Bytronic

Process Control Module

Flow Control
PCT-M1
Version 1.5

Bytronic Limited
24 Anglesey Court,
Towers Business Park,
Rugeley, Staffordshire.
WS15 1UL
United Kingdom
Tel: +44 (0) 8456 123 155
Fax: +44 (0) 8456 123 156
Email: [email protected]
Website: www.bytronic.net
 Table of Contents
Requirements ......................................................................................................................................................................1

Introduction.........................................................................................................................................................................1

Software Installation ..................................................................................................................................................... 3 – 5

Connecting the PCT-M to the PC. ...........................................................................................................................6

Software Manual Flow Control...........................................................................................................................................7

Software PID Flow Control ...............................................................................................................................................8

The Principles of Flow Control ...................................................................................................................................9 - 22

Flow Measurement Methods...............................................................................................................................................9
Mechanical Flow Meters. ........................................................................................................................................9
Bucket-And-Stopwatch ................................................................................................................................9
Turbine Flow Meter......................................................................................................................................9
Paddle Wheel Meter .....................................................................................................................................9
Piston Meter/Rotary Piston...........................................................................................................................9
Variable Area Meter .....................................................................................................................................9
Woltmann Meter...........................................................................................................................................9
Single Jet Meter ............................................................................................................................................9
Multiple Jet Meter ........................................................................................................................................9
Pelton Wheel .............................................................................................................................................. 10
Oval Gear Meter ......................................................................................................................................... 10
Nutating Disk Meter ................................................................................................................................... 10
Pressure-Based Meters........................................................................................................................................... 10
Venturi Meter ............................................................................................................................................. 10
Orifice Plate................................................................................................................................................ 10
Dall Tube.................................................................................................................................................... 11
Pitot Tube ................................................................................................................................................... 11
Multi-Hole Pressure Probe ......................................................................................................................... 11
Optical Flow Meters .............................................................................................................................................. 11
Open Channel Flow Measurement......................................................................................................................... 11
Level to Flow.............................................................................................................................................. 11
Area / Velocity ........................................................................................................................................... 11
Dye Testing ................................................................................................................................................ 12
Acoustic Doppler Velocimetry ................................................................................................................... 12
Thermal Mass Flow Meters ................................................................................................................................... 12
Vortex Flowmeters ................................................................................................................................................ 12
Electromagnetic, Ultrasonic And Coriolis Flow Meters........................................................................................ 12
Magnetic Flow Meters................................................................................................................................ 12
Ultrasonic (Doppler, Transit Time) Flow Meters ....................................................................................... 13
Coriolis Flow Meters .................................................................................................................................. 14
Laser Doppler Flow Measurement......................................................................................................................... 14
Turbine Flow Meter (Further Reading)............................................................................................................................. 15
General Performance Characteristics ................................................................................................................................ 16
Theory ..................................................................................................................................................................17 - 19
Calibration, Installation, and Maintenance........................................................................................................................ 20
Design and Construction ................................................................................................................................................... 21
Control Methods .............................................................................................................................................................. 23
Modern Control Theory ........................................................................................................................................ 23
Topics in Control Theory....................................................................................................................................... 23
Main Control Strategies ........................................................................................................................................ 24
Adaptive Control .................................................................................................................................................. 24
Hierarchical Control ............................................................................................................................................. 24
Intelligent Control ................................................................................................................................................. 24
Optimal Control .................................................................................................................................................... 24
Robust Control ...................................................................................................................................................... 24
Stochastic Control ................................................................................................................................................. 24

Control Principles ............................................................................................................................................................ 25

1st Order Systems .......................................................................................................................................... 25 – 26
Transfer Functions ................................................................................................................................................ 26
Block Diagrams ............................................................................................................................................ 27 – 30
Assessment of System Performance ..................................................................................................................... 31
Transient Response ....................................................................................................................................... 31 – 33
Control System Instability .................................................................................................................................... 33
Final Value Theorem ............................................................................................................................................ 34
The Routh-Hurwitz Test ............................................................................................................................... 35 – 36
Bode Plots ..................................................................................................................................................... 37 – 38
Bode Phase Lag versus Frequency Plot ................................................................................................................ 39
Nyquist Plots ................................................................................................................................................. 40 – 41
Process Modelling ................................................................................................................................................ 42
Process Models from Step Data Tests ................................................................................................................... 42
Process Models from Frequency Response Tests ................................................................................................. 42
Process Models from Time Domain Tests ............................................................................................................ 43
PID Controllers ..................................................................................................................................................... 43
Proportional Control Term .................................................................................................................................... 44
Integral Control Term ........................................................................................................................................... 44
Derivative Control Term ....................................................................................................................................... 44
Multi Term Control ............................................................................................................................................... 45
Ziegler Nichols Tuning ................................................................................................................................. 46 – 47
Analysis of Digital Systems .................................................................................................................................. 48
Block Diagrams for Digital Systems .................................................................................................................... 49
Pulse Transfer Functions .............................................................................................................................. 50 – 51
Z Transform Initial and Final Value Theorems .................................................................................................... 52
Stability of Sampled Data Control Systems .......................................................................................................... 53
Inverse Z Transformations ............................................................................................................................ 54 – 55
Digital Three Term Controller ...................................................................................................................... 56 – 58
Open Loop Control ............................................................................................................................................... 58

Exercises ....................................................................................................................................................................59 -62

Exercise 1. Proportional Control ........................................................................................................................... 59
Exercise 2. Proportional and Integral Control ....................................................................................................... 60
Exercise 3. Saturation and Integral Windup ......................................................................................................... 61
Exercise 4. Three Term or PID Control ................................................................................................................ 61
Exercise 5. Ziegler Nichols Tuning ....................................................................................................................... 62

Glossary of Terms..................................................................................................................................................... 63 – 65

Figure 1. Turbine Flow Meter.............................................................................................................................................................. 15

Figure 2. Turbine Flow Meter.............................................................................................................................................................. 15

Figure 3. Axial Turbine Rotor.............................................................................................................................................................. 18

Figure 4. Axial Turbine Linearity Graph ............................................................................................................................................. 19

Requirements
PCT-M Software requires the following minimum PC configuration.

 Pentium Processor
 8Mb free system RAM
 4Mb HDD space
 VGA Graphics Display
 Windows XP or later
 USB Port
 CD Drive

The PCT-M3 unit is supplied with Windows compatible PID software that allows the following main areas of
control to be studied; Level.

Introduction
The PCT-M is controllable by a PC using a USB Interface cable.

The unit has been designed to demonstrate control of one particular aspect of Process control.

The Software provided covers all four of the PCT-M units (Flow, Level, Pressure and Temperature) and detects
the units plugged into the PC.

The USB device has drivers from FTDI1 provided and can be used with other control software such as National
Instruments LabVIEW1 or MATLAB1.

The Manual discusses methods of control techniques, using the software supplied and suggested experiments.

A pressure transducer is used for the measurement of level and an overview of pressure transducers is provided
for further discussion.

The PCT-M1
The PCT-M1 consists of a reservoir tanks; a pump; a control valve; a flow sensor and Flow Meter. Control is
achieved using a pump, solenoid valve and a flow transducer to measure the results. A controller fitted to the unit
is linked to the PC using a USB connection.

The objective is to control the flow in the system using the pump and valve and the measured value from the
transducer.

1. All trademarks and copyrights are recognised.

© Bytronic Ltd Page 1 PCT-M1

© Bytronic Ltd Page 2 PCT-M1
Software Installation
On the CD accompanying your units is a SETUP.EXE file. This should start automatically when the CD is
inserted into the CD drive on the PC, if auto run does not start then, run the file either by double clicking on it, or
by select Run from the Start Menu and typing D:\SETUP.EXE into the box. (Replace D:\ with the drive letter of
your CD/DVD-ROM drive). Now click the OK button.

The software will go through the different steps of installation detailed below.

Click the Next button to continue.

© Bytronic Ltd Page 3 PCT-M1

Click the Install button to continue.

This screen will show that the files are being copied to your system. The time will be depending on the
specification of your PC.

© Bytronic Ltd Page 4 PCT-M1

This screen shows that the FTDI Drivers have been successfully installed. The FTDI drivers are the drivers that
communicate between the software and the hardware. If these drivers do not install correctly then the software
will not be able to control the hardware units.

Finally, this screen shows that the software has been completely installed on your machine. You should click the
Finish button to finalise the installation.

© Bytronic Ltd Page 5 PCT-M1

Connecting the PCT-M to the PC
You must first ensure that the software supplied with the PCT-M has been installed on the PC and that the
drivers for the units were also installed correctly. During the installation, a box will pop up on the screen
showing the FTDI CDM Drivers have been successfully installed.

Ensure the power supply is plugged into the mains supply and connected to the PCT-M, with the unit switched
off. Using the supplied USB cable, connect one end to the PCT-M and the other end to the PC.

Switch on the PCT-M. The computer will detect the device and install the necessary drivers. You will see a
message that your hardware has been installed and ready for use (this will only appear the first time you plug the
PCT-M into the PC).

Start the software on your computer. A short cut will have been installed on the Start Menu  Bytronic Ltd 
PCT-M.

It is possible to either control the device manually or PC control of the PCT-M using PID control. There are two
menu options in the software, PID Control and Manual Control and there are options for each of the PCT-M
units.

If this is the first time that you have connected the device then you will be informed that the USB Interface needs
to be configured/installed and a box will appear.

In the drop down box, select the unit connected and click OK. You will now be able to control that unit using the
software.

The software starts off with a simple menu structure. The units can be controlled manually or
using PID control.

© Bytronic Ltd Page 6 PCT-M1

Manual Mode

Selecting the Flow Control Unit in manual mode gives the above screen. There are readings for flow and
controls for the pump and the valve.

 When the pump is on, it turns green. When the pump is off, it turns red.
 When the valve is open, it turns green. When the valve is closed, it turns red.

The pump and valve are controlled using the sliding controls. If you slide the control to the right, you output a
higher value to the pump or valve. Sliding the control to the left turns off the pump or closes the valve.

Clicking the close button closes this form and resets the Flow Control unit.

© Bytronic Ltd Page 7 PCT-M1

PID Control Mode
In PID Control Mode, the computer can control the units using Proportional, Integral and Derivative control
methods.

Flow Control Unit

 Selecting the Flow Control Unit in PID mode gives the above screen.
 You can alter any of the PID values by clicking on them and entering a new value.
 Clicking the Start button starts the PID control and initialises the graph.
 Clicking the Stop button stops the PID control.
 Clicking the Menu button returns to the main screen and resets the Flow Control Unit.

Select Flow Control and click START to see the unit working with the default controller. This represents
automatic control of the rate of flow of water through the impeller type flow meter on the unit. Over the duration
of the experiment vary the set point several times by clicking the SP box and observe the effects. If the set point
is increased however, the pump output is increased to increase the flow rate in the system as quickly as possible.
As the flow rate approaches the new set point the control algorithm is used to progressively control the pump so
that there is no overshoot. The resulting graph may be manipulated to change the time base or look back at the
results.

Sample Time/Period is an important factor that affects the performance of a three-term controller. It is the time
interval between successive measured values. If the sampling frequency is too low then important high
frequency information will be lost. As the time between consecutive samples is effectively a dead time, the
closed loop stability of the system may be reduced.

Set Point is the process variable desired value that the controller is trying to achieve and maintain. It can either
be a fixed value or a square, saw tooth, ramp or sinusoidal waveform for the experiments.

Trends show the results from the experiment session. To display more data on the screen adjust the slider to the
top right of the trend. This slider varies the time base of the trend. The arrows to the left of the trend allow the
data to be scrolled left or right. Direct numerical input equivalents for these two features are also provided.

Start/Stop starts or stops the simulated process. Once the process has begun, the data will be captured and
displayed on the trend. Once stopped, data can be saved or printed.

© Bytronic Ltd Page 8 PCT-M1

The Principles of Flow Control
Flow Measurement Methods

Mechanical Flow Meters.

Bucket-And-Stopwatch
Perhaps the simplest way to measure volumetric flow is to measure how long it takes to fill a known volume
container. A simple example is using a bucket of known volume, filled by a fluid. The stopwatch is started when
the flow starts, and stopped when the bucket overflows. The volume divided by the time gives the flow. The
bucket and stopwatch method is an off-line method, meaning that the measurement cannot be taken without
interrupting the normal flow.

Turbine Flow Meter

The turbine flow meter (better described as an axial turbine) translates the mechanical action of the turbine
rotating in the liquid flow around an axis into a user-readable rate of flow (gpm, lpm, etc.). The turbine tends to
have all the flow traveling around it. The turbine wheel is set in the path of a fluid stream. The flowing fluid
impinges on the turbine blades, imparting a force to the blade surface and setting the rotor in motion. When a
steady rotation speed has been reached, the speed is proportional to fluid velocity. Turbine flow meters are used
for the measurement of natural gas and liquid flow.

Paddle Wheel Meter

This is similar to the single jet meter, except that the impeller is small with respect to the width of the pipe, and
projects only partially into the flow, like the paddle wheel on a Mississippi riverboat.

Piston Meter/Rotary Piston

Because they are used for domestic water measurement, piston meters, also known as rotary piston or semi-
positive displacement meters, are the most common flow measurement devices in the UK and are used for
almost all meter sizes up to and including 40 mm. The piston meter operates on the principle of a piston rotating
within a chamber of known volume. For each rotation, an amount of water passes through the piston chamber.
Through a gear mechanism and, sometimes, a magnetic drive, a needle dial and odometer type display are
advanced.

Variable Area Meter

The variable area (VA) meter, also commonly called a rotameter, consists of a tapered tube, typically made of
glass, with a float inside that is pushed up by fluid flow and pulled down by gravity. As flow rate increases,
greater viscous and pressure forces on the float cause it to rise until it becomes stationary at a location in the tube
that is wide enough for the forces to balance. Floats are made in many different shapes, with spheres and
spherical ellipses being the most common. Some are designed to spin visibly in the fluid stream to aid the user in
determining whether the float is stuck or not. Rotameters are available for a wide range of liquids but are most
commonly used with water or air. They can be made to reliably measure flow down to 1% accuracy.

Woltmann Meter
The Woltmann meter comprises a rotor with helical blades inserted axially in the flow, much like a ducted fan; it
can be considered a type of turbine flow meter. They are commonly referred to as helix meters, and are popular
at larger sizes.

Single Jet Meter

A single jet meter consists of a simple impeller with radial vanes, impinged upon by a single jet. They are
increasing in popularity in the UK at larger sizes and are commonplace in the EU.

Multiple Jet Meter

A multiple jet or multijet meter is a velocity type meter which has an impeller which rotates horizontally on a
vertical shaft. The impeller element is in a housing in which multiple inlet ports direct the fluid flow at the
impeller causing it to rotate in a specific direction in proportion to the flow velocity. This meter works
mechanically much like a single jet meter except that the ports direct the flow at the impeller equally from
several points around the circumference of the element, not just one point; this minimizes uneven wear on the
impeller and its shaft.

© Bytronic Ltd Page 9 PCT-M1

Pelton Wheel
The Pelton wheel turbine (better described as a radial turbine) translates the mechanical action of the Pelton
wheel rotating in the liquid flow around an axis into a user-readable rate of flow (gpm, lpm, etc.). The Pelton
wheel tends to have all the flow traveling around it with the inlet flow focused on the blades by a jet. The
original Pelton wheels were used for the generation of power and consisted of a radial flow turbine with
"reaction cups" which not only move with the force of the water on the face but return the flow in opposite
direction using this change of fluid direction to further increase the efficiency of the turbine.

Oval Gear Meter

An oval gear meter is a positive displacement meter that uses two or more oblong gears configured to rotate at
right angles to one another, forming a tee shape. Such a meter has two sides, which can be called A and B. No
fluid passes through the center of the meter, where the teeth of the two gears always mesh. On one side of the
meter (A), the teeth of the gears close off the fluid flow because the elongated gear on side A is protruding into
the measurement chamber, while on the other side of the meter (B), a cavity holds a fixed volume of fluid in a
measurement chamber. As the fluid pushes the gears, it rotates them, allowing the fluid in the measurement
chamber on side B to be released into the outlet port. Meanwhile, fluid entering the inlet port will be driven into
the measurement chamber of side A, which is now open. The teeth on side B will now close off the fluid from
entering side B. This cycle continues as the gears rotate and fluid is metered through alternating measurement
chambers. Permanent magnets in the rotating gears can transmit a signal to an electric reed switch or current
transducer for flow measurement.

Nutating Disk Meter

This is the most commonly used measurement system for measuring water supply. The fluid, most commonly
water, enters in one side of the meter and strikes the nutating disk, which is eccentrically mounted. The disk
must then "wobble" or nutate about the vertical axis, since the bottom and the top of the disk remain in contact
with the mounting chamber. A partition separates the inlet and outlet chambers. As the disk nutates, it gives
direct indication of the volume of the liquid that has passed through the meter as volumetric flow is indicated by
a gearing and register arrangement, which is connected to the disk. It is reliable for flow measurements within 1
percent.

Pressure-Based Meters
There are several types of flow meter that rely on Bernoulli's principle, either by measuring the differential
pressure within a constriction, or by measuring static and stagnation pressures to derive the dynamic pressure.

Venturi Meter
A Venturi meter constricts the flow in some fashion, and pressure sensors measure the differential pressure
before and within the constriction. This method is widely used to measure flow rate in the transmission of gas
through pipelines, and has been used since Roman Empire times. The coefficient of discharge of Venturi meter
ranges from 0.93 to 0.97.

Orifice Plate
An orifice plate is a plate with a hole through it, placed in the flow; it constricts the flow, and measuring the
pressure differential across the constriction gives the flow rate. It is basically a crude form of Venturi meter, but
with higher energy losses. There are three type of orifice: concentric, eccentric, and segmental.

Orifice Plate

© Bytronic Ltd Page 10 PCT-M1

Dall Tube
The Dall tube is a shortened version of a Venturi meter, with a lower pressure drop than an orifice plate. As with
these flow meters the flow rate in a Dall tube is determined by measuring the pressure drop caused by restriction
in the conduit. The pressure differential is typically measured using diaphragm pressure transducers with digital
readout. Since these meters have significantly lower permanent pressure losses than orifice meters, Dall tubes are
widely used for measuring the flow rate of large pipeworks.

Pitot Tube
A Pitot tube is a pressure measuring instrument used to measure fluid flow velocity by determining the
stagnation pressure. Bernoulli's equation is used to calculate the dynamic pressure and hence fluid velocity.

Multi-Hole Pressure Probe

Multi-hole pressure probes (also called impact probes) extend the theory of pitot tube to more than one
dimension. A typical impact probe consists of three or more holes (depending on the type of probe) on the
measuring tip arranged in a specific pattern. More holes allow the instrument to measure the direction of the flow
velocity in addition to its magnitude (after appropriate calibration). Three holes arranged in a line allow the
pressure probes to measure the velocity vector in two dimensions. Introduction of more holes, e.g. five holes
arranged in a "plus" formation, allow measurement of the three-dimensional velocity vector.

Optical Flow Meters

Optical flow meters use light to determine flow rate. Small particles which accompany natural and industrial
gases pass through two laser beams focused in a pipe by illuminating optics. Laser light is scattered when a
particle crosses the first beam. The detecting optics collects scattered light on a photodetector, which then
generates a pulse signal. If the same particle crosses the second beam, the detecting optics collect scattered light
on a second photodetector, which converts the incoming light into a second electrical pulse. By measuring the
time interval between these pulses, the gas velocity is calculated as V = D / T where D is the distance between
the laser beams and T is the time interval.

Laser-based optical flow meters measure the actual speed of particles, a property which is not dependent on
thermal conductivity of gases, variations in gas flow or composition of gases. The operating principle enables
optical laser technology to deliver highly accurate flow data, even in challenging environments which may
include high temperature, low flow rates, high pressure, high humidity, pipe vibration and acoustic noise.

Optical flow meters are very stable with no moving parts and deliver a highly repeatable measurement over the
life of the product. Because distance between the two laser sheets does not change, optical flow meters do not
require periodic calibration after their initial commissioning. Optical flow meters require only one installation
point, instead of the two installation points typically required by other types of meters. A single installation point
is simpler, requires less maintenance and is less prone to errors.

Optical flow meters are capable of measuring flow from 0.1 m/s to faster than 100 m/s (1000:1 turn down ratio)
and have been demonstrated to be effective for the measurement of flare gases, a major global contributor to the
emissions associated with climate change.

Open Channel Flow Measurement

Level To Flow
The level of the water is measured at a designated point behind a hydraulic structure using various means
(bubblers, ultrasonic, float, and differential pressure are common methods). This depth is converted to a flow
rate according to a theoretical formula of the form Q = KHX where Q is the flow rate, K is a constant, H is the
water level, and X is an exponent which varies with the device used; or it is converted according to empirically
derived level/flow data points (a "flow curve"). The flow rate can then integrated over time into volumetric flow.

Area / Velocity
The cross-sectional area of the flow is calculated from a depth measurement and the average velocity of the flow
is measured directly (Doppler and propeller methods are common). Velocity times the cross-sectional area yields
a flow rate which can be integrated into volumetric flow.

© Bytronic Ltd Page 11 PCT-M1

Dye Testing
A known amount of dye (or salt) per unit time is added to a flow stream. After complete mixing, the
concentration is measured. The dilution rate equals the flow rate.

Acoustic Doppler Velocimetry

Acoustic Doppler velocimetry (ADV) is designed to record instantaneous velocity components at a single point
with a relatively high frequency. Measurements are performed by measuring the velocity of particles in a remote
sampling volume based upon the Doppler shift effect.

Thermal Mass Flow Meters

Thermal mass flow meters generally use combinations of heated elements and temperature sensors to measure
the difference between static and flowing heat transfer to a fluid and infer its flow with a knowledge of the fluid's
specific heat and density. The fluid temperature is also measured and compensated for. If the density and specific
heat characteristics of the fluid are constant, the meter can provide a direct mass flow readout, and does not need
any additional pressure temperature compensation over their specified range.

Temperature at the sensors varies depending upon the mass flow

Technological progress has allowed the manufacture of thermal mass flow meters on a microscopic scale as
MEMS sensors; these flow devices can be used to measure flow rates in the range of nanolitres or microlitres per
minute.

Thermal mass flow meter technology is used for compressed air, nitrogen, helium, argon, oxygen, and natural
gas. In fact, most gases can be measured as long as they are fairly clean and non-corrosive. For more aggressive
gases, the meter may be made out of special alloys (e.g. Hastelloy), and pre-drying the gas also helps to
minimize corrosion.

Vortex Flowmeters
Another method of flow measurement involves placing a bluff body (called a shedder bar) in the path of the
fluid. As the fluid passes this bar, disturbances in the flow called vortices are created. The vortices trail behind
the cylinder, alternatively from each side of the bluff body. This vortex trail is called the Von Kármán vortex
street after von Kármán's 1912 mathematical description of the phenomenon. The frequency at which these
vortices alternate sides is essentially proportional to the flow rate of the fluid. Inside, atop, or downstream of the
shedder bar is a sensor for measuring the frequency of the vortex shedding. This sensor is often a piezoelectric
crystal, which produces a small, but measurable, voltage pulse every time a vortex is created. Since the
frequency of such a voltage pulse is also proportional to the fluid velocity, a volumetric flow rate is calculated
using the cross sectional area of the flow meter. The frequency is measured and the flow rate is calculated by the
flowmeter electronics using the equation f = SV / L where f is the frequency of the vortices, L the characteristic
length of the bluff body, V is the velocity of the flow over the bluff body, and S is the Strouhal number, which is
essentially a constant for a given body shape within its operating limits.

Electromagnetic, Ultrasonic And Coriolis Flow Meters

Modern innovations in the measurement of flow rate incorporate electronic devices that can correct for varying
pressure and temperature (i.e. density) conditions, non-linearities, and for the characteristics of the fluid.

Magnetic Flow Meters

The most common flow meter apart from mechanical flow meters is the magnetic flow meter, commonly
referred to as a "mag meter" or an "electromag". A magnetic field is applied to the metering tube, which results
in a potential difference proportional to the flow velocity perpendicular to the flux lines. The physical principle
at work is Faraday's law of electromagnetic induction. The magnetic flow meter requires a conducting fluid, e.g.
water, and an electrical insulating pipe surface, e.g. a rubber lined nonmagnetic steel tube.

© Bytronic Ltd Page 12 PCT-M1

 Industrial magnetic flowmeter

Ultrasonic (Doppler, Transit Time) Flow Meters

Ultrasonic flow meters measure the difference of the transit time of ultrasonic pulses propagating in and against
flow direction. This time difference is a measure for the average velocity of the fluid along the path of the
ultrasonic beam. By using the absolute transit times both the averaged fluid velocity and the speed of sound can
be calculated. Using the two transit times tup and tdown and the distance between receiving and transmitting
transducers L and the inclination angle α one can write the equations:

and

where v is the average velocity of the fluid along the sound path and c is the speed of sound.

Ultrasonic flow meters are used for the measurement of natural gas flow. One can also calculate the expected
speed of sound for a given sample of gas; this can be compared to the speed of sound empirically measured by
an ultrasonic flow meter and for the purposes of monitoring the quality of the flow meter's measurements. A
drop in quality is an indication that the meter needs servicing. Recently, Ultrasonic flow meters are also being
used for measurement of LNG flow.

Schematic view of a flow sensor.

Measurement of the Doppler shift resulting in reflecting an ultrasonic beam off the flowing fluid is another
recent innovation. By passing an ultrasonic beam through the tissues, bouncing it off a reflective plate, then
reversing the direction of the beam and repeating the measurement, the volume of blood flow can be estimated.
The frequency of the transmitted beam is affected by the movement of blood in the vessel and by comparing the
frequency of the upstream beam versus downstream the flow of blood through the vessel can be measured. The
difference between the two frequencies is a measure of true volume flow. A wide-beam sensor can also be used
to measure flow independent of the cross-sectional area of the blood vessel.

For the Doppler principle to work in a flowmeter it is mandatory that the flow stream contains sonically
reflective materials, such as solid particles or entrained air bubbles.

A related technology is acoustic Doppler velocimetry.

© Bytronic Ltd Page 13 PCT-M1

Coriolis Flow Meters
Using the Coriolis effect that causes a laterally vibrating tube to distort, a direct measurement of mass flow can
be obtained in a coriolis flow meter. Furthermore a direct measure of the density of the fluid is obtained. Coriolis
measurement can be very accurate irrespective of the type of gas or liquid that is measured; the same
measurement tube can be used for hydrogen gas and bitumen without recalibration. Coriolis flow meters can be
used for the measurement of natural gas flow.

Laser Doppler Flow Measurement

Laser-doppler flow meter.

Blood flow can be measured through the use of a monochromatic laser diode. The laser probe is inserted into a
tissue and turned on, where the light scatters and a small portion is reflected back to the probe. The signal is then
processed to calculate flow within the tissues. There are limitations to the use of a laser Doppler probe; flow
within a tissue is dependent on volume illuminated, which is often assumed rather than measured and varies with
the optical properties of the tissue. In addition, variations in the type and placement of the probe within identical
tissues and individuals result in variations in reading. The laser Doppler has the advantage of sampling a small
volume of tissue, allowing for great precision, but does not necessarily represent the flow within an entire organ.
The flow meter is much more useful for relative rather than absolute measurements.

© Bytronic Ltd Page 14 PCT-M1

Turbine Flow Meter (Further Reading)
The axial turbine flowmeter, is a reliable device capable of providing the highest accuracies for both liquid and
gas volumetric flow measurement. These devices are the result of decades of intensive innovation and
refinements to the original axial vaned flowmeter principle first credited to Woltman in 1790, and applied at that
time to measuring water flow. Today, axial turbine flowmeters of differing and often proprietary designs are
used for a variety of applications where accuracy, reliability, and rangeability are required in numerous major
industries including water, natural gas, oil, petrochemical, chemical process, cryogenics, milk and beverage,
aerospace, and biomedical.

Figure 1 Figure 2

Figure 1. The key components of an axial turbine flowmeter can be seen in this longitudinal section view. The
relative size of the annular flow passage at the rotor varies among different designs. Magnetic reluctance pickups
(shown). Other types include mechanical and modulated- carrier pickups.

Figure 2 is a schematic longitudinal section through the axis of symmetry showing the key components of a
typical meter. The meter is an inline sensor that incorporates a single turbine rotor concentrically mounted on a
shaft within a cylindrical housing through which the flow passes. The shaft or shaft bearings are located by end
supports inside suspended upstream and downstream aerodynamic structures called diffusers, stators, or simply
cones. The flow thus passes through an annular region occupied by the rotor blades. The blades, which are
usually flat but may be slightly twisted, are inclined at an angle to the incident flow velocity and hence
experience a torque that drives the rotor. The rate of rotation, which can be up to several tens of thousands of
rpm for smaller meters, is detected by a pickup, usually a magnetic type, and registration of each rotor blade
passing implies the passage of a fixed volume of fluid.

© Bytronic Ltd Page 15 PCT-M1

General Performance Characteristics
Axial turbines perform best when measuring clean, conditioned, steady flows of gases and liquids with low
kinematic viscosities (below about 10–5 m2s–1, or 10 cSt, although they are used up to 10–4 m2s–1, or 100 cSt), and
are linear for subsonic, turbulent flows. Under these conditions the inherent mechanical stability of the meter
design provides excellent repeatability. Except for the special case of water meters, which are described later, the
main performance characteristics are:
 Sizes (i.d.): 6–760 mm (1/4–30 in.)
 Maximum measurement capacities: 0.025–25,500 Am3/hr (0.015–15,000 ACFM) for gases, where A
denotes actual; 0.036–13,000 m3/hr (0.16–57,000 gpm or 82,000 barrels/hr) for liquids
 Typical repeatability: ±0.1% of reading for liquids; ±0.25% for gases, with up to ±0.02% for high
accuracy meters
 Typical linearities (before electronic linearization): 0.25% to ±0.5% of reading for liquids; ±0.5% to
±1.0% for gases. (High-accuracy meters have linearities of ±0.15% for liquids and ±0.25% for gases,
usually specified over a 10:1 dynamic range below maximum rated flow. Traceability to NIST is
frequently available, allowing the estimation of a flowmeter's overall absolute accuracy under specified
conditions. Under ideal conditions, absolute accuracies for optimum designs and installations can
approach the accuracy capabilities at NIST, which are stated as ±0.13% for liquid flows and ±0.25% for
air.)
 Rangeability, when defined as the ratio of flow rates over which the linearity specification applies:
typically between 10:1 and 100:1
 Operating temperature ranges: –270°C to 650°C (–450°F to 1200°F)
 Operating pressure ranges: from coarse vacuum to 414 MPa (60,000 psi)
 Pressure drop at the maximum rated flow rate: ~0.3 kPa (0.05psi) for gases;
~70 kPa (10 psi) for liquids

© Bytronic Ltd Page 16 PCT-M1

Theory
The current literature discusses two techniques for analyzing axial turbine performance. The first describes the
fluid driving torque in terms of momentum exchange; the second describes it in terms of aerodynamic lift via
airfoil theory. The advantage of the former technique is that it readily produces analytical results describing basic
operation, some of which have not appeared via airfoil analysis. Although the latter approach allows more
complete descriptions using fewer approximations, it is mathematically intensive and leads rapidly into
computer-generated solutions. One prominent pioneer of the momentum approach is Lee, who went on to invent
one of the few currently successful dual-rotor turbine flowmeters. Thompson and Grey, provided one of the most
comprehensive models based on the airfoil technique, taking into account, for instance, the effects of blade
interference. The following discussion uses the momentum exchange approach as a way to highlight the basic
concepts of the axial turbine flowmeter.

In a hypothetical situation, where there are no forces acting to slow down the rotor, it will rotate at a speed that
exactly maintains the fluid flow velocity vector at the blade surfaces. Figure 2 is a vector diagram for a flat-
bladed rotor with a blade pitch angle equal to β. assuming that the rotor blades are flat and that the velocity is
everywhere uniform and parallel to the rotor axis, then referring to Figure 2:

Eliminating the time dimension from the left-hand side quantity reduces it to the number of rotor rotations per
unit fluid volume, which is essentially the flowmeter K factor specified by most manufacturers. Hence,
according to Equation (2), in the ideal situation the meter response is perfectly linear and determined only by
geometry. (In some flowmeter designs the rotor blades are helically twisted to improve efficiency. This is
especially true of blades with large radius ratios, R/a. If the flow velocity profile is assumed to be flat, then the
blade angle in this case may be described by tanb = constant 3r. This is sometimes called the "ideal" helical
blade). In practice, there are instead a number of rotor retarding torques of varying relative magnitudes. Under
steady flow, the rotor assumes a speed that satisfies the following equilibrium:

fluid driving torque (3)

= rotor blade surfaces' fluid drag torque
+ rotor hub and tip clearance fluid drag torque
+ rotation sensor drag torque
+ bearing friction retarding torque

Referring again to Figure 2, the difference between the actual rotor speed, rω, and the ideal rotor speed, r i, is

the rotor slip velocity due to the combined effect of all the rotor retarding torques as described in Equation (3).
As a result, the fluid velocity vector is deflected through an exit or swirl angle,θ. Denoting the radius variable by
r, and equating the total rate of change of angular momentum of the fluid passing through the rotor to the
retarding torque, one obtains:

© Bytronic Ltd Page 17 PCT-M1

 Figure 3.

Figure 3. As shown in this vector diagram for a flat-bladed axial turbine rotor, the difference between the ideal
(ωi) and actual (ω) tangential velocity vectors is the rotor slip velocity and is caused by the net effect of the rotor
retarding torques. This gives rise to linearity errors and creates swirl in the exit flow. Identified here are V,
incident fluid velocity vector; VE, exit fluid velocity vector; θ, exit flow swirl angle due to rotor retarding
torques; β, blade pitch angle, same as angle of attack for parallel flow; ω, rotor angular velocity vector; r, rotor
radius vector; F, flow-induced drag force acting on each blade surface; c, blade chord; s, blade spacing along the
hub; c/s rotor solidity factor.

The trends evident in Equation (6) reflect the characteristic decline in meter response at very low flows and the
reason that lower friction bearings and lower drag pickups tend to be used in gas rather than liquid applications
and small-diameter meters. In most flowmeter designs, especially those for liquids, the latter three of the four
retarding torques described in Equation (3) are small under normal operating conditions compared to the torque
due to induced drag across the blade surfaces. As shown in Figure 2, the force, F, due to this effect acts in a
direction along the blade surface and has a magnitude given by:

where:
CD = drag coefficient
S = blade surface area per side

Using the expression for drag coefficient corresponding to turbulent flow selected by Pate et al. and others, this
force may be estimated by:

© Bytronic Ltd Page 18 PCT-M1

where:

Re = flow Reynolds number based on the blade chord shown as dimension c in Figure 2. Assuming θ is small
compared with β, after integration the magnitude of the retarding torque due to the induced drag along the blade
surfaces of a rotor with n blades is found to be:

Combining Equations (9) and (6), and rearranging yields:

Figure 4.

Figure 4. A typical single-rotor axial turbine linearity error, or calibration, curve is plotted for a low-viscosity
fluid showing the principal alternative presentations in current use. Higher accuracy specifications usually
correspond to a 10:1 flow range down from Qmax, while extended operating ranges usually correspond to reduced
accuracies. The hump in the curve is a characteristic feature caused by flow velocity profile changes as Re
approaches the laminar region. This feature varies in magnitude from one meter to the next. Sensitivity and
repeatability degrade at low Re. Percent registration is used only with meters that have mechanical pickups. All
other meters have a K factor. UVC and Re calibrations remain in effect at different known media viscosities,
provided Re or f/n stays within the specified range. Re is referenced to the connecting conduit diameter and is
less within the flowmeter.

The Re range shown is therefore approximate and can vary by an order of magnitude depending on the meter.
Linearity error may also be expressed in terms of Strouhal number (fD/V) vs. Re (VD/v) or Roshko number
(fD2/v), when instead D is a flowmeter reference diameter.

Equation (10) is an approximate expression for K factor because it neglects the effects of several of the rotor
retarding torques as well as a number of meter design and aerodynamic factors such as rotor solidity and flow
velocity profile. Nevertheless, it reveals that linearity variations under normal, specified operating conditions are
a function of certain basic geometric factors and Reynolds number. These results reflect general trends that
influence design and calibration. Additionally, the marked departure from an approximate ρV2 (actually
ρ0.8V1.8µ–0.2 via Re in Equation (8) dependence of the fluid drag retarding torque on flow properties under
turbulent flow, to other relationships under transitional and laminar flow, gives rise to major variations in the K
factor vs. flow rate and media properties for low-flow Reynolds numbers. This is the key reason why axial
turbine flowmeters are generally recommended for turbulent flow measurement.

© Bytronic Ltd Page 19 PCT-M1

Calibration, Installation, and Maintenance
Axial turbine flowmeters have a working dynamic range of at least 10:1 over which the linearity is specified.
The maximum flow rate is determined by design factors related to size vs. maximum pressure drop and
maximum rotor speed. The minimum of the range is determined by the linearity specification itself. Due to
small, unavoidable, manufacturing variances, linearity error curves are unique to individual meters and are
normally provided by the manufacturer. Although recommended where possible, however, the conditions of the
application cannot usually and need not necessarily duplicate those of the initial or even subsequent calibrations.
This has pivotal importance in applications where actual operating conditions are extreme or where the medium
is expensive or difficult to handle. Figure 4 shows a typically shaped calibration curve of linearity vs. flow rate
expressed in terms of multiple alternative measures, various combinations of which are in current use. The
vertical axis thus represents one of the following: the linearity error as a percentage of flow rate, a K factor
expressed in terms of the number of pulses from the rotation sensor output per volume of fluid, or the deviation
from 100% registration, the latter applying only to flowmeters with mechanical pickups. The horizontal axis may
be expressed in terms of flow rate in volume units/time, Reynolds number, or pulse frequency (from the rotation
sensor for non-mechanical) divided by kinematic viscosity, (f/ν), in units of Hz per m2s–1, (Hz/cSt or Hz/SSU;
10–6 m2s–1 = 1 cSt ≃31.0 s Saybolt Universal), and where kinematic viscosity is the ratio of absolute viscosity
(m) to density. Calibrations are preferably expressed, vs. Re or f/ν, which is proportional to Re.

The hump shown in the curve is a characteristic frequently observed at lower Re and is due to velocity profile
effects. K factor vs. f/ ν calibration curves are specifically called universal viscosity curves (UVC), and for most
meters are available from the manufacturer for an extra charge. A key utility of UVC is that where media type
and properties differ significantly from those of the original calibration, accuracies much greater than the overall
linearity error can still readily be obtained via the flowmeter's UVC if the kinematic viscosity of the application
is known. An alternative, advanced calibration technique is to provide response in terms of Strouhal number vs.
Re or Roshko number. This approach, although not widely used, is gaining acceptance in certain applications,
and it is particularly relevant to high accuracy and extreme temperature applications because it further allows
correct compensation for flowmeter thermal expansion errors.

The features of the flowmeter shown here are an electropolished super passivated rotor and support that help
ensure clean flow conditions. Rulon bearings provide long service life in non-lubricating sanitary applications.
The meter has the standard Tri-Clover end fittings, and in the cutaway view can be seen flush holes for the
bearings to satisfy 3-A sanitary crack and crevice standards.

The accuracy of axial turbine flowmeters is reduced by unconditioned flow, especially swirl. An installation
incorporating flow conditioners along with specific upstream and downstream straight pipe lengths is generally
recommended. Some axial turbine flowmeters can be purchased with additional large flow straighteners that
mount directly ahead of the flowmeter body with or conditioning plates that are integral to the body. The
manufacturer is the first source of information regarding installation.

Errors due to flow velocity pulsations are another concern, particularly in certain gas installations, but no
standard technique for effectively counteracting this source of error has yet been adopted. Periodic maintenance,
testing, and recalibration are required because the calibration will shift over time due to wear, damage, or
contamination. For certain applications, especially those involving custody transfer of oil and natural gas,
national and international standards and other recommendations exist that specify the minimum requirements for
turbine meters with respect to these aspects.

© Bytronic Ltd Page 20 PCT-M1

Design and Construction
The many, often proprietary, flowmeter designs incorporate variations in rotors, bearings, pickups, and other
components in formats and materials that are tailored to different applications. Meter bodies are available with a
wide range of standard end-fittings. Within application constraints, the primary objective is usually to optimize
the overall mechanical stability and fit in order to achieve good repeatability. Performance, application, and
manufacturing considerations affect every internal component, but most of all the rotor with respect to blade
shape and pitch, blade count, and balance and rigidity vs. drag, stress, and inertia; bearings with respect to
precision vs. friction, speed rating, and durability; and rotation pickup vs. performance and drag.

Most low-radius-ratio blades are machined flat, while high-ratio blades tend to be twisted. The blade count
varies from about 6 to 20 or more depending on the pitch angle and blade radius ratio so that the required rotor
solidity is achieved. Rotor solidity is a measure of the "openness" to the flow such that higher solidity rotors are
more highly coupled to the flow and achieve a better dynamic range. The pitch angle, which primarily
determines the rotor speed, is typically 30°–45° but may be lower in flowmeters designed for low-density gas
applications. Rotor assemblies are usually a close fit to the inside of the housing. In large-diameter meters the
rotor often incorporates a shroud around the outer perimeter for enhanced stability. Also, since large meters are
often used for heavy petroleum products, the fluid drag created by the wall clearance gap is often designed to
offset the tendency at high media viscosities for the meter to speed up at lower Reynolds numbers. The materials
of construction range from nonmagnetic to magnetic steels to plastics.

Stainless steel ball bearings tend to be used for gas meters and low-lubricity liquids such as cryogenic liquids
and freon; combination tungsten carbide or ceramic journal and thrust bearings are often considered best for
many other liquid meters depending on the lubricity of the medium. Fluid bearings (sometimes called "bearing-
less" designs) are often used in conjunction with the latter, but sometimes with gases as well, for reducing the
drag. They operate by various designs that use flow-induced forces to balance the rotor away from the shaft
ends. Bearing lubrication is either derived from the metered medium, or an internal or external system is
provided. The more fragile, jewelled pivot bearings are also used in certain gas applications and small meters.
Sanitary meters may incorporate flush holes in the bearing assembly to meet 3A crack and crevice standards.
The most common types of rotation sensor are magnetic, modulated carrier, and mechanical, but optical,
capacitive, and electrical resistance are also used. In research, a modulated nuclear radiation flux rotation sensor
for use in certain nuclear reactors has also been reported.

Mechanical pickups, which sometimes incorporate a magnetic coupling, are traditional in some applications.
They can have high resolution and offer the advantage of requiring no electrical power, but the pickup drag tends
to be high. The magnetic and modulated carrier types incorporate at least a coil in a pickup assembly that screws
into the meter housing near the rotor.

In magnetic inductance types, which are now less common, the shroud or blades carry magnetized inserts and
signals are induced in the coil by the traversing magnetic fields. In the more prevalent magnetic reluctance type,
the coil is wrapped around a permanent magnet or magnet pole piece in the pickup assembly, which is mounted
next to a high-magnetic- permeability bladed rotor (or machined shroud). The latter is then typically made of a
magnetic grade of stainless steel such as 416, 430 or 17-4Ph. As the rotor turns, the reluctance of the magnetic
circuit varies, producing signals at the coil. In the more expensive modulated carrier types, the rotor need only be
electrically conductive. The coil is part of an RF oscillator circuit and the proximity of the rotor blades changes
the circuit impedance, giving rise to modulation at a lower frequency that is recovered. The RF types have much
lower drag and higher signal levels at low flow, and can operate at temperatures above the Curie point of typical
ferromagnetic materials. They are preferred for wide dynamic range and high-temperature applications.
Bidirectional flowmeters usually have two magnetic pickups to determine flow direction. This configuration is
useful, for example, for monitoring the container filling and emptying operations often encountered in sanitary
applications. Multiple magnetic pickups are also used in some designs to provide increased measurement
resolution. As for output, various pulse amplifiers, totalizers, and flow computers for gas pressure and
temperature correction, along with 4–20 mA and other standard interface protocols, are available to suit
particular applications. As an example of advanced transmitters, at least one manufacturer provides a real-time,
miniature, reprogrammable, smart transmitter that is integrated into the pickup housing along with a meter body
temperature sensor, for full viscosity compensation and UVC linearization. These transmitters are designed for
use in dedicated applications such as airborne fuel management, where the medium viscosity-temperature
relationship is known.

© Bytronic Ltd Page 21 PCT-M1

© Bytronic Ltd Page 22 PCT-M1
Control Methods
Modern Control Theory
In contrast to the frequency domain analysis of the classical control theory, modern control theory utilizes the
time-domain state space representation, a mathematical model of a physical system as a set of input, output and
state variables related by first-order differential equations. To abstract from the number of inputs, outputs and
states, the variables are expressed as vectors and the differential and algebraic equations are written in matrix
form (the latter only being possible when the dynamical system is linear). The state space representation (also
known as the "time-domain approach") provides a convenient and compact way to model and analyze systems
with multiple inputs and outputs. With inputs and outputs, we would otherwise have to write down Laplace
transforms to encode all the information about a system. Unlike the frequency domain approach, the use of the
state space representation is not limited to systems with linear components and zero initial conditions. "State
space" refers to the space whose axes are the state variables. The state of the system can be represented as a
vector within that space.

Topics in Control Theory

The stability of a general dynamical system with no input can be described with Lyapunov stability criteria. A
linear system that takes an input is called bounded-input bounded-output (BIBO) stable if its output will stay
bounded for any bounded input. Stability for nonlinear systems that take an input is input-to-state stability (ISS),
which combines Lyapunov stability and a notion similar to BIBO stability. For simplicity, the following
descriptions focus on continuous-time and discrete-time linear systems.

Mathematically, this means that for a causal linear system to be stable all of the poles of its transfer function
must satisfy some criteria depending on whether a continuous or discrete time analysis is used:
 In continuous time, the Laplace transform is used to obtain the transfer function. A system is stable if
the poles of this transfer function lie strictly in the open left half of the complex plane (i.e. the real part
of all the poles is less than zero).
 In discrete time the Z-transform is used. A system is stable if the poles of this transfer function lie
strictly inside the unit circle. i.e. the magnitude of the poles is less than one).

When the appropriate conditions above are satisfied a system is said to be asymptotically stable: the variables of
an asymptotically stable control system always decrease from their initial value and do not show permanent
oscillations. Permanent oscillations occur when a pole has a real part exactly equal to zero (in the continuous
time case) or a modulus equal to one (in the discrete time case). If a simply stable system response neither
decays nor grows over time, and has no oscillations, it is marginally stable: in this case the system transfer
function has non-repeated poles at complex plane origin (i.e. their real and complex component is zero in the
continuous time case). Oscillations are present when poles with real part equal to zero have an imaginary part not
equal to zero.

Differences between the two cases are not a contradiction. The Laplace transform is in Cartesian coordinates and
the Z-transform is in circular coordinates, and it can be shown that:
 the negative-real part in the Laplace domain can map onto the interior of the unit circle
 the positive-real part in the Laplace domain can map onto the exterior of the unit circle

© Bytronic Ltd Page 23 PCT-M1

Main Control Strategies
Every control system must guarantee first the stability of the closed-loop behaviour. For linear systems, this can
be obtained by directly placing the poles. Non-linear control systems use specific theories (normally based on
Aleksandr Lyapunov's Theory) to ensure stability without regard to the inner dynamics of the system. The
possibility to fulfil different specifications varies from the model considered and the control strategy chosen.
Here a summary list of the main control techniques is shown:

Adaptive Control
Adaptive control uses on-line identification of the process parameters, or modification of controller gains,
thereby obtaining strong robustness properties. Adaptive controls were applied for the first time in the aerospace
industry in the 1950s, and have found particular success in that field.

Hierarchical control
A Hierarchical control system is a type of Control System in which a set of devices and governing software is
arranged in a hierarchical tree. When the links in the tree are implemented by a computer network, then that
hierarchical control system is also a form of Networked control system.

Intelligent control
Intelligent control uses various AI computing approaches like neural networks, Bayesian probability, fuzzy
logic, machine learning, evolutionary computation and genetic algorithms to control a dynamic system.

Optimal control
Optimal control is a particular control technique in which the control signal optimizes a certain "cost index": for
example, in the case of a satellite, the jet thrusts needed to bring it to desired trajectory that consume the least
amount of fuel. Two optimal control design methods have been widely used in industrial applications, as it has
been shown they can guarantee closed-loop stability. These are Model Predictive Control (MPC) and Linear-
Quadratic-Gaussian control (LQG). The first can more explicitly take into account constraints on the signals in
the system, which is an important feature in many industrial processes. However, the "optimal control" structure
in MPC is only a means to achieve such a result, as it does not optimize a true performance index of the closed-
loop control system. Together with PID controllers, MPC systems are the most widely used control technique in
process control.

Robust control
Robust control deals explicitly with uncertainty in its approach to controller design. Controllers designed using
robust control methods tend to be able to cope with small differences between the true system and the nominal
model used for design. The early methods of Bode and others were fairly robust; the state-space methods
invented in the 1960s and 1970s were sometimes found to lack robustness. A modern example of a robust
control technique is H-infinity loop-shaping developed by Duncan McFarlane and Keith Glover of Cambridge
University, United Kingdom. Robust methods aim to achieve robust performance and/or stability in the presence
of small modelling errors.

Stochastic control
Stochastic control deals with control design with uncertainty in the model. In typical stochastic control problems,
it is assumed that there exist random noise and disturbances in the model and the controller, and the control
design must take into account these random deviations.

© Bytronic Ltd Page 24 PCT-M1

Control Principles
To predict the way a system will respond to a particular disturbance or change of set point an attempt might be
made to model it mathematically. This usually means representing physical device characteristics such as inertia,
damping force, capacitance, and resistance etc. by terms in a differential equation. Although idealised conditions
which cannot be achieved in practice often have to be assumed, valuable insights into a system’s performance
may be gained this way. The ideal system is linear in that it may be represented by a linear differential equation.
Systems are often classified by the order of the equation which represents their behaviour e.g. 1st. order, 2nd.
order etc.

1st. Order Systems

The behaviour of a 1st. order system can be described by a 1st. order linear differential equation. For example,
consider a temperature probe with the following response:

Where:
1 = Temp of medium being measured (ºC)
2 = Temp of probe (ºC)
M = Mass of probe (kg)
Cp = Specific heat capacity of probe (J.kg-1.ºC-1)
U = Overall heat transfer coefficient (J.m-2.ºC-1.s-1)
A = Surface area (m2)
t = Time (s)

Applying unsteady state "energy balance", the rate of heat transfer to the probe must equal the rate of heat
accumulation at the probe. Therefore assuming no losses:

d 2
Where: = rate of change of the probe temperature
dt

To comply with the standard format the forcing function must be on the right hand side and the response on the
left hand side. Using the convention that operator D means d/dt we therefore have:

(UA  MCpD )  2  UA1

MCpD
(1  )  2  1
UA

© Bytronic Ltd Page 25 PCT-M1

Comparing this with the standard form for a 1st. order system:

(1  TD )  2  1 (Equation 1)

We see that for this example the time constant T is MCp/UA. Using K as a constant of integration, the solution
for this particular differential equation is:

In a similar manner it can be shown that a 2nd. order system may be represented by:

(MD 2  fD  K )  2  K1

Laplace form: (ms 2  fs  K )  2  K1

These linear differential equations provide a complete system description and for a given input the output may be
determined by solving the equations with integrating factor or 'D' operator methods etc. However these methods
can be cumbersome as well as difficult and therefore it is useful to consider the transfer function concept as
described latter.

Transfer Functions
The transfer function (TF) of a linear system is defined as the ratio of the Laplace transform of the output to the
Laplace transform of the input, with all initial conditions assumed to be zero.

If differential equations are linear the Laplace transform may be regarded as a method of converting them into
algebraic equations to allow more convenient manipulation. This method involves converting functions of time
(t) in the differential equation to a function of the Laplace variable (s) by applying the transform:

Extensive listings of these transforms have been compiled and it is rarely necessary to evaluate the integral. The
transformed equation may be manipulated with standard algebraic techniques to yield a solution in the variables.
The final step is to inversely transform the solution in back into the time domain, again using the transform
tables.

As control systems usually involve multiple ‘blocks’, s-plane manipulation can be quite complex. However it
can be simplified by the use of block diagram reduction techniques before applying the inverse transformation to
the time domain.

© Bytronic Ltd Page 26 PCT-M1

Block Diagrams
In general these include blocks for the controller, plant and perhaps feedback signal processing plus take off
points, summing junctions and directional arrows.

To ease algebraic manipulation a short hand notation for the transfer function Gn and feedback signals Hn may
be used, where n is a suitable subscript. A typical closed loop control block diagram is shown in figure below.

Typical Closed Loop Block Diagram

The block diagram of a practical system is often quite complicated but by applying systematic block diagram
reductions, multiple loop systems may be simplified.

Example of Block Diagram Reductions

© Bytronic Ltd Page 27 PCT-M1

Having obtained the reduced block diagram the overall system TF may be determined by replacing the short
hand notations Gn, Hn with the full forms of the individual block TFs. Inversely transforming the overall TF
from the s-plane to the time domain will then allow the model's response to be calculated at specific times.

As an example consider the servo control system shown diagrammatically below. The set point is determined by
a single turn manual potentiometer and feedback is provided by another single turn potentiometer linked to the
output shaft. The error between the set point and the measured value is amplified and used to control the motor
to regulate the shaft position. When the feedback position equals the set point the error will be zero and the
position will be maintained. For example we will assume that the amplifier has a gain of K1, the motor has a TF
of K2/s and the TF for the input potentiometer equals that of the feedback transducer. For the potentiometer one
complete rotation will cause the output voltage to vary from zero to its maximum value. Therefore:

(Equation 2)

Where: A = Angle of rotation in radians (input)

V = Supply voltage
The potentiometer TF is therefore:

© Bytronic Ltd Page 28 PCT-M1

 Servo Control Loop

Using the block diagram reduction techniques referred to earlier this reduces to:

Reduced Block Diagram for the Servo Control System

The block diagram can be reduced in several stages to that shown. Therefore the overall TF is:

If a step input is applied then from rule 2 of the Laplace transform Table, R(s) = A/s therefore the system output
(given by R(s) multiplied by the TF) will be:

We then use inverse transforms to transform from the s-plane back to the time domain. From rule 9 using laplace
transform table inversely:

Therefore:

© Bytronic Ltd Page 29 PCT-M1

The output response c(t) may be calculated by inserting the values K1, K2, V, A and the time (t). For example
take K1 = 2, K2 = 5, V = 5V and the set point A = 0.5 radians. Substituting these values into the above equation
produces the following results:

Time (t) Seconds c(t) Response (Radians)

0.05 0.1641
0.10 0.2744
0.15 0.3484
0.20 0.3982
0.25 0.4316
0.30 0.4541
0.35 0.4691
0.40 0.4793
0.45 0.4860
0.50 0.4906

By plotting c(t) against time, the system response may be evaluated for a step input of 0.5 radians. The result
may either be plotted by hand or the system may be simulated (open loop) using a control system design and
simulation software package. The plot is shown in figure below.

Reduced Block Diagram for the Servo Control System

System Response of the Servo System

© Bytronic Ltd Page 30 PCT-M1

Assessment of System Performance
It is sometimes necessary to assess the performance of a process control loop, (plant, sensor, controller and
actuator) in order to ascertain whether it may be improved. The best method is to obtain a response curve,
(sometimes called a process reaction curve), for the type of input disturbance expected in normal operation.
Typically the input disturbance might be a step change e.g. increase the flow-rate set point by 15% or reduce the
temperature set point by 10%. In general a control loop’s response to a disturbance includes both transient and
steady-state components. Given enough time, the transient component will eventually die away completely and
the difference between the measured value of the process variable before the disturbance and after the transient
has faded, represents the steady-state element of the response. (Transient and steady-state responses, which are
determined by alternative solutions to the underlying differential equation, are superimposed upon each other).

Transient Responses
Transients are caused by sudden or discontinuous changes in a variable upon which the measured value depends.
Depending upon the tuning of the controller, the transient response will be under damped, over damped or
critically damped. Typical curves for these three are shown in figure below.

In the under damped case the transient response of the system to a disturbance is oscillation about the set point.
In the over damped case the transient response is an excessively long period in which the measured value
increases until it reaches the set point. In the critically damped case the response is an optimally rapid increase to
the set point without oscillation.

In order to compare system performance other parameters may be considered including peak overshoot, rise
time, settling time, period and transport delay. see Figure below. The definitions that follow may be applied
equally to open or closed loop systems.

 Peak overshoot is the maximum amount by which the response exceeds the final steady state value of
the process variable. It is sometimes expressed as a percentage of the final steady state value.
 Rise time is the time taken for the response to increase from 10% of its final steady state value to 90%
of its final steady state value.
 Settling time is the time taken for the response to reach its final steady state value, within some
specified tolerance. System Performance Parameters Figure shows the settling time for a 5% tolerance.
 Periodic time (or period) is the duration of one complete cycle of oscillation. It can therefore be
measured as the interval between alternate crossings of the final steady state value or the interval
between successive peaks or successive troughs on the response curve.
 Frequency is the reciprocal of the period, i.e. the number of cycles per second which is expressed in
Hertz (Hz). Sometimes the frequency is expressed in radians per second and the relationship between
the two units is that radians per second equals 2 times the frequency in Hertz.
 Transport delay is the period during which there is no change in the process variable after a step change
has been made to the set point.

Completely separate from the transient, steady state error may also be present in the overall response, see Steady
State Errors Figure. This is the deviation between the measured value and set point once the system has
stabilised and the transient has faded. This may be evaluated mathematically for a system where the transfer
function is known, using the final value theorem discussed in a latter section.

Typical System Response

© Bytronic Ltd Page 31 PCT-M1

 System Performance Parameters

Ramp Input

Steady State Errors

© Bytronic Ltd Page 32 PCT-M1

 Final Value Theorem Example

G K
NOTE - Reduction Rule where G =
1+GH s(1+0.3s)

Control System Instability

If the output of a system increases even though the input is not demanding such an increase the system is said to
be unstable and is of no practical use. Closed loop control relies upon negative feedback to eliminate error i.e.
measured value is subtracted from set point to determine error and this is used to calculate control action. If
parameter values are such that positive feedback occurs then errors are amplified and the system becomes
unstable. A control system should be designed to be rigidly stable within its intended operating range.
Mathematical techniques involving Routh arrays, Bode plots or Nyquist plots may be used to assess system
stability. These methods are suitable for assessing stability of the process 6 control loops.

© Bytronic Ltd Page 33 PCT-M1

Final Value Theorem
The final value theorem allows the steady state value of a time domain function to be evaluated from its Laplace
transform.

The final value theorem states that:

If: L[f(t)] = E(s)

Then: ess = lim f(t) = lim sE(s)

t s0

Where: f(t) is the error as a function of time, E(s) is the error transfer function
and ess is the steady state error.

For an example refer to figure 7.5. Using block diagram reduction techniques, the system transfer function is:

(1)

We now wish to apply the final value theorem for a ramp input. Firstly it can be seen that:

E(s) = R(s) - C(s) (2)

We know the input R(s) and we are interested in the error E(s) so rearranging:

C(s) = R(s) - E(s) (3)

From equation (1): C(s)(s(1 + 0.3s) + K) = R(s)K

Substituting equation (3) for C(s):

(R(s) - E(s))(s(1 + 0.3s) + K) = R(s)K

R(s) s(1 + 0.3s) + R(s)K - E(s) s(1 + 0.3s) - E(s)K = R(s)K

Re-arranging:
E(s) s(1 + 0.3s) + E(s)K = R(s) s(1 + 0.3 s)

E(s)( s(1 + 0.3s) + K ) = R(s) s(1 + 0.3s)

Therefore:

Applying the final value theorem: ess = lim f(t) = lim sE(s)
t s0

The steady state error is:

However, R(s) is a ramp input which is 1/s2 (from rule 5 in Appendix 8) therefore:

As: s  0: ess = 1/K

Therefore the greater the value of K the lower will be the steady state error in this control loop.

© Bytronic Ltd Page 34 PCT-M1

The Routh-Hurwitz Test
A Routh array may be used to determine the stability of a linear system directly from its transfer function (TF),
without any knowledge of the input signal. This is possible because stability is governed by the denominator of
the closed loop transfer function. Equating the denominator to zero gives the ‘characteristic equation’ of the
closed loop system. The nature of the roots of this equation indicates stability or instability.

C( s) 1
For example, transfer function: 
R( s ) s (1  0 . 3s )  K

Characteristic equation: s(1+ 0.3s) + K = 0

Firstly write down the characteristic equation in descending order of the powers of "s":

If any coefficient is missing or negative then the system is definitely unstable. If all of the coefficients are
positive the next step is to construct the Routh array as shown below:

COLUMNS

1 2 3 4

1 a0 a2 a4 a6

2 a1 a3 a5 a7

3 b1 b2 b3

ROWS 4 c1 c2 c3

5 d1 d2 d3

6 e1

7 f1

The first two rows are obtained from the descending order power characteristic equation. The remaining rows are
calculated from the elements in the first two rows as follows:

The construction of the array terminates only when zeros are obtained. The Routh-Hurwitz criterion is that the
first column must have no sign changes in order for the system to be stable.

© Bytronic Ltd Page 35 PCT-M1

 C( s) 1
Example 1 
R( s ) 10s  2s  3
2

Characteristic Equation:0 = 10s2 + 2s + 3

Routh Array:
10 3 0
2 0 0
60
3 0
2
0

There is no sign change in first column therefore the system is STABLE.

C( s) 3K
Example 2 
R( s ) 10s  s ( s  7 )  K
4

Characteristic Equation: 0 = 10s4 + s2 + 7s + K

There is a missing s3 term therefore the system is UNSTABLE.

C( s) 1
Example 3 
R( s ) 5s  4s  10s  20
3 2

Characteristic Equation: 0 = 5s3 + 4s2 + 10s + 20

Routh Array:
5 10 0
4 20 0
-15 0
20

There is a sign change in the first column therefore the system is UNSTABLE.

Example 4

The Routh array method may be used to find the limiting gain value for a controller (K), whilst avoiding
instability. e.g.: Characteristic equation s4 + 6s3 + 11s2 + 6s + K = 0 :

Routh Array:
1 11 K
6 6 0
10 K 0

0
K

For the system to be stable then 60 must be greater than 6K therefore K must be less than 10. Note that the Routh
array method does not indicate the degree of stability, this may be examined using either Bode or Nyquist plots
which are discussed in latter sections.

© Bytronic Ltd Page 36 PCT-M1

Bode Plots
The phrase ‘frequency response’ refers to the overall relationship between a system’s input and output signals
over a range of input signal frequencies. In general we are interested in the way the system’s gain and phase lag
(between input and output) vary with frequency. (Gain is the ratio of output signal amplitude to input signal
amplitude). The Bode plot method of assessing stability analyses the open loop frequency response of a system
in order to predict the degree of stability for the closed loop. A complete Bode plot comprises graphs of gain
magnitude |G| and phase lag angle θ plotted against input signal frequency ω. Determination of the degree of
stability is by a straightforward inspection of the graphs.

Bode plots may be drawn using experimentally determined open loop data which are often quite easy to obtain.
This offers the great advantage that no knowledge of the differential equations underlying the system or its
transfer function, are required. Alternatively, if the TF is known then an approximate Bode plot may be drawn
without carrying out any experiments. (This would be desirable for certain industrial systems which might suffer
damage if driven beyond their design limits with experimental test signals). However the open loop Bode plot is
produced, if it is reasonably accurate it will give a good guide to the degree of closed loop stability. Before
considering how to interpret a Bode plot it is worth digressing a little to discuss how to sketch one, given a
known TF.

10 (1  0.1s )
As an example consider the open loop transfer function =
s (1  0. 5s )(1  0 . 04s )

This may be considered as the product of several standard TF elements. If the Bode plots for each element are
plotted individually on logarithmic scales, the overall Bode plot for the whole TF may be obtained by simple
addition and straight line approximation. (Since multiplication is equivalent to adding logarithms).

Bode Gain versus Frequency Plot:

With reference to the standard functions in Appendix 4, the transfer function may be broken down into the
following elements:

Gain |G| in decibels may be plotted against frequency (ω) in radians per second. Refer to Bode Plot Example
Figure.

NOTE: It is best to plot gain against frequency and phase angle against frequency on the same log paper since
cross referencing is necessary to assess system stability.

© Bytronic Ltd Page 37 PCT-M1

 Bode Plot Example

Determining the Gain and Phase Margins from a Bode Plot

© Bytronic Ltd Page 38 PCT-M1

Bode Phase Lag versus Frequency Plot:
With reference to the standard functions, the transfer function may be broken down into the following elements:

type K no effect
type 1+ Ts (1 + 0.1s) has a break point at 10 rad/s
type 1/s lag of -90o

1
type has break points at 2 rad/s and 25 rad/s
1  Ts

Phase lag angle (θ) may be plotted against frequency (ω) in radians per second. Refer to Bode Plot figure.

For both graphs in Bode Plot figure the line for the overall system response may be obtained by adding the
values of the individual elements at each point along the frequency axis.

For closed loop feedback systems instability occurs if an open loop gain of 1 coincides with an open loop phase
lag of -180º. (This leads to positive feedback because closing the loop introduces a further 180º phase lag when
the feedback is subtracted from the input). The relative stability of a closed loop system is therefore defined
using two criteria, see Determining the Gain and Phase Margins from a Bode Plot figure:

Gain Margin. This is the gain of the system in decibels (dB) when the phase lag is -180º. It expresses the factor
by which the gain could be increased before the system becomes unstable (or more accurately, marginally
unstable).

Space Margin. This is 180º less the phase lag angle at the frequency when the gain is 0 dB. It expresses the
additional phase lag that would make the system unstable (or more accurately, marginally unstable).

Typical values for a stable system are a gain margin of more than 10 dB and a phase margin of more than 45º.

For our original Bode plot example the gain margin is 60 dB and the phase margin is 50º so this system would be
stable in closed loop mode.

© Bytronic Ltd Page 39 PCT-M1

Nyquist Plots
The Nyquist stability criterion is another method of assessing the closed loop stability of a system by reference
to its open loop frequency response. A Nyquist plot represents the open loop frequency response as a polar plot
of the gain magnitude |G| and the phase lag <G over a frequency range from zero to infinity. The information is
plotted using vectors from the origin and an open loop system is represented by a series of such vectors, each one
at a different frequency. By joining the extremities of the vectors a frequency locus may be obtained, see
Principle of Nyquist Plot figure.

As with the Bode plot method, data for drawing the curve may be obtained experimentally or from analysis of a
known transfer function. These days a simulation package would normally be used to display the curve for a
particular TF but it could be done manually if necessary. In the latter case the first step is to evaluate the gain
magnitude |G| and phase lag <G. This is done by subdividing the system’s open loop transfer function into the
standard elements.
5
As an example consider the open loop transfer function = .
s (1  0. 5s )(1  0.166s )
The standard elements within this TF are:

To evaluate |G| multiply the standard gain elements from Appendix 12:

To evaluate <G add the standard phase lag elements from Appendix 12:

(2)

Now substitute values for ω into (1) and (2) and calculate |G| and <G to obtain:
ω (rad/s) |G| (dB) <G (º)
----------------------------------------------------
1.5 2.58 -140.8
2 1.68 -153.4
3 0.74 -173
4 0.46 -187
8 0.09 -219
16 0.013 -242

© Bytronic Ltd Page 40 PCT-M1

Plot the Nyquist diagram
The assessment of closed loop system stability reduces to an observation of whether or not the open loop
Nyquist plot encloses the point on the real axis where the gain is -1. If it does then the system is unstable. The
relative stability of the closed loop system is again defined using the two criteria:

Gain Margin. This is the number of decibels (dB) by which the magnitude of the open loop gain falls short of
unity when the phase angle is -180º.

Phase Margin. This is the angle by which the open loop phase lag falls short of -180º when the gain magnitude
is unity.

As was stated earlier, typical values for a stable system are a gain margin of more than 10 dB and a phase margin
of more than 45º. the gain margin in dB for our example is -20 log10 0.6201 = 4.151dB and the phase margin
just over 12º. Since these values are below those recommended for a stable closed loop system the system would
be regarded as unstable, or at best, marginally stable.

Principle of Nyquist Plot

© Bytronic Ltd Page 41 PCT-M1

Process Modelling
In the previous sections we have seen that given the open loop transfer function it’s possible to determine
whether the closed loop system will be unstable or not using the Routh test. It is also possible to determine the
degree of stability with a Bode or Nyquist plot. However in many situations we do not have any idea of the
transfer function for a particular open loop system. This section touches on different types of tests which may be
used to collect data, from which an approximate transfer function may be derived.
There are three types of experimental approaches:

 Step Data Tests

 Frequency Response Tests
 Time Domain Test (Online Method)

Process Models from Step Data Tests

This technique subjects a process operating under steady-state open loop conditions to a step change input and
records the resulting transient response, which is called a process reaction curve. Data taken from this curve are
used to estimate the mathematical model (i.e. transfer function). In the open loop context, the word ‘input’ refers
to the actuator setting which directly affects the process variable. (Clearly the phrase ‘set point’ has no meaning
in an open loop situation). In industry step data tests the following points should be observed:

 Step size must be carefully chosen so that the process reaction curve is distinguishable from plant
noise without the system being driven beyond its linear range.
 The process should be free of load changes for the duration of the tests.
 Tests should be repeated several times and the average values of parameters should be used to
estimate the transfer function.

From the process reaction curve several methods have been suggested for determining the parameters used in the
transfer function. Three important approaches are:

 Ziegler, Nichols and Miller describe a simple method for finding the parameters of a first order plus
dead time model.
 Caldwell, Oldenbourg and Sartorius et al suggest a procedure to determine the parameters of a
second order model.
 Sundaresan et al developed a method for evaluating second order models with dead time.

Process Models from Frequency Response Tests

Another way to estimate the TF of an open loop system is to examine its frequency response across a large range
of frequencies. There are several ways to do this but the most straightforward is to drive the system with a series
of sinusoidal input signals. (These are often referred to as forcing functions). These inputs should have the same
amplitude but different frequencies and in each case the gain and phase lag (between input and output signals)
should be measured. Once the frequency response tests have been completed a Bode plot should be drawn from
the data.’ Once these are known the open loop TF may be written down.

The use of frequency response tests to determine the transfer function for a process depends upon various
factors:

 Is the production down time for the period required to conduct the tests, affordable?
 What type of driving signal is permissible? (Sinusoidal signals are ideal but they may be completely
unacceptable for some industrial processes).
 What magnitude of driving signal is acceptable? (Components should not be stressed beyond the level
for which they were designed).
 The degree of signal noise in the system.

The validity of any transfer function estimated using this technique may be checked by simulating the open loop
response using a simulation package and comparing this to the system's actual response.

© Bytronic Ltd Page 42 PCT-M1

Process Models from Time Domain Tests
In many industrial processes the parameters vary with time, for example the build up of sediment in pipe work or
reaction vessels might increase the lag between input and output signals. In such cases it might be advisable to
identify the transfer function from time to time and update the controller settings if necessary. This leads to the
more complex approach of auto-tuning and on-line system identification both of which are beyond the scope of
this manual.

Once you have a reasonable idea of a system’s transfer function it is often possible to design a control algorithm
which uses measured value feedback to maintain the output at the set point. Normally an industrial controller
would be designed to accept set point changes and drive the output to the new value smoothly and quickly
without oscillation or dramatic overshoot. A good controller will also eradicate small to medium sized
perturbations due to changes in other parameters, e.g. upstream pressure, fluid density etc. Even without a
detailed mathematical model of the system it is often still possible to design such a controller, see Ziegler
Nichols tuning. In many cases the controller of choice is the so called ‘three term algorithm’. This takes its name
from the fact that the control output is determined from the error signal (set point minus measured value) by
means of a calculation involving up to three distinct mathematical operations.

PID Controllers
PID Control uses ‘three term control’ to calculate the output signal which is used to drive the controlled element
of the unit. Essentially a three-term controller takes a measured value from the sensor (pressure transducer for
level sensor) and compares it against the set point (desired value). The discrepancy between the measured and
desired values, called the error, is used to determine the control output signal. This is all represented in the
following block diagram.

The line running from the block labelled ‘Measurement’ to the circular ‘summing junction’ in the diagram is
known as the feedback signal and any process that contains such a signal is a closed loop system. The three
components of a three-term controller each behaves in a different way but when combined they can provide a
very accurate degree of control. The three elements of the controller are the Proportional term, the Integral term
and the Derivative term, hence the other name by which this type of algorithm is known, ‘PID control.

A three term controller consists of elements which are proportional to the:

 Magnitude of the error signal – ‘proportional term’
 Time integral of the error signal – ‘integral term’
 Time derivative of the error signal – ‘derivative term’

It has been found that such a controller can give excellent results, taking into account as it does the absolute
magnitude of the error, the history of the error and the current rate of change of the error. The three terms are
individually explained.
error (t )  SP (t )  MV (t )

© Bytronic Ltd Page 43 PCT-M1

Proportional Control Term
The output of a proportional controller (or the portion of the output of a two or three term controller contributed
by its proportional term) is proportional to the error between the set point (SP) and the measured valve (MV).
Proportional control may be expressed as either:

1. Proportional Gain (PG): Control output (MP) equals proportional gain multiplied by the error.

Mathematically: MP = PG x (SP – MV) + C = PG e(t) + C

Where: C = Controller output with zero error
e(t) = Error as a function of time

2. Proportional Band (PB): With proportional control saturation occurs at a certain value of error when the
control output reaches 100%. Thereafter further increases in error do not produce further increases in control
output. This same effect occurs when the output drops to 0%. The error band where the control output is between
0% and 100% is called the proportional band, thus the higher the gain, the smaller the proportional band.

Unfortunately a simple proportional controller rarely produces adequate results. The main problem is a
phenomenon called proportional offset which is a steady state error between SP and MV. A proportional
controller cannot eliminate this offset - there must be some error in order for there to be an automatically
determined component of the controller output. The only value of SP for which there is no offset is the value for
which C was initially chosen. (C is often called the manual contribution because it is set from outside the control
loop [usually by a human operator] and is not under the controller’s influence.

Integral Control Term

The integral control term is often used to remove proportional offset errors. It determines a component of
controller output (MI) based upon the history of the error. It is calculated by multiplying the net area under the
error curve, e(t), by PG divided by the integral action time (IAT) in seconds.

IAT is defined as the time taken for the integral action to duplicate the proportional action of the controller, if the
error were to remain constant during the period.

Derivative Control Term

The derivative control term is often used to reduce the response time of the system. It determines a component of
controller output (MD) based upon the current rate of change of the error. It is calculated by multiplying the
gradient of the error, e(t), by PG times the derivative action time (DAT) in seconds.

DAT is defined as the time it would take the proportional action of the controller to duplicate the instantaneous
output of the derivative term. The rate of change of error may be approximated by taking the difference between
two values of error and dividing by the time between the two values.

Where: e(t) = error at time t

e(t0) = error at time (t0)
The derivative control mode is never used alone since there is no controller output corresponding to a zero rate
of change. An important point to remember about derivative action is that it can exaggerate any high frequency
noise in the system so it should always be used cautiously.

© Bytronic Ltd Page 44 PCT-M1

Multi Term Control
In many industrial control situations the three terms, P I and D are combined to form two or three term
controllers:

© Bytronic Ltd Page 45 PCT-M1

Ziegler Nichols Tuning
How does one select appropriate proportional, integral and derivative values for any given process? Simply
plucking numbers out of the air and trying them in a PID controller might be acceptable for very small systems
but if you were trying to establish a control algorithm for a full sized industrial process then this would not be a
very good strategy!

This method could time consuming which might bring about financial losses due to the plant downtime. This
approach might also risk causing damage to actuators or sensors - if it was to drive them beyond their intended
range of operating values for instance. A more scientific approach to finding a reasonable set of PID terms is
required.

Ziegler Nichols tuning is a semi-empirical method of controller tuning that gives a reasonable guide to the
parameters PG, IAT and DAT. Once the values have been determined it might be necessary to apply some fine
tuning for optimum performance but in general, Ziegler Nicholls tuning gives acceptable results for many
systems. There are two techniques and most industrial processes will be amenable to one or other.

1. Closed Loop Continuous Cycling Method.

The Continuous cycling method requires that the integral and derivative control actions are reduced to zero and
the controller gain (PG) is gradually increased until the onset of permanent oscillations. At this point the gain
(kp) and the period of oscillation (T) are recorded. The recommended control parameter settings are then:

PG IAT DAT
PI Controller: 0.45 kp 0.83T
PID Controller: 0.6 kp 0.5T 0.125T

This method can be made to oscillate permanently with a proportional only controller, as illustrated by the
following graph.

This closed loop method does assume that the system may be made unstable using only proportional control.
For some processes this might not be true or it might be entirely unacceptable to push the process into instability
because of the risk of equipment damage or operator injury. The second method uses parameters obtained from
an open loop process reaction curve to calculate the required controller terms. A typical process reaction curve is
shown below:

© Bytronic Ltd Page 46 PCT-M1

 2. Open Loop Process Reaction Curve Method

The second technique requires that an open loop step response curve be produced showing a measurable
‘transport delay’ or ‘dead time’ (Transport delay is the period during which there is no change in the process
variable after the controller output has been stepped up or down).

The following graph shows a typical process reaction curve.

For a process which reacts in this way the recommended control parameter settings are then:

PG IAT DAT
P Controller: u
RL
PI Controller: u
0. 9
RL 3.3L
PID Controller: u
1.2
RL 2L 0.5L

Where: Δu represents the step input used to obtain the reaction curve, as a fraction of the total possible input
range.

 R is the maximum slope of the process reaction curve

 L is as defined on the graph above

© Bytronic Ltd Page 47 PCT-M1

Analysis of Digital Control Systems
The use of laplace transforms for analysis and design of continuous linear control systems. and z transforms may
used to represent and manipulate discrete time systems. The z transformation transforms functions of discrete
time, f(n), into functions of the variable z. As for laplace transforms, tables of z transforms have been evaluated.

Where: The capital Z and braces indicate the z transformation operator

The subject of z transform techniques is very advanced and well beyond the scope of this documentation. A brief
summary of some of the idea is given in the remainder of this section.

Several steps are required to analyse a digital control loop:

 Starting with the block diagram of the system use block diagram reduction techniques to reduce its
complexity.
 Determine the pulse transfer function of the reduced block diagram.
 The stability of the system may be assessed from the pulse transfer function.
 The response of the system may be assessed in the time domain by determining the inverse z transform.

The algorithm implemented in a digital control loop might be a digital form of a traditional three term controller
or a specialised digital controller such as a Dahlin or Deadbeat type.

Diagram of a Sampled Signal

Signal Reconstruction using Digital-to-Analogue Conversion and a Zero-Order Hold Device

© Bytronic Ltd Page 48 PCT-M1

Block Diagrams for Digital Systems
On a block diagram a sampled signal is denoted by a switch symbol with a directional arrow, as shown below.

There is a significant difference between the reduction of continuous block diagrams and discrete block
diagrams. If there is no sampling between two continuous elements then they may be combined. Thus:

R(s) C(s)
G1(s) G2(s)

Simplifies to:

However if sampling is present between two continuous elements they cannot be combined.

The block diagram above cannot be simplified, each block must be considered as an individual element.

Pulse Transfer Functions

In continuous (analogue) control systems the laplace transform of the output function is related to the laplace
transform of the input function by the transfer function of the system. Similarly, when dealing with digital
systems the z transform of the pulsed output may be related to the z transform of the pulsed input by the Pulse
Transfer Function of the system.

For example to find the pulse transfer function of the open loop system below:

1 1
G( s )  
( s  1)( s  3 ) ( s  a)( s  b)

© Bytronic Ltd Page 49 PCT-M1

With a closed loop system the analysis becomes slightly more complicated, for example:

Note that asterisks have been used to indicate the discrete time forms of continuous signals.

From the diagram: C*(s) = Gp(s).G0h(s).U*(s)

Let: G(s) = Gp(s).G0h(s)

Therefore: C*(s) = G(s).U*(s) (1)

And: U*(s) = D(z).E*(s) (2)

And: E*(s) = R*(s) – C*(s) (3)

Substitute (2) into (1): C*(s) = G(s).D(z).E*(s) (4)

Substitute (3) into (4): C*(s) = G(s).D(z).[R*(s) – C*(s)]

C*(s) = G(s).D(z).R*(s) – G(s).D(z).C*(s)

C*(s).[1 + G(s).D(z)] = G(s).D(z).R*(s)

The pulse transfer function is:

Where: G(z) = Gp.G0h(z)

By introducing some values this may be further reduced. Taking D(z) as a pure gain controller with a gain value
of 5; Gp(s) as 1/(s+1); sampling time T as 0.2 seconds and G0h(s) as (1-e-Ts)/s and using capital Z and square
brackets to indicate the z transformation operator:

Since: Z e  Ts  z 1

© Bytronic Ltd Page 50 PCT-M1

 (5)

Now to evaluate Z[1/s(s+1)] see Appendix 2:

(1  e  T ) z ( z  1)
Substitute this into (5): G ( z)  T
( z  1)( z  e ) z

1 e  T
G( z)  (6)
z  eT

© Bytronic Ltd Page 51 PCT-M1

Z Transform Initial and Final Value Theorems
These theorems provide a quick method of evaluating the initial and final values of a system represented in a z
transform format, without having to calculate the inverse z transform.

The initial value theorem states that if the z transform of f(t) is F(z) then:

lim f(t) = lim F(z)

t0 z 

The final value theorem states that if the z transform of f(t) is F(z) then:

lim f(t) = lim [F(z) ( 1 - z-1 )]

t  z1

For example: f(z) =

Initial value theorem:

lim f(t) = lim F(z) = 2/2 = 1
t0 z

Final value theorem:

z
= = 1/0.5 = 2
( z  0. 5 )

© Bytronic Ltd Page 52 PCT-M1

Stability of Sampled Data Control Systems
A common method used to determine the stability of a digital system is to map the variable z to the W plane.
This change of variable is known as the bi-linear transformation, mathematically it is described thus:

1+ W
P( z ) =
1 - W

This method is best explained in the form of an example. Consider the effect of the mapping upon the pulse
transfer function:

By multiplying the numerator and denominator by (1 - W)2 we can simplify this to:

We now apply the Routh test to the characteristic equation to give the array:

1 2

1 +11 +1

2 +20 0

3 +1

Since there are no sign changes in column 1 the system is stable. If a sign change occurs in this column the
system would be unstable. For stability, all of the poles of C(z)/R(z) (i.e. the roots of its denominator) must fall
within the unit circle around the origin of the z plane.

© Bytronic Ltd Page 53 PCT-M1

Inverse Z Transformations
Once the response of a digital system to a given input has been evaluated using z transform techniques, we will
generally wish to revert back to the time domain to determine the actual output. The inverse z transformation
gives the discrete time function f(n), but not the continuous time function f(t). This is because a digital system
has discrete output steps.

Mathematically: f(n) = Z-1{F(z)}

Where: The capital Z-1 and braces indicate the inverse z transformation operator. (As
opposed to the symbol z-1, (z transform final value theorem), which means 1/z).

Common methods of determining the inverse transform are the use of inverse z transform tables, long division or
partial fractions. The simplest method is the use of tables but since these are limited, the other methods are often
required. Consider the z transform:

z2 z2
F(z) = = 2
( z 1)( z  0.15 ) z  1.15 z  0.15

The use of inverse z transform tables:

A table of inverse z transforms will include the following entry:

z2 an 1  bn 1
Z-1{ } =
( z  a)( z  b ) ab

Where: n is the sample index of the discrete time function, a sequence of integers (0 to ∞).

In our example a = 1 and b = 0.15, therefore:

1n 1  0.15 n 1
f(n) = = 1.1765 – (1.1765 x 0.15n+1)
1  0.15

Clearly therefore f(n) has a steady state value of 1.1765. The actual values of f(n) at each discrete time interval
may be calculated by substituting values for n in the expression above.

n = 0: f(0) = 1.1765 – (1.1765 x 0.15) = 1.0

n = 1: f(1) = 1.1765 – (1.1765 x 0.152) = 1.15
n = 2: f(2) = 1.1765 – (1.1765 x 0.153) = 1.1725
n = 3: f(3) = 1.1765 – (1.1765 x 0.154) = 1.1759
n = 4: f(4) = 1.1765 – (1.1765 x 0.155) = 1.1764

Long Division:
By definition, the z transform of a discrete time sequence f(n) may be written as:

F(z) = f(0) + f(1)z-1 + f(2)z-2 + f(3)z-3 + …

Therefore if F(z) can be manipulated into this form, the values of f(n) may be determined for each value of n by
inspection. Long division may be used to manipulate F(z) into the required form.

Considering our original example once again:

z2
F(z) =
z  1.15 z  0.15
2

Multiply the numerator and denominator by z-2 to give:

© Bytronic Ltd Page 54 PCT-M1

 1
F(z) =
1  1.15 z  0 .15 z  2
1

Apply long division to simplify this:

Therefore: F(z) = 1 + 1.15z-1 + 1.1725z-2 + 1.1759z-3 + …

Therefore from inspection of F(z), we can say f(0) = 1; f(1) = 1.15; f(2) = 1.1725; f(3) = 1.1759.

Partial Fractions:
Considering our original example once again:

z2
F(z) =
( z 1)( z  0.15 )

Use of partial fractions can yield terms that may be inversely transformed using the tables. To ensure the result
is in the form of the tables it is necessary to calculate F(z)/z.

F ( z) z A B
= = + (1)
z ( z  1)( z  0.15 ) ( z  1) ( z  0.15 )

Therefore: z = A(z - 0.15) + B(z - 1)

z = Az - 0.15A + Bz - B

Equating coefficients: z0: 0 = -0.15A - B (2)

Z1: 1 = A+B (3)

Therefore: A = 1.176 and B = -0.176

Substitute into (1):

F ( z) 1.176 0 .176
= –
z ( z  1) ( z  0.15 )

1.176 z 0 .176 z 1.176 0.176

Therefore: F(z) = – =  –
( z  1) ( z  0.15 ) 1 z 1
1  0.15 z 1

From inverse z transform tables:

f(n) = 1.176 – 0.176(0.15)n

Substituting specific values for n: f(0) = 1; f(1) = 1.15; f(2) = 1.172; f(3) = 1.175; f(4) = 1.176.

© Bytronic Ltd Page 55 PCT-M1

Digital Three Term Controller
The symbol en is used in this section to indicate the error at the nth. sample interval, the total controller output in
that interval would be denoted by Mn. The three elements of the digital Proportional, Integral and Derivative
controller are similar to those discussed but they are calculated in a discrete interval of time:

Proportional Component
The equations used for the output of the proportional component of the controller are the same as those used for
continuous systems:

Since digital systems are usually analysed in the z plane, these become:

Integral Component
There are several ways of defining this since are several numerical integration techniques available. However, if
the sampling time is small the simplest method is shown in the following diagram which shows a step input at
time equals zero seconds.

Derivative Component
The derivative component is a constant (PG x DAT) multiplied by the rate of change of the error. This rate of
change may be approximated by using the values of the error at the nth. and (n - 1)th. sampling instants, as shown
below:

© Bytronic Ltd Page 56 PCT-M1

 en  en 1
The approximate slope at the nth. sample interval is =
T

Therefore the transfer function of the derivative controller component is:

The Effects of Sampling Time

When a digital control algorithm is designed a suitable sampling interval must first be chosen. However, as the
sampling interval is increased some potentially degrading effects become significant.

Destabilising effect
Dead-time has a dramatic and destabilising effect on a closed loop system due to the phase shift that it can
introduce.

Information loss
If the sampling frequency is too low (i.e. sample interval too long), vital high frequency information may be lost
as shown in figure below, It is clear from the top section of the diagram that higher frequency information
(shown circled) is missed by the computer because it occurs in between consecutive sampling events. The lower
section shows a possible reconstruction of the sampled signal. It is quite accurate where the sampled signal
varied slowly by hopelessly inaccurate where it varied at a higher frequency.

Information Loss - Sampling Frequency Too Low

Aliasing
Closely related to this type of information loss is the risk of aliasing when the sampling frequency is too low.
The sampling frequency must be at least twice the highest frequency present in the sampled signal to prevent
aliasing. (This principle is known as Shannon’s sampling theorem). When the sampling rate complies with this
principle, we can assume that the control algorithm will respond to the signal being sampled. However, if the
sampling rate is less than twice the highest frequency of the measured value, the controller will respond to a
lower frequency ‘alias’ of that signal. (In essence the computer will ‘see’ a lower frequency pseudo-signal rather
than a sampled version of the true signal). Even in the case where the sampling rate is exactly twice the
frequency of the measured value it can be seen that square or triangular waves would give exactly the same
samples. A worst case example would be where the measured value of a sinusoidally varying process variable is
sampled at exactly the same frequency. The controller would effectively interpret the data as a constant
(unchanging) signal because the samples would occur at corresponding points on subsequent wavelengths,

© Bytronic Ltd Page 57 PCT-M1

 Aliasing - Sampling Frequency Too Low

Use of Simulation
Simulation can be an important element in the design process of control systems. It is also invaluable as a
teaching aid for demonstrating ideas without confusing the student with a large amount of mathematics early on
in a course.

Open Loop Control

Open loop control allows a user-determined function, either a step, ramp or sine wave etc. to be applied to the
pump. The system’s open loop response is then drawn on the screen. A block diagram of the control loop is
shown below:

© Bytronic Ltd Page 58 PCT-M1

Exercise 1.
Proportional Control
Use the PCT-M software to implement ‘proportional only control’ of the flow loop with the following pairs of
SP and PG values. Record the eventual ‘steady state’ flow rate values in litres/minute in the table below, once
the initial oscillations have decayed.

SP (l/min) PG Steady State SP (l/min) PG Steady State

2.0 0.5 0.6 1.0
2.0 1.0 1.0 1.0
2.0 1.5 1.4 1.0
2.0 2.0 1.8 1.0
2.0 4.0 2.2 1.0
2.0 6.0 2.6 1.0
2.0 8.0 3.0 1.0
2.0 10.0 3.4 1.0

What conclusions about the nature of ‘proportional only control’ may be drawn from your observations? You
should be able to see that for a given SP the final ‘steady state’ value increases as PG is increased. However
there is no value of PG for which the steady state value is exactly equal to the SP because with proportional
control there must always be some error in order for there to be a controller output. As PG is increased there is
also an increase in the magnitude and duration of the initial oscillations and if PG is too high many systems will
oscillate continuously and never settle to a steady state.

For a proportional only controller with a given PG value although the set point is never reached, the resulting
steady state value increases as the set point is increased. This shows that it is possible to use simpler proportional
only control by setting SP a suitable amount higher than the truly desired value, effectively ‘deceiving’ the
controller. This strategy might be acceptable in a situation where the set point is not going to change at all or
where external disturbances are minimal but if either of these conditions are not true then a more sophisticated
controller is generally required. To prove this run an experiment with PG = 1 and SP = 3 which means that we
should actually obtain a steady state value of 1.8 litre/minute and if this is the truly desired value, and there are
no disturbances, then all will be fine.

© Bytronic Ltd Page 59 PCT-M1

Exercise 2.
Proportional and Integral Control
Start a flow loop experiment using only proportional control initially and then adding in an element of integral
action after about ten seconds. Repeat this process several times and each time increase the amount of integral
action according to the following table. In order to turn off the integral action completely you can either remove
the tick from the box to the left of the I term or you can leave this in and set the I term to 999. (Note that the
amount of integral action is inversely proportional to the integral action time as it is specified in the software and
on the table below). Record the final steady state flow value (if the flow actually does settle) and your main
observations as to the nature of the response.

SP (l/min) PG I Steady State Observations

2.0 1.0 999
2.0 1.0 100
2.0 1.0 50
2.0 1.0 10
2.0 1.0 3
2.0 1.0 1
2.0 1.0 0.5
2.0 1.0 0.2
2.0 1.0 0.1

What conclusions about the effects of integral action upon the nature of a ‘PI controller’ may be drawn from
your observations? As in labwork 1, a controller with no integral action (I = 999) is characterised by a response
which exhibits a constant offset between the SP and the steady state of the process variable. This is referred to as
the ‘proportional offset’. When a relatively small amount of integral action is added (I = 100, 50, 10) the flow
rate increases gently and produces a trace which is similar to a capacitor charging curve. Clearly the integral
action takes account of the recent history of the error whereas the proportional action only reacts to the current
value of the error. PI control can be very effective if the terms are chosen appropriately. Increasing the integral
action above that which is suitable for a given system can lead to instability and the possibility of gross
oscillations in the value of the process variable.

With the integral action time set to around 0.8 seconds the response is ‘lightly damped’ which means that whilst
the control loop is not unstable, (i.e. continuous large amplitude and/or growing oscillations), the controller is
not optimally ‘tuned’ for critical or near critical damping.

The Overshoot is the maximum amount by which the response exceeds the final steady state value of the
process variable. It is sometimes expressed as a percentage of the final steady state value.

The Rise Time is the time taken for the response to increase from 10% of its final steady state value to 90% of
its final steady state value.

The Settling Time is the time taken for the response to reach its final steady state value, within some specified
tolerance. The diagram above shows the settling time for a 5% tolerance.

The Periodic Time or Period is the duration of one complete cycle of oscillation. It can therefore be measured
as the interval between alternate crossings of the final steady state value or the interval between successive peaks
or successive troughs on the response curve.

The Frequency is the reciprocal of the period, i.e. the number of cycles per second which is expressed in Hertz
(Hz). Sometimes the frequency is expressed in radians per second and the relationship between the two units is
that radians per second equals 2 times the frequency in Hertz.

The Transport Delay is the period during which there is no change in the process variable after a step change
has been made to the set point.

© Bytronic Ltd Page 60 PCT-M1

Exercise 3.
Saturation and Integral Windup
Run a flow loop experiment using PI control with P = 1 and I = 3; and if you make changes to SP, both up and
down, you will see that the response is rapid.

Why does this effect occur? This phenomenon, which is called ‘integral windup’ (or ‘reset windup’), is due to
the fact that the integral term within the PI controller generates a component of the control output which is based
upon the recent history of the error.

In more elaborate industrial controllers there is often a feature called ‘anti reset windup’ which can be used to
eliminate this problem so that after a saturation episode the process variable will be returned to the SP as soon as
the physics of the hardware allows.

In industry, saturation might occur because an actuator, is not powerful enough or the pipes are too small or
accumulated detritus has reduced the effective diameter of the pipe at some point in the circuit.

Exercise 4.
Three Term or PID Control
In this labwork you will investigate the effect of derivative action. Run a flow experiment with SP = 1, PG = 1
and the I and D terms turned off. After a few seconds set I = 0.35. The result will be similar to that seen in
labwork 2, a proportional offset whilst there is no integral term followed by permanent oscillations once the
integral term is added. Clearly too much integral action was added to the controller!

To eliminate the oscillation we could simply reduce the integral action by increasing the I term to 1 but it would
be useful if we could retain the rapid response which the higher integral action confers without pushing the
system into unstable oscillations. While this is not always possible with a PI controller, a PID controller is
usually capable of eliminating instability and providing a fast response.

Run another experiment with the same initial settings and after a few seconds add the same integral term of 0.35
and a derivative term of 1 second. After the initial sudden rise the flow trace does oscillate about six or seven
times but there is a smooth decay of the amplitude and it soon settles at the SP. A little more derivative action
will improve the response, try D = 1.9 second to prove this. The following diagram summarises the phenomena
which you should have observed.

Beware that despite the benefits, derivative action may give rise to detrimental effects in some situations,
particularly if there is a significant amount of high frequency ‘noise’ in the measured, and hence error, value.
Derivative action can dramatically amplify this noise and degrade the performance of the controller.

© Bytronic Ltd Page 61 PCT-M1

Exercise 5.
Ziegler / Nichols Tuning
How do we select appropriate proportional, integral and derivative values for any given process? Simply trying
numbers in a PID controller might be acceptable for very small systems but if you are trying to establish a
control algorithm for a full sized industrial process then this is not a very good strategy! This method could be
time consuming which might bring about financial losses due to the plant downtime. This approach might also
risk causing damage to actuators or sensors - if it was to drive them beyond their intended range of operating
values. A more scientific approach to finding a reasonable set of PID terms is required.

Ziegler/Nichols tuning is a popular semi-empirical method of obtaining approximate PID values which can be
applied successfully to many different types of processes. The method provides a reasonable starting point and
the experienced control engineer might wish to adjust the calculated PID terms slightly to improve the nature of
the response. There are two techniques the first called the ‘continuous cycling method’ assumes that the closed
loop system can be made to oscillate permanently with a proportional only controller, as illustrated by the
following graph.

The continuous cycling method requires that the gain of a proportional only controller is increased a small
amount at a time until the onset of permanent oscillations occurs. At this point the value of the gain (kp) together
with the period of the resultant oscillation (T) are noted. The recommended two and three term controllers are
then given by:

PI: PG = 0.45kp IAT = 0.83T

PID: PG = 0.6kp IAT = 0.5T DAT = 0.125T

Run some flow experiments similar to those of labwork 1 with SP = 2. Increase the PG value from 1.0 to about
3.5, initially in increments of 0.5 or so. Try to estimate the lowest PG value which produces permanent
oscillations in the flow rate. (It may be necessary to adjust the PG value by 0.1 or 0.2 when you get close to this
‘ultimate proportional gain’). Record the PG value (kp) together with the period of the oscillation (T) and use
them in the expressions quoted above for PI and PID algorithms. Apply these control algorithms to the flow
simulation and note your observations in the table below.

Algorithm Observations
PI
PID

Many systems cannot be made unstable (i.e. caused to oscillate) by means of a proportional only controller and
for some industrial systems it might be undesirable to do this because of the risk of damage referred to above.

© Bytronic Ltd Page 62 PCT-M1

Glossary of Terms
Actuator Device by means of which control output effects process variable, e.g. pump, valve.
Algorithm See control algorithm.
Amplitude For a sinusoidal waveform, half the peak-to-peak value.
Anti reset windup Available on industrial controllers that eliminates reset windup (integral windup).
Antiphase Situation in which the maxima of an input signal occur simultaneously with the minima of the
output signal and vice versa
Automatic control Regulation of a process variable by means of an algorithm implemented by a controller that is
usually computer or PLC.
Batch volume control Control of supply of a specific volume of liquid at a specific temperature over a specific period
of time.
Block diagram Visual method of describing a control loop based upon process, controller and feedback blocks
together with signal paths and (a) summing junction(s).
Bode plots Graphs showing system response in terms of gain and phase angle plotted against frequency.
Closed loop Any control loop in which the error value is used by the controller algorithm to determine the
control output.
Continuous cycling method Ziegler/Nichols tuning method that requires that a closed control loop can be made unstable by
a proportional only controller.
Control algorithm Rule or set of rules by which controller generates control output(s), based upon the error signal
in a closed loop system.
Control cycle Phrase sometimes used to refer to execution of a control experiment.
Control loop System comprising process, sensor(s), error feedback and source of control output signal. If
error feedback is used to calculate control output then it is a closed loop, alternatively it is an
open loop.
Control output Signal generated by controller. Calculated from the error according to control algorithm in a
closed loop.
Controller Device or software routine which applies a control algorithm to determine a control output
from the error signal in a closed loop.
Critical damping Damping such that measured value reaches set point as quickly as possible without any
overshoot.
D See derivative, derivative action, derivative term.
Damping Force(s) that oppose the tendency of a system or control loop to oscillate, e.g. friction, air
resistance or the effect of controller.
Decay Successive reduction in amplitude of control loop oscillations due to damping force(s).
Derivative, Third component in a PID or three-term control algorithm that takes account of the time rate of
Derivative action change of error. Derivative action is based upon the time rate of change of error multiplied by
Derivative term a constant called the derivative action time D.
Error Function of time defined as the difference between set point (SP) and measured value (MV).
Feedback Technique of using error signal to determine control output signal that causes error to be
minimised in a well tuned system.
Flat Description of set point that remains at a certain fixed value until changed by human
intervention.
Free integrator A concept from Laplace transform/transfer function theory that refers to the way certain
devices integrate an input with respect to time thus obviating the requirement for integral
action within any associated controller. e.g. a tank integrates net inward fluid flow to produce
fluid level.
Frequency The rate of repetition of any periodic waveform, i.e. the number of cycles of that waveform
that occur in a unit of time. Usually quoted in cycles per second, hertz (Hz) or radians per
second.
Frequency response Overview of the way a system responds to input signals across the complete range of expected
frequencies. Input signals include set point, noise and external perturbations. Often displayed
as Bode graphs, i.e. gain and phase angle plotted against frequency.
Gain The ratio of the amplitude of the output signal to the amplitude of the input signal, i.e. output
amplitude divided by input amplitude.
I See integral, integral action, integral term.
In phase Situation in which maxima of input and output signals occur simultaneously and minima of
input and output signals occur simultaneously.
Instability Characteristic of control loop that continues to oscillate with constant or increasing amplitude.
Integral, Second component in a PI, PID or three-term control algorithm that takes account of the recent
Integral action, history of the error. Integral action is calculated by dividing the net area under the error curve
Integral term against time by the integral action time I.
Integral windup See reset windup
Laplace transform Result of a mathematical operation that converts functions of time e.g. input and output
signals, to functions of the Laplace variable that is universally assigned the letter ‘s’. This

© Bytronic Ltd Page 63 PCT-M1

 allows more convenient mathematical manipulation for advanced work on control theory.
Level control Control of level of a fluid in a particular vessel.
Lightly damped Description of a system where the damping forces are less than they would be for critical
damping. Lightly damped systems all oscillate and the lighter the damping the more cycles of
oscillation which are produced.
Linearity Characteristic of a system that exhibits increases in its output that is proportional to increases
in its input. e.g. sinusoidal input and output amplitudes in an open loop flow cycle.
Manual control Situation in which a human operator determines the control output(s) based upon the required
value and current measured value of the process variable
Measured value Function of time returned by the sensor and signal conditioning which represents the value of
the process variable.
MV See measured value
Noise Normally random non-periodic signal superimposed upon one or more of the signals in a
control loop, having an amplitude which is small in comparison to the magnitude of the signals
of interest.
Non linearity Characteristic of a system that does not exhibit increases in its output that is proportional to
increases in its input. e.g. flow of current through a diode as voltage is increased from
-ve. to +ve.
Normalise Process of dividing all elements in a set of data by a chosen value, (sometimes the value of one
particular element, perhaps the largest) in order to render the whole set more convenient for
assimilation or for comparison with other data sets.
Off-line Beyond the particular time period pertaining to real-time control of a system or process.
Open loop control Any control loop in which the error value is not used by the controller to determine the control
output.
Open loop response A description of the way a system responds to certain input signals such as sine waves, step
changes or square waves etc. These input signals may originate from manual/human
intervention or from an open loop control output. Obtaining an open loop response to a series
of well-chosen sine waves may be used to determine the frequency response in the form of
Bode graphs.
Optimal tuning Tuning of the PID controller to produce the best possible results for a given control system.
Oscillation, Periodic pattern, almost always sinusoidal, observed in the response of a system.
Oscillatory
Overshoot Maximum amount by which measured value exceeds final steady state value of process
variable. Often expressed as percentage of final steady state value.
PB See proportional band.
PCT trend update utility Software utility for converting old DOS based PCT trend files to PCT-M compatible format.
Period, Duration of one complete cycle of oscillation i.e. interval between alternate crossings of the
Periodic time final steady state value.
PG See proportional gain.
Phase angle In the context of sinusoidal input and output signals, the amount by which the input leads the
output expressed in degrees or radians.
PID control Automatic control based upon an algorithm that includes proportional, integral and derivative
contributions to the final control output.
Process reaction curve method Ziegler/Nichols tuning method that requires that data from an open loop response to a step
change be acquired.
Proportional, First component in a PI, PID or three-term control algorithm that generates a component of the
Proportional term control output which is directly proportional to the magnitude of the error.
Proportional band The error band where the control output is between 0% and 100%. Proportional band equals
100 divided by the proportional gain.
Proportional gain A constant multiplied by the error value to give a control output.
Proportional offset Steady state discrepancy between set point and measured value that is characteristic of all
systems which use proportional only control and which do not have an intrinsic ‘free
integrator’.
Ramp Description of set point which changes up or down gradually at a fixed rate.
Rate of flow Velocity at which fluid moves along an open or closed channel, usually expressed in litres per
minute.
Reset windup Accumulation of controller integral action ‘debt’ after a saturation episode. This offsets
measured value from set point for longer than the physics of the process would strictly require.
Response, Shape of the trace produced by plotting process variable measured value against time.
Response curve
Rise time Time taken for the response to increase from 10% of its final steady state value to 90% of its
final steady state value.
Sample time Period between successive samples of measured value.
Sampling frequency Rate at which samples of measured value.
Saturation Description for the situation when, even with 100% control output, the measured value cannot

© Bytronic Ltd Page 64 PCT-M1

 be brought to the set point.
Sawtooth Description of set point which changes up and/or down periodically, gradually and at a fixed
rate.
Set point Desired value of the process variable.
Settling time Time taken for the process variable measured value to reach its final steady state value within
some specified tolerance.
SP See set point.
Stability Characteristic of control loop in which all oscillations ultimately decay completely after a step
change.
Steady state The condition of a control loop once all transients have decayed completely, after a step
change. The steady state might be a fixed measured value or an oscillation with a fixed
amplitude.
Step, Step change Description of set point which changes up and/or down instantaneously.
Summing junction Point in a control loop block diagram symbolised by a crossed circle where the measured value
feedback is subtracted from the set point to provide an error signal input for the controller.
System frequency response See frequency response.
Three term control Automatic control by means of a PID control algorithm.
Time constant Parameter that characterises the speed of response of a control loop or an element within a
control loop.
Trace Graph of process variable measured value, plotted against time.
Transfer function Mathematical model (or description) of control loop block (e.g. process or controller).
Specifically the ratio of Laplace transform of output signal to Laplace transform of test input
signal.
Transport delay Period during which there is no change in the process variable after a step change has been
made to the set point (in closed loop) or the open loop control output.
Trend Graph of process variable measured value plotted against time in. This term is used to refer to
complete graph.
Tuning Selection of PID terms to be used within control algorithm.
Waveform The shape of one cycle of a periodic signal.
Wavelength The length of one cycle of a periodic signal.
Ziegler/Nichols tuning A pair of techniques for estimating appropriate PID values for a given control loop. See
continuous cycling method and process reaction curve method.

© Bytronic Ltd Page 65 PCT-M1

© Bytronic Ltd Page 66 PCT-M1
 BYTRONIC LIMITED
124 Anglesey Court
Towers Business Park
Rugeley. WS15 1UL
Staffordshire
United Kingdom
Tel : +44 (0)8456 123155
Fax : +44 (0)8456 123156
Email : [email protected]
Website : http://www.bytronic.net