameter labeled “Levl Noise: 0=No: 1=Yes”. (Block 64) Change this parameter value to 0.0.
Select the parameter labeled 'BLK 57 LEAD TIME". (Block 57) Change the value of this parameter from
1.5 to -3.0.
Press Clear.
Press Steplncr. This simulates an increase in steam flow from the boiler drum.
Observe the PV (drum level). What does it initially do? ________________
(This is the drum level "swell” effect. If the steam flow had have been decreased, rather than increased,
we would have seen the opposite effect, or the “shrink”.)
What does the level eventually do? ________________
Page 7.1.6
Suppose a feedback level controller had been in Automatic and controlling the feedwater rate (drum
input). Upon sensing the initial change In level, would the level controller Increase or decrease the
feedwater rate?
Would this change be in the proper direction for long-term correction?
Laboratory Exercise 23, Drum Level Control, will demonstrate a control technique widely used in steam
generation applications for overcoming this problem.
5.0 FREQUENCY RESPONSE

One means of characterizing a process is by its response to a sinusoidal (sine wave) input at various
frequencies. At each frequency, the relevant data is: the ratio of the amplitudes of the input and output
sine waves; the phase shift between input and output signals.
After a number of data points are taken, a Bode plot of the data can be constructed. This portion of this
laboratory exercise will determine a few data points for a Bode plot.
Through the Process | Select Model menu, read in ‘Generic1’ process model. Be sure you are
using the FEEDBACK control strategy (read the right hand side of the display title bar).
Check to see that the controller is in Manual.
Select Control | Control Options.
Scroll down until you see ‘Enable Sinusoidal Output.’ Click on ‘YES’
Set the following parameter values:
Period, minutes: 60
Amplitude, Peak-to-Peak: 10
Press ‘ENTER’ to set value.
Press Clear then press On on the controller to initiate sinusoidal testing of the process.
Record the following:
Period, controller output and measurement: 60 minutes
Frequency, controller output and measurement, 0.0167 cycles/min
Controller Output Amplitude, Peak-to-Peak: 10%
Process Variable Amplitude, Peak-to-Peak (in % of full scale) __________
Amplitude Ratio, PV/Controller Output: __________
Time lag between controller output and PV (minutes): __________
Time lag (minutes)

Phase shift =
Period (minutes)
∗ 360°
__________
Page 7.1.7
Repeat this test for periods of 30, 15, 7.5 and 3.75 minutes. (You may have to increase the
amplitude of the input signal at the shorter periods. This is alright, since it is only the ratio of the
output and input amplitudes that you are seeking.)
Plot these values on the graphs on the following page.
(Note: On many Bode plots, decibels, rather than amplitude ratio, is plotted on a linear scale.
The conversion equation from amplitude ratio to decibels is:
Decibels = 20 Log10(Amplitude Ratio)
Also, frequency is usually plotted on a logarithmic scale reading in radians per time unit, rather than
cycles per time unit. A complete Bode plot is beyond the scope of this laboratory exercise, however.)
Page 7.1.8
Page 7.1.9
CONTROL VALVE CHARACTERISTICS

OBJECTIVE: To demonstrate the relation between valve stem position and the fluid flow through a
control valve, for both linear and equal percentage valves. Also, to demonstrate the difference between
manufactured and installed valve characteristics.
PREREQUISITES: Completion of Exercise 1, Process Dynamic Characteristics.
BACKGROUND: A control valve is the most common type of final control element for process control
loops. The position of the valve stem (typically somewhere between 0 and 100% open) sets the flow rate
of fluid through the valve. The response is typically non-linear, however; an incremental change in valve
position will produce a different change in fluid flow, depending upon the initial position of the valve. Two
factors which cause the non-linear response are the type of valve (i.e., the valve characteristics) and the
effect of the rest of the process piping network on the pressure drop through the valve.
Two common forms of valve characteristics are equal percentage and linear. These characteristics are
determined by the physical shaping of the valve internals by the valve manufacturer; hence they are
termed the manufactured characteristics. If the pressure drop across the valve were constant at all
valve positions, then the flow rate would vary with valve position according to the manufactured
characteristics. However, in most practical applications, as the flow rate increases other portions of the
process piping begin to take some of the available system pressure drop; hence the pressure drop across
the valve will decrease. This gives rise to the term installed characteristics, which describes how the
flow rate varies with valve position in a particular installation.
The variation in pressure drop across the valve can be characterized by a factor termed the ‘pressure
drop ratio’ (PDR), where:
Minimum pressure drop across valve

PDR =
Maximum pressure drop across valve
Pressure drop with valve fully open

PDR =
Pressure drop with valve closed
A PDR of 1.0 indicates that the valve takes all of the available pressure drop at all valve positions; the
balance of the components in the piping network take none of the pressure drop. In other words, this
indicates that there is a relatively small valve in a very large line. If the PDR becomes smaller
(approaching but never reaching zero), then the indication is that the line is smaller relative to the size of
the valve. The usual practice is to size the valve and the other components in the piping network so that
the valve will take 25% to 50% of the available system pressure drop when it is wide open. This
corresponds to a PDR of 0.25 and 0.5, respectively.
1. STARTING THE PROGRAM

Start Windows.
Run PC-ControLAB.
Regulatory Control and PID loop tuning Exercise 2. Control valve characteristics
2. PREPARATION
Confirm that the Feedback control strategy is being used.
Select View | Horizontal Grid Scale | Seconds. (Note that the grid now displays the last 60 seconds of
operation, not the last 60 minutes.) Most flow loops are very fast, so it is better to run the chart on a fast
time scale for this exercise.
Select Process | Select Model. Highlight ‘flow.mdl’ and press Open.
This process exhibits flow measurement noise, which is typical of many real flow processes. For this
exercise, however, we will use a noise-free measurement. Therefore, select Process | Change
Parameters. Then highlight ‘Meas Noise: 0=Off; 1=On’ (Block 40). Enter a new value of 0.0, then press
OK then Clear.
Note from the coordinates at the right hand side of the trend display that the maximum flow rate is 50
GPM.
For this exercise, we will use percent, rather than engineering value. Select View | Display Range |
Percent of Span. Notice that the scale is now 0 . 100%.
For convenience in reading numerical values of flow rate, in percent of span, click on View | Data
Monitor. Leave the Data Monitor on display for the rest of this exercise.
3. FLOW RATE VS CONTROLLER OUTPUT

3.1 Equal Percentage Valve
Select Process | Change Parameters and check or set the following process parameter:
Pressure Drop Ratio 1.0 (Block 34)
Valve Pos: 0=No; 1=Y 1.0 (Block 26)
Valve: 0=EqPct; 1=Linear 0.0 (Block 35.2)
Valve Cv: 11.18 (Block 35.1)
Meas Noise: 0 = Off; 1=On 0.0 (Block 40)
For three different cases (three different pressure drop ratios) determine the flow rate at a series of
controller output values (with the controller in Manual). Make the settings and record the flow data in the
Table 1.
(Note that the theoretical flow rate for an equal percentage valve does not go to zero at zero valve
position. This is one reason why is it not good practice to use a the control valve as a shut-off valve.)
Page 7.2.2
Plot the data from Table 1 for each of these cases on the graph, Figure 1, below:
Page 7.2.3
NOTE: These curves are based on a .theoretically ideal. equal percentage valve. A real valve with equal
percentage characteristics will approximate these curves.
The data with a pressure drop ratio of 1.0 represents the manufactured characteristics ofthe valve. The
other two data columns, with a pressure drop ratio of 0.3 and 0.1represent installed characteristics of the
valve, since the pressure drop ratio (if it is less than 1.0) is a function of the actual line characteristics into
which the valve is installed.
3.2 Linear Valve:
Select Process | Change Parameters and check or set the following process parameter:
Pressure Drop Ratio 1.0 (Block 34)
Valve Pos: 0=No; 1=Y 1.0 (Block 26)
Valve: 0=EqPct; 1=Linear 1.0 (Block 35.2)
Valve Cv: 11.18 (Block 35.1)
Meas Noise: 0 = Off; 1=On 0.0 (Block 40)
For three different cases (three different pressure drop ratios) determine the flow rate at a series of
controller output values (with the controller in Manual). Make the settings and record the flow data in
Table 2 below.
Page 7.2.4
Plot the data for each of the cases in Table 2 on the following graph.
NOTE: These curves are based on a .theoretically ideal. linear valve. A real valve with linear
characteristics will approximate these curves.
3.3 Interpretation of Data:
For any of these cases, the process gain,

∆ flow
∆ valve position
is represented by the slope the curve. If the slope is relatively constant, then the control loop response
should be relatively constant at all operating points. If the slope varies greatly, then unless thecontroller is
retuned the control loop response will also vary.
(This is perhaps oversimplified, because it only considers the steady state gain; in actuality, the dynamics
of the loop may also vary with operating point.)
Suppose that our operation covers the upper 75% of the flow range (25 . 100%). From the data taken, if
the valve has a typical pressure drop ratio of 0.3, which appears to be the better selection of valve
characteristics in order to maintain approximately the same response (without changing the tuning
parameters) over most of the range of the valve?
________________________________________
Page 7.2.5
If the application happens to be such that a relatively constant pressure drop is maintained across the
valve, which appears to the better selection of valve characteristics?
________________________________________
3.4 Confirmation of Interpretation:

We will test 4 of the previous cases to see how control loop responds at a low and a high operating point,
with no change in controller tuning. The following table gives settings that should be made for each of
these test cases. Sub-section number refers to the subsection activities listed following the table.
For all sub-sections, set or check the following process parameters.
Valve Pos? 0=No: 1=Yes 1.0 (Block 26)
Meas Noise: 0=Off; 1=On 0.0 (Block 40)
1. Set the process model and controller tuning parameters as listed above for column 1. Put the controller
in Auto and set the set point at 40%. Then make set point changes of 10%, within the range of 40% to
90%. (For comparison purposes it might help to ‘PAUSE’ between each measurement).
Does the closed loop response appear to be approximately the same at all points within this range?
__________
Set the set point at 40%, then make 10% set point changes to 30%, 20% and 10%. Does the loop appear
to get more sluggish at lower flow rates?
__________
Could this be predicted from the graph, Figure 1, for case 2-EP? __________
Yes. The curve you plotted for Case 2-EP shows a relatively constant slope (process gain) in the upper
half of the range, but a lower slope in the lower half.
2. Set the process model and controller tuning parameters as listed above for column 2.
Page 7.2.6
With a the controller in Auto and the same controller tuning as used above, start with a set point of 10%.
Make 10% set point changes to 20%, 30%, 40%, 50%,60%, 70%, 80%. The control loop should first
appear overly sluggish (for a flow loop), then less sluggish, finally more and more oscillatory as the set
point is increased. Do you observe this?
__________
Does this indicate that for best tuning the controller gain should be increased or decreased at the higher
set points
____________________
(For this situation, the controller gain should be decreased to compensate for the
higher process gain at higher set points.)
3. We will now repeat parts 1 and 2 of .Confirmation of Interpretation., but this time with a linear valve. Set
the process model controller tuning parameters as listed for column 3.
Put the controller in Auto and set the set point to 20%. When the loop has settled out, change the set
point to 30% and observe the response.
Now change the set point to 70%. When the loop has settled out, change the set point to 80% and
observe the response.
For a 10% set point change, where was the loop the most sluggish, at low operating point or high
operating point?
____________________
Is this predictable from the graph for Case 2-L? __________
4. Set the process model and controller tuning parameters listed for column 4. Put the controller in Auto
and set the set point to 20%. When the loop has settled out, change the set point to 30% and observe the
response.
Now change the set point to 70%. When the loop has settled out, change the set point to 80% and
observe the response.
For a 10% set point change, where was the loop the most sluggish, at low operating point or high
operating point, or about the same?
____________________
Is this predictable from the graph for Case 1-L? __________
Page 7.2.7
VALVE AND POSITIONER RESPONSE

OBJECTIVE: To explore the way sticking of a valve stem, due to friction at the packing gland, prevents
precise valve positioning, and to demonstrate the effects of adding a valve positioner to overcome the
effect of stem friction.
PREREQUISITE: Completion of PC-ControLAB tutorial ( under Help | Tutorial ) or an equivalent

BACKGROUND: Frequently the process control valve actuator consists of an air-filled chamber
closed on one side by a diaphragm which is attached to the valve stem, either directly or through some
type of mechanical linkage. If a valve positioner is not being used, the air pressure in the diaphragm is
that of the control signal to the valve. This pressure creates a force equal to the pressure times the
diaphragm area. Opposing this force is a force due to the compression of a spring; the amount of
compression is a function of the valve stem position. Thus, the net force on the stem is the difference
between the force due to signal to the valve and the spring compression due to valve position. (There
may also be a force due to the pressure drop of the process fluid through the valve.)
Valve sticking (sometimes called “stiction”) is due to static friction between the stem and packing gland. It
determines the net force on the stem (difference between signal to the valve and the actual stem position)
required to initiate movement of the stem from a resting condition. Deadband (or hysteresis) is due to
dynamic friction between the stem and packing gland. It is the amount of signal reversal required to cause
the stem to move in the opposite direction. Collectively, these two phenomena are usually called the
"dead band" of the valve.
Thus, friction between the stem and the packing gland manifests itself in two ways:
1) If there is a reversal in the direction of change of the signal to the valve, there will be a deadband, or
finite amount of reversal of signal before the valve stem exhibits movement in the opposite direction. Until
the signal reversal equals or exceeds this amount, the valve stem remains motionless.
2) If the signal to the valve is changing in a constant direction, and if the valve stem is currently not in
motion, the stem will “stick” at that position until there is sufficient unbalance in forces to cause the stem
to “slip” to a new position.
If a valve positioner is installed, it provides a local feedback system, comparing actual valve stem position
with desired valve position (signal to valve). One of its purposes is to overcome the effect of stem friction.
However, the combination of the stem friction and the high gain of a positioner will often cause the stem
positioning response to a change in control signal to exhibit a second order underdamped response. This
will be a fairly fast response and is of no consequence in relatively slow loops, like temperature or level.
In fast loops such as flow, however, the response of the positioner-actuator-stem combination and the
response of the flow itself are both relatively fast and may occur at approximately the same frequency. In
that case, the interaction between the positioner and the process flow itself may result in undesirable loop
behaviour.
If a positioner is not being used, then due to the volume of the air chamber and the restriction air signal
line, the actual diaphragm pressure may lag valve signal. If a positioner is used, this lag will not be
present, due to the small size of the receiving bellows in the positioner.
Regulatory Control and PID loop tuning Exercise 3. Valve and positioner response

Start Windows.
Run PC-ControLAB.
2. PREPARATION
If the Feedback strategy is not already up (check the top line of the display, right hand end), then select
Control | Select Strategy | Feedback. We will not actually be using Feedback control. We select this
strategy simply because the strip chart provides recording of thee process variables, PV-1 (valve
position), PV-2 (diaphragm pressure) and PV-3 (signal to valve).
Select View | Horizontal Grid Scale | Seconds. (The chart now displays the last 60 seconds of
operation rather than the last 60 minutes.)
Select Process | Select Model. Choose the VALVE.mdl and press Open.
Select View | Variable Plot Selection. Select YES for PV-1, PV-2 and PV-3. Select NO for all other
variables.
Select Grid background. Select Dark.
Press Clear.
3. SIMULATED EQUIPMENT
The laboratory equipment model (see the figure) consists of a ramp generator which generates a
triangular wave signal to the valve, the valve itself, and finally a recording of the signal to the valve (PV-3,
cyan trace), the diaphragm pressure (PV-2, magenta trace) and the actual valve stem (PV- 1, red trace).
Two switches permit selection of the modes of operation. One switch selects the signal source to be the
ramp (triangular wave) generator or a manually set value. The other signal selects the destination of the
signal as the valve positioner or directly to the valve, bypassing the positioner. The feedback controller is
not used in this exercise.
Page 7.3.2
4. EXERCISE
Note from the chart record the triangular wave signal to the valve (PV-2). The actual stem position (PV-1)
is lagging the signal to valve, and changes in steps.
Select Process | Change Parameters. Note the following:
Ramp Generator, On or Off? (Block 30) __________
Positioner being used, Yes or No (Block 80) __________
Value of Deadband? (Block 95.1) __________
Value of Slip-stick? (Block 95.2) __________
Observe the chart record when the signal is approximately in the middle of either an increase or decrease
cycle. When the valve stem moves, approximately how much does it move?
__________
When the signal reverses, about how much does it have to change in order to get the initial stem
movement following the reversal?
__________
If you think some other values of deadband and slip-stick may be more reasonable, change these
values. (Select Process | Change Parameters. Highlight either Deadband or Stick-slip and
enter a new value.) Does the response look reasonable?
__________
Select Process | Change Parameters. Highlight “0=No Pos; 1=Pos” (Block 80). Enter 1.0.
Does the actual valve stem position now accurately track the signal to valve?
__________
We have been examining the static response of the valve stem to relatively slow changes in
signal. We will now make step changes in signal, to see the dynamic response of the valve stem,
both with and without a positioner.
If you changed the values for deadband and stick-slip from their original values, re-enter:
Deadband = 5%
Stick-slip = 2%
Also, take the positioner OFF the valve (“0=No Pos; 1=Pos”. Enter 0.0), and turn the ramp
generator off (‘Ramp Gen: 0=Off; 1=On’. Enter 0.0)
Page 7.3.3
Note that the signal to the valve is now a constant value. Due to stem friction and the absence of
a positioner, the actual stem position probably will not match the signal. True or False?
_________
Without a positioner, the actuator dynamics can be approximately modelled as a first order lag with a fast
time constant. The effect of stem friction is still present, however.
In the “Change Parameters” list, select “Manual Signal Source” (Block 40) and enter a value of 70%.
Except for the stick-slip effect, does the valve stem respond approximately as a first order lag?
__________
What is the (approximate) time constant? (Remember that the chart width represents 0 - 60
seconds.)
__________
In the “Change Parameters” list, note the ‘Actuator time constant (minutes)’ (Block 80)
__________
Now put the positioner on the valve. (“0=No Pos; 1=Pos”. Enter 1.0).
With a positioner, a valve positioner-actuator-stem system can act as an underdamped second order
system. This is due to the fact that if there is a tendency of the valve to stick at one position, the
positioner puts out increasingly higher air pressure until the stem moves. At that time, the air pressure
may be too high, so the valve stem will overdrive the requested position, causing the positioner to
decrease the air to the actuator. The net effect is a slight oscillating response, characteristic of a second
order, underdamped system.
Change the “Manual Signal Source” to 50%. Is the response what is expected (underdamped, second
order type of behavior)?
__________
Approximate period of oscillation (seconds): __________
WHAT YOU SHOULD HAVE OBSERVED IN THIS EXERCISE:
Without a positioner, stem friction can cause valve sticking, limiting the resolution of valve travel
and causing an excessive amount of signal reversal to initiate stem movement in the opposite
direction. This can be detrimental in any type of feedback loop. With a positioner, the effect of
stem friction is largely overcome in the steady state. However, a dynamic effect is produced,
approximating that of a relatively high frequency, damped second order system. If the valve is in
a slow loop (e.g., temperature), this is probably of no consequence. However, if this is in a fast
loop, such as a flow loop, this in itself may have a detrimental effect on loop behaviour.
Exercise 4, Flow Control Loop Characteristics, will illustrate this.
Page 7.3.4
FLOW CONTROL LOOP CHARACTERISTICS

OBJECTIVE: To demonstrate various aspects of flow control loops, including typical speed of
response, measurement noise, and the effect of sticking valves.
PREREQUISITE: Completion of the following exercises:

2 Control Valve Characteristics
3 Valves and Positioners
BACKGROUND: Exercise 2 covered the relation between valve stem position and flow rate, for both
equal percentage and linear valve characteristics, at varying pressure drop ratios. Exercise 3 covered
valve actuators and their tendency to “stick” if not well maintained. Also covered in Exercise 3 was the
effect of adding a valve positioner.
This exercise demonstrates additional characteristics of a flow loop, including measurement noise which
is often present, typical speed of response and the effect of a sticking valve in a control loop.

Start Windows.
Start PC-ControLAB.
2. PREPARATION
Confirm that the Feedback control strategy is being used.
Select View | Horizontal Grid Scale | Seconds.
Select Process | Select Model. Highlight “Flow.mdl” and press Open.
Enter the following tuning parameters:
Gain: 0.8
Reset: 0.05 minutes/repeat
3. FLOW CONTROL LOOP RESPONSE

Most flow loops exhibit some amount of measurement noise. The severity usually depends upon
the type of measuring element used. Does this simulation exhibit measurement noise?
________
Select Process | Change Parameters and highlight “Meas Noise Maximum” (Block 40.2).
Change the value from 1.0 to 0.5. Does that reduce the amplitude of the noise?
________
Regulatory Control and PID loop tuning Exercise 4. Flow control loop characteristics
Select Process | Change Parameters and highlight “Meas Noise Correlation” (Block 26).
Change its value from 0.8 to 0.95. Does that tend to smooth out the noise?
________
This action was roughly equivalent to filtering the signal at the transmitter.
Select Process | Change Parameters and highlight “Valve Pos: 0=No; 1=Y” (Block 40.2). Change the
value
from 0.0 to 1.0.
Put the controller in Auto and change the set point to 30 GPM.
Is there any overshoot of the set point? ________
How much time elapsed between the changing of the set point and when the PV first
crossed the set point?
________
This demonstrates the relatively fast nature of most flow loops.
Before proceeding, change the set point back to 25 GPM.
4. STICKY VALVE
The simulation realistically exhibits flow measurement noise. However, because that tends to
masks the points we wish to illustrate, we will remove it. We will also remove the positioner.
Select Process I Change Parameters and highlight "Meas Noise: 0=No; 1=Yes” (Block 40.1).
Enter 0.0.
Select Process | Change Parameters and highlight “Valve Pos: 0=No; 1=Y” (Block 26).
Change the value from 1.0 to 0.0.
Enter or confirm the following tuning values for the controller:
Gain: 1.0
Reset 0.05 minutes per repeat
Put the controller in Auto. Then change the set point 30 GPM.
Observe the response. Both the controller output and the PV are moving up and down more or
less like triangular waves. This type response is sometimes called “oscillation,” although it does
not appear to be the sinusoidal wave typical of oscillation due to poor controller tuning.
Record the following:
Is the period regular or irregular? ____________________
Average (approximate) period of “oscillation”: ____________________
Peak-to-peak amplitude of controller output swing: ____________________
Page 7.4.2
Note: If you did not change the horizontal scale to seconds, you will see both the PV and the controller
output changing quite rapidly. This is due in part to the change of time scale, but possibly also due to
numerical instability due to the attempt to simulate a fast process at a slow time scale. Better to be on the
seconds scale for this exercise.
Suppose you (erroneously) interpreted the cause of the “oscillation” as improper tuning. Your action might
be to reduce the gain of the controller. Change the gain from 1.0 to 0.5. Then record the following:
Average period of “oscillation”: ____________________
Peak-to-peak amplitude of controller output swing: ____________________
Note that the amplitude of oscillation did not change appreciably, but the period got longer.
REASON. Suppose the flow rate is below set point. The integral action of the controller will gradually
increase the controller output. However, the valve itself will not respond until there is a sufficient
difference in the signal to the valve and the spring force corresponding to the valve stem position. When
there is a sufficient difference in force, the valve will move in a jump to a new position, consequently
causing a jump in flow rate. If the flow rate is then above set point, the integral action will begin
decreasing the signal to the valve, and the action repeats, except in the opposite direction. (Review the
results of Exercise 3, Valves and Positioners.)
To see what the valve stem itself is actually doing, select View I Variable Plot Selection, then click on
“yes” for PV-2. This is the signal which would be displayed if there were a valve position transmitter
installed on the valve.
The amplitude of oscillation is really determined by the amount of “stiction” in the valve itself. Reducing
the gain (or lengthening the reset time), merely slows down the rate of change of the controller output
hence increases the period of oscillation. This is the wrong solution to the problem.
A proper solution might be to perform maintenance on the valve to reduce the stem friction. Or add a
positioner to the valve. The positioner, however, while overcoming the effect of packing and stem friction,
can introduce a dynamic problem of its own. Before we install a positioner, let’s see the best that could be
achieved under ideal conditions, that is, with no stem friction and no measurement noise.
Go through Process I Change Parameters and change both “Deadband” and “Stick-slip” to 0.0.
This simulates an ideal valve with no stem friction.
Change the tuning parameters back to Gain = 1.0; Reset = 0.05 minutes/repeat.
Did this appear to cure the problem? __________
Start with a set point of 25 GPM, then increase the set point by 5 GPM. Is the response acceptable?
__________
Put the set point back at 25 GPM. When the loop comes to equilibrium, put the controller in Manual.
Now add a positioner. (Select Process I Change Parameters. Highlight “0=No Pos; 1=Pos”. Enter 1.0.)
Change the controller output by 10%. (You should still have the stem position record on display.)
How did the stem position respond to a step change in signal to the valve?
Overdamped: ________ Underdamped: ________ No dynamic effect: ________
Page 7.4.3
Both the flow loop without a positioner and the positioner-stem-actuator combination are responding as
slightly underdamped systems at approximately the same frequency. When we close the loop with a
positioner on the valve, these two responses may interact, causing a “ringing” of the response (e.g.,
continuing oscillation, with very slight damping).
Put the controller in Auto. Set the set point at 25 GPM. When the loop settles down, change the set point
to 30 GPM.
Is the loop more oscillatory than it was before adding the positioner? ________
To compensate for this, reduce the gain from 1.00 to 0.8
Change the set point change back to 25 GPM.
Is the loop behaviour more acceptable? ________
Page 7.4.4
TEMPERATURE CONTROL LOOP CHARACTERISTICS

OBJECTIVE: To demonstrate typical characteristics of temperature control loops, primarily, the
nonlinear effect of increasing load.
PREREQUISITES: Completion of the following:

Exercise 1 Process Dynamic Characteristics
Exercise 3 Valves and Positioners
Exercise 4 Characteristics of Flow Control Loops
BACKGROUND: Temperature control loops often have a number of similar characteristics. These
include:
Very limited amount of (or no) process measurement noise;
A relatively large time constant-to-dead time ratio;
If the primary load disturbance is a change in process throughput (for example, feed rate to a
process heater), then the process characteristics change with throughput. Specifically, as the
process throughput decreases, the process gain, time constant and dead time all tend to
increase.
Due to this last characteristic, it is often beneficial to combine ratio and cascade control, where the
Primary controllers output does not directly set the set point of the Secondary, but rather sets the required
ratio of secondary PV to process flow. For example, a process heater outlet temperature controller may
set the required fuel-to-feed ratio.

Start Windows.
Run PC-ControLAB.
2. PREPARATION
Confirm that FEEDBACK control strategy is being used. (See top line of the display.)
Select Process I Select Model. Highlight ‘Heater.mdl’ and press Open.
Select View I Variable Plot Selection. Select ‘Yes’ for PV-2. Leave all others the same. Press Clear.
The label above the controller, reading ‘Temperature (PV-1)’, should be highlighted in red. If it is not, click
on It. Note the range of heater outlet temperature measurement. (Read scale at right of grid.)
From ________ °F to ________ °F. Span: ________ °F

Regulatory Control and PID loop tuning Exercise 5. Temperature control loop characteristics
Click on the label above the grid that reads ‘Feed Rate’. Note the range of feed rate measurement.
From ________ GPM to ________ GPM Span: ________ GPM.
Click on the label above the grid that reads ‘Fuel Flo (PV-2)’. Note the range of fuel flow
measurement.
From ________ KCFH to ________KCFH. Span: ________ KCFH.
Return to the ‘Temperature (PV-1)’ label.
Press Process I Change Parameters. Note the type of valve characteristics and the pressure drop ratio.
Type of valve (Block 19.2) ____________________
Pressure drop ratio (Block 18) ____________________
For a gaseous fuel, a pressure regulator upstream of the valve would maintain a constant upstream
pressure. The valve acts as a variable choked nozzle, hence the flow characteristics are as if there were
a constant pressure drop across the valve, even though the burner back pressure will vary.
3.0 TESTING THE PROCESS

What we will be doing:
We will test the process for process gain and dynamic characteristics at four different process
flow rates. Our tests will always be such that the temperature remains in the vicinity of its normal
operating region, 275 - 300 °F. We will determine the process gain in two ways.
Temperature (%)
KP1 =
∆ Signal to valve (%)
This is the process gain seen by the controller when a simple feedback control loop is used (that
is, when the controller output goes directly to the valve).
Temperature (%)
KP1 =
∆ Fuel flow rate (%)
This is the process gain seen by the controller when a cascade loop is used (that is, when the
temperature controller output sets the set point of a fuel flow controller).
3.1 35% Process Flow.
With the controller in Manual, set the controller output at 10%.
Click StepDecr 8 times to decrease the feed rate (Designated ‘LOAD’ on grid) to 140 GPM = 35%.
(You can read the scale on the LH side of the grid.)
At 35% process flow and 10% controller output, read and record in the table below the temperature (PV-
1), fuel flow rate (PV-2), and signal-to-valve (controller output). (Use PAUSE to read the values.)
Page 7.5.2
Then change the controller output to 13% and read and record the same variables. Also estimate and
record the dead time and time constant due to the change in controller output.
Then calculate ΔTemp (%), ΔFuel Flow (%) and the two process gains listed above.
3.2 55% Process Flow
Click StepIncr 4 times to increase the process flow to 220 GPM (55%).
Repeat the procedure above for the controller output at 25%, then at 28%.
Click Steplncr 4 times to increase the process flow to 300 GPM (75%).
Click StepIncr 4 times to increase the process flow to 380 GPM (95%).
Page 7.5.3
What you should have observed.
The process gain, as determined by the ratio of temperature change to change in fuel flow rate,
decreases with increasing process flow rate, or conversely, increases with decreasing process flow
rate. Thus, if a Cascade control loop were applied (Temperature Controller setting the set point of
a Flow Controller), and if the process is operated at widely varying flow rates, then there could be a
problem with controller tuning. If the Temperature Controller were tuned properly for a high process
flow rate, it would probably be overly aggressive at a low process flow rate. If it were tuned properly
for a low rate, it would probably be too sluggish at a high process flow rate.
(Don't give up on Cascade control just yet, however. There are other reasons why it is beneficial.
When we explore Ratio and Feedforward control, we will see how we can overcome the non-
linearity problem.)
The process gain, as determined by the ratio of temperature change to valve position change,
stayed relatively constant as the process load increased. This is due to the fact that we are using
an equal percentage valve with a constant pressure drop. The inherent non-linear characteristics of
the valve (see Exercise 2) offsets the non-linear characteristics of the process. If we had used a
linear valve, rather than an equal percentage valve, then we would have seen the same type of
decrease in process gain as the process flow rate increased.
You should also have observed the changing dynamics (decrease in dead time and time constant)
as the process flow is increased. Since the ratio of these does not very nearly as widely as does
the process gain, this is not of as serious of consequences as is the gain variation, however. It may
still cause a different control loop response at high and low process flows.
4. CONFIRMATION OF OBSERVATIONS
This section pertains to a process heater with a simple temperature controller on the heater outlet.
The following is a process and instrumentation diagram for sub-sections 4.1 and 4.2.
Page 7.5.4
4.1 Feedback Control, Equal Percentage Valve
Set the following, which represents the ‘normal’ operating point:
Process flow 300 GPM (75%)
Temperature set point 275 °F
Controller output 35 %
Enter the following controller tuning parameters:
Gain: 1.0
Put the controller in Auto.
Change the set point to 300 °F. Acceptable response? ________
Change the set point to 250 °F. Acceptable response? ________
Increase the process flow to 380 GPM (95%).
When the loop settles out, change the set point to 300 °F. Compare the response to the previous
response when the load was 300 GPM:
About the same ________
A bit more sluggish ________
Much worse ________
Decrease the process flow to 220 GPM (55%).
When the loop settles out, change the set point to 250 °F. Compare the response to
the previous response when the load was 300 GPM:
Much worse ________
Decrease the process flow to 140 GPM (35%).
When the loop settles out, change the set point to 300 °F. Compare the response to
the previous response when the load was 300 GPM:
Much worse ________
Page 7.5.5
What you should have observed:
Due to changing dynamics at lower process flow rates, the control loop becomes slightly more
aggressive. However, we can probably live with it at all flow rates.
4.2 Feedback Control, Linear Valve
Put the controller in Manual
Set the following:
Set Point 275 °F
Controller Output 35%
Select Process I Change Parameters and change the following:
‘Valve Cv-Max’ (Block 19.1) from 104.69 to 22.36
‘Valve Type: 0=EqPct’ (Block 19.2)1=Lin From 0.0 to 1.0
Gain 1.2
Reset 12.00 minutes/repeat
Put the controller in Auto and change the set point to 300 °F.
Acceptable response? ________
Change the set point to 250 °F?
Put the controller in Manual
Set the following:
Set Point 275 °F
Controller Output 16.31%
Put the controller in Auto and change the set point to 300 °F.
What is happening to the control loop? ________________________________________
Page 7.5.6
With the linear valve and an effective constant pressure across the valve, the nonlinearity of the
process unit dominates and causes the loop to become unstable at low process flow rates.
4.3 Cascade Control

The following is a process and instrumentation diagram for sections 4.3 and 4.4:
Click on Control | Select Strategy I Cascade
Select the Fuel Flow controller and enter the following tuning parameters:
Gain 0.5
Reset 0.15 minutes/repeat
Select the Temperature controller and enter the following tuning parameters:
Gain 1.0
Reset 15 minutes/repeat
Put the Fuel Flow controller in the cascade mode (press Casc) and the Temperature controller in the
Automatic mode (press Auto).
Page 7.5.7
With the Temperature controller selected, make a series of set point changes:
From 275 to 300 °F
From 300 to 250 °F
From 250 to 275 °F
With the Temperature controller selected, press StepIncr 4 times to increase the process flow to
380 GPM (95%). When the loop settles out, make the same series of set point changes as
above.
How does the response compare with the response when the process flow was 300 GPM (75%)?
About the same: ________
Slightly more sluggish: ________
With the Temperature controller selected, press StepDecr one step at a time to reduce the process flow
rate, eventually to 140 GPM (35%). Let the control system settle out between each step. What happens
to the response as the process load is decreased?
Becomes increasingly more aggressive ________
Becomes increasingly more sluggish ________
Due to the increasing process gain (KP2 in section 3) at lower process flows, the Cascade control
becomes increasingly more aggressive, and eventually becomes unstable, as the process flow is
decreased.
But don't give up on Cascade yet; there are other reasons to use it. And when we look at Ratio and
Feedforward control (Exercises 14, 16 and 17) we will see what can be done to overcome the
nonlinearity.
Page 7.5.8
PRESSURE CONTROL LOOP CHARACTERISTICS
OBJECTIVE: To demonstrate the characteristics of pressure control loops as exemplified by several

application examples.
PREREQUISITE: Completion of Exercise 1, Process Dynamic Characteristics
BACKGROUND: Pressure control loops can be characterized in several ways, including by the state
of the flowing medium and by the relative position of the pressure sensor and the control valve.
Liquid phase pressure loops are controlled by maintaining a flow balance into and out of an enclosed
volume. The loops are fast and have somewhat the characteristics of flow control loops.
Two-phase (gas and liquid) pressure loops are usually controlled by the addition or removal of heat to
vaporize liquid into the gas phase or condense gas into the liquid phase. An example is the control of
steam header pressure leaving a steam generator. Two phase pressure control loops are relative slow
and noise free; hence they have characteristics similar to temperature control loops.
Pure gas phase pressure control loops, with the exception of pressure control for a long pipeline, are
typically fast, noise free and exhibit essentially no dead time. Consequently they can be controlled with
high gain proportional controller with less emphasis placed on reset action.
In terms of the relative position of the sensor and control valve, if the sensor is located downstream of the
control valve, the application is usually called a "pressure regulator" or "pressure reducer station." If the
sensor is located upstream of the control valve, the application is usually called a "back pressure
regulator."
The process models in this laboratory exercise simulate a gas pressure reducing station and a pressure
regulator for a gas separator. Neither process model exhibits significant dead time. Process disturbances
are the upstream pressure (accessible as a load change), downstream flow demand and enclosed
volume. These latter effects are set via the "Change Parameters" feature of the process model.

Start Windows
Run PC-ControLAB
2. PREPARATORY
Be sure you are using the FEEDBACK control strategy. (Check the top line of the display.)
Select Process | Select Model, then highlight “pressure.mdl” and press Open.
Regulatory Control and PID loop tuning Exercise 6. Pressure control loop characteristics
Click on Load (in the Menu bar), then in the sub-panel labelled “Auto Load Change – Random Walk”,
click in the data field by the label “Correlation”.
Press Esc on your keyboard, then key in a new value of 0.97. Press the Enter button, then click on
Clear.
3. EXERCISE – GAS PRESSURE REDUCING STATION

Change the controller output from 35% to 45%. Does the pressure go up or down? __________
Observe that the process response is fast with relatively no dead time. It can be controlled with a
controller with high gain proportion-only control action (i.e., no integral action).
Return the controller output to 35%.
Select Control | Control Options. In the first section of the drop-down list of options, ‘Control Algorithms’
select “Proportional Only.” Then press Clear.
Enter the following feedback controller tuning parameters.
Gain: 10.0
Manual Reset: 35%
Press AutoLoad.
You are seeing a randomly varying load demand. Note that the pressure remains fairly
close to set point even though the controller is proportional-only (no reset).
Press AutoLoad again to discontinue the randomly varying load.
If the load (read gray trace vs. coordinates on left hand side of grid) is above 50% when AutoLoad
was discontinued, press StepDecr until the load is below 50%.
Press SP and change the set point from 50 to 54 psig. Describe the initial response of the
controller output:
___________________________
Because of the high gain, a step change in set point produces a very large proportional response. Very
quickly, the feedback action removes a large proportion of this proportional response. Even so, to avoid a
proportional "spike" on the output, we should probably, it possible, avoid making step changes in set
point. There are several alternative possibilities available with various makes of commercial systems (all
of these are available with PC-ControLAB):
• Use the configuration option "Proportional Mode on Measurement." Then on a set point change,
the integral action gradually moves the output to a new value.
• If set point jog buttons are available, use these for making gradual set point changes.
• The set point can be ramped (changed gradually at a user selectable rate) to new set point value.
In practice, pressure control loops, probably more so than many other types of loops, frequently
operate for long periods of time without encountering a set point change
Page 7.6.2
Use the down jog button (on the right hand side of the vertical scale on the controller) to gradually move
the set point down to about 48 psig. Do you see the large change in controller output that you saw when
you made a step change to the set point?
__________
Press Process | Change Parameters. Highlight “Supply Pressure-psig” and change it from 75.0 to 95.0
Does this tend to cause the valve to close (after equilibrium is reached)?
__________
As the upstream pressure rises, the valve must move toward closed to maintain approximately the same
downstream pressure.
Change the Supply pressure to 55.0 psig. Does this tend to cause the valve to open? __________
As the upstream pressure falls to just above the required output pressure, the valve must move wider and
wider open to accommodate the demand. Hence with proportional-only control, we see a larger deviation
from set point when the upstream pressure is low and the controller output is correspondingly high.
Press AutoLoad to reactivate random load changes. The random variations of the demand will cause the
controller output to move up and down. Observe it at some time when the controller output is
approximately 35%. At that time, what is the relation of the PV to the set point?
(Press StepIncr or StepDecr if necessary to force the load into a region you want to observe.)
PV is above SP ________ PV is about equal to SP _________ PV is below SP _________
When the controller output is above 35% (higher demand), what is the relation of the PV and SP?
When the controller output is below 35% (lower demand), what is the relation of the PV and SP?
What is the significance of “35%”? __________________________________________________
These later demonstrations have illustrated the gas pressure control loop’s tolerance to both demand
changes and set point changes, even though a simple proportional controller without reset action is being
used. In fact, this entire exercise has demonstrated that a gas pressure control loop is a fairly easy loop
to control, being relatively insensitive to load changes as well as parameters within the control loop itself.
There are undoubtedly exceptions to this statement, but for most gas pressure control loops, these
characteristics should hold true.
Page 7.6.3
4. EXERCISE – GAS SEPARATOR
We will now read in a process model for a gas separator. The inflow is a two-phase stream, with the gas
phase exhausted through a pressure control valve to a vent header. The valve is failopen; in the event of
a signal failure to the valve, the valve opens and relieves the vessel pressure to the vent header.
Select Process | Select Model. Highlight “Pressur1.mdl” and press Open.
When the model opens, press Pause. While in Pause, make the following changes:
Press Control | Control Options.
In the sub-panel labeled “Control Action” select Direct.
Scroll down until you see a sub-panel labelled “Reverse Output”. Select Yes.
Press Clear to remove the Control Options dialog box.
Press Load (in the Menu bar). In the sub-panel labelled “Auto Load Change” click in the field adjacent to
the label “Correlation”. Press the ESC key on you keyboard, then key in a value of 0.97. Press Enter on
your keyboard, then click on Clear.
Press Out and change the controller output to 35%.
Press Tune and enter the following tuning parameters.
Gain: 10.0
Press Run then put the controller in Auto.
The controller PV should come to a set point of 50.0 psig.
Change the set point to 48 psig. Satisfactory response? __________
Press AutoLoad. The simulation now represents a varying amount of gas flow into the separator.
Still a satisfactory response? __________
The controller output that is being trended represents the percent open condition of the valve. This is
NOT the same as the signal to the valve, however. Since the valve is failopen, then a 0% signal to the
valve causes the valve to be fully open, and a 100% signal to the valve causes the valve to be fully
closed. To achieve this, we reversed the output when we were in Control Options, above.
To see the actual signal going to the valve, press View | Variable Plot Selection. Select “Yes” for PV-2.
You should see a magenta trace which is the mirror image of the controller output (blue) trace.
OK? __________
Page 7.6.4
In many applications of this type, in the event of failure of the signal to the valve, the valve would go fully
open, relieving the gaseous contents of the vessel to the vent header. We will now simulate a signal
failure to the valve.
Press View | Discrete Controls. In the sub-panel labelled “D-Out#1”, select Off.
Observe that the magenta trace immediately goes to 0, indicative of a break in the signal line. The valve
opens fully (the actual valve position is unseen), causing the vessel pressure to drop. Because the
controller is still in Auto, it tries (in vain) to close the valve to correct for the low pressure, hence its output
drops.
Did you observe and understand all of these actions? __________
You have observed a gas pressure control loop for a large vessel which does not respond as rapidly as
the pressure station in Section 3. You have observed a typical configuration for a fail-open valve, and you
have observed the need for a direct acting controller, where, on an increase in PV, the controller output
must increase. (If you wish to continue experimenting with the pressure control system, select On in the
Discrete Control Panel, to reconnect the signal.)
Page 7.6.5
LEVEL CONTROL LOOP CHARACTERISTICS

OBJECTIVE: To illustrate the characteristics of level control loops, in particular the non-self-regulating
character, and methods of obtaining data useful in controller tuning.
PREREQUISITE: Completion of Exercise 1, Process Dynamic Characteristics
BACKGROUND: Whereas most processes exhibit some degree of self-regulation, level control
applications are usually non-self regulating. That is, unless the inflow and outflow are exactly equal, the
level will continue to rise or fall until the vessel overflows or becomes empty. For most self-regulating
processes, three parameters (process gain, dead time and time constant) can be used to approximate the
process dynamic characteristics. For most liquid level loops, one, or at most two, parameters characterize
the process. Instead of process gain, a relevant parameter is the tank residence time, TR. For a vertical
tank, the residence time of the vessel is given by:
Quantity of fluid between upper and lower level taps

TR =
Maximum flow rate through vessel
π d2 h
TR =
4Fm
where:
d = diameter of vessel (feet)
h = distance between upper and lower level taps(feet)
Fm = maximum controllable throughput through vessel (cu ft/min)
Notes: (1) 1 cu ft = 7.48 gallons.

(2) The residence time could also be computed using any compatible units, such as
metric units.
If the level controller is cascaded to a flow controller, then Fm is simply the maximum reading of the flow
transmitter, and d and h come from tank geometry.
If the level controller output goes directly to a valve, then TR for the current operating point can be found
by making a process test. In addition, another parameter, KV, the valve gain, can also be found from the
same test.
Test procedure:
1. When the tank is in a steady state condition, put the controller in Manual.
2. Change the controller output a slight amount.
3. The level should start to change at a constant rate. Before the level reaches an extreme
(either 0% or 100%), put the controller output back to its original value. The level should stop
changing. Your chart or trend record should look something like this:
Regulatory Control and PID loop tuning Exercise 7. Level control loop characteristics
4. Note the following:
ΔV = the amount you changed the controller output (signal to valve), %
ΔL = the amount the level changed, % of full scale
ΔF = the amount the flow changed, % of full scale
ΔT = the time duration of the test, minutes
5. Calculate the following:
Tank residence time:

∆T ∙ ∆F
TR =
∆L
Valve gain:
∆F
KV =
∆V

Start Windows.
Run PC-ControLAB.
Be sure the control strategy is Feedback and that the controller is in Manual.
Select Process I Select Model. Highlight ‘Level.mdl’ and click Open.
Select View I Variable Plot Selection. Select .Yes. for PV-2. Press Clear.
This displays the outflow as a magenta trace. It may not be visible immediately, since it is the same as
the inflow and is covered up by the grey trace labelled “Load”.)
Page 7.7.2
Click on the label above the trend chart reading ‘Inflow’. Note the maximum flow rate of inflow from the
scale at the right of the trend graph.
Maximum inflow, GPM: ________
Then click on the ‘Liquid Level (PV-1)’ label above the controller.
2. TESTING THE PROCESS

Decrease the controller output by 5% from its initial value.
Before the level reaches a limit, change the controller output back its original value. Then press Pause.
If you weren't fast enough, select Process I Initialize so you can try again. You will have to again select
View I Variable Plot Selection and select “Yes” for PV-2. This time, after you have made the initial
change to the controller output, press Pause. Then you can change the controller output back to its
original value while the program is suspended. Press Run to resume program operation.
With the controller output at the decreased value, does it appear that the level will ever come to
an equilibrium?
________
This illustrates the non-self-regulating nature of liquid level processes.
From your test, determine the following:
ΔT = the time duration of the test, minutes ________
ΔV = the amount of change in the controller output, % ________
ΔL = the amount the level changed, % of full scale ________
ΔF = the amount the flow changed, % of full scale ________
(Be sure to convert your flow reading to %. See the maximum flow factor determined at the end of
Section 1.)
Calculate the following:

Tank residence time:
∆T ∙ ∆F
TR =
∆L
____________ mins.
Valve gain:
∆F
KV =
∆V
_________________
Page 7.7.3
3. CALCULATING RESIDENCE TIME FROM TANK DIMENSIONS

Select Control I Select Strategy I Cascade.
With the Outflow controller selected (if the label reading ‘Outflow (PV-2)’ is NOT highlighted in red, click
on the label, or press Sel on the controller), note the maximum outflow rate (PV scale at right hand side of
trend chart).
________GPM.
The dimensions of this vertical tank are:
Diameter: 3.0 feet
Distance between level sensor taps: 5.0 feet
Calculate the tank residence time. (See equation in ‘Background’)
π d2 h
TR =
4Fm
T R = _________mins.
(Your answer should be close to the tank residence time you determined experimentally in Section 2. If
not, recheck your results, either here or in Section 2.)
The tank residence time, TR, and the valve gain, KV, will be used to calculate the level controller tuning
parameters in Exercise 12.
Page 7.7.4
PID CONTROLLER CHARACTERISTICS

OBJECTIVE: To demonstrate the characteristics of the proportional, integral (reset) and derivative
control modes in open loop operation, including definitions of the tuning parameters.
PREREQUISITE: Completion of PC-ControLAB tutorial (under Help | Tutorial ) or an equivalent


Start WINDOWS
Run PC-ControLAB
2. PROPORTIONAL MODE DEMONSTRATION
2.1 Set Up
Check the top line of the display. Be sure that you are using the GENERIC process model, the
FEEDBACK control strategy. Be sure that the controller is in Manual.
From the Menu Bar, select Control | Control Options.
For Control Algorithm, select Proportional Only.
While you still have the Control Options box on display, be sure the following options are selected:
Control Action REVERSE
Set Pt Tracking NO
Press Clear to remove the Control Options box.
Select Control | Measurement Options.
Select Yes for “Use substitute value instead of value from process sensor?”
Press Clear to remove the Measurements Option box.
Select View | Display Range | Percent of Span
Select View | Display Proportional Band. (An auxiliary bar display will appear between the strip chart
and the instrument faceplate. Its use will be explained later.
Regulatory Control and PID loop tuning Exercise 8. PID controller characteristics
2.2 Gain and Proportional Band Features

Press TUNE. Set or confirm the following values:
GAIN 2.0
MANUAL RESET 35
Click on the Options tab and select “Display proportional tuning parameter as” Prop Band
Return to the Tuning tab.
What value and what name does the Tuning Display now show in place of GAIN?
Value ________ Name ________________________________
Recall that:
100
Prop Band =
Gain
100
Gain =
Prop Gain
Press Clear to remove the Tuning dialog box.
Select View | Data Monitor.
Record the following present values
Process variable: ________
Set point: ________
Controller output: ________
For a Reverse Acting controller, the controller error is calculated as:
ERROR = SP - PV
and the top and bottom of the proportional band can be calculated from the present values of set point
(SP), manual reset (MR) and gain (or PB).
Calculate the PBtop and PBbot and confirm the figures from viewing the PB bar (CYAN coloured) at the
right of the strip chart.
PBtop Calculated: ________ Observed: ________
Pbbot Calculated: ________ Observed: ________
Page 7.8.2
Put the controller in AUTO. Change the Set Point to 65. Did the PB bar move as expected? ________
Calculate the theoretical controller output from:
OUTPUT = GAIN * ERROR + MR ________
Does the theoretical output agree with the observed output? ________
From an observation of the Proportional Band bar, what value of PV would cause the controller output to
go:
to zero? ________
to 100%? ________
(Recall that the Proportional Band is the range through which the PV must travel to cause the
controller output to change by 100%. For a reverse acting controller, as the PV drops from PBtop
to PBbot , the controller output goes from 0% to 100%.)
Select Control | Measurement Options. In ‘Substitute Value’ press ESC and then enter a substitute PV
value of 82.5 and then ENTER.
(At this point it is difficult to tell whether the controller output has been calculated at 0.0, or
whether it is actually some lower value, but "pegged out" at 0.0. Therefore, back off on the PV
slightly by entering a substitute value of 82. That should produce a controller output just slightly
above 0.0.)
Now enter a substitute PV value of 32.5.

(You can raise the substitute PV value slightly, say to 33, to demonstrate that the output value
was truly at 100.0, not just pegged out at the top of the chart.)
Select Control | Measurement Options. Click “Yes” for “Use substitute value ..., then enter a substitute
value of 70.0
Observe the PB bar and the present value of the PV. Make a visual estimate of the fraction of PB down
from PBtop to the PV value.
________
Does this (approximately) agree with the controller output? ________
2.3 Closed Loop Proportional Mode Response

Put the controller in MAN. Change the Set Point to 55. Change the controller output to 35.0. Change the
substitute PV value to 55. Then select Control | Measurement Options and select NO for “Use
substitute value … ?” (Now we will be using process feedback.) Now put the controller in AUTO.
Page 7.8.3
Change the Set Point to 65. When the control loop comes to equilibrium, note:
PV ________
SP ___ 65___
Controller output ________
Calculate the error, then calculate the theoretical controller output from the equation given previously.
ERROR = SP - PV = 65% - 70% = - 5%
OUTPUT = GAIN * ERROR + MR

= 2.0 * - 5% + 35% = 25.0%
Does this agree with the observed output? YES
The PV and SP do not agree. Which way, increase or decrease, should the Manual Reset be adjusted to
bring the PV into agreement with the SP? ________
Experimentally adjust the Manual Reset until the PV matches the SP.
Final value of Manual Reset: ________
Final value of Controller output: ________
Did the PB bar move as you adjusted the Manual Reset? ________
1. The relationship between controller Gain and Proportional Band width.

2. The relationship between the Set Point, Manual Reset and Proportional Band position.
3. The relationship between Proportional Band position, PV and controller output, when the controller
is in AUTO.
4. That Proportional-Only control will not (usually) cause the PV and SP to agree, unless Manual
Reset is adjusted.
5. That adjusting the Manual Reset (in AUTO) is equivalent to shifting the Proportional Band.
3. INTEGRAL MODE DEMONSTRATION

3.1 Set Up
Select View and click on Display Proportional Band (to remove it).
Put the controller in MAN.
Select Control | Control Options and set or check the following options:
Control Algorithm: PID, Non-Interacting
Action, Direct or Reverse: Reverse
Set Point Tracking? No
Deriv on Error or Meas: Meas
Page 7.8.4
Prop on Error or Meas: Error
Press CLEAR
Go to Control | Measurement Options and set:
Use substitute value: YES
Substitute measurement value: 50
Press Tune | Options and set the following:
Display the proportional tuning parameter as GAIN
Return to the Tuning tab and set
Gain: 1.0
Reset, Min/Repeat: 5.0
Deriv, Mins: 0.0
Set the following:

Set Point: 50
Controller Output: 40
Change to AUTO.
(You should now see the control loop simulation in stable operation with measurement
and set point both at 50% and the valve signal at 40%.)
3.2 Integral Mode Response

Go to Control | Measurement Options and enter:
Substitute measurement value 60
Observe:
The controller output signal immediately drops from 40% to 30%.
(Reason: The controller is reverse acting, so a measurement increase causes a controller output
decrease. The GAIN is 1.00, so the change in controller output due to proportional action is 1
times the measurement change.)
The controller output then begins to ramp downward, at the rate of 2% per minute.
(Reason: The proportional mode change was 10%)

A 2%/minute change is equivalent to 5 minutes per 10% change, or 5 minutes per repeat of the
proportional response.
Note that the ramping action continues until the output reaches a saturation limit (0%). This
condition is called “reset windup.”
Page 7.8.5
3.3 Integral Mode Response, A Second Example

Enter a substitute measurement value of 50.
Enter a controller output of 40.
Put the controller in AUTO.
Be sure the controller is in a steady state condition, with the measurement at 50% and controller output
signal at 40%
Now, enter a substitute measurement value of 55%.
Observe:
The proportional response is now only 5%.
The controller output ramps downward at 1%/minute. This rate is equivalent to 5

minutes for 5% change, or again, it is 5 minutes per repeat of the proportional
response.
4. DERIVATE MODE DEMONSTRATION
4.1 Set Up
In the Tune | Options tab, set the following:
Reset Action Off
Derivative Gain 100.0
Enter the following tuning values:
Gain: 0.5
Reset, Min/Repeat: Not applicable (it’s OFF)
Deriv, Min: 0.0
Enter the following operational values:
Set Point: 50
Controller output: 40
In the Control | Measurement Options, select Yes for “Use substitute value …?”, then enter a substitute
measurement value of 50
Change the controller mode to AUTO.
Verify that the controller is in stable operation with set point and process variable at 50, and controller
output signal at 40.
Page 7.8.6
4.2 Ramp Response with Proportional Mode Only

Under Control | Measurement Options, select Yes for “Use ramped value instead of value from process
sensor?”
Enter a ramp rate of 4%/minute.
Select Up.
Observe:
The process variable ramping upward from its initial value of 50% at the rate of 4%/minute.
The controller output ramping downward from its initial value of 40% at 2%/minute, stopping at
15% when the measurement reaches 100%.
(Reason: It is a reverse acting controller, so as measurement increases, the controller output

decreases. The Gain is 0.5, so the controller output drops half as fast as the measurement rises.)
4.3 Ramp Response with Proportional and Derivative Modes

In Control | Measurement Options select No for “Use ramped value …?”
Change the controller mode to MAN.
Enter a controller output signal of 40
In Control | Measurements Options enter a substitute measurement value of 50
Change the controller mode back to AUTO.
Press Tune and change the Deriv setting to 5 minutes.
Verify stable operation at a measurement of 50 and controller output signal of 40
In Control | Measurement Options, enter the same ramp parameters as before:
Rate: 4%/minute
Direction: UP
then select Yes for Use ramped value…?”
Observe:
The measurement ramping as before.
The controller output signal makes an almost immediate change from 40% to 30% then ramps
downward at the rate of 2% per minute to 5%, when the measurement has reaches 100%. As
soon as the measurement ramp stops, the controller output signal makes an almost immediate
change from 5% to 15%, then remains constant.
Page 7.8.7
Reason: The derivative contribution is:
- Gain x Deriv time x Meas ramp rate
- 0.5 x 5minutes x 4%/minute
= 10 %
If the controller had no derivative (as in the first trial), you can estimate approximately
how long, from the initiation of the ramp, was required for the controller output signal to
change from its initial value (40%) to some chosen value (say, 20%). Your estimate
should be approximately 10 minutes, with proportional control only.
With derivative (as in this trial), estimate the time required for the controller output signal
to change from its initial value (40%) to the same point (20%). Your estimate should be
approximately five minutes.
Conclusion: With a derivative time of 5 minutes, the controller output signal should lead
(or get to the same chosen point) by 5 minutes its behavior with proportional control
only.)
Repeat this part of this laboratory exercise, this time with a measurement ramp rate or a derivative time of
your choice. You should observe:
The “ – “ sign is due to the fact that the derivative contribution is always in a
direction which opposes the direction of measurement change.
4.4 Derivative Gain and Set Point Changes Using Derivative (Optional)
A block diagram for the classical form of PID controller is shown below.
Many commercial manufacturers of controllers, however, place a filter, in the form of a first order lag, in
the derivative portion of the controller. This results in the modified block diagram shown below:
Page 7.8.8
The usual practice is for the time constant of the first order lag to be some fraction of the derivative time
setting. This is accomplished by dividing the time constant by a factor known as “derivative gain” (DG). In
commercial controllers, the value of DG is usually in the range of 8 to 15.
The filter in the derivative portion has two purposes:
1. If there is measurement noise, the filtering action minimizes the amount of noise amplification
placed on the controller output by the derivative unit.
2. With the classical form of PID, a step change in set point will result in a spike in the controller
output caused by the derivative unit. Theoretically this spike will be of infinite height and
infinitesimal width. As a practical matter, the spike height is limited by the maximum or minimum
controller output. With the filter added, the spike height following a step set point change will be
the value of DG times the proportional response. The output will then decay to normal with the
time constant of the first order lag.
Select Process | Initialize (This returns all options to their original settings.)
Select Control | Control Options and set Derivative on Error. (The default setting is Derivative on
Measurement.)
Gain 2.0
Reset, Minutes/Repeat(reset) 8.0
Derivative, Min 2.0
Through the Tune | Options tab make the following setting:
(This large of value for derivative gain essentially eliminates the filtering action on the derivative.)
Change the controller mode to AUTO.
The controller should be in stable operation with the following values:
Set Point 55
Process Variable 52.5
Controller Output 35.0
Change the set point to 57
Observe:
Even with this moderate set point change, the controller output “spikes” all the way to 100%, then
very quickly drops back to a normal control range. In a real world situation, you probably would
not want to send a severe shock like that to your process.
Now change the set point back to 55
Page 7.8.9
Through the Tune | Options tab, set:
Change the set point to 57.
Observe:
This time the controller output spike is limited. The immediate change in controller output should
be:
SP Change x Controller Gain x Deriv Gain
2 x 2.0 x 10 = 40%
(Quantization error in the digital algorithm may cause some deviation between the actual value and the
theoretical value):
Also note that the return to the normal operating range is not as abrupt.
Change the set point back to 55
Change the controller to MAN
Select Control | Control Options and set Derivative on Measurement
Change the controller to AUTO
Change the set point to 57
Observe:
With derivative on measurement, the immediate response of the controller output is the
proportion response only – there is no derivative spike. Now one of the purposes of the filter on
the derivative component has been removed, but the filter is still useful in minimizing the effect of
measurement noise being amplified by the derivative unit.
Page 7.8.10
PID TUNING FROM OPEN LOOP TESTS

OBJECTIVE: To provide practice in open-loop testing, estimation of process parameters and
calculation of tuning parameters from the open loop test data.
PREREQUISITE: Completion of Exercises

1. Process Dynamic Characteristics
2. PID Controller Characteristics or equivalent level of instruction or experience

Start Windows.
Start PC-ControLAB.
If PC-ControLAB is already running, then re-read the “GENERIC” process model to initialize the program:
From the menu bar, select Process | Select Model
Highlight “Generic.mdl” and press Open
Note: The Generic model will not produce similar tuning results from Open Loop testing (this Laboratory
Exercise) and from Closed Loop testing (Laboratory Exercise 10.) If similar results are desired, use
Generic2 model rather than Generic.
Confirm the following:
Process: GENERIC (see the top line, left hand side)
Control Strategy FEEDBACK (see the top line, right hand side)
2. OPEN LOOP TESTING

This program begins operation with a PV of 275 °F, a controller output of 35% and a load variable (feed
rate) of 300 GPM. Assume that this is the normal operating point for this process. If the right hand scale
of the grid is not in engineering units, then select View | Display Range | Engineering Units
With the controller in Manual, change the output to 45%. (On a real process, you may not be able to
make that much change in controller output. Estimate (see Figures 1 and 2 at the back of this exercise for
methods of estimating process parameters.)
∆PV (%)
Process gain Kp (%) =
∆OP (%)
___________
Dead time (θ) ___________
Time constant (τ) ___________

Regulatory Control and PID loop tuning Exercise 9. PID tuning from open loop tests
Return the controller output to 35%.
Calculate tuning parameters for a P, PI and PID controller, using the Ziegler-Nichols equations.
(See Table 1 at the back of this Exercise.) Enter these in the table below.
For each type of controller, enter the parameters, put the controller in Auto and test the loop for a set
point change. (Suggestion: Change the set point, either up or down, by 50 °F. On the job, you probably
cannot make that large of change.)
NOTE: For the Proportional Only controller, with the controller in Manual, select Control | Control
Options, and select the Proportional only control algorithm. For the PI and PID controllers, with the
controller in Manual, select Control | Control Options, and select PID-Non-interacting control algorithm.
Calculate or measure the decay ratio, period and (for the PI controller only) the period-to-integral time
ratio. (This will be used in a subsequent Laboratory Exercise.)
For each type of controller, make a 5% (40 GPM) load change. Press StepIncr or StepDecr.
Which controller type, PI or PID, do you prefer, for:
Set point response __________
Load (disturbance) response __________
PID control appears to be better than PI for both a set point change and a load change. However, what
you have seen so far is a perfectly noise-free process. Real processes usually have some noise on the
measurement.
Page 7.9.2
Enter the PID tuning parameters from the table above, then, with the controller in Auto:
Select Process | Change Parameters
Use the scroll bar to scroll down until you see
“PV#1 Meas Noise 0-N; 1-Y”
Highlight that, then enter “1"
Press OK then Clear
Describe what you see:________________________________________________________________
___________________________________________________________________________________
Frequently the measurement signal is filtered to “hide” the effect of the noise. (The noise is still there -
you just can’t see it!) We will put a filter on the measurement. This is equivalent to implementing a
software filter on the analog input block on a DCS. EXCEPT: Due to the relatively slow sampling rate of
this simulation, as compared with the typical sampling rate of a DCS, we will use a larger filter time
constant here than would be used in real life.
Select Control | Measurement Options in the drop down menu
Click in the circle “Yes” in the section “Measurement Filter”
Enter “1.0" (minutes) for the filter time constant
Press Clear
Describe the effect:__________________________________________________________________
__________________________________________________________________________________
Page 7.9.3
∆PV (%)
Process gain Kp (%) =
∆ Valve (%)
Draw a tangent at the point of steepest rise. Be sure it intersects the initial equilibrium line; it is not
necessary to carry the tangent all the way until it intersects the final equilibrium line.
Dead Time: θ = Time from valve change to intersection of tangent with initial
equilibrium line.
Time Constant: τ = Time, from end of dead time (as determined above) until the process
changes by 63.2% of its final amount.
Page 7.9.4
PID TUNING FROM CLOSED LOOP TESTS

OBJECTIVE: To provide familiarity with closed loop testing, determination of process data and the
calculation of tuning parameters from the closed loop data.
PREREQUISITE: Completion of Laboratory Exercise 9, PID Tuning from Open Loop Tests
BACKGROUND: Tuning by closed loop process testing involves putting the controller in
AUTOMATIC, removing all Reset and Derivative, and setting the Gain just high enough to cause a
sustained process oscillation. From this test, the relevant parameters are the period of oscillation in
minutes, and the Gain which ultimately caused the sustained oscillation. These are called the “ultimate
Period” and “ultimate Gain”, respectively. From this data, the tuning parameters can be calculated.
NOTE: The open-loop (the subject of Exercise 9) and the closed-loop methods (the closed-loop method
is the subject of this exercise) may or may not produce similar tuning values, even when using the same
process model. The “Generic” process model is used in these exercises; there is some difference in the
results. For demonstration purposed, is you wish to obtain similar results from t he two methods, use the
Generic2” model.

Start WINDOWS
Run PC-ControLAB
2. TUNING BY CLOSED LOOP PROCESS TESTS.
2.1 Setup
Process: GENERIC (see the top line, left hand side)
Control Strategy FEEDBACK (see the top line, right hand side)
If PC-ControLAB has already running, you may have to do any or all of the following:
Select Control | Select Strategy | Feedback
Select Process | Select Model
Highlight “Generic.mdl” and press Open
Select Process | Initialize to initialize the process model.
Press Tune then select the Options tab. Select Reset Action OFF.
Regulatory Control and PID loop tuning Exercise 10. PID tuning from closed loop tests
2.2 Process Testing

Set: Gain: 1.0
Reset: Off
Deriv: 0.0 minutes
Make a set point change of 5% of full scale. (Press StepIncr once.)
If there is no oscillation, or if the oscillation dies out, increase the Gain (or decrease the Proportional
Band) and repeat the set point change. (The Gain can initially be changed approximately 50% of its
present value, or the PB can be changed to one-half of its present value. As the response gets closer to
sustained oscillation, smaller changes should be made.) You should not have to observe the response for
more than three cycles to determine whether or not the oscillation is decaying or not.)
When sustained oscillation is ultimately achieved, record the following:
Gain required for sustained oscillation Kcu = __________
Period of sustained oscillation Pu = __________
Use the table for the closed-loop Ziegler-Nichols method (Table 1 at the back of this exercise) to calculate
tuning parameters for a P, PI and PID controller. Enter these in the table below:
(First calculate Gain (KC), Integral time (TI) and Derivative (TD) from the equations. Then, if your system
uses PB rather than Gain, or Reset Rate rather than Reset Time, calculate those values.)
Before testing for the closed loop response, go to Tune | Options tab and set Reset Action ON.
For each type of controller, enter the parameters, put the controller in Auto and test the loop for a 10% (of
full scale) set point change.
Calculate or measure the decay ratio, period and (for PI controller only) the period-to-integral time ratio.
(This will be used in a subsequent exercise.)
Also, for each type of controller, make a 5% load change. (Press StepIncr or StepDecr.) Mark which
controller type has the best, and the worst, response to a load change.
Page 7.10.2
Regulatory Control and PID loop tuning Exercise 10. PID tuning from closed loop tests
Page 7.10.3
IMPROVING “AS FOUND” TUNING

OBJECTIVE: To provide practice in improving “as found” tuning for proportional plus integral (PI)
control of a self-regulating process.
PREREQUISITES: Completion of Exercises

9 PID Tuning from Open Loop Tests
10 PID Tuning from Closed Loop Tests
BACKGROUND: Quite often a control systems engineer or instrumentation technician is called upon
to improve the behaviour of a loop that is currently in operation, but without resorting to either the open
loop or closed loop testing methods. Assuming that the loop is not behaving acceptably at present, and
that process and equipment problems (e.g., sticking valve) have been eliminated, then most persons
resort to “trial and error” tuning. For novice tuners, this is often simply an exploratory procedure; “How
about changing this knob in this direction and see what happens”.
This laboratory exercise presents a method for directed trial and error tuning, where each tuning
parameter change is made for a deliberate reason. The objective is to go from the current unacceptable
behaviour to acceptable behaviour as efficiently as possible; i.e., in the fewest number of tuning
parameter changes.
This method is based upon the premise that if a PI controller, controlling a self-regulating process, is well
tuned (that is, exhibiting a slightly underdamped oscillation with a quarter wave decay), then there will be
a predictable relationship between the period of oscillation (P) and the integral time (TI). This relationship
(stated in three different ways) is:
P
1.5 ≤ ≤2.0
TI
1.5 TI ≤ P ≤2.0 TI
1.5 P ≤ TI ≤2.0 P
I
This premise leads to the following rule-based procedure:
1. If the loop is not oscillating, decrease the gain, say by 25 to 50%.
2. If the loop is oscillating then:
2.1. If the Period is between 1.5 and 2.0 times the integral time (or the period-to-reset ratio is
between 1.5 and 2.0), then either increase or decrease the gain as required to obtain the desired
decay ratio (such as quarter wave damping)
2.2. If the Period is greater than 2.0 times the integral time (or the period-to-reset ratio is greater
than 2.0), then choose a new integral time according to the criterion:
0.5P ≤ T I ≤ 0.67P
Regulatory Control and PID loop tuning Exercise 11. Improving ‘as found’ tuning
2.3. If the Period is less than 1.5 times the integral time (or the period-to-reset ratio is less than
1.5), then:
2.3.1. If the decay ratio is greater than 1/4, then decrease the gain, say by 25 to 50%,
depending upon how much the decay ratio exceeds 1/4.
2.3.4. If the decay ratio is less than, or approximately equal to, 1/4, then choose a new
integral time, using the criterion given in 2.2.
3. After each adjustment, make a slight set point change to test the response to the latest
combination of tuning parameters.

Start Windows
Start PC-ControLAB
2. LOOP TUNING
Select Control | Retrieve Strategy and Tuning. Highlight “Feedbck1” (not “Feedback”) and
press Open.
Observe from the top row that this opens the normal Feedback control strategy, as well as the Generic
process model which you have worked with in previous laboratory exercises. The thing that is different
here is that the loop has already been tuned – for better or for worse. Press Tune and note the existing
tuning parameters.
If the PV scale is not in engineering units, select View | Display Range | Engineering Units.
Put the loop in AUTO.
Change the set point to 300 °F.
NOTE: The following two procedures will NOT product the same response:
Make the set point change first, then put the loop into AUTO.
Put the loop in AUTO first, then make the set point change.
For the purpose of the procedure described in this laboratory exercise, it is important for you to see the
set point response with the loop already in AUTO. Therefore, the correct procedure is to put the loop into
AUTO then make the set point change.
Observe the response. Suppose that this is the behavior of the loop when you are asked to
make tuning parameter changes. In other words, this is your “as found” condition.
Does the loop need to be retuned? ________
Page 7.11.2
If so, list the “as found” conditions in the top line of the table below. Then use the procedure listed
in “BACKGROUND”, or use the flowchart, to make tuning parameter changes. Keep track below
of each tuning change you make. (Suggestion: Use set point values of 275 °F and 300 °F.)
3. A COMPARISON
Many (novice) loop tuners, faced with the “as found” condition, would simply reduce the controller Gain
until acceptable damping (e.g., quarter-wave decay) was achieved. We will demonstrate why that may not
be a good idea.
3.1 Re enter the original tuning parameters.
With the loop in Automatic, set the set point at 275 °F and let the loop come to equilibrium. Working
between set point values of 275 and 325 °F, adjust the Gain until quarterwave damping is achieved. Do
not change the reset.
Gain: ________
Change the set point to 275. When the loop comes to equilibrium, make a 10% disturbance (load change)
by clicking twice - rapidly - on StepIncr.
Record:
Maximum deviation of PV from set point: ________
How long before loop “settles down” to within 1/2% (± 2.5 Deg) of SP ________
Page 7.11.3
3.2 Enter your final tuning parameters from part 2.0. (These should have produced a quarter wave decay
response following a set point change, and a period-to-reset ratio which meets the criterion stated in
“BACKGROUND”.
With the loop in Automatic, set the set point at 275 °F and let the loop come to equilibrium.
Make a 10% disturbance (load change) by clicking twice - rapidly – on StepIncr.
Record:
Maximum deviation of PV from set point: ________
How long before loop settles down to within 1/2% of SP ________
3.3 Which tuning combination produces the best response to a load change?
Page 7.11.4
TUNING LIQUID LEVEL CONTROL LOOPS

OBJECTIVE: To illustrate the characteristics of level control loops, in particular the non-self-regulating
character, and to illustrate several different tuning philosophies.
PREREQUISITES: Completion of the following exercise:

7 Level Control Loop Characteristics
BACKGROUND: Whereas most processes exhibit some degree of self-regulation, level control
applications are usually non-self regulating. That is, unless the inflow and outflow are exactly equal, the
level will continue to rise or fall until the vessel overflows or becomes empty.
For most self-regulating processes, three parameters (process gain, dead time and time constant) can be
used to approximate the process dynamic characteristics. For most liquid level loops, one, or at most two,
parameters characterize the process. Instead of process gain, a relevant parameter is the tank residence
time, TR. For a vertical tank, the residence time of the vessel is given by:
Quantity of fluid between upper and lower level taps

TR =
Maximum flow rate through vessel
π d2 h
TR =
4Fm
where:
d = diameter of vessel (feet)
h = distance between upper and lower level taps(feet)
Fm = maximum controllable throughput through vessel (cu ft/min)
Notes: (1) 1 cu ft = 7.48 gallons.

(2) The residence time could also be computed using any compatible units, such as
metric units.
If the level controller is cascaded to a flow controller, then Fm is simply the maximum reading of the flow
transmitter, and d and h come from tank geometry.
If the level controller output goes directly to a valve, then TR for the current operating point can be found
by making a process test. In addition, another parameter, KV, the valve gain, can also be found from the
same test.
The significance of these parameters is that from them plus the tuning parameters, gain (KC) and integral
time (TI ), the behaviour of the level control loop can be predicted. Alternatively, if a certain behavior is
desired, such as a particular decay ratio, then the required tuning parameters can be calculated.
Regulatory Control and PID loop tuning Exercise 12.Tuning liquid level control loops
Specifically, the parameter group
K C K V TI
[2]
TR
for applications where the level controller output connects directly to a valve, or
K C TI
[3]
TR
for applications where the level controller is cascaded to a flow controller.
To avoid duplication, we will use equation [2] for both circumstances, with the understanding that for
cascade applications, KV should be set equal to 1.0.
KC KV TI
If > 4.0, the level control loop will be overdamped.
TR
KC KV TI
If = 4.0, the level control loop will be critically damped.
TR
KC KV TI
If < 4.0, the level control loop will be underdamped.
TR
.
Two special cases.
KC KV TI
If = 0.185 (or approximately 0.2), the level control loop will respond with a
TR
quarter-decay
KC KV TI
If TR
= = 0.743 (or approximately 0.75), the level control loop will respond with
a 1/20 decay ratio (the second peak overshoot will be 0.05 times the first overshoot).
In most liquid level control applications, the set point is rarely changed. Instead, the purpose of the
control loop is to maintain the level at (or acceptably near) set point in the presence of load changes.
Hence we will focus on the response to a load change, rather than to a set point change in this exercise.

Start Windows.
Run PC-ControLAB.
2. WHY LIQUID LEVEL CONTROL IS DIFFERENT

Press Control I Retrieve Strategy, Model and Tuning. Highlight .Level.stg. and press Open.
Be sure you followed those instructions exactly, else redo them.
Page 7.12.2
The simulation is of a liquid level controller cascaded to a flow controller. All we will work with in this
exercise is the level controller. Both controllers are initially tuned.
Select View I Variable Plot Selection. Select No for Load-2.
Press Sel on the Level controller (left hand controller).
Select Load from the Menu Bar.
Click in the field labeled Min. Load, then press Esc.
Type in .20.0. and press Enter on the keyboard. Then click Clear.
Press Tune. Note the initial tuning parameters for the level controller.
Gain = __________
Reset = __________ minutes/repeat
Record the value for tank residence time and valve gain found from Exercise 7.
TR __________ minutes
KV _______________
Press Casc on the Flow controller (right hand controller).
Press Auto on the Level controller (left hand controller.)
Press Sel on the Level controller, then press StepIncr twice in rapid succession. This causes a 10%
step change in inflow.
Observe the response of the level. Does this look like approximately a quarter-decay? __________
From the tank residence time and controller tuning, what type of response should we expect? (Since in
this application, the level controller is cascaded to a flow controller, use Kv = 1.0, rather than the value
found in Exercise 7.)
Page 7.12.3
____________________________________________________________________________________
Suppose you want less oscillation? A conventional approach would be to reduce the gain. With the Level
controller still Selected, change the controller gain from 1.0 to 0.5.
Press StepDecr twice in rapid succession. (Same size load change, only in the opposite direction.)
Did that decrease the oscillation? __________
Reduce the gain by half again. Change the Level controller gain from 0.5 to 0.25, then press StepIncr
twice in rapid succession.
Did that decrease the oscillation? __________
What happened to the response as the gain was decreased? ______________________________
This has illustrated the fact that rules of thumb often used for other types of loops are counter-intuitive
when applied to liquid level loops.
3. CALCULATION OF TUNING PARAMETERS

Suppose the maximum anticipated step change in inflow (ΔFin) is 10%. Also suppose that the allowable
change in level (ΔL) is 5%. Use Table 1 below to calculate tuning parameters for each type of response,
critically damped, 0.05 decay ratio and quarter decay ratio. (Use Kv = 1.0.)
DECAY RATIO KC TI
Critically Damped ________________ ________________
0.05 ________________ ________________
0.25 ________________ ________________
Page 7.12.4
NOMENCLATURE
ΔFin = Maximum change in inflow, % of full scale
ΔFo = Maximum change in outflow, % of full scale
KC = Controller gain
KV = Valve gain (for cascaded loops, use KV = 1.0
ΔL = Maximum allowable level deviation from set point, % of full scale
P = Period of oscillation, minutes
TaF = Outflow arrest time, minutes. Time of maximum deviation of outflow from inflow
TaL = Level arrest time, minutes. Time when level first reaches maximum deviation from set point
TI = Controller reset time (minutes per repeat)
TR = Vessel residence time
4. TESTING THE RESPONSE
For each type of response, enter the parameters you calculated. When the loop is in equilibrium, make a
10% step change in inflow by pressing Steplncr or StepDecr twice in rapid succession.
(DON’T go above 85% load, to avoid saturating the outflow.)
Use Table 2 to spot check a few of the parameters below, to see if the predicted value comes close to
what you actually observe:
Decay Level Arrest Period of Outflow Arrest Max Outflow

Ratio Time Oscillation Time Change
Crit. Damped ____________ ____________ ____________ ____________
0.05 ____________ ____________ ____________ ____________
0.25 ____________ ____________ ____________ ____________
Considering both performance of the level control and the effect on outflow (this may be potential
disturbance to a downstream process unit), which form of response do you like best?
Crit. Damped ________ 0.05 Decay Ratio ________ 0.25 Decay Ratio _________
We will now go to a control strategy without the cascaded flow controller.
Page 7.12.5
Select Control I Select Strategy I Feedback.
Select Control I Control Options. Select Direct Acting for Control Action.
Select Load.
In the sub-panel labeled .Auto Load Change., click in the field adjacent to the label ‘Correlation’.
Press ESC on your keyboard.
Key in 0.90 and press Enter on your keyboard.
Press Clear.
From Table 1, calculate parameters for the 0.05 decay ratio. Use KV value found in Exercise 7.
Gain ____________
Reset ____________ minutes/repeat
Press AutoLoad. The simulation now exhibits a randomly changing load. Observe this for at least two full
screen changes. What is the maximum deviation from set point?
________
Press AutoLoad again to discontinue random load changes.
5. NON-LINEAR LEVEL CONTROL

Another level controller tuning approach favored by many is non-linear, often called “error squared”
control. By a slight modification to the control algorithm, the controller appears to have a low gain when
the level is near set point, with an increasing gain as the deviation from set point increases.
Put the controller in Manual.
Select Control I Control Options. For the option ‘Error Squared’ select Yes.
Press Clear.
The modified error is computed as
before being passed to the PID section of the algorithm.
Enter (or retain) the tuning parameters for the 0.05 decay ratio from the previous section.
Press AutoLoad to activate random load variations.
Page 7.12.6
After observing the response for about an hour (simulated time) estimate the following:
Maximum deviation from set point: ________
Note that there is much more deviation in level. You should also observe that there is a considerably
lower frequency activity in the controller output. If the outflow were the feed rate to a downstream process
unit, then the disturbances to that unit would be much less severe than with ordinary PI control. This is
accomplished at the expense of greater fluctuation in the level measurement. If the deviation is too great,
you can increase the controller gain and shorten the reset time, recognizing that there will be an increase
in the outflow activity, consequently an increased disturbance to the downstream process unit.
Enter the following:
Gain: 10.0
Reset: 30 minutes/repeat
After observing the response for an adequate amount of time, estimate the maximum deviation
from set point:
________
Put the controller in Manual and press AutoLoad to deactivate random load changes.
Press Pause.
6. AVERAGING LEVEL CONTROL

Still another approach to the tuning of liquid level control loops is called "averaging. control. No attempt is
made to keep the level at set point. Rather, the level is maintained within bounds, and is guaranteed not
to exceed those bounds.
Select Control I Control Options.
In the .Control Algorithm. sub-panel, select .Proportional Only..
In the .Error Squared. sub-panel, select .No..
Suppose we want to maintain the level within ± 10% of set point (that is, between 40 and 60%).
That would require a proportional band of 20%, or a gain of 5.
Press Tune and set:
Gain: 5
Manual Reset: 50
Press Run.
Put the controller in Auto. (AutoLoad should now be OFF.)
Record the following: Set Point: __________%
Process Variable: __________%
Error( = PV - SP): __________%
Page 7.12.7
Controller Output: __________%
Calculate the theoretical controller output from the Proportional-Only equation:
Output = Gain x Error + Manual Reset __________%
Does the theoretical value agree with the actual output value? ____________
Press StepIncr until the load (inflow) is just below 100%. What is the PV? __________%
Press StepDecr until both the load and the controller output are above 0%. PV? __________%
Press AutoLoad to activate random load changes. Observe the response for some time.
Does the PV remain at set point? ____________
Maximum value of PV? ___________% Minimum value of PV? __________%
We are guaranteed of keeping our level within the desired bounds of 40 . 60%. The penalty we pay is
that, even with a constant load, the level will not be at set point. This control scheme shows up very well
for surge tank level control, where we can tolerate more variation in level. Suppose we allow the level to
vary ± 25% (or from 25% to 75%). This would require a proportional band of 50%, or a controller gain of
2.0.
Change the gain to 2.0.
With AutoLoad on, observe the response for some time. Does the level remain within the
bounds of 25% to 75%?
__________
Page 7.12.8
SCHEDULED TUNING
OBJECTIVE: To demonstrate the benefits of scheduled tuning for non-linear processes.
PREREQUISITES: Completion of Exercises

11 Improving .As Found. Tuning
BACKGROUND: If a process is highly non-linear and is operated over a wide operating region, then
either the tuning parameters must be changed for different operating conditions, or else the controller will
be tuned for the .worst case. condition, resulting in less than optimum tuning at other conditions. One
means of coping with this is to issue instructions to the operator to change tuning parameters for different
operating conditions. Another possibility is to develop a set of tuning parameters for each operating
condition and employ some mechanism which applies the correct tuning parameter set for each condition.
This is known as .scheduled tuning. (also sometimes called .adaptive gain.). A basic premise is that the
operating condition can be indexed by a single variable, such as the measured process variable, the set
point, controller output, or a disturbance variable to the loop. This variable can then become the key to
finite regions of the operating zone, each with its own set of tuning parameters. A slight variation of the
above is to assign nominal parameters, then apply a unique multiplying factor to each of the parameters
for each defined region. This is the approach used by PC-ControLAB.

Start Windows
Start PC-ControLAB
2. INITIAL TUNING
2.1 Preparatory
Confirm that you are running the Feedback control strategy.
Select Process | Select Model. Highlight .Temp.mdl. and press Open.
2.2 Tuning for Base Conditions

Notice that the PV scale is 0 to 100 °C, that the set point is at 60 °C, and that the load variable (process
flow rate) is 0 to 100 m3/hr. For this exercise, we will assume that the base conditions (normal operating
point) is close to the current set point value of 60 °C with a load variable of 75% of full scale. The load
variable can vary, however, between 50 and 100% of its full range.
We’re going to be making fairly small set point changes, so press Zoom and change the vertical scale of
the PV from 0 . 100 to 55 . 65 °C. (Why small SP changes? To avoid having the controller output hit a
limit when we are working at high loads.)
Regulatory Control and PID loop tuning Exercise 13.Scheduled tuning
With the controller in Manual, change the controller output from 48% to 53% (a 5% change). Use the
open loop testing method (Laboratory Exercise 9) to determine an initial set of tuning parameters for a PI
controller.
Enter these parameters. Put the controller in AUTO. Fine tune these parameters if necessary, using the
‘Improving As Found Tuning’ method (Laboratory Exercise 11).
Make your set point changes between 60 and 62 °C.
When you have determined satisfactory tuning parameters for this condition (load variable = 75 m3/hr),
record them below:
Process Flow: 75 m3/hr. Gain: ________ Reset: ________ min/rpt.
Now make load changes of 5% (Press StepDecr once. After the PV settles out, press StepIncr once.)
and observe the response.
Return the Set Point to 60 °C.
2.3 Other Operating Points

Change the process load to 95% of full scale. (Press StepIncr 4 times).
Change the SP to 62 °C, then to 60 °C. Observe the responses.
Also make load changes of 5% (press StepDecr once. After the PV settles out, press StepIncr once.)
Observe the responses.
.
How do these responses (to SP and load changes) compare with what you observed when the load was
at 75%?
_______________________________________________________________________
Change the process load to 55% of full scale. (Press StepDecr 8 times).
Change the SP to 62 DegC, then to 60 DegC. Observe the responses.
Also make load changes of 5% (Press StepDecr once. After the PV settles out, press StepIncr once.)
Observe the responses.
How do these responses (to SP and load changes) compare with what you observed
when the load was at 75%?
_______________________________________________________________________
What you should have observed so far: Starting with the controller tuned acceptably when the load is
75%, a significant load increase with the same tuning parameters causes the control loop to be slightly
more sluggish. A significant load decrease with the same tuning parameters causes the loop to be to
aggressive.
Page 7.13.2
2.4 Tuning for Other Operating Points

While we are at a load of 55%, adjust either the Gain or the Reset, or both, until you get approximately
the same type of response for a SP change from 60 to 62 as you did when the load was 75%. Record
your tuning parameters.
Now go to 95% load and adjust the tuning parameters to get approximately the same type
of response as you did when the load was 75%.
2.5 Calculation of Tuning Schedule

Suppose we divide the operating zone into 3 regions, based upon the load value.
Region 1 Load < 65%. (Actually, we never use a load less than 50%.)
Region 2 Load between 65% and 85%.
Region 3 Load ≥ 85%.
Since we said that in our base case (See Section 2.2) the load was 75%, then Region 2 is our base
operating region. We will enter tuning parameters for that region as our nominal parameters, and adjust
the parameters for the other regions by multiplicative factor. (The multiplicative factor for Region 2 will be
1.00.)
Calculate the Gain multiplier and Reset (Min/Rpt) multiplier for Region 1 and Region 3.
Gain determined at 55% load (Section 2.4)

Region 1 Gain Multiplier = = ________
Gain determined at 75%load (Section 2.2)
Reset determined at 55% load (Section 2.4)

Region 1 Reset Multiplier = = ________
Reset determined at 75%load (Section 2.2)
Gain determined at 95% load (Section 2.4)

Region 3 Gain Multiplier = = ________
Gain determined at 75%load (Section 2.2)
Reset determined at 95% load (Section 2.4)

Region 3 Reset Multiplier = = ________
Reset determined at 75%load (Section 2.2)
Enter these values into the table below. (Note that the Deriv multipliers can be left at the default value of
1.0, since the base Deriv value is 0.0. Also note that Region 2 represents our base case, so those
multipliers are 1.0.)
Page 7.13.3
Press Tune | Schedule tab.
Enter, or confirm, that the Number of Regions is set at 3.
Select PV-4 for the Key variable. (This process model has PV-4 connected directly to Load-1, which
represents the process flow disturbance.)
Fix up the table to resemble columns 2 - 6 listed above.
Select Scheduled Tuning ON.
Return to the Tuning tab. Re-enter the tuning values you determined for the base case (Section 2.2).
Press Clear to remove the Tuning dialog box.
2.6 Testing of Scheduled Tuning

Change the load to 55%.
For each load value between 55% and 95%, make set point changes to 62, then to 60.
Observe the response. Then increment (press StepIncr) the load to the next level and repeat.
Ideally, you should see a fairly satisfactory response at each load level. If not, consider what changes you
might make to the schedule. Consider changing one or more of the following:
Number of regions
Break points between regions
Multiplying factors.
Record your final tuning schedule in the table below:

Number of Regions: _________
Page 7.13.4
CHARACTERISTICS OF RATIO CONTROL
OBJECTIVE: To demonstrate the behaviour of ratio control.
PREREQUISITE: Completion of PC-ControLAB tutorial ( under Help | Tutorial ) or an equivalent

BACKGROUND: Ratio control is used when it is necessary to maintain a certain ratio between the
flow rate of two streams. One stream, called the “wild” flow, is measured only. It is the pacing stream. The
other stream, the controlled flow, is controlled so as to maintain a specified ratio between the two. There
are two general types of ratio control. In one type, illustrated in this laboratory exercise, the ratio is
manually set. In the other type, the ratio is automatically set by the output of a Primary feedback
controller.
That type of ratio control is illustrated by Laboratory Exercise 17, Multiplicative Feed forward Control. A
typical configuration of ratio control is shown in the following figure. For illustrative purpose, this
laboratory exercise assumes that the controlled flow is Steam Flow, with a measured range of 0 – 1000
P/hr. The “wild” flow is a process stream, with a measured range of 0 – 400 gpm.

Start Windows.
Run PC-ControLAB.
Click on Control | Select Strategy | Ratio.
Click on Process | Select Model; highlight “RatioFlo.mdl” and press Open

Regulatory Control and PID loop tuning Exercise 14. Characteristics of ratio control
2. CONTROLLER SET UP
Press Tune and enter the following tuning parameters (representative of a flow loop).
Gain: 0.5
Reset: 0.15 min/rpt
Deriv: 0.00 min
Observe:
Present value of process variable, in engineering units ________
Present value of process variable, in percent of measurement span: ________
Present value of “wild” variable, in engineering units (gray trace, labeled LOAD) ________
(Click on the label “GPM” to change the vertical scale from the PV to the wild flow.)
Present value of “wild” flow, in percent of measurement span: ________
PV % of span
Ratio : ________
Wild flow % of span
Click on the label “Steam Flow (PV-2)” above the faceplate.
Click on the button labeled R on the controller faceplate to access the present calculated ratio, as well as
the set ratio. Calculated ratio: ________
Set ratio: ________
In concept, the set ratio back could be calculated from the present ratio of measured variables; this would
provide bumpless transfer from the AUTO mode to the RATIO mode. This would probably not be wise,
however, since the required ratio is normally set by process conditions or product specifications. The
operator, not the system, should be the one to enter the required ratio.
Therefore, the set ratio is not back-calculated in this program or in most commercial systems. When the
ratio is set by the output of a feedbacfk controller (for instance, in the Multiplicative Feedforward control
strategy – see Exercise 17), the ratio is back calculated, thus producing the output for the Primary
controller. This provides for bumpless transfer from Manual to Auto for the Primary.
If you change the controller to the Ratio mode now, where do you think the controller set point will go?
________
Try it! Put the controller in Ratio. What is the SP? ________
You probably observed that that made quite a “jolt” to the flow loop. A wise operator probably would not
change from Auto to Ratio when there was that much difference between the existing ratio and the ratio
setting. Instead, he/she would probably set the ratio setting to whatever the application or product
specifications required, then adjust flow set point until actual ratio met the required ratio. Then the switch
to Ratio mode would be made.
Page 7.14.2
Regulatory Control and PID loop tuning Exercise 14. Characteristics of ratio control
Press R and enter a set ratio of 0.5.
After the loop stabilizes, observe:
Present value of PV, in engineering units ________
Present value of PV, in percent of measurement span: ________
Present value of “wild” variable, in engineering units ________
Present value of “wild” flow, in percent of measurement span: ________
PV % of span
Ratio ________
Wild flow % of span
Press StepIncr twice to simulate an increase in “wild” flow.
Did the controlled flow (PV) follow it at the set ratio? ________
Press StepDecr twice, then press AutoLoad. This simulates a randomly varying wild flow.
Does the controlled flow (PV) follow it at the set ratio? ________
Page 7.14.3
CHARACTERISTICS OF CASCADE CONTROL

OBJECTIVE: To demonstrate the advantage of cascade control over non-cascade control for some
types of load upsets.
PREREQUISITE: Completion of Exercises

10 PID Tuning from Closed Loop Tests
11 Improving “As Found” Tuning or equivalent level of experience.
BACKGROUND: With cascade control, the output of one feedback controller (the “primary”) sets the
Set-point of a second feedback controller (the “secondary”). A significant advantage is that disturbances
to the inner loop, controlled by the secondary, have minimal effect on the primary process variable.

Start Windows.
Run PC-ControLAB.
2. CASCADE LOOP SET UP

After the title display, the initial operational display contains the Feedback control strategy.
You can call in a CASCADE control strategy with the controllers already tuned for a particular process
model. Select Control | Retrieve Strategy, Model and Tuning.
Highlight “Cascade.stg” and press Open.
The default configuration of the Cascade control strategy is that of cascade control loop, with the
Secondary controller (right hand controller on display) controlling a relatively fast process (Flow) and the
Primary Controller (left hand controller on display) controlling a slower process (Temperature). The
measurement range for the Primary controller is 0 to 500 DegF. For the Secondary Controller, the
measurement range is 0 – 1000 P/hr of steam flow. The simulation represents control of temperature
leaving a heat exchanger. A disturbance to the primary loop is feed rate to the exchanger; a disturbance
to the secondary loop is the steam pressure to the control valve. Disturbances (load upsets) can be
introduced separately into either the secondary loop or primary loop.
Press SEL on the Secondary Controller to enable accessing the Secondary Controller tuning parameters.
(The label above the Secondary Controller should be red with white letters when it is selected.)
Press TUNE.
Regulatory Control and PID loop tuning Exercise 15. Characteristics of cascade control
Note the following tuning values have already been entered for the Secondary Controller:
Gain: 0.5
Deriv: 0.00 minutes
Press Clear.
Confirm acceptable operation of the Secondary Controller. Change the Secondary Controller to
Automatic, and change the set point to 400 P/hr.
Observe the secondary loop response:
Is it appropriately fast, as a flow loop should be? ________
Return the Secondary set point to 280, then place the Secondary Controller in Cascade. (Note the
change in color of the LEDs on the Primary Controller. This signifies that the Primary Controller can now
send a signal to the Secondary.) Use the controller output buttons on the Primary to make slight changes
in its output. You should observe changes to the Secondary set point.
Press SEL on the Primary controller, then press Out and change the Primary controller output to 28%.
This returns the Secondary set point to 280 P/hr.
With the Primary Controller SELected, press TUNE.
Note that the following tuning values have already been entered for the Primary Controller:
Gain: 2.5
Deriv: 0.0 minutes
Press Clear to remove the Primary Controller tuning dialog box.
Check the Secondary Controller. Is it in Cascade? (It should be, if you followed instructions.) If so, put the
Primary Controller in Automatic. Increase its set point to 325. (Select the Primary Controller.
Press SP and key in in the new value in the dialog box.)
Observe the response:
Is it approximately quarter decay? ________
Period of oscillation? ________
Page 7.15.2
3. CASCADE LOOP OPERATION

With the Secondary in Cascade and the Primary in Auto, set the set point for the Primary to 275.
When the loops have stabilized, read and record the values for the following variables:
Primary process variable: ________
Primary Controller output or Secondary Controller set point ________
Secondary Controller output (valve signal): ________
In order to have a better display of the process response in the following sections, you may want
to SELect the Temperature controller and zoom the PV scale to 250 – 300 degF.
Select the Primary Controller. (Very important that the PRIMARY is selected!) Press StepIncr (above the
faceplates) once, to make a load change on the primary loop. (This simulates an increase in process
flow.) The primary control loop load variable affects the outer loop process, but has no effect on the inner
loop process.
Observe the response of the primary process variable:
Did the primary PV eventually return to set point? ________
Primary Controller output (secondary controller set point) ________
Secondary Controller output (valve): ________
You have observed a load change on the outer loop. To compensate for this, the Primary Controller
increased the demand on the inner loop; i.e., it changed the set point of the Secondary Controller. To
meet this additional demand, the Secondary Controller output also changed.
With the Primary Controller still selected, press StepDecr to return to the load to its original value.
When the control loops have stabilized, select the Secondary Controller. (Very important here
that the SECONDARY is selected!) Press StepDecr once to cause a disturbance to the
secondary loop. (This simulates a drop in steam pressure.)
Observe the response and record the following:
Maximum deviation from set point of primary PV: ________
Does the primary PV eventually return to set point? ________
Primary Controller output (secondary set point) after inner loop load change: ________
Secondary Controller output after inner loop load change: ________
You have just observed a load change on the inner loop. Since the load on the outer loop has not
changed, the long term demand of the Primary Controller on the inner loop (i.e., the set point of the
secondary controller) does not change - it may undergo some fluctuation in the short term. But to
compensate for the changed load on the inner loop, the Secondary Controller output had to change.
Page 7.15.3
The most significant observation is that the secondary process load change was contained within the
inner loop. The effect on the primary process variable was fairly slight.
Before going on, while the Secondary Controller is still selected, press StepIncr to return the secondary
load to its original value.
4. NON-CASCADE CONTROL OPERATION

This section of the exercise will make the same type of load changes, but the control structure will be a
simple feedback controller. The Secondary Controller will be eliminated, and the Primary Controller output
will go directly to the valve.
First, switch both controllers to Manual. (In a real control system, you can’t make changes to the
configuration of a controller that is in Automatic. The same thing applies here.)
With either controller selected, select Control | Control Options. Use the scroll bar at the left of the
options table to scroll to the end of the table. Choose the option YES for "Bypass Secondary Controller."
You have just “softwired” around the Secondary Controller. The Primary Controller output now goes
directly to the valve. This operation was performed without bumping the process, since the primary
controller’s output was initialized to be the same as the output of the Secondary Controller before the
switch was made. The display still shows both faceplates, although the Secondary Controller is inactive.
Select the Primary Controller and press TUNE.
Are the tuning parameters the same as previously set for this controller? ________
Change the Primary Controller to Automatic.
Was there a bump in the process variable? ________
Although the control structure has been altered, the process itself has not. There is still a
secondary process which feeds a primary process, with independent load disturbances to
each.
With the Primary Controller Selected, press StepIncr once to increase the load (increase the
process flow rate) on the primary loop.
Observe the response and record the following:
Is this response approximately the same or significantly different from the response with cascade
present?
________
Observation: With a disturbance to the primary process, the presence or absence of a

cascade loop makes very little difference.
Press StepDecr to return the primary process load to its original value.
When the control loop is stabilized, select the Secondary controller and press StepIncr to cause a load
change (drop in steam pressure) on the secondary process.
Page 7.15.4
Observe and record the following:
Is the response to this load change approximately the same or significantly different from the
response to the same load change with cascade present?
________
Is the response to this load change approximately the same or significantly different from the
response to a primary load upset, both without cascade control being present?
________
You have observed load upsets on both the secondary and primary processes, without cascade control
present. Since there is no closed inner loop to compensate for the load upset on the secondary process,
the responses are approximately the same. Comparing the response to a secondary load upset both with
and without cascade, you should observe that there is a very significant improvement when cascade is
present.
Page 7.15.5
Page 7.15.6
DEMONSTRATION EXERCISE 16
ANTI-RESET WINDUP (BATCH SWITCH)

OBJECTIVE: To demonstrate the behaviour of anti-reset windup (also called a “batch switch”) in
certain types of applications, such as batch process start-ups and recovery from out-of-control situations.
NOTE: In a process control seminar, it is recommended that this exercise be conducted as an

instructor-led demonstration exercise, rather than student exercise
BACKGROUND: Frequently encountered

situations in the process industries are start-ups,
grade changes, and recovery from out of control
conditions. If the controller contains the integral
mode and is in Automatic during these situations, the
likely result will be a significant overshoot (or
“undershoot”) of the set point. Many manufacturers
incorporate a special feature in their controllers (it
may be an option, not necessarily a standard feature)
variously called “anti-reset windup” or “batch switch.”
There are variations between manufacturers in the
implementation details; this program demonstrates
one of the most common forms of implementation,
wherein once the controller output hits a limit, the
proportional band is forced in the other direction. The
objective is that, on the next start-up or control
recovery, there will be a significant reduction in the
overshoot.
Figure 1. Process and control configuration used for application example
An application scenario around which the demonstration exercise is written involves a batch
reactor, or other batch operated process unit, in which there is a repetitious heating-cooling cycle. Figure
1 shows a possible process and control configuration.
Suppose the temperature controller is always left in automatic, with a constant set point. Suppose further,
that at the end of the heating cycle, the heat source (say, steam) is removed simply by closing a remote
operated block valve in series with the temperature control valve. The effect will be for the temperature to
drop. The controller, still being in AUTO, will sense the drop in temperature and drive the control valve
wide open.
On the next heating cycle, when the block valve is reopened, the controller will be in AUTO with its
measurement far below its set point. The control valve will be wide open. Under these circumstances, full
steam will be applied to the process. Without anti-reset windup, the proportional band will have shifted to
the top of the scale, so that the temperature will have to rise a significant amount before the steam valve
even begins to cut back. The result will be a significant temperature overshoot of the set point.
Regulatory Control and PID loop tuning Demonstration exercise 16. Ant-reset windup
Following the same scenario, but with anti-reset windup in the controller, the proportional band will initially
rise to the top of the scale, then forced downward by the action of the anti-reset windup function. When
the next heating cycle commences (when the block valve is opened), full steam will initially be applied to
the process as before. However, with a depressed proportional band, the valve will begin cutting back as
soon as the temperature begins to rise. The result will be a significant reduction in temperature overshoot,
as compared with the non-anti-reset windup situation. This exercise demonstrate the scenario described
above.

Start Windows.
Run PC-ControLAB.
2. SET UP
Verify that the control strategy is FEEDBACK, and that the GENERIC process model is in use.
Select Control | Control Options. In the “Anti-Reset Windup” panel, Verify that NO is selected.
Then press Clear.
Press TUNE, then enter the following tuning values:
Gain: 3.0
Deriv: 2.00 minutes
When all values have been entered, press OK to clear the tuning dialog box.
Set Point: 275
PV: 275
Controller Output: 35.0
Process Load (Grey trace on chart) 75.0 % of full scale
Select View | Display Proportional Band. Observe the proportional band (light blue bar) extends from
about 182 °C to 325 °C, or a span of 143 °C.
100 100
Proportional Band = = = 28.6
Gain 3.5
28.6 x 500°C = 143°C
Page 7.16.2
Put the controller in Automatic.
The control loop should now be in normal operation, with the set point and process variable at 275 °C and
the controller output at 35%. Confirm this by making a set point change to 325. When the loop stabilizes,
return the set point to 275. When the process restabilizes, proceed.
3. DEMO OF SHUT-DOWN/START-UP WITHOUT ANTI-RESET WINDUP.
Select View | Discrete Controls.
In the panel labeled “System Heat”, notice that the current state of the shutdown system is ON. In
the panel labeled “Sol Valve”, notice that the solenoid valve is now ON. Click OFF for the
System Heat, and notice that the solenoid valve goes to OFF.
This simulates blocking the heat source from the control valve. Observe:
The process variable (temperature) drops;
The valve goes to a wide open position;
The proportional band bar rises toward the top of the scale, stopping when the output
reaches a limit, then remains in a fixed position.
When the process variable has fully decayed, select System Heat: ON to reopen the block valve.
Observe:
The process variable (temperature) begins to rise almost immediately; Since the proportional
band is approximately in the middle of the scale, the process variable must rise a considerable
distance before the valve even begins to cut back. (The valve begins to cut back before the
measurement actually reaches the proportional band, due to the presence of derivative action in
the controller.)
Note the amount that the process variable overshoots the set point.
Maximum temperature 305 °C
Amount of Overshoot 30 °C
Observe the length of time after the process variable begins to rise before the valve starts to cut back.
3 mins
Page 7.16.3
4. DEMONSTRATION OF SHUT-DOWN/START-UP WITH ANTI-RESET WINDUP

When the control loop is stable at set point, put the controller in Manual.
Select Control | Control Options.
Select YES for “Anti-Reset Windup.”
Select Control | Control Options. In the “Anti-Reset Windup” panel, select YES. Then press Clear.
Select View | Discrete Controls. Select OFF for “System Heat.
Observe:
The process variable (temperature) drops;
The valve goes to a wide open position;
The proportional band bar initially rises, but when the controller output reaches the top of the
scale, then the anti-reset windup feature causes the proportional band to shift downward.
(Colloquially, this is called "dumping the reset.") The proportional band comes to rest when the
top of the PB is the “Preset” (percent of proportional band) below the set point.
When the process variable has fully decayed to the bottom of the scale, reopen the block valve.
Observe:
The process variable (temperature) begins to rise almost immediately;

Since the proportional band is now nearer the bottom of the scale, the process variable must rise
a shorter distance before the valve begins to cut back. (The controller output appears to begin to
cut back at about the same instant that the measurement begins to rise.)
The proportional band begins to shift as the controller nears the set point.
Maximum temperature: 285 °C
Maximum overshoot of the process variable. 10 °C
Observe the length of time after the process variable begins to rise before the valve starts to cut back.
1.5 mins
5. OTHER FACETS OF ANTI-RESET WINDUP

Two parameters available through the Control Options display govern the behavior of the antireset
windup function:
“Output maximum” limits the maximum controller output. In addition, when the controller output reaches
this point, the anti-reset windup function begins to shift the proportional band downward. “Anti-reset
windup preset” limits the amount by which the proportional band can be shifted downward. For example,
suppose the controller is Reverse Acting, with a PB of 40% (GAIN = 2.5) and the set point is at 60%. Also
suppose that the anti-reset windup preset is set for -25%. On a shut down, the anti-reset windup would
force the PB down to where the top of the PB is at 50% of the PV scale. That's 10% below the set point,
where 10% of PV scale is 25% of the PB width. (0.25 x 40% = 10%)
Page 7.16.4
If the anti-reset windup preset is set too high, the proportional band will not be shifted downward far
enough, resulting still in an unwanted amount of overshoot.
On the other hand, if the anti-reset windup preset is set to low, the proportional band will be shifted
downward an excessive amount, resulting in the process variable beginning to cut back before reaching
the set point. In essence, the anti-reset windup preset becomes an additional tuning parameter that must
be matched to process dynamics and load conditions for optimum response.
Put the controller in Manual.
Select Control | Controller Options. Change the anti-reset windup preset from its present value (0.0) to
-25.
Put the controller in Auto and repeat the shut down-start up procedure. This time you should see
essentially the optimum response, meaning minimum overshoot.
Record the amount: 0%
6. DISCUSSION
Some manufacturers achieve a similar function in different manners. For instance: Some manufacturers
implement two-sided anti-reset windup. (This program only demonstrated one-sided anti-reset windup,
effective when the controller output reached an upper limit.)
One manufacturer leaves the proportional band at the top of the scale, but then when the measurement
begins to recover, the integral time is reduced by a factor of 16, thus shifting the proportional band
downward 16 times as fast as would a normal PID controller. This minimizes overshoot in start-up or
recovery from loss of control. Some manufacturers simply limit the contribution to the controller output by
the integral mode and refer to this as “anti-reset windup.” This should not be confused with the form of
anti-reset windup demonstrated here.
Page 7.16.5
Page 7.16.6

Complete PID Manual PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Complete PID Manual PDF

Uploaded by

Copyright:

Available Formats

Section 1.

© 2019 Crabtree Controls Ltd. All rights reserved 1

 Identify some of the components that contribute to

© 2019 Crabtree Controls Ltd. All rights reserved 2

 One of the major problems in regard to the field of PID

© 2019 Crabtree Controls Ltd. All rights reserved 3

The four basic elements of any control system comprise:

© 2019 Crabtree Controls Ltd. All rights reserved 4

Transducer (sensor + transmitter) Final control element

© 2019 Crabtree Controls Ltd. All rights reserved 5

Opening the valve …

© 2019 Crabtree Controls Ltd. All rights reserved 6

 However, new value only reached some time later

© 2019 Crabtree Controls Ltd. All rights reserved 7

 Capacitance of a process is a measure of its ability to

© 2019 Crabtree Controls Ltd. All rights reserved 8

 Consider two tanks – both having the same liquid

© 2019 Crabtree Controls Ltd. All rights reserved 9

Capacity and capacitance

 In other words, Tank A has twice the liquid volume

© 2019 Crabtree Controls Ltd. All rights reserved 10

 Paraffin in Tank A requires about 4kJ of heat energy to raise

Heat energy required = 4 kJ Heat energy required = 7.7 kJ

© 2019 Crabtree Controls Ltd. All rights reserved 11

 Heat transfer cannot be instantaneous and heat energy

Steam out Fluid out

© 2019 Crabtree Controls Ltd. All rights reserved 12

 Resistance may be likened to the ‘opposition to flow’

Steam out Fluid out

© 2019 Crabtree Controls Ltd. All rights reserved 13

© 2019 Crabtree Controls Ltd. All rights reserved 14

 Deadtime is often given the symbol ()

© 2019 Crabtree Controls Ltd. All rights reserved 15

 Controlled variable is called the Process Demand

Transducer (sensor + transmitter) Final control element

© 2019 Crabtree Controls Ltd. All rights reserved 16

 If PD subject to a step change…

© 2019 Crabtree Controls Ltd. All rights reserved 17

 Process Gain given by:

© 2019 Crabtree Controls Ltd. All rights reserved 18

 Assume a step change in the PD of 10%

© 2019 Crabtree Controls Ltd. All rights reserved 19

 Generally, the process gain for a self-regulating process

© 2019 Crabtree Controls Ltd. All rights reserved 20

 Therefore if the valve were adjusted from 0-100% of the full

© 2019 Crabtree Controls Ltd. All rights reserved 21

Setting span too high

● Setting the transmitter span too wide is common practice

© 2019 Crabtree Controls Ltd. All rights reserved 22

 One disadvantage of using an oversized valve is that in

© 2019 Crabtree Controls Ltd. All rights reserved 23

 If a transmitter is spanned too wide, resulting in a process

© 2019 Crabtree Controls Ltd. All rights reserved 24

 Again it should be noticed that the process gain will never

© 2019 Crabtree Controls Ltd. All rights reserved 25

 If PD subject to 30% step change…

© 2019 Crabtree Controls Ltd. All rights reserved 26

As the liquid level (h) increases,

© 2019 Crabtree Controls Ltd. All rights reserved 27

First order lag

The rate of change of level is proportional to the difference

© 2019 Crabtree Controls Ltd. All rights reserved 28

Substituting for Q2: