D EPARTMENT

OF

E LECTRICAL AND E LECTRONIC E NGINEERING

EE208: Part II Control Systems

C. P. Jobling 15 February 1997 Version 0.99 A printable version of these lecture notes are available for downloading in PDF and Zipped PostScript formats.

D EPARTMENT OF E LECTRICAL AND E LECTRONIC E NGINEERING

EE208
Slide 1

Part II Control Systems

CONTENTS

Contents
1 Introduction to Control Systems 1.1 Control Systems Applications . . . . . . . 1.1.1 Brief History . . . . . . . . . . . . 1.1.2 The Challenges of Control Systems 1.1.3 An Example The Space Shuttle . 1.1.4 The Benets of Studying Control . 1.2 Describing Control Systems . . . . . . . . 1.2.1 The Control System . . . . . . . . 1.2.2 Description of Input and Output . . 1.2.3 Advantages of Control Systems . . 1.2.4 Open-Loop Systems . . . . . . . . 1.2.5 Closed-Loop Systems . . . . . . . 1.2.6 Computer-Controlled Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 9 9 9 9 18 18 18 18 21 22 24 27 29 29 29 30 30 30 30 36 36 36 39 39 39 42 42 42 51 51 52 52 56 57 59 62 67 68 70 70 70 71 73

Introduction to the Analysis and Design Process 2.1 Control Systems Analysis and Design Objectives . . . . . . 2.1.1 Transient Response . . . . . . . . . . . . . . . . . . 2.1.2 Steady-state accuracy . . . . . . . . . . . . . . . . . 2.1.3 Stability . . . . . . . . . . . . . . . . . . . . . . . . 2.2 An Example of a Closed-Loop System . . . . . . . . . . . . 2.2.1 Antenna Azimuth Position Control . . . . . . . . . . 2.2.2 Block Schematic Diagram . . . . . . . . . . . . . . 2.2.3 Transient Performance . . . . . . . . . . . . . . . . 2.2.4 Steady-state error . . . . . . . . . . . . . . . . . . . 2.3 The Design and Analysis Sequence . . . . . . . . . . . . . . 2.3.1 Determine a Physical System from the Requirements 2.3.2 Transform the Physical System into a Schematic . . 2.3.3 Mathematical Models for the Schematic . . . . . . . 2.3.4 Block Diagram Reduction . . . . . . . . . . . . . . 2.3.5 Analysis and Design . . . . . . . . . . . . . . . . . Modelling the Azimuth Position Control System 3.1 The Schematic for the Plant . . . . . . . . . . . 3.2 Mechanical Side . . . . . . . . . . . . . . . . . 3.3 Effect of the Gearbox . . . . . . . . . . . . . . 3.3.1 An Ideally Matched Gearbox . . . . . . 3.4 Electrical Side . . . . . . . . . . . . . . . . . . 3.4.1 Armature-Voltage Controlled DC Motor 3.5 Coupling Constants . . . . . . . . . . . . . . . 3.6 The Rest of the Control System . . . . . . . . . 3.6.1 Position sensors . . . . . . . . . . . . . 3.6.2 Velocity sensors . . . . . . . . . . . . 3.6.3 Pre-amplier . . . . . . . . . . . . . . 3.6.4 Power Amplier . . . . . . . . . . . . 3.6.5 Block Diagram of Plant . . . . . . . . 3.7 Velocity control system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CONTENTS
4 Evaluation of System Response 4.1 Poles and Zeros and System Response . . . . . . . . . . . . . . . 4.2 First-Order System Responses and Specications . . . . . . . . . 4.2.1 The signicance of a . . . . . . . . . . . . . . . . . . . . 4.2.2 Time Constant . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Rise Time Tr . . . . . . . . . . . . . . . . . . . . . . . . 4.2.4 Settling Time Ts . . . . . . . . . . . . . . . . . . . . . . 4.3 Types of Second-Order System Responses . . . . . . . . . . . . . 4.3.1 Overdamped Response . . . . . . . . . . . . . . . . . . . 4.3.2 Underdamped Response . . . . . . . . . . . . . . . . . . 4.3.3 Undamped Response . . . . . . . . . . . . . . . . . . . . 4.3.4 Critically Damped System . . . . . . . . . . . . . . . . . 4.4 The General Second-Order Response . . . . . . . . . . . . . . . . 4.4.1 Denitions . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Derivation of Formulae . . . . . . . . . . . . . . . . . . . 4.4.3 Pole-zero locations . . . . . . . . . . . . . . . . . . . . . 4.4.4 Further analysis for underdamped second-order systems . 4.5 The Specication of Second-Order Response . . . . . . . . . . . 4.5.1 Evaluation of Tp . . . . . . . . . . . . . . . . . . . . . . 4.5.2 Evaluation of OS . . . . . . . . . . . . . . . . . . . . . 4.5.3 Evaluation of Ts . . . . . . . . . . . . . . . . . . . . . . 4.6 Relating Response Specications to Pole Locations in the s-Plane 4.6.1 Effect of moving poles along design curves . . . . . . . . Analysis and Design of Feedback Systems 5.1 Interpretation of the generalised closed-loop transfer function 5.2 Unity-gain feedback . . . . . . . . . . . . . . . . . . . . . . 5.3 Closed-loop transient performance . . . . . . . . . . . . . . 5.4 The Root-Locus: A Preview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 76 77 84 84 85 85 86 89 89 90 93 93 98 98 99 100 102 105 106 107 108 111 112 122 123 123 124 128 133 133 134 134 136 138 138 138 141 141 144 145 146 147 148 150 151

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Stability 6.1 What is stability? . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 How do we determine if a system is stable? . . . . . . . . . . . 6.2.1 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.2 Instability . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.3 Marginal stability . . . . . . . . . . . . . . . . . . . . . 6.3 Testing for Stability . . . . . . . . . . . . . . . . . . . . . . . . 6.4 The Hurwitz Criterion . . . . . . . . . . . . . . . . . . . . . . . 6.5 The Routh-Hurwitz Stability Criterion . . . . . . . . . . . . . . 6.5.1 The Routh array . . . . . . . . . . . . . . . . . . . . . 6.5.2 The Routh-Hurwitz Test . . . . . . . . . . . . . . . . . 6.6 Special Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.6.1 A Zero in the rst column . . . . . . . . . . . . . . . . 6.6.2 An Entire row is zero . . . . . . . . . . . . . . . . . . . 6.6.3 Interpretation of a row of zeros . . . . . . . . . . . . . . 6.7 Use of the Routh-Hurwitz Criterion for Control Systems Design 6.8 Relative Stability . . . . . . . . . . . . . . . . . . . . . . . . .

CONTENTS
7 Steady-state errors 7.1 Forms of Steady-State Errors . . . . . . . . . . . . . . 7.2 Block Diagram Perspective . . . . . . . . . . . . . . . 7.3 Steady-State Errors for Unity-Gain Feedback Systems . 7.3.1 Step input . . . . . . . . . . . . . . . . . . . . 7.3.2 Ramp input . . . . . . . . . . . . . . . . . . . 7.3.3 Parabolic input . . . . . . . . . . . . . . . . . 7.4 Denition of Static Error Constants and System Type . 7.4.1 Static error constants . . . . . . . . . . . . . . 7.4.2 System type number . . . . . . . . . . . . . . 7.5 Steady-State Error Specications . . . . . . . . . . . . Dynamic Compensation 8.1 The trouble with gain compensation . 8.2 Velocity Feedback Compensation . . . 8.3 Other forms of dynamic compensation 8.4 PID Cascade Compensators . . . . . . 8.5 Design of cascade compensators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4 154 156 159 162 163 163 164 165 165 165 169 172 172 175 177 178 180 186 186 187 189 197 201 206 206 206 207 209 210 215 215 218 220 220 222 226 226 228 229 232 232 236 236 237 239

. . . . . . . . . . . . . . .

. . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

The Root Locus 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 9.1.1 The Control System Problem . . . . . . . . . . . . 9.1.2 Complex numbers and their vector representations 9.1.3 Dening the Root Locus . . . . . . . . . . . . . . 9.2 Properties of the Root Locus . . . . . . . . . . . . . . . . 9.3 Rules for sketching the root-locus . . . . . . . . . . . . . 9.3.1 Number of branches . . . . . . . . . . . . . . . . 9.3.2 Symmetry . . . . . . . . . . . . . . . . . . . . . . 9.3.3 Real-axis segments . . . . . . . . . . . . . . . . . 9.3.4 Start and end-points . . . . . . . . . . . . . . . . 9.3.5 Asymptotic behaviour . . . . . . . . . . . . . . . 9.4 Rening the root locus diagram . . . . . . . . . . . . . . . 9.4.1 Real-axis break-away and break-in points . . . . . 9.4.2 Calculation of j! -Axis crossing . . . . . . . . . . 9.4.3 Angles of Departure and Arrival . . . . . . . . . . 9.5 Plotting and calibrating the root-locus . . . . . . . . . . . 9.5.1 Transient response from the root-locus . . . . . . .

10 Frequency Response Techniques 10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.2 The Open-Loop Frequency Response . . . . . . . . . . . . . . . 10.2.1 An example . . . . . . . . . . . . . . . . . . . . . . . . . 10.2.2 Effect of Gain . . . . . . . . . . . . . . . . . . . . . . . . 10.3 The Simplied Nyquist Criterion . . . . . . . . . . . . . . . . . . 10.4 Gain and Phase Margins . . . . . . . . . . . . . . . . . . . . . . 10.4.1 Gain margin . . . . . . . . . . . . . . . . . . . . . . . . 10.4.2 Phase margin . . . . . . . . . . . . . . . . . . . . . . . . 10.4.3 Design considerations of the use of gain and phase margin

CONTENTS
A Solutions to Problems B Second-Order Responses B.1 Overdamped system response . . . B.2 Underdamped response . . . . . . B.3 Undamped response . . . . . . . . B.4 Critically damped system response B.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5 246 272 272 274 276 277 278

CONTENTS

Syllabus
EE208: Control Systems is concerned with the development of techniques for modeling linear dynamic systems with feedback and analysing their performance and stability. The topics to be included are as follows: Introduction: Control system applications; a brief history; an example. Describing control systems: inputs and outputs; advantages; open- and closedloop control systems; computer controlled systems. Analysis of design objectives for control systems: transient response; steadystate response; stability. An antenna dish azimuth position control system (course case-study1 ). The design and analysis sequence. Mathematical modeling. Time response analysis and design. Evaluation of system response: poles and zeros and system response; the characteristic equation; rst-order system responses amd specications; Second order system responses: general second-order system; natural frequency; damping ratio; Transient performance specications for second-order systems. Analysis and design of feedback systems: closed-loop transient performance; introduction to the root-locus. drill problem. Stability: the Hurwitz criterion; the Routh-Hurwitz criterion; Steady-state errors: steady-state errors for unity-gain feedback systems; static error constant and error type; specications of steady-state errors; A taste of feedback compensation: velocity feedback. Second drill problem. Root locus: the control system problem; complex numbers and their vector representation; dening the root locus; properties of the root locus; rules for sketching the root locus. Third drill problem. Frequency response analysis and design. Nyquist: stability criterion; simplied Nyquist stability criterion. Relative stability: gain and phase margin; assessing closed-loop performance from bode diagrams. Closed-loop frequency response: M - and N - circles; Hall and Nichols charts; Mmax . Final drill problem.
1 http://www-ee.swan.ac.uk/Courses/level2/ee208/drillp/

CONTENTS

Supporting studies
The course is supported by two experiments in the Part II Laboratory Course. The rst is an examination of a position control system, a scaled-down version of the course case-study2 . The second makes use of M ATLAB and S IMULINK to simulate a model of the same system. I strongly encourage you to make use of M ATLAB and S IMULINK (which are installed on the PC network) to help you to model, simulate and understand the dynamics of control systems. There are manuals in the technicians ofce at the back of room 502 but there are now several books such as the ones by Saadat [21], Shahian and Hassul [22], Leornard and Levine [16] and Bishop [2] which provide good tutorial introductions to the use of M ATLAB for control systems design. In addition, many standard textbooks have been recently updated to include tutorial and reference material concerned with the use of M ATLAB in control systems analysis and design. For those of you who own your own PC, the Student Editions of M ATLAB [23] and S IMULINK [24] are available and should be sufcient to support the analysis methods covered in the course. As an alternative to M ATLAB, Golten [11] is a basic text on control systems analysis and design which makes extensive use of the computer program called CODAS included in the price. There is also an electronic handbook for Mathcad which is part of a new venture by McGraw & Hill and Mathsoft to supplement selected titles in the the Schaums Outline Series. The title of this package, which includes the book [13], the software and a run-time version of Mathcad is Interactive Feedback and Control Systems [14].

Reading List
The reference section of these notes is an annotated booklist for the control course. The course notes are based on Nise [18] which is therefore the recommended text. Dorf [8], which was the main text in the past, is still worth considering as an alternative to Nise particularly because it is supported by a solutions manual [7] and a M ATLAB tutorial [2]. For a cheaper alternative you might wish to consider DiStefano et al in the Schaum Series [13] (or the electronic handbook version for Mathcad [14]) which is both a good reference and supply of worked problems. Those looking for more comprehensive coverage of the course material might consider DAzzo and Houpis [5]. The other books included in the list fall into the category of alternative titles or further reading. There are many books on control that also cover much of the material that is taught in Part II and Part III, for example you might prefer one of [10, 19, 20]. Serious students of the topic might consider [15], and those with less interest might just consider [4, 17, 3] which are all cheap and cheerful. For those with an interest in the practicalities of control system design might care to look at Bateson [1] which is very much an applied control textbook/reference manual for control systems practitioners. Doeblin [6] is a text with a similar aim which also emphasises computer simulation. Additional background reading can be found in [9] which emphasizes computer modeling and simulation and [12], which is a good tutorial introduction to M AT2 http://www-ee.swan.ac.uk/Courses/level2/ee208/drillp/

CONTENTS
LAB.

Other Sources of Information

Aside from the on-line version of the course material which is to be found at http://www-ee.swan.ac.uk/Courses/level2/ee208/ee208.html this list of additional resources may prove useful. For M ATLAB, the main tool used to support this course, you will nd a wealth of material on the home-page of the MathWorks3 . Included there is an extensive list of books4 that use M ATLAB, a repository5 of contributed software, Frequently asked question6 and lots of other information. For on-line information about other software tools which could be used in the support of this course you should try the web home pages of MathSoft7 (developers of Mathcad), Wolfram Researchs MathSource8 for information on Mathematica, and MapleSoft9 for information on Maple (the symbolic processor that is inside M ATLABs Symbolic Toolbox and Mathcad).

3 http://www.matworks.com/ 4 http://www.mathworks.com/BookList.html 5 http://www.mathworks.com/ftpindex.html 6 http://www.mathworks.com/faq.html 7 http://www.mathsoft.com/ 8 http://www.wri.com/WWWDocs/mathsource/ 9 http://www.maplesoft.com/

1 INTRODUCTION TO CONTROL SYSTEMS

Lecture 1: Intro

Introduction to Control Systems

Introduction to Control Systems

Aims The purpose of this lecture is to: Introduce the Topic Slide 2 To give an idea of the many applications of the subject To give an insight into its history To highlight its advantages To demonstrate the depth and breadth of the subject To illustrate its usefulness as a subject worth studying Dene some Simple Terms

1.1 Control Systems Applications

Control systems are an integral part of modern society. There are numerous applications in industry (see Slide 3), they also exist in nature (see Slide 4) and there are even some non-physical control systems (see Slide 5). 1.1.1 Brief History Because they are found in nature, control systems can be considered to have been in use from the earliest times. However, in terms of Human endeavour, slides 10 and 11 illustrate the main milestones in the development of man-made control systems. 1.1.2 The Challenges of Control Systems

1.1.3 An Example The Space Shuttle Just taking the case of ight control: the navigation functions are illustrated in Slide 16.

1 INTRODUCTION TO CONTROL SYSTEMS

10

Some Control System Applications

Slide 3

Space shuttle Automatic machine tools Automatic parts delivery in a factory

Control Systems in Nature

Pancreas regulates blood sugar Slide 4 Adrenelin automatically generated to increase heart-rate and oxygen intake in times of ight Eyes able to follow a moving object Hand able to pick up an object and place it at a predetermined location

1 INTRODUCTION TO CONTROL SYSTEMS

11

Some Articial Applications of Control

Modern Economies A Model of Student Performance Slide 5 Input is available study time. Output is performance/exam mark. Such a model could be used to predict time required to improve the grade. With such a scheme you could decide whether it is worth spending more effort to pass the Control Systems Exam!

Control Systems Provide Power Application

Control systems allow us to move large pieces of equipment with precision Slide 6 Radio telescopes can be accurately pointed at far reaches of the universe. Lifts stop at the right oor. We could not perform these tasks ourselves. Motors provide the power and control systems regulate the position and speed.

1 INTRODUCTION TO CONTROL SYSTEMS

12

Control nds Applications in Transportation

Engine regulation, active suspension systems and anti-lock braking systems in automobiles Slide 7 Steering of missiles, planes, aircraft and ships at sea For example, modern ships use a combination of electrical, mechanical and hydraulic components to develop rudder commands in response to desired heading commands. The rudder commands, in turn, produce a rudder angle, which steers the ship.

Control nds Applications in Process Industries

In the process industries control is used to regulate level, pressure and temperature of chemical renery vessels. In a steel rolling mill, the position of the rolls is controlled according to the measured thickness of the steel coming off the nishing line.
Move

Slide 8

Rollers Strip
measure thickness No OK?

1 INTRODUCTION TO CONTROL SYSTEMS

13

Control Systems in the Home

CD Players the position of the laser spot in relation to the microscopic pits in a Compact Disc is controlled Slide 9 Video Recorders the tracking of the record and play-back heads is controlled by controlling the velocity of the tape Central heating systems use thermostats to measure and control the temperature in the room Washing machines use sequencing controls to provide a variety of wash cycles and temperature controls to avoid damage to delicate fabrics

Historical Development of Control Systems

Ancient Greece (circa 3000 BC): water clocks, automatic oil lamps; special effects in temples. Slide 10 17th Century: Cornelis Drebbel temperature control for an egg incubator 18th Century: James Watt Flyball governer for steam engine Late 19th Century to mid-20th Century development of classical control theory 1960s present modern control theory

1 INTRODUCTION TO CONTROL SYSTEMS

14

Heroes and Milestones in the Development of Control Systems

Late 19th Century: Fathers of Stability Theory J. C. Maxwell, E. J. Routh and A. M. Lyapunov Slide 11 Late 1920s mid 1930s: Bell Telephone Labs USA. Discovery of negative feedback (Black), frequency response analysis (H. W. Bode), stability theory (H. Nyquist). 1948 invention of the Root Locus method (W. R. Evans) 1960s development of state-space methods (Kalman and others)

Control Engineering is Challenging

Slide 12 It is a multi-disciplinary subject cuts across numerous engineering disciplines covers numerous functions within a discipline

1 INTRODUCTION TO CONTROL SYSTEMS

15

Control Engineering is Challenging

It Covers all Aspects of a Project from High to Low Level from conception through to system requirements; Slide 13 subsystem functions; interconnection of functions; interfaces between functions; hardware and software design; right up to test plans and procedures.

Control Engineering is Challenging

It is Broad and Diverse Control engineers typically need to work closely with Slide 14 biologists, chemical, mechanical and electrical engineers, mathematicians and physicists. They get involved with sensors and actuator technology, electronics, pneumatics and hydraulics and and computers.

1 INTRODUCTION TO CONTROL SYSTEMS

16

The Space Shuttle

The space shuttle would be impossible to y without control systems. All the shuttles many control systems are controlled by on-board computers on a time-shared basis. Slide 15 The main control systems in the shuttle are: Flight control Orbit control Life support

Flight Control in the Shuttle

Slide 16

Navigation functions take in data to estimate the shuttles position and velocity. The position and velocity data is used to steer the shuttle: In space by use of pulsed jets of gas; In the Earths atmosphere by adjusting the geometry of the shuttles air-surfaces.

1 INTRODUCTION TO CONTROL SYSTEMS

17

There are numerous subsystems in the shuttle and many engineering disciplines are needed to make it y (see slide 17)

Subsystems and Disciplines Represented in the Shuttle

Numerous subsystems Slide 17 ight elevon controls to counteract wind disturbances life support systems; power systems; heating. Many disciplines orbital mechanics; propulsion; aerodynamics; electrical engineering; mechanical engineering; hydraulics; temperature and pressure control, etc., etc.

1 INTRODUCTION TO CONTROL SYSTEMS

1.1.4 The Benets of Studying Control

18

What will I get out of this course?

Control is a top-down engineering subject. Such subjects are rare in engineering: Slide 18 most engineering courses are taught bottom up they start with components develop circuits assemble circuits into products. The reason for this is that top-down courses are difcult to teach because of the high-level of mathematics needed for a systems approach.

1.2 Describing Control Systems

In this section we shall describe what a control system is, its characteristics, and its advantages. We also present two important classications of control systems. 1.2.1 The Control System

1.2.2 Description of Input and Output The input represents a desired response. The output is the actual response. For example, the inputs and outputs of a lift control system are illustrated in Slide 23. The fourth-oor button is pressed on the ground oor. The lift-car rises to the fourth oor with a speed and oor levelling accuracy designed for passenger comfort. The fourth oor button is the input shown by a step command. The lift does not mimic the input this would be undesirable for passenger comfort as well as impossible with nite power supplied by motor. Instead, the input represents the position we would like the lift to be in when the lift has stopped moving. The lift itself follows the lift response curve. Two factors make the output different from the input.

1 INTRODUCTION TO CONTROL SYSTEMS

19

Top down design in Control Systems

design high-level system requirements Slide 19 choose functions and hardware to implement system to meet requirements. Control works from the big picture. It unies many other elements. This is part of the difculty of the subject, it is also the challenge. Recognition of the unication, that is being able to use lessons learned in other courses, will help you to master this course material.

Taking Stock
So far in this lecture we have introduced the Topic of Control. In the next part of the lecture we shall dene some of the terms used to describe control systems. In this section we will give: Slide 20 a denition of a control system a description of typical inputs and outputs for control systems highlight some advantages of control systems an account of the difference between open-loop and closed-loop control and an introduction to computer controlled systems

1 INTRODUCTION TO CONTROL SYSTEMS

20

A Control system consists of subsystems and processes (or plants) assembled for the purpose of controlling the output of the processes. A central heating boiler is a process that produces heat as a result of a ow of fuel. This process is assembled from subsystems called fuel valves. Fuel valve actuators regulate the temperature of a room by controlling the ow of fuel into the boiler. Other subsystems, such a thermostats, act as sensors, to measure the room temperature.

Slide 21

Input; Stimulus
Slide 22

Control System

Output; Response Actual Response

Desired Response

1 INTRODUCTION TO CONTROL SYSTEMS

21

Floor
desired response --- input

4
lift response --- output steady-state error

Slide 23

transient

steady-state

Time

First consider the instantaneous change in the input against the gradual change in the output. Physical entities cannot change their position or velocity instantaneously. The state changes through a path dictated by the physical devices and the way it aquires and dissipates energy. The lift undergoes a gradual change as it moves from the ground to fourth oor called the transient response. After the transient response is complete, the physical system approaches its steady-state response which is an approximation to the commanded or desired response. This occurs when the lift reaches the fourth oor. The accuracy of the lifts nal level is the second factor that makes the output different from the input. The difference is called the steady-state error. Steady-state error may also be a feature of the system being controlled and it is one of the features that the control engineer considers when specifying the desired behaviour. For example, when tracking a satellite, some error may be tolerated provided that the satellite stays close to the centre of the tracking radar beam. However, if a robot is inserting a chip into a PCB the steady-state error must be zero. 1.2.3 Advantages of Control Systems We tolerate the differences between desired response and actual response because of the many advantages of control systems. Power Amplication: Satellite dish can be positioned by a low power knob at the input but requires large power to rotate. Power gain is one good reason for building control systems. Dangerous Applications remote control of a robot arm for handling nuclear material.

1 INTRODUCTION TO CONTROL SYSTEMS

22

Advantages of Control Systems

Power Amplication Slide 24 Dangerous Applications Compensation for Human Deciencies Convenience by Change of Form of Input Compensation of Disturbances

Compensation for Human Deciencies e.g. to help handicapped people or the exo-skeleton used by Ripley in Aliens. Convenience by change of form of Input Temperature control is by the position of a dial on a thermostat, output is heat. Compensation of Disturbances Typical control variables are temperature, position and velocity, voltage, current or frequency. The control system must yield the correct output even in the presence of disturbances. Consider Slide 25. The satellite tracking antennas position and azimuth angle is controlled. Internal noise or wind gusts disturb the position. What corrects for the disturbance? disturbances must be measured measurements must be used to reposition the dish.

1.2.4 Open-Loop Systems An open-loop control system is illustrated in Slide 26. A subsystem called the controller drives the process. the input is called the reference. the output is called the controlled variable.

1 INTRODUCTION TO CONTROL SYSTEMS

23

Wind Distu rb

anc

Slide 25

Open-Loop Control System

Disturbance 1 Disturbance 2

Slide 26

Input or Reference

+ Controller + Process +

Output or Controlled Variable

1 INTRODUCTION TO CONTROL SYSTEMS

24

Description of an Open-Loop Temperature Control System

Slide 27

Process is a boiler, input is fuel, output is heat. Controller is electronics, valves, etc. that control fuel ow into furnace. Input is thermostat position.

other signals, such as disturbances are shown added to the controller or process outputs via summing junctions. Open-loop systems cannot compensate for disturbances added to the controllers driving signal (disturbance 1) such as amplier noise which is just added to the process demand. Output is also corrupted by disturbances at the output (e.g. wind in the tracking system). These cannot be corrected either. Open-loop systems do not correct for disturbances and are simply commanded. E.g. an electric toaster: time is input, output is colour but colour is not measured. 1.2.5 Closed-Loop Systems In Slide 28 the architecture of a closed-loop system is shown. Note that the input transducer is now shown explicitly. It converts the input which is the desired output into the form required by the controller. An output transducer measures the actual output or response of the plant and also converts it to the form required by the controller. The rst summing point substracts the output from the input the result is called the actuating signal and if there is any difference the controller drives the process. If there is no difference the plant is not driven since its output is already at the desired value. Closed-loop control systems are accurate because they tend to reduce the difference between input and output. They are also less sensitive to disturbances. Transient performance and steady-state errors can be controlled more conveniently and with greater exibility than with open-loop sytems often by simple

1 INTRODUCTION TO CONTROL SYSTEMS

25

Closed-Loop Control System

Disturbance 2 Disturbance 1

Slide 28
Input Input Transducer

+ +
-

+ Process +

Controller

Output

Output Transducer or Sensor

Description of Closed-Loop Temperature Control System

Slide 29

Input temperature dial position converted into a voltage by a potentiometer. Output temperature converted to a voltage by a thermistor. Differencing circuit subtracts output from input result is actuating signal controller drives the plant only if there is a difference. Closed-loop systems are less sensitive to disturbances

1 INTRODUCTION TO CONTROL SYSTEMS

26

adjustment of gains in the loop, and sometimes by redesign of the controller (called compensation). On the other hand, closed-loop control systems are more complex, and therefore more expensive than open-loop systems, so the designer must balance the cost when desciding what to use.

1 INTRODUCTION TO CONTROL SYSTEMS

1.2.6 Computer-Controlled Systems

27

Computer-Controlled Systems
The controller or compensator is a computer Slide 30 many loops can be controlled by time sharing. adjustment of controller parameters are in software rather than hardware. supervisory functions such as scheduling, data logging, error and fault monitoring, can also be done.

In the next lecture we look at the design objectives for control systems.

1 INTRODUCTION TO CONTROL SYSTEMS

28

Summary
In this lecture we have Introduced the Topic of Control and Given a denition of a control system a description of typical inputs and outputs Slide 31 an introduction to the terms steady-state error and transient performance some advantages of control systems an illustration of the difference between open-loop and closed-loop control an introduction to computer controlled systems

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

Lecture 2: Analysis and Design

29

Preamble
In the last lecture we introduced the idea of a control system and examined the basic features of such systems. Control Systems are dynamic systems, they respond to an input by undergoing a transient response prior to reaching a steady-state response that generally resembles the input. In this lecture we discuss the importance of transient and steady-state response and then establish our analysis and design objectives. We also introduce a new concept called stability. We then present an example of a closed-loop control system to further illustrate some of these concepts. We nally present the control systems analysis and design sequence and nish with an overall summary of the introductory lectures.

Introduction to the Analysis and Design Process

Introduction to the Analysis and Design Process

Control systems analysis and design objectives Transient response Slide 32 Steady-state accuracy Stability An example of a closed-loop system The design and analysis squence Summary of introductory lectures

2.1 Control Systems Analysis and Design Objectives

2.1.1 Transient Response Another example is a hard-disk drive for a computer. The transient is related to the read-write time. Read/write cannot take place until the head is in place over

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

30

Transient Response (1)

Transient response is very important. Taking the lift example Slide 33 if the lift moves too slowly the passengers would get impatient if on coming to rest, the lift was to oscillate for more than about a second the passengers would get worried. if the lift moved too quickly there may be structural damage caused to the building.

the correct track of the disk. So speed of the read/write head over the surface of the disk from one track to another will be important for the control of the hard-disk drive. 2.1.2 Steady-state accuracy We also focus control systems analysis and design on this aspect of performance. (Slides 35,36,37) 2.1.3 Stability See slides 3842.

2.2 An Example of a Closed-Loop System

We now introduce an example closed-loop control system which will serve as a case study for this course. The aim in introducing this example here is to: to see a physical example to see how it works to see how we can affect its performance 2.2.1 Antenna Azimuth Position Control The output is the azimuth angle o t follows the input i t of potentiometer. A model of this system is used in the lab.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

31

Analysis and Design Objectives: Transient Response

In this course: Slide 34 we shall use quantitative measures of transient response we analyse a systems existing transient response we seek to adjust the design parameters to yield a desired transient response.

Steady-State
Steady-state is concerned with the state of a system after it arrives at the desired output lift system: when the lift car reaches the fourth oor; hard disk controller: when the read-write head is over the correct track on the hard disk.

Slide 35

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

32

Steady-State Accuracy
We are concerned with the accuracy of the steady-state. Slide 36 The oor of the lift must be sufciently level with the oor of the corridor to allow passengers to safely enter or leave the car; the read-write head would yield disk errors if it was not positioned correctly over a track on the disk surface; a satellite tracking system must keep the satellite within its beam width.

Analysis and Design Objectives: Steady-State Accuracy

Slide 37 In this course: We shall dene quantitative measures for steady-state accuracy We shall design corrective measures to reduce steady-state errors.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

33

Stability
Slide 38 Transient response and steady-state error are moot points if the system does not have the characteristic of stability.

What is Stability?
Recall that the total system response is equal to the sum of the natural response and the forced response Slide 39

Total Response = Forced Response + Natural Response

natural response is obtained from the homogeneous solution of a differential equation forced response is obtained from the particular solution.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

34

Natural and Forced Response

Slide 40

The natural response (or homogeneous solution) describes the way a system acquires or dissipates energy. The form and nature of the natural response depends only on the system, not its inputs. The form or nature of the forced response (or particular solution) depends on the input.

Instability
In some systems, the natural response grows without bound rather than diminishing or oscillating. Eventually the natural response is so much bigger than the forced response that the system becomes out of control. This condition is called instability. It could lead to self destruction.

Slide 41

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

35

Analysis and Design Objectives: Stability

Slide 42 Control systems must be designed to be stable.

Summary of Analysis and Design Objectives

The objectives for a control system are to achieve: Slide 43 stability steady-state error transient performance (where the order indicates priorities for the designer). The question to be answered in the course is how do we achieve these?

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

36

o ( t )

Azimuth Angle

antenna

Slide 44

cable

i ( t )

Antenna Azimuth Position Control

potentiometer

2.2.2 Block Schematic Diagram A block schematic diagram of the system is shown in Slide 45. The system normally operates to drive the error to zero. When the input = the output there is no actuating signal, the motor is not driven. The motor is only driven when the input 6= the output. The bigger the error the faster the motor. 2.2.3 Transient Performance What happens if the gain of the signal amplier is increased? The motor is driven harder, but the actuating signal is still zero when input = output. The difference will be in transients motor driven harder so will move faster. The increased speed leads to increased momentum so the system may overshoot the nal value and be forced by the system to reverse its direction. The result may be a diminishing oscillation. (Slide 46) 2.2.4 Steady-state error In Slide 46 there is no steady-state error. In some systems there is, and the increase in gain will tend to reduce its value. This leads to a trade-off between transient performance and steady-state error. To combat this extra components may need to be added to the system to allow both the gain and the transients to be adjusted. This is called compensation.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

37

Input: Desired Azimuth Angle

i (t )

Potentiometer

Voltage proportinal to input +

-

Error or Actuating Signal

Controller
Signal & Power Amplifier

Plant or Process
Motor, Load & Gears

Output: Azimuth Angle

o (t )

Slide 45
Voltage proportinal to output

Sensor
Potentiometer

Block Schematic Diagram

Response

high gain input

Slide 46

low gain

Time

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

38

Control Systems Analysis and Design Objectives

Slide 47

The design objectives revolve around transient, steady-state accuracy and stability. gain adjustments can affect performance and lead to trade-offs among performance criteria compensation may be used to achieve performance without trade-offs.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

39

2.3 The Design and Analysis Sequence

The design and analysis sequence is illustrated in Slide 48. It consists of ve basic stages: 1. Determine a physical system from the requirements. 2. Transform the physical system into a schematic. 3. Construct a mathematical model. 4. Perform block-diagram reduction. 5. Analysis and design. Although the sequence is shown to be linear, it need not be and in practice there will be iterations between stages. Also, it is often the case that the control engineer does not have inuence over the rst stage and may have to design controllers for existing plant. Another point to be aware of is that the early stages are often quite difcult! In the next sections we review each of the stages in the sequence.

Slide 48

Determine a physical system from the requirements

Transform the physical system into a schematic diagram

Mathematically model the schematic as a block diagram

Reduce the block diagram to a single block or closedloop system

Analyze or design the system to meet the requirements

The Control Systems Design and Analysis Sequence

2.3.1 Determine a Physical System from the Requirements (Slides 49 and 50) 2.3.2 Transform the Physical System into a Schematic (Slides 51 to 53)

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

40

Azimuth Position Control System Example

able to position antenna azimuth angle from a remote location Slide 49 weight of antenna physical dimensions desired transient performance desired steady-state accuracy etc.

Functional Description
It is sometimes useful to develop a functional diagram of the system which will help to dene the required hardware. Here is a block-schematic diagram of the azimuth position control system.
Input: Desired Azimuth Angle

Slide 50

i (t )

Potentiometer

Voltage proportinal to input +

-

Error or Actuating Signal

Controller
Signal & Power Amplifier

Plant or Process
Motor, Load & Gears

Output: Azimuth Angle

o (t )

Voltage proportinal to output

Sensor
Potentiometer

Block Schematic Diagram

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

41

Transform the Physical System into a Schematic

Makes relationships more concrete Enables decisions to be made about what can be neglected in formulating the mathematical model. Assumptions made can be easily reviewed and schematic and/or model adjusted as necessary. Should be kept as simple as possible: Checked by analysis and simulation Phenomena added if results do not agree with observed behaviour.

Slide 51

Schematic Diagram of Azimuth Position Control System

i (t )
+ Amplifiers
Potentiometer

Motor
Armature resistance

+ Differential and Power Amplifier

Slide 52
-

Inertia Gear Viscous Damping Gear

d

o (t )

Fix e

fie

ld

+
Potentiometer Gear

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

42

Some simplifying assumptions

Neglect friction and inertia of potentiometers (no dynamics). Slide 53 Neglect dynamics of signal ampliers whose responses will be rapid compared to the motor. Assume pure gain K . Armature voltage controlled motor assume inductance is negligible. Load can be modelled as an inertia plus bearing resistance. There are no losses in the gearbox.

2.3.3 Mathematical Models for the Schematic (Slides 54 and 55) 2.3.4 Block Diagram Reduction (Slide 56) 2.3.5 Analysis and Design (Slides 5760 and Table 1)

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

43

Mathematical Models for the Schematic

Use of basic physical laws. Slide 54 Construction of circuit diagrams and mechanical mobility diagrams. Differential equations. Block diagrams with transfer functions. Covered in Dynamic Systems Course Result is a block diagram in this case.

Alternative Mathematical Models for the Schematic

Slide 55 State-space models could be used. For these nth order differential equations are converted into n rst-order differential equations which are themselves represented by matrices.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

44

Block Diagram Reduction

Slide 56 This is used to obtain canonical representations of the system which is the same as those used to develop the theories used to analyse and design the system.

Analysis and Design

Performance characteristics such as Slide 57 stability steady-state accuracy, and transient performance are determined.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

45

Input

Table 1: Test Waveforms Used in Control Systems Function Description Sketch

f (t )

Use

Impulse

t

t = 1 for 0, t 0+ = 0 elsewhere

( t)

Z 0+
0,

tdt = 1

Transient response modelling

Sinusoid

sin !t
f (t )

t
Transient response modelling; Steadystate error

Step

ut

ut = 1 for t  0 = 0 elsewhere

f (t )

Transient response; Steadystate error

Ramp

tut

tut = t for t  0 = 0 elsewhere

f (t )

Steady-state error

Parabola

1 t2 ut 2

1 t2 ut 2

= 1 t2 for t  0 2 = 0 elsewhere

Steady-state error

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

46

Use of simple test inputs.

Impulse, sinusoid, step, ramp and parabolic inputs are used to excite the system in order to determine the actual response of the system. Impulse is used to inject energy into the system so that its natural response may be obtained. This can be used to determine what the transfer function of an unknown system is. Sinusoid is used to determine the steady-state and transient behaviour from frequency response measurements. It can also be used to determine the transfer function of an unknown system. Step is used to analyse steady-state and transient performance Ramp and Parabolic inputs are used to determine steady-state accuracy..

Slide 58

Choosing Components
The choice of components to be assembled into a control system depends on factors such as speed and power. The system must be analysed to see if the requirements can be met. If the requirements cannot be met then the designer may need to design additional components or make adjustments to the systems parameters in order to meet the requirements.

Slide 59

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

47

Other Considerations

Slide 60

Sensitivity analysis may need to be performed in order to determine how changes in system parameters will affect the performance of the system. Systems must be built to withstand small changes in parameters due to causes such as temperature, pressure, etc. Once the design is complete, time response analysis may need to be done again to verify that the control system meets all the requirements.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

48

Recap of Introductory Lectures

Summary of Introduction (1)

Control systems are dynamic systems with feedback: Slide 61 contribute to many aspects of modern society exist naturally and in elds such as economics are used where power gain, remote control or conversion of input is required.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

49

Summary of Introduction (2)

A control system has an input, a process and an output. Slide 62 Open-loop systems do not monitor or correct the output for disturbances; they are simpler and less expensive then closed-loop systems. Closed-loop systems monitor the output and compare it to the desired output (input). If an error is detected, the closed-loop system corrects the output; they can correct the effects of disturbances.

Summary of Introduction (3)

Control systems design focuses on Slide 63 transient response steady-state response stability. During analysis and design, the engineer tries to achieve stability, transient performance and steady-state accuracy requirements.

2 INTRODUCTION TO THE ANALYSIS AND DESIGN PROCESS

50

In the next lecture we shall develop the mathematical model of the azimuth position controller.

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

Lecture 3: Mathematical Modelling (1)

51

Preamble
In the last lecture we considered the analysis and design sequence for control systems. An important part of that sequence is the creation of a mathematical model, in the form of a block diagram, for the system to be analysed and designed. In the next two lectures we shall present the mathematical model of the Antenna Azimuth Control System introduced in the last lecture. The presentation is split into two parts. In part 1, a block diagram model of the plant and actuator components is presented. In part 2 the additional sensing, comparison and amplication components needed to complete the closed-loop control system are considered. Both parts assume that a sufcient level of modelling knowledge has been acquired in the preand co-requisite courses EE106: Dynamic Systems and EE206: Dynamic Systems

Modelling the Azimuth Position Control System

3.1 The Schematic for the Plant

Consider the azimuth control system we have already seen. The plant for this is illustrated in Slide 64. It is an armature-voltage controlled DC motor which drives the load through a gearbox.

Armature Controlled DC Motor Schematic Diagram

motor
Slide 64

N1 = 10 J1 = 1 kg m2
N2 = 20 J2 = 2 kg m 2
N3 = 10 J3 = 2 kg m 2

N 4 = 20

Zl ( s )

We shall build up to a full block diagram in stages starting from the mechanical side, introducing the gearbox and nally adding in the electrical side.

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

52

3.2 Mechanical Side

The rotor of the electric motor has inertia Jm kg m2 and bearing resistance Rm N m/(rad/s). Assume that we can lump the load effects into a single rotational mechanical impedance

Ql s = Z s l l s

then, ignoring the gearbox for the moment, a schematic diagram of a driven shaft is that shown in Slide 65.

Generalised Driven Rotational Shaft

Slide 65

qe (t ) Rm

Jm

qm

ql

zl
load impedance

l (t )

driving shaft bearing driving shaft resistance Inertia

The impedence of the driving shaft is Zm s = sJm + Rm and Qe s = Qm s + Ql s, m s = l s hence a block diagram for the driven shaft is that shown in Slide 66, which has Qe s as input and l s as output. Reducing the block diagram gives

l s =
or

1 Zm s + Zl s Qe s

Z s Ql s = Z sl+  s Qes Zl m

3.3 Effect of the Gearbox

A gearbox is an example of a transformer: a passive element which couples two systems of the same kind by transmitting energy without loss from one system to the other. The element is a sink in one system and a source in the other. The roles of

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

53

Block Diagram

Qe ( s )
Slide 66

Qm (s )

+Ql ( s )

1 Z m (s )

m ( s) = l ( s)

Z l ( s)

A Gearbox
driving shaft impedence

qe ( t )

qm m

zm
Slide 67

N1
idler gear

ql

l (t )

zl
N2
load impedence

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

54

source and sink are interchangeable and depend on the direction of the net energy ow. A side view of a typical gearbox is shown in Slide 67. In a simple gearbox. the driving shaft is connected to a gear-wheel that has N1 teeth. This drives a second gear-wheel that is attached to the driven shaft. The second gear has N2 teeth. One input rotation of the driving gear produces N1 =N2 rotations of the driven shaft in the opposite direction. If the direction of rotation is to be in the same direction as the driving shaft, an idler gear is placed between the driving and driven gear-wheels (as in the illustration). The ratio N2 =N1 is called the gear-ratio r and is always 1. Thus:

l = !l = N1 = 1 : m !m N2 r

(1)

We assume that there are no losses in the gearbox1 so that the energy into the gearbox equals the energy out (2):

!m qm = !l ql
and hence

(2) (3) (4)

!m = r!l ql = !m = r: qm !l

The purpose of a gearbox, a pure rotational transformer, is: 1. to change the speed of a power source, i.e. a motor, to meet the need for a different output speed, e.g. car wheels. 2. to change the torque of a power source to meet the need for a different output torque. In most cases, the power source is a high-speed low-torque device and the load is a low-speed high-torque device. So in general r 1 and often r 1. When developing a mathematical model for a gearbox, it is important to recognise that the energy equations (2) and (4) are fundamental. As a result of this equation there are a pair of constraints on the driven and driving speeds and torques that must be satised. In block diagram terms this means that the gearbox is either represented by the two blocks shown on the left of Slide 68 or, alternatively as the two blocks shown at the right of Slide 68 (the equations represented by these block diagrams are easily derived from (4)). To derive equations for the gearbox coupled mechanical system illustrated in Slide 67, we note that

Ql s = Zl s l s 1 m s = Z s Qm s m

(5) (6)

1 there are actually several losses in practical gearboxes, including friction, backlash, gear-wheel inertia, etc.

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

55

Alternative Block Diagrams of a Gearbox

Qm ( s)

Slide 68
l ( s)

r r

Ql ( s )

Ql ( s )

1 r

Qm ( s)

m ( s)

m ( s)

1 r

l(s )

Motor speed - load torque

Load speed - motor torque

Block Diagram of a Gearbox Coupled Drive and Load

Qe ( s )
Slide 69

Qm ( s)

+-

1 Zm ( s )

m ( s)

1 r

l (s )

1 r

Ql ( s )

Zl ( s )

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

56 (7) (8)

1 s r m Qe s = Qms + 1 Ql s r

l s =
A block diagram representing these equations is shown in Slide 69. From (6) and (7) we have

l s =
and from (5) and 8):

rZms Qms

(9)

 Qm s = Qes , Zlr s l s  Qes = rZm s + Zlr s l s Qes = Z s + Zl s m r2 m s

(10)

Hence: (11) (12)

Equation (12) is of particular interest since it tells us that the impedance of the load shaft as seen at the input shaft of the gearbox is reduced by r2 . Thus for a motor driving a load through a gearbox we have:

1 m s = Qes Zm s + Zl s=r2

(13)

which should be compared with the results derived in Section 3.2. The interpretation of this result is that a gearbox allows quite large transfers of energy with modest torques. For example, a gearbox with a gear ratio of 20 : 1 is able to move an inertia of 400 times larger than the directly connected driven-shaft intertia with the same amount of input effort. 3.3.1 An Ideally Matched Gearbox Transformers are often used to match parameters in one system to those of another, normally to optimize some aspect of performance. A gearbox could be used to maximise the rotational acceleration d! t in one system achievable by a given dt torque in the other. Taking the model of driving shaft and driven load that is represented by the schematic in Slide 64, then, assuming that Zm s = sJm + Rm and Zl s = sJl + Rl , from equation (13) we have:

m s = sJ + R  + sJ + R =r2 Qe s m m l l Qes = rsJm + Rm + sJl + Rl =r l s = rJm + Jl =rs l s + rRm + Rl =r s

(14) (15)

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

57

Thus to minimize the torque due to acceleration (s l s), we must minimize rJm + Jl =r with respect to r. That is:

d rJ + J =r = 0 dr m l , r12 Jl + Jm = 0 r =

(16) (17) (18)

rJ

l Jm

Such an ideally matched gearbox ensures that the inertia in one side matches p the inertia in the other side because each contributes Jl Jm .

3.4 Electrical Side

The equations for the electrical side are derived from the the basic laws for a DC motor which are:

qe t = Kmia tif t (19) ve t = Kmif t!mt (20) where if t is the eld circuit current; ia t is the armature circuit current; !m t is the rotational speed of the rotor of the motor; qe t is the electrically generated

torque applied to the rotor shaft by the interactions of the electrical elds produced by the eld and armature coils; ve t is the back-emf generated across the brushes of the motor when the rotor rotates and which opposes the armature circuit voltage; and Km is an electromagnetic coupling constant. These equations are nonlinear. To make them linear, either the eld current or the armature current is kept constant and the motor speed is then controlled by the current owing in the other circuit. We thus have four basic congurations for the DC motor. If the armature circuit current is kept constant then the motor is said to be eld-controlled. The basic equation of motion becomes:

qe t = Kmf if t

(21)

where Kmf = Km ia = constant is the eld-circuit controlled electromagnetic ccoupling constant which has units N m/A. If the eld current is used to control the motor the motor is said to be eld-current controlled and (21) sufces. If the eld voltage is used to control the motor we need an extra equation to take into account the eld circuit impedance which is taken to be the eld coils inductance and resistance in series. The motor is then said to be eld-voltage controlled. If the eld circuit current is kept constant then the motor is said to be armaturecontrolled. The basic equations of motion become:

qe t = Kmaia t ve t = Kma!m t

(22) (23)

where Kma = Km if = constant is the armature-circuit controlled electromagnetic coupling constant which has units N m/A. If the armature current is used to control

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

58

the motor the motor is said to be armature-current controlled and (22) sufces. If the armature voltage is used to control the motor we need extra equations to take into account the armature circuit impedance and the back e.m.f. (23). The motor is then said to be armature-voltage controlled. Block diagrams for the possible DC motor congurations are easy to derive and they are all illustrated in Slide 70. Note that we have used the driven-load equations derived in Section 3.2 to model the mechanical side of the motor. You should be comfortable with deriving models for all these congurations of motor, and to that end, Exercises 31 to 34 are provided to give you some practice.

Possible DC Motor Congurations

Field-current controlled
If (s )

Field-voltage controlled
Vf ( s)
1 sLf + Rf

Kmf

Qe (s )

1 Zm ( s ) + Zl ( s )

m ( s)

I f ( s)

Kmf

Qe (s )

1 Zm (s ) + Zl ( s )

m ( s )

Slide 70

Constant Armature Current Armature-current controlled

Ia ( s )
Kma
Qe (s )

Armature voltage controlled

Va ( s )

1 Zm ( s) + Zl (s )

m ( s)

+Ve (s )

1 sLa + Ra

Ia ( s)

K ma

Qe (s )

1 Z m ( s ) + Z l ( s)

m ( s)

K ma

Constant Field Current

As an aside, the same basic equations are used to derive models for electrical generators. In that case, the input is the rotor speed !m t and the output is the back e.m.f ve t. The model is linearised by either keeping the rotor speed or the eld current constant.

DC Motors
Exercises
31 A eld voltage controlled motor drives a load with resistance R N m=rad=s and inertia J kg m2 . Determine the transfer function relating the load speed to the eld voltage. 32

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

59

An armature current controlled motor drives a load with negligible resistance and inertia J kg m2 through a long shaft with compliance C rad=N m. Determine the transfer function relating the load speed to the armature current. 33 An armature voltage controlled motor drives a load with resistance R N m=rad=s and inertia J kg m2 . Obtain an electrical network for which the input impedance is the same as the input impedance of the armature circuit. 34 In a eld voltage controlled motor with eld resistance 1 , eld inductance 5 H, rotor resistance 0:5 N m=rad=s and rotor inertia 2 kg m2 , the electromechanical coupling constant relating torque to eld current is 10 N m=A. If the motor drives the load with resistance 0:5 N m=rad=s and inertia 8 kg m2 , determine the output speed following a step input of 20 V applied to the eld circuit when the motor is at rest.

We are now ready to put together these results to construct a model for the DC motor used to control the azimuth position of the antenna. 3.4.1 Armature-Voltage Controlled DC Motor For the azimuth position control system we shall use the armature-voltage controlled DC motor shown in schematic form in Slide 64. By putting together all we know so far, the block diagram for this kind of motor is shown in Slide 71. [This block diagram is available for download as a Simulink model1 ] The gearbox loop can be reduced to that shown in Slide 72 from which it is clear that the motor speed is related to the electrically generated torque by the transfer function

m s = sJm +Rm sJl +Rl 1 Qe s 1 + r2 sJm +Rm

1 = sJ + J =r2 + R + R =r2  m l m l

m s = Qes

(24)

A shorthand for (24) is

1 sJe + Re

(25)

where Re = Rm + Rl =r2 and Je = Jm + Jl =r 2 are the effective resistance and inertia as seen at the motor shaft. sJl + Rl =r 2 is called the reected load impedance.
1 http://www-ee.swan.ac.uk/Courses/level2/ee208/models/avcdcm.m

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

60

Block Diagram Model

gearbox motor torque load impedance

1 r

Ql ( s)

sJl + Rl

Slide 71
Va (s )

armature circuit admittance

electromagnetic coupling

+Ve (s )

1 sLa + Ra

I a ( s)

Kma

Qe ( s )

motor admittance

gearbox

Qm (s )

1 sJ m + Rm

m (s )

1 r

l ( s)

back e.m.f.

Kma
electromagnetic coupling

Gear-box Loop
sJ l + Rl r2
m ( s)

Slide 72
Va (s )

+Ve (s )

1 sLa + Ra

I a ( s)

Kma

Qe ( s )

gearbox

Qm ( s)

1 sJ m + Rm

1 r

l ( s)

Kma

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

61

The block diagram has now been reduced to that shown in Slide 73 [This block diagram is available for download as a Simulink model1 ] from which it is easy to show that

Kma m s = 2 Va s Ra + sLa sJe + Re + Kma

(26)

Reduced Block Diagram

gearbox

Slide 73

Va (s )

+Ve (s )

1 sLa + Ra

I a ( s)

Kma

Qe ( s )

1 sJe + Re

m ( s )

1 r

l (s )

Kma

Now m s = sm s where m s is the transformed motor shaft position m t. If we neglect the armature inductance La then:

 R sJ + R  + K 2 a e e ma sm Va s = Kma R 

=
a sJ + R  + K ma Kma e e Kma m s =
R
aJe Kma  : Va s s s + J1 Re + R2 e a

sm

(27)

Hence

(28)

Equation (28) is a relatively simple transfer function of the form:

T s = ssK  +
of which, more later.
1 http://www-ee.swan.ac.uk/Courses/level2/ee208/models/avcdcm2.m

(29)

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

62

3.5 Coupling Constants

[This section is not examinable, but is included for completeness] How do we determine the coupling constants and hence nd a suitable motor for a given load? Recall that for the motor

Ve s = Kma m s = Kmass

and Hence, for the armature circuit

(30) (31) (32)

Qe s = KmaIa s: V Ia = sLa , Ve a + Ra Ra + sLaIa = Va , Ve = Va , Kmasm Q Va = Ra + sLa  K e + Kmasm : ma Assuming that the armature windings inductance, La , is negligible then R Va = K a Qe + Kmasm : ma

(33)

(34)

In the time domain:

If a constant DC voltage va is applied to a given motor, the motor will run at a constant speed !m with a constant torque qe , hence in the steady-state

R va t = K a qe t + Kma !m t:

ma

(35)

Solving for qe :

R va = K a qe + Kma!m :
ma

(36)

in Slide 74. We use a dynameter to measure this torque-speed characteristic for a given va using a set-up as shown in Slide 75 When the speed !m is zero, the curve intercepts the torque axis at a value that is called the stall-torque qstall .

2 qe = , Kma !m + Kma va : Ra Ra This is a straight-line relationship qe versus !m as shown

(37)

When the torque is zero we have a speed called !no,load

qstall = Kma va : R
a

(38)

v !no,load = K a

ma

(39)

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

63

Torque-Speed Relationship for a DC Motor

q ( Nm )

qstall
Slide 74

Torque

va2

va1
no load m ( rad / s)
Speed

Dynameter

Slide 75

Tension

spring balance

A Dynameter
torque
The disk is attached to the motor under test. The belt is held against the edge of the disk under tension. As the motor rotates, the friction between the disk and the belt due to the torque generated by the motor causes the tension in the belt to increase. This increase is measured by the spring balance. At the same time the speed of the disk can be measured by using a tachometer or a stroboscope.

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

Hence the electrical constants are:

64

Kma = qstall Ra va Kma = ! va no,load

Example

(40) (41)

For the motor with the torque-speed characteristic shown in Figure 1 nd the transfer function l s=Va s for an armature-voltage controlled DC motor which drives a load with inertia 700 kg m2 and bearing resistance 800 N m/(rad/s) through a gearbox with gear ratio r = 10. The rotor inertia of the motor is 5 kg m2 and bearing resistance is 2 N m/(rad/s).
Torque (Nm)

500

v a = 100 V

50 Speed (rad/s)

Figure 1:

Solution

Re = Rm + Rl =r2 = 2 + 800 = 10 102 qstall = 500 N m, !no,load = 50 rad/s, va = 100 V. Hence Kma = qstall = 500 = 5: Ra va 100 Kma = ! va
no,load

Je = Jm + Jl =r2 = 5 + 700 = 12 102

= 100 = 2: 50

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

Given that

65

Kma : 1 m s =

Ra Je  Va s s s + J1 Re + Kma Kma Ra e
1 5: 2 m s = ,
1 Va s s s + 12 10 + 2  5 0 = ss +:417 : 1:1667

then

To nd l s=Va s we note that !l

= !m=r hence l = m =r so l s = 0:0417 : Va s ss + 1:1667

A More Difcult Problem

Exercises
This problem is similar to the example except that the gearbox has non-negligible inertia. 35 An armature voltage controlled DC motor whose torque-speed characteristics are shown in Figure 2 drives a load with inertia 16 kgm2 and bearing resistance 32 N m/(rad/s) through the gearbox, illustrated in Figure 3, in which some of the gears have non-neglible inertia. Find the transfer function relating the load speed to the armature voltage. Hint: reect all the inertias and resistances of the gear-wheels and the load shaft to the drive shaft using the rule equivalent impedance = impedance=r2 .

Recap
We have now derived a mathematical model, in block diagram form, of the DC motor and load which forms the actuator and plant of the azimuth position control system. We are still some way from a mathematical model of the complete closedloop control system. To create this we need to add sensors for actual and demanded position (and perhaps velocity), signal and power ampliers. This will be the topic of the next lecture at the end of which we shall be able to describe, in block diagram form, suitable control systems for both position and speed control.

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

66

Torque (Nm)

va = 5 V

600
Speed (RPM)

Figure 2: Torque-speed characteristics

motor
N2 = 20 J2 = 2 kg m 2

N1 = 10 J1 = 1 kg m2
N3 = 10 J3 = 2 kg m 2

N 4 = 20

Zl ( s )

Figure 3: Gear-box

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

Lecture 4: Mathematical Modelling (2)

67

Preamble
In the last lecture we developed a block diagram model for the armature voltage controlled DC motor that is used in the antenna azimuth position control system. The block diagram that resulted is shown in Fig. 4 and the corresponding transfer function is:

Kma m= 2 Va Ra + sLa sJe + Re + Kma

I a ( s)
Kma
Qe ( s )

gearbox

Va (s )

+Ve (s )

1 sLa + Ra

1 sJe + Re

m ( s )

1 r

l (s )

Kma

Figure 4: Block diagram of armature-voltage-controlled DC motor If the armature circuit inductance La is assumed negligible, and it is noted that m = sm (where m is the motor shaft position), an equivalent transfer function relating the motor-shaft position to the armature voltage is:

m = Kma 2 Va sRasJe + Re + Kma K = Je = ss + 1=J maRRaK2 =R  e  e + ma a

which is of the form

Gs = ssK  +

In this lecture, we will add sensing elements and ampliers to this model to create block-diagram models of the closed-loop control systems for position control (called a servomechanism) and speed control (called a velodyne). We will also show that the steady-state response of these two types of system are different. The position control system has zero steady-state error whilst the velocity control system has a non-zero steady-state error which can be made small if there is sufcient gain in the system.

3.6 The Rest of the Control System

The motor ts into the azimuth position control system as shown in Fig. 5. Note that the potentiometers, ampliers, gearbox, inertias and bearing resistances have been given physical values according to Table 2 [A Matlab script le that can be

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

68

used to load this data into the workspace for the purposes of simulation of one of the Simulink models is available1 .]
i ( t )
Pot. 10 turn

+10 Pre-Amp. -10 v ( t ) 1

+ -

Power Amp.

v2 (t )

100 s + 100

Jm = 0 . 02 kg m2

N1 = 25
Jl = 1 kg m2

Rm = 0. 01 N m rad s

o (t )

N 2 = 250
Rl = 1N m rad s

+10

-10
Pot. 10 turn

N3 = 250

Figure 5: Position Control System Schematic The various new components are described below and then a complete mathematical model of the system, in block diagram form, will be developed. 3.6.1 Position sensors A potentiometer (Fig. 6) produces vout

in . There are no dynamics. vout = vmax , vmin : in 2n

In this case 10 turns produces 20 V hence

vout = 20 = 1 V=rad in 20 

in ( t )

vmax

vmin
vout (t )
n turn pot.

Figure 6: Potentiometer
1 http://faith.swan.ac.uk/Courses/level2/ee208/models/ee208dat.m

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

69

Name Load inertia Load bearing resistance Motor shaft gear wheel Load shaft gear wheel Gear ratio Motor rotor inertia Motor shaft bearing resistance Equivalent inertia Equivalent resistance Electromagnetic coupling constant (forward) Electromagnetic coupling constant (feedback) Armature winding inductance (neglected in some models) Armature winding resistance Ten turn potentiometer supplied with 10 V Tacho generator Voltage pre-amplier gain Power voltage amplier (unity DC gain)

Symbol

Jl Rl N1 N2
r

Jm Rm Je = Jm + Jl =r2 Re = Rm + Rl =r2 Kma Kma La Ra Kp Kt K 100=s + 100

N2 =N1
0.02 0.01

Value 1 1 25 250

0.5 0.5 0.8 8

Units kg m2 N m/(rad/s) teeth teeth none kg m2 N m/(rad/s) kg m2 N m/(rad/s) N m/A V/(rad/s) H

V/rad V/(rad/s) none none

Table 2: Component Parameters for the Position Control System

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

3.6.2 Velocity sensors

70

A tacho-generator (Fig. 7) is used to sense the speed of a motor. vout !in . Provided that the load-circuit impedance is high and the tacho-generator is physically small with respect to the driven shaft then the device can be assumed to have no dynamics.

vout = K V=rad=s: T !in

in (t )

vout ( t )

Figure 7: Tacho-generator

3.6.3 Pre-amplier The pre-amplier is assumed to be a small current differencing op-amp circuit as shown in Fig. 8.

vp t = Rf fvi t , vo tg Ri = K fvi t , vo tg

Rf
vo (t )

Ri
Ri

vi (t )

v p ( t)

Rf

Figure 8: Differencing pre-amplier

3.6.4 Power Amplier This produces a high-current voltage output from a low-current voltage input. Such an amplier will often have dynamics which cannot be neglected. In this case we

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

assume

71

indicating that the amplier has unity DC gain and a time constant of 1=100 seconds. This means that the amplier would reach 63 of its nal output voltage in 0:01 seconds following a step change in the input voltage (see Fig. 9).

Va s = 100 Vp s s + 100

Amplifier response: 1 V demand

63% final value

1/100 sec Time (seconds)

Figure 9: Response of the power amplier

3.6.5 Block Diagram of Plant For the system shown in Fig 5 we have derived a block diagram for the DC motor, gearbox and load which on reduction becomes that shown in Fig. 10.
Va ( s )
K ma ( Ra J e ) K2 1 s s+ Re + ma Je Ra

m ( s)

1 r

l ( s)

Figure 10: Reduced block diagram of armature volatge controlled DC motor used in position control system. Output of gearbox is load shaft position l The given system parameters are listed in Table 2. Some still need to be calculated. The motor-load gearbox ratio is

r = N2 = 250 = 10: N1 25
The equivalent motor-load inertia as reected back to the motor shaft is

Jl 1 Je = Jm + r2 = 0:02 + 1  100 = 0:03:

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

The equivalent bearing resistance is:

72

1 Re = Rm + R2l = 0:01 + 1  100 = 0:02: r

Plugging these values into transfer function for the DC motor and load gives:

= m s =
Kma
RaJe   2 Va s s s + J1e Re + Kma Ra m s = , 0:5=8  0:03

, Va s s s + 0:103 0:02 + 0:852 :083 = ss2+ 1:71
The transfer function relating the load position l s = o s is

(42)

o s = 1
Kma
RaJe   = 2 Va s r s s + J1 Re + Kma Ra e 0 2083 = ss :+ 1:71 :

(43)

The load-pot gearbox has unity gear-ratio so that the pot moves at the same speed as the load. So, putting everything together we end up with the block diagram shown in Fig. 11. [This block diagram is available for download as a Simulink model2 ]
pot pre-amp power amp motor & load gearbox

i ( s)

Vi ( s)
+

Vp ( s )

K
-

100 s + 100

Va ( s )

2. 083 s( s + 1. 71)

m ( s)

1 r

o ( s)

Vo ( s)

1
pot

Figure 11: Complete position control system. It is convenient to reduce the block diagram to the unity-gain feedback canonical form shown in Fig. 12 which we can do because the potentiometer in the forward path and the feedback path have the same gain. The open-loop transfer function is then:

66:3 Go s = oss = ss + 100K+ 1:71 ; E s

2 http://faith.swan.ac.uk/Courses/level2/ee208/models/srvomech.m

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

i ( s)

73
o ( s)

+
-

E (s )

Vp ( s )

100 s + 100

Va ( s )

2. 083 s( s + 1. 71)

m ( s)

1 r

Figure 12: Reduced block diagram and the closed-loop transfer function is

Note that in the steady-state, i.e. when all terms in s are removed

s K Gc s = os = s3 + 101:7s66:3171s + 66:3K : 2+ 

i

i = o :
So the steady-state position error is zero! Another name for this type of control system is a servomechanism.

3.7 Velocity control system

A velodyne is a load-speed control system. A velodyne can be made from the components seen so far if a tacho-generator with gain KT v/(rad/s) is used in the feedback loop and a potentiometer with gain KT is used as a demanded speed sensor. Such a set up is shown in Fig. 13. [This block diagram is available for download as a Simulink model3 ]
pot pre-amp power amp motor & load gearbox

i(s )

Vi ( s)

KT

Vp ( s )

K
-

100 s + 100

Va ( s )

2. 083 s + 1. 71

m ( s)

1 r

o (s )

Vo ( s)

KT
tacho

Figure 13: A velocity control system (Velodyne). On manipulating this block diagram we get the unity-gain feedback control system shown in Fig. 14. The closed-loop transfer function for the velodyne is

s 66: KK Gc s = os = s2 + 101:71s +3171 T 66:3KK  + i T

In the steady-state

!o = 66:3KKT !i 171 + 66:3KKT

3 http://faith.swan.ac.uk/Courses/level2/ee208/models/velodyne.m

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

i(s )

74
o (s )

+
-

E (s )

KT

100 s + 100

2. 083 s + 1. 71

1 r

Figure 14: Reduced block diagram that is !o 6= !i but !o  !i if 66:3KKT 171. This can be achieved if the pre-amplier gain K can be made sufciently large. 36 In a control system for rotating a radar aerial assembly, an electric motor with inertia 0:05 g m2 and resistance 0:02 N m=rad=s is required to drive the inertia with inertia 500 kg m2 and resistance 50 N m=rad=s through a gearbox. Determine the gear ratio that minimises the torque needed to accelerate the load, and the transfer function relating aerial speed to motor torque if such a gearbox is used. What is the motor power required to rotate the aerial at 10 rev=min? 37 In a servomechanism using a eld voltage controlled motor, the ratio of the motor torque to the error between the demanded and actual load position is 100 N m=rad. If the inertia and resistances refererred to the motor shaft are 20 kg m2 and 3 N m=rad=s, and the motor drives the load through a gearbox with ratio 50:1, determine the overall transfer function of the system. 38 A position control system is illustrated in Figure 38. Evaluate the transfer function of each subsystem and determine the closed-loop transfer function o s=i s.
i ( t )
Pot. 1 turn

+10 Pre-Amp. -10 v ( t ) 1

+ -

Power Amp.

v2 (t )

150 s + 150

Kma = 1 N m A

J m = 0 . 05 kgm 2

N1 = 50
J l = 5 kg m 2

Rm = 0. 01 N m rad s

o (t )

N 2 = 250
Rl = 3 N m rad s

+10

-10
Pot. 1 turn

N3 = 250

(Adapted from Chapter 2 Objective Problem, Nise, Exercise 48, page 109).

Recap
In this lecture we have developed block diagram models for a position control system and the speed control system created around an armature voltage controlled

3 MODELLING THE AZIMUTH POSITION CONTROL SYSTEM

75

DC motor. The position control system (servomechanism) was found to have zero steady-state error which means that it is ideal for the antenna azimuth position control system. The same mechanism, when used in a speed control system (velodyne), was found to have nite non-zero steady-state error which can be decreased by increasing the pre-amplier gain. In the next sequence of lectures called collectively Evaluation of System Response, we shall turn our attention to transient response. Beginning in the next lecture with an evaluation of how poles and zeros effect the system response and building up to an understanding of rst and second-order system response and measures which can be used to evaluate and ultimately control them.

4 EVALUATION OF SYSTEM RESPONSE

Lecture 5: Evaluation of System Response (1)

76

Preamble
In the last lecture we developed block diagram models for a position control system and a speed control system created around an armature voltage controlled DC motor. The position control system (servomechanism) was found to have zero steadystate error which means that it is ideal for the antenna azimuth position control system. The same mechanism, when used in a speed control system (velodyne), was found to have nite non-zero steady-state error which can be decreased by increasing the pre-amplier gain. In this sequence of lectures called collectively Evaluation of System Response, we shall turn our attention to the transient response of systems. We begin with an evaluation of how poles and zeros effect the system response and build up to an understanding of rst-order and second-order system response and measures which can be used to evaluate and ultimately to control them.

Evaluation of System Response

This section of the course is devoted to the evaluation of the transient response from the system model. After introducing the useful concept of poles and zeros, we begin analyzing models to nd the step response of rst and second order systems (where the term order refers to the polynomial order of the denominator of the transfer function).

Evaluation of System Response

Objectives Given the antenna azimuth-angle control system developed in the section on modelling, at the end of this section we shall be able to: Slide 76 Predict, by inspection, the form of the open-loop angular velocity response of the load to a step-voltage input to the power amplier; Find the damping ratio and natural frequency of the open-loop system; Derive the complete analytical expression for the open-loop response using transfer functions.

4 EVALUATION OF SYSTEM RESPONSE

77

4.1 Poles and Zeros and System Response

The output response of a linear system is the sum of the forced response and the natural response

ct = cf t + cn t:

Solving differential equations or using inverse Laplace transforms allow us to evaluate the output response. But this is laborious and time consuming. By having an analysis technique that is quick and easy to apply we will increase productivity. Knowledge of the effects of poles and zeros gives us a means of qualitatively evaluating the response of a system by inspection. Poles of a Transfer Function: The poles of a transfer function are those values of the transfer function variables that cause the transfer function to become innite. Zeros of a Transfer Function: The zeros of a transfer function are those values of the transfer function variables that cause the transfer function to become zero. Let Gs be a transfer function

m + b sm,1 + + b + Gs = as = s n + am,1sn,1 +





+ ab1ss+ ab0 s s n,1

1 0
The poles of Gs are solutions of

as = sn + an,1 sn,1 +


+ a1 s + a0 = 0 that is, they are zeros of the denominator polynomial as. If the denominator of Gs is zero then Gs = 1. The zeros of Gs are solutions of bs = sm + bm,1sm,1 +


+ b1s + b0 = 0 that is, they are zeros of the numerator polynomial bs. If the numerator of Gs is zero then Gs is also 0.
Example 4.1 Determine the step response of a transfer function. Solution: The step response is

C s = sss+ 25 + = A + sB5 s + = = 2=5 + s3+55 s

thus

ct = 2 + 3 e,5t : 5 5

4 EVALUATION OF SYSTEM RESPONSE

78

Example G s

+2   = ss+5
j

s plane

Slide 77

R( s ) =

1 s

s +2 s+5
poles at s = 5 zeros at s = 2

C( s )

The step response of

is shown in Slide 78. A Plot of jGsj for s = x + jy , x = ,6; : : : ; y = ,3; : : : ; 3, indicating why a pole is called a pole is shown in Slide 79. Discussion

Gs = s + 2 s+5

0;

1. A pole in the input function generates the form of the forced response cf t (a pole at the origin generated a step function in the output). 2. A pole of the transfer function generates the form of the natural response (i.e. a pole at s = ,5 generated cn t = e,5t ). 3. A pole on the real-axis in the s-plane generates an exponential response of the form e, t where , is the pole location on the real axis. Thus the further left a pole is on the negative axis, the faster the exponential transient decays to zero. 4. The zero helps generate the amplitudes for both the steady-state and transient performances. (This can be seen from the calculation of A and B in the partial fraction expansion.)

4 EVALUATION OF SYSTEM RESPONSE

79

Step Response for G(s) = (s + 2)/(s + 5) 1 0.9 0.8 0.7 0.6 c(t) Matlab command: step([1 2],[1 5])

Slide 78

0.5 0.4 0.3 0.2 0.1 0 0

0.1

0.2

0.3

0.4 0.5 0.6 time t seconds

0.7

0.8

0.9

Poles and Zeros

35 30 | G(s) | = |(s + 2)/(s + 5)| 25 20 15 10 5 0 4 2 0 -2 Im{s} -4 -5 -6 Re{s} -3 -4 -1 -2 0

Slide 79

4 EVALUATION OF SYSTEM RESPONSE

80

Contributions of Poles and Zeros

j
input pole system zero

j
system pole

Slide 80

C(s ) =

25 35 + s s+5

c (t ) =

2 3 5t + e 5 5
natural response

forced response

4 EVALUATION OF SYSTEM RESPONSE

81

Example 4.2 Use poles to evaluate the system response of the system shown in Fig. 15 by inspection.
R( s ) = 1 s

s+3 ( s + 2)( s + 4)( s + 5)

C( s )

Figure 15: Poles and System Response

Solution:

C s =

forced response

K1 | sz

+ sK22 + sK34 + sK45 + + +

natural response

ct = |Kz1 + K2e,2t + K3ez,4t + K4e,5t : |

cf t cn t
Transfer Functions in Matlab There are two forms of transfer function representation in Matlab. The most obvious is the polynomial form where

b s2 s Gs = as = s3 + 4+22+ +s3+ 6 s s 5

is entered as two row vectors with the polynomial coefcients entered in the order of descending powers of s. >> b = [1, 2, 3]; >> a = [1, 4, 5, 6]; Missing coefcients, must be entered as zero: so 4 + s2 + 1 are entered as s >> q = [1, 2, 0]; >> r = [1, 0, 2, 0, 1]; An alternative form of representing transfer functions is the factored polynomial, for example

qs = s2 + 2s and rs =

Gs = sss+ 1ss+ 3 + 2 + 4

The advantage of this formulation is that the roots or zeros of the numerator and denominator polynomials are obvious by inspection so it is often used in the preliminary analysis of the performance of a dynamic system. The poles of this transfer function are therefore s = 0; ,2; ,4 and the zeros are s = ,1; ,3. In Matlab, this form of transfer function is specied by a column vector of the zeros and a column vector of the poles:

4 EVALUATION OF SYSTEM RESPONSE

>> z = [-1; -3]; >> p = [0; -2; -4];

82

A third parameter, the overall gain K , completes the denition of the so called pole-zero-gain form of transfer function. In this case K = 1 >> k = 1; Transformations Matlab supports the transformation of transfer function between forms. For example to convert a transfer function from expanded form to pole-zero-gain form the command tf2zp is used: >> [z1,p1,k1] = tf2zp(b,a) z1 = -1.0000 + 1.4142i -1.0000 - 1.4142i p1 = -3.0000 -0.5000 + 1.3229i -0.5000 - 1.3229i k1 = 1 To convert from zero-pole-gain form to expanded form we use the function zp2tf: >> [b1,a1] = zp2tf(z,p,k) b1 = 0 a1 = 1

Partial Fraction Expansions Matlab also provides a command called residue that returns the partial fraction expansion of a transfer function. That is, given

m+b sm,1 + + + Gs = s n + am,1sn,1 +





+ ab1ss+ ab0 s n,1

1 0
it returns

R1 + R2 +


Rn + K s s + p1 s + p2 s + pn

4 EVALUATION OF SYSTEM RESPONSE

83

where pi are the poles of the transfer function, Ri are the coefcients of the partial fraction terms (called the residues of the poles) and K s is a remainder polynomial which is usually empty. To use this, the starting point must (rather perversely) be the expanded form of the transfer function. Thus given

5s + C s = ss + 3s2 10 +

we obtain the partial fraction expansion using the Matlab command sequence: >> k = 5; z = [-2]; p = [0; -3; -10]; % zero-pole-gain form >> [num,den] = zp2tf(z,p,k) num = 0 0 5 10 den = 1 13 30 0 (Note that the leading terms in num are zero). >> [r,p,k] = residue(num,den) r = -0.5714 0.2381 0.3333 p = -10 -3 0 k = [] which we interpret to mean

C s = 0:3333 + 0s:2381 , 0s:5714 : s +3 +5

If C s represents the step response of the system

5s + 2 Gs = s + 3s + 10

then the step response is, by inspection,

ct = 0:3333 + 0:2381e,3t , 0:5714e,10t:

You can check this with the command: >> step([5, 10],[1, 13, 30])

4 EVALUATION OF SYSTEM RESPONSE

84

(where the 1=s term has been eliminated because step provides the forcing function itself). This should give exactly the same results as: t = 0:.05:1.5; % time vector c = 0.3333 + 0.2381 * exp(-3*t) - 0.5714 * exp(-10*t); plot(t,c) Exercise Use Matlab to determine the actual coefcients of the partial fraction expansion for the previous example.

4.2 First-Order System Responses and Specications

A rst-order system is illustrated in Slide 81. It has the transfer function

which is also called a rst-order lag. The system has one pole at s = ,a as shown in the pole-zero diagram. For the step response, the input signal transform is

a Gs = s + a Rs = 1 s

so the step response is

C s = GsRs = ss a a : +

In the time domain the step response is:

ct = cf t + cn t = 1 , e,at :

The forced response cf t = 1 is generated by the pole of 1=s. The natural response cn t = e,at is generated by the pole of a=s + a. 4.2.1 The signicance of a

The parameter a is very important for specifying the performance of a rst-order system. It is signicant because

e,at t=1=a = e,1 = 0:37:

Or, alternatively, the step response

ct = 1 , e,at t=1=a = 0:63:

That is, we can relate the shape of the time response to the parameter a.

4 EVALUATION OF SYSTEM RESPONSE

85

First-Order System

s plane

Slide 81

R( s ) =

1 s

a s+a

C( s )

4.2.2 Time Constant The parameter = 1=a is called the time constant of the rst-order response. It is the time taken for e,at to decay to 37% of its initial value or the time taken for the step response to reach 63% of its nal value. has units s. a has units s,1 or frequency. a is called the exponential frequency. The derivative of e,at at t = 0 is ,a, so a is the initial rate of change of the exponential at t = 0. Thus the time constant can be considered a transient response specication for a rst-order system since it is related to the speed of response. The time constant can be evaluated directly from the pole-zero plot. The pole is s = ,a. The further to the left of s-plane, the larger a hence the smaller the time constant = 1=a and the faster the response. A plot of a typical rst-order response is shown in Slide 82. The parameters due to a are marked on the plot. Some additional time-response parameters that are in common use can also be identied. These are rise-time and settling-time. 4.2.3 Rise Time Tr

The rise-time (symbol Tr units s) is dened as the time taken for the step response to go from 10% to 90% of the nal value. For a rst-order system it is rather easily derived by solving ct = e,at for ct0:1  = 0:1c nal and ct0:9  = 0:9c nal from which

31 11 : Tr = 2:a , 0:a = 2a2 :

4 EVALUATION OF SYSTEM RESPONSE

86

First-Order System Response

c(t)
100% 98% 90% initial slope = 1/(time constant) = a 2%

Slide 82

63%

Ts
10%

Tr
1 a

2 a

3 a

4 a

t (seconds)

4.2.4 Settling Time Ts

We shall dene the settling-time (symbol Ts units s) to be the time taken for the step response to come to within 2% of the nal value of the step response.1 From the denition,

cTs  = 0:98c
4 Ts = a :

nal

which gives

In other words the 2% settling-time for a rst-order system is 4 the time constant.

Recap
In this lecture, we have seen that poles determine the nature of the time response: the poles of the input determine the nature of the forced response; the poles of the transfer function determine the nature of the natural response. For the latter, we could say that the poles of a transfer function determine the character of the natural response. For this reason, the equation

as = sn + an,1 sn,1 +


+ a1 s + a0 = 0

is called the characteristic equation. Some further observations are:
denitions are common. For example, some authorities prefer to use 5% of the nal value or 1% of the nal value. To be absolutely clear we shall use the term x% settling-time to indicate what band we are using.
1 Other

4 EVALUATION OF SYSTEM RESPONSE

87

The zeros of a transfer function contribute to the amplitude of the component parts of the total response. Poles on the real axis generate exponential responses. We have also examined the response of the rst-order system (or rst order lag)

a Gs = s + a
and discovered that: System has a real pole at s = ,a Step response:

ct = 1 , e,at . The parameter a is called the exponential decay frequency (units s,1 ) Time constant = 1=a (units s) is time for step response to reach 63% of the
nal value. The rise time is time taken for step response to rise from 10% to 90% of the nal value. The rise time Tr = 2:2=a = 2:2 s The 2% settling time Ts response of the system.

= 4=a = 4 s. The larger the value of a the smaller the value of

and hence the faster the

Second-order systems exhibit a wider range of responses than rst-order systems. In the next lecture we shall describe the types of responses exhibited by second-order systems.

4 EVALUATION OF SYSTEM RESPONSE

Lecture 6: Evaluation of System Response (2)

88

Preamble
Second-order systems exhibit a wide range of responses which must be analysed and described. Whereas for a rst-order system, varying a single parameter changes the speed of response, changes in the parameters of a second order system can change the form of the response. For example, a second-order system can display characteristics much like a rst-order system or, depending on component values, display damped or pure oscillations for its transient response. In this lecture we shall use the general second-order transfer function shown in Slide 83. to explore the range of responses that are possible.

General second-order transfer function

Slide 83

1 R( s ) = s

b s 2 + as + b

C (s )

2 poles. No zeros.

There are four types of response possible for a second order system. They are Overdamped Response Underdamped Response Undamped Response Critically Damped Response To explore these we consider the second order system

with b = 9 and a selected to illustrate each type of response.

b Gs = s2 + as + b

4 EVALUATION OF SYSTEM RESPONSE

89

4.3 Types of Second-Order System Responses

4.3.1 Overdamped Response IF a = 9 we have the system shown in Slide 84.

Overdamped system

Slide 84

R( s ) =

1 s

9 s 2 + 9s + 9

C (s )

2 poles. No zeros.

The step response1 will be

9 C s = ss2 +9 s + 9 = ss + 7:854s + 1:146 : 9

The pole at s = 0 comes from the input (forced response) and there are two real poles at s = ,7:854 and s = ,1:146 respectively. Thus the step response will be of the form

ct = K1 + K2 e,7:854t + K3e,1:146t:

(Slide 85). The response consists of the some of two rst-order responses. The response due to the largest pole at s = ,7:854 has a time constant of fast = 0:127 s. Left to itself, this pole would have a settling time of Tsfast = 0:51 s. The response due to the smallest pole at s = ,1:146 has a time constant of slow = 0:873 s. This pole has a settling time of Tsslow = 3:49 s. Settling time Tsslow is about 4 times slower than Tsfast . Thus the slow pole dominates the later stages of the response while the fast pole dominates the early part of the response. In addition, the initial slope is zero.
1A

full analysis of the step response for a general second-order system is given in an Appendix.

4 EVALUATION OF SYSTEM RESPONSE

90

Overdamped response
s plane

c(t)

Slide 85
7. 854

1. 146

4.3.2 Underdamped Response IF a = 3 we have the system shown in Slide 86. The step response will be

C s = ss2 +9 s + 9 : 3
The pole at s = 0 comes from the input (forced response) and there are two complex poles at s = ,1:5  j 2:598. The step response of this system will be of the form

ct = K1 + e,1:5t K2 cos 2:598t + K3 sin 2:598t:

(Slide 87). Comparing the time response with the pole locations, we see that the exponential decay term is related to the real-part of the complex pole pair. Real part: d = 1:5; exponential term e,1:5t = e, d t . The frequency of the oscillatory term is related to the imaginary part of the complex pole pair. Imaginary part: !d = 2:598; sinusoidal term: K2 cos 2:598t + K3 sin 2:598t. This information is summarised in Slide 88). The frequency of oscillation !d = 2 (rad/s) is called the damped natural T frequency. The time constant of the decay is = 1= d .

4 EVALUATION OF SYSTEM RESPONSE

91

Underdamped system

Slide 86

R( s ) =

1 s

9 s 2 + 3s + 9

C (s )

2 poles. No zeros.

Underdamped response
s plane

j
+ j2. 598

c(t)

Slide 87
1. 5
j2. 598
t

4 EVALUATION OF SYSTEM RESPONSE

92

Decaying oscillation
c(t)
exponential decay generated by real part of complex pole pair

Slide 88

T sinusoidal oscillation generated by imaginary part of complex pole pair

4 EVALUATION OF SYSTEM RESPONSE

93

Example 4.3 Determine by inspection the form of the step response of the system shown in Fig. 16.

R( s ) =

1 s

200 s 2 + 10 s + 200

C (s )

Figure 16: Example underdamped system

Solution: The poles are: response is

s = ,5  j 13:23 and the input is a step hence the step

where from the partial fraction expansion

ct = K1 + e,5t K2 cos 13:23t + K3 sin 13:23t = K1 + K4e,5t cos 13:23t ,  p 2 2 = tan,1 K3=K2 and K4 = K2 + K3 : K1, K2 and K3 are obtained K1 + K2 s + K3 : s s2 + 10s + 200

The proof is left as an exercise. 4.3.3 Undamped Response IF a = 0 we have the system shown in Slide 89. The step response will be

C s = ss29+ 9 :
The pole at s = 0 comes from the input (forced response) and there are two imaginary poles at s = j 3. The step response of this system will be of the form

ct = K1 + K2 cos 3t:

(Slide 90). The frequency of oscillation of the undamped response is called the natural frequency !n . In this case the natural frequency !n = 3 rad.s,1 . 4.3.4 Critically Damped System IF a = 6 we have the system shown in Slide 91. The step response will be

9 C s = ss2 +96s + 9 = ss + 32 :

4 EVALUATION OF SYSTEM RESPONSE

94

Undamped system

Slide 89

R( s ) =

1 s

9 s2 + 9

C (s )

2 poles. No zeros.

Undamped response
s plane

j
+ j3

c(t)

Slide 90

j3
t

4 EVALUATION OF SYSTEM RESPONSE

95

Critically Damped System

Slide 91

R( s ) =

1 s

9 s 2 + 6s + 9

C (s )

2 poles. No zeros.

The pole at s = 0 comes from the input (forced response) and there are two real and equal poles at s = ,3. The step response of this system will be of the form

ct = K1 + K2e,3t + K3te,3t :

(Slide 92).

4 EVALUATION OF SYSTEM RESPONSE

96

Critically Damped Response

s plane

c(t)

Slide 92

4 EVALUATION OF SYSTEM RESPONSE

97

Recap
The natural or characteristic response of a second-order system is governed by the poles of the denominator of the system transfer function Gs. These characteristics are: 1. Overdamped. Poles: two real at , 1 and , 2 . Transient response: two exponentials with time response equal to 1= . cn t = K1 e, 1 t + K2 e, 2 t .

2. Underdamped. Poles: two complex at , d  j!d . Transient response: damped sinusoid with exponential envelope whose time constant is 1= d and whose frequency is equal to !d rad/s. cn t = K1 e, d t cos!d t + .

3. Undamped. Poles: two imaginary at j!1 . Transient response: undamped oscillation with frequency equal to !1 rad/s. cn t = K1 cos!1 t + : 4. Critically damped. Poles: two real and equal at cn t = K1e, 1 t + K2te, 1 t .

1 . Transient response:

These results are summarised in Slide 93. It is clear from this picture that critical damping is the fastest possible response without overshoot.

Second-order responses
c(t) underdamped
Slide 93

undamped

critically damped

overdamped

In the next lecture we will further generalise the second-order transfer function so that we can determine the form of the step response by inspection without the need to nd the poles of the characteristic equation s2 + as + b = 0.

4 EVALUATION OF SYSTEM RESPONSE

Lecture 7: Evaluation of System Response (3)

98

Preamble
In the last lecture we discussed the forms of characteristic response that the secondorder system can exhibit. In summary, these are: 1. Overdamped when the system has two real distinct poles; 2. Underdamped when the system has two complex conjugate poles; 3. Undamped when the system has two imaginary poles; and 4. Critically damped when the system has two real but equal poles. We now seek to generalize the discussion and establish a set of quantitative specications which will enable the response to be described to a designer without the need for a sketch of the response. We develop two parameters that describe second-order response in a similar way to the way time constants can be used to describe a rst-order response. The two new quantities are called 1. natural frequency, 2. damping ratio.

4.4 The General Second-Order Response

4.4.1 Denitions Natural frequency !n is dened as the frequency of oscillation of a second-order system without damping. E.g. the frequency of oscillation of an RLC circuit with R shorted. It has units of rad.s,1 . Damping ratio
is dened by

= Exponential decay frequency Natural frequency Natural period = 21 Exponential time constant 
The damping ratio has no units. That is the damping ratio of a system that decays to zero after 3 oscillations in 1 ms is the same as that of a system that decays to zero in three oscillations in 1 hour. It is independent of the speed of response or the rate of the oscillation.

4 EVALUATION OF SYSTEM RESPONSE

4.4.2 Derivation of Formulae Let us derive formulae for these quantities from their denitions, given that

99

Without damping, the term as = 0 and we have

b Gs = s2 + as + b : Gs = s2 b b : + !n = b; 2 b = !n : s = ,a: 2

The poles are imaginary, and the frequency of oscillation of this system is !n by denition. Thus p

Assuming an underdamped system. The complex poles have a real part given by

The magnitude of the real part is the exponential decay frequency d . Thus d a=2 and from the denition

= !d = a=2 ; !n n a = 2
!n :

hence

In general then, the second order system has the canonical form:

!2 Gs = s2 + 2
!n s + !2
n

(44)

Knowledge of the parameters !n and

may be used to determine the type of motion of any particular second-order system. Example 4.4 Find the natural frequency !n and damping ratio with transfer function:

for the system

Gs = s2 + 436s + 36 : :2

Solution: Comparing to the standard form

2 36 !n Gs = s2 + 4:2s + 36 = s2 + 2
! s + !2 n n 2 !n = 36 ! !n = 36 = 6:

2
!n = 4:2 4: :
= 2!2 = 2426 = 0:35: n

4 EVALUATION OF SYSTEM RESPONSE

4.4.3 Pole-zero locations

100

Having dened !n and

, let us relate these quantities to the pole locations in the s-plane. Solving the characteristic equation

2 s2 + 2
!n s + !n = 0
gives

The various types of response for a given value of natural frequency !n are a function of
and may be summarised as shown in Table 3. Example 4.5 Describe the nature of the second-order system response via the value of the damping ratio for the systems with transfer function:

s1;2 = ,
!n  !n
2 , 1:

12 Gs = s2 + 8s + 12 16 Gs = s2 + 8s + 16 20 Gs = s2 + 8s + 20 :

Solution: In all cases the transfer function is of the form

so a = 2
!n and !n For

= b hence
= a=2 b. a = 8 in all cases. 12 Gs = s2 + 8s + 12 p p b = 12 hence
= 8=2 12 = 2= 3 1: system response is overdamped. For 16 Gs = s2 + 8s + 16 p b = 16 hence
= 8=2 16 = 1: system response is critically damped. For 20 Gs = s2 + 8s + 20 p p b = 20 hence
= 8=2 20 = 2= 5 1: system response is underdamped.

b Gs = s2 + as + b

4 EVALUATION OF SYSTEM RESPONSE

101

2 !n :

Table 3: The relationship between damping ratio, pole location and transient re2 sponse for second-order systems with transfer function Gs = !n =s2 + 2
!n s + Pole Locations
s plane

Type of Response

j
+ j n

j n

=0

Oscillatory
s plane

j
+ j n 1 2

j n 1 2

Underdamped
s plane

=1

Critically Damped
s plane

n n 2 1
n + n 2 1

Overdamped

4 EVALUATION OF SYSTEM RESPONSE

4.4.4 Further analysis for underdamped second-order systems The step response for the general second-order system.

102

!n C s = ss2 + 2
! s + !2  n n K1 + K2s + K3 = s s2 + 2
! s + !2
n

where it is assumed that

1 (the underdamped case) is now obtained 1 Expanding by partial fractions yields

2 1 + s +
!n  + p1,
2 !n 1 ,
C s = s 2 s +
!n 2 + !n 1 ,
2
from which

ct = 1 , e,
!n t cos !n 1 ,
2 t + p

p = 1 , p 1 2 e,
!n t cos!n 1 ,
2 t +  1,

= tan,1 p
2 : 1,

d =
!n

1 ,
2

sin !n 1 ,
2 t

!
(45)

where

If we dene the exponential decay frequency to be d (the size of the real-part of the complex pole pair) then its value is

. Similarly, if we dene a second quantity, to be called the decayed natural frequency

!d

, to be the frequency of the decayed sinusoid, then its value is

!d = !n 1 ,
2 rad.s,1
and equation (45) may be simplied further to:

where is now tan,1  d =!d . Slide 94 shows the development of the second-order response from overdamped to oscillatory response as
is decreased from 2 to 0. The graph is plotted against normalized time !n t which means that the time-axis is independent of the natural frequency. For an actual second-order system, the response for a given damping ratio will be a scaled version of the response that is shown here.
1A

ct = 1 , !n e, d t cos!d t +  !d

(46)

full analysis of the step response for a general second-order system is left as an exercise.

4 EVALUATION OF SYSTEM RESPONSE

103

The Family of Step Responses for Second-Order Systems

c ( n t )
= 1 10
=1 4

Slide 94

=1 2

=1

=1

=0
= 3 2

=2

nt

Recap
In this lecture we have developed a general formula for the second-order transfer function

2 Gs = s+ 2
!!n + !2 s

in which the form of the response is determined by the damping ratio: 1. Overdamped when

response, the

1, two real poles at s = ,

!n  !n
2 , 1; 1, two complex conjugate poles at s = ,
!n  2. Underdamped when
p 2 j!n 1 ,
; 3. Undamped when
= 0, two imaginary poles at s = j!n ; 4. Critically damped when
= 1, two real and equal poles at s = ,
!n . The natural frequency !n governs the speed of response. For the underdamped
exponential decay frequency is d

=
!n s,1 p damped natural frequency is !d = !n 1 ,
2 rad s,1. The closed-loop poles are s = , d  j!d and the time response formula (45) can be simplied to ct = 1 , !n e, d t cos!d t +  !d
where

= tan,1 !d :
d

4 EVALUATION OF SYSTEM RESPONSE

104

In the next lecture we will develop some performance measures based on a typical underdamped response curve.

4 EVALUATION OF SYSTEM RESPONSE

Lecture 8: Evaluation of System Response (4)

105

Preamble
In the last lecture we developed a general formula for the second-order transfer function

!2 Gs = s2 + 2
!n s + !2
n

in which the form of the response is determined by the damping ratio: 1. Overdamped when
2. 3. 4.

1; Underdamped when
1; Undamped when
= 0; Critically damped when
= 1.

We also stated that the natural frequency !n governs the speed of response and dened the terms exponential decay frequency ( d ) and damped natural frequency (!d ). We also considered the location of the poles of the second-order system and discovered, for underdamped systems, s = , d  j!d . Finally, we developed a formula for the step response:

ct = 1 , p 1 2 e,
!n t cos!n 1 ,
2 t +  1,
!n e, d t cos! t +  = 1, ! d d

(47)

In this lecture, based on a typical underdamped response curve, we will develop equations for the performance measures, percentage overshoot, settling time, and rise time in terms of the generalised second-oder parameters.

4.5 The Specication of Second-Order Response

In order to specify a second-order response we need to dene some performance measures based on a typical underdamped response curve. The specications are: 1. Peak Time Tp : the time taken to reach the rst, or maximum, peak. 2. Percent overshoot OS : the the amount that the waveform overshoots the steady state, or nal value at the peak time, expressed as a percentage of the steady-state value. 3. Settling time Ts : the amount of time required for the transients damped oscillations to stay within 2 of the nal value. 4. Rise time Tr : the amount of time required to go from 10% to 90% of the nal value.

4 EVALUATION OF SYSTEM RESPONSE

106

Second-Order Response Specications

c(t)

cmax
%OS
Slide 95
100%

102% 98% 90%

2%

cfinal

Ts
10%

Tp

t
Tr

A summary of these specications is illustrated in Slide 95. Rise time, settling time and peak time yield information about the speed and quality of the transient response. They can help the designer achieve a given speed of response without excessive overshoot or oscillations. It should be noted that the last two specications (Tr and Ts ) are the same as those used for rst-order systems and they may also be used for overdamped and critically damped second-order systems too. In fact, these specications may also used for systems with order higher than two, provided that the response is of the same approximate shape. However, analytical formulae relating the timeresponse specication parameters to pole-and-zero locations can only be developed for second order systems. In the next sections, relationships between Tp , OS , Ts and Tr and !n and
are developed. Following that, they are related to the location of poles in the s-plane. 4.5.1 Evaluation of Tp In order to nd the value of the peak overshoot we must differentiate the time response with respect to time and then nd the maximum value. This can be done using Laplace transforms since

! Lfc_tg = sC s = s2 + 2
!n s + !2
n
Completing squares in the denominator gives:

n p!,
2 !n 1 ,
2 1 Lfc_tg = s +
! 2 + !2 1 ,
2  n n

4 EVALUATION OF SYSTEM RESPONSE

So

107

c_t = p !n 2 e,
!n t sin !n 1 ,
2 t: 1,
!n 1 ,
2t = n

Setting the derivative to zero gives

or

t=

Each value of n yields the value for a local maxima or minima. Letting n = 0 gives t = 0 which corresponds to the starting point of the step response. Setting n = 1 gives the time at which the response reaches its rst peak, that is Tp , thus:

!n 1 ,
2

n p

Tp =
4.5.2 Evaluation of OS

!n 1 ,
2

p

(48)

From Slide 95, the percent overshoot, OS , is given by

cmax is found by substituting t = Tp into Equation (47). Thus using (48) we have cmax = 1 , e

OS = cmax , c c
nal

nal

 100

(49)

p1,
2   ,
= p

cos  + p

= 1 + e,
= 1,
2 
For the unit step used in (47)

1 ,
2

sin 

!
(50)

nal

=1

(51)

Therefore, substituting (50) and (51) into (49) gives:

p OS = e,
= 1,
2   100 :

(52)

Note that the percentage overshoot is a function only of the damping ratio! The inverse of equation (52) allows us to nd the value of the damping ratio
that yields a given OS thus:

= q , lnOS=100 : 2 + ln2 OS=100

A plot of the relationship between OS and
is given in Slide 96.

(53)

4 EVALUATION OF SYSTEM RESPONSE

108

Relationship between

OS and

Slide 96

Percent overshoot, %OS

Damping ratio,

4.5.3 Evaluation of Ts

In order to nd the settling time we need to nd the time for which ct reaches and stays within 2 of c nal . From Slide 95, an estimate of Ts is the time for which the decaying sinusoid in equation (47) reaches an amplitude of 0.02, or

e,
!n t p 1 2 = 0:02: 1,

This is a conservative estimate because we are assuming that cos!n  = 1 at t = Ts. Nevertheless, solving for t gives

1 ,
2 t ,
(54)

2 Ts = , ln0:02 1 ,

!

p
n

It is easy to verify that the numerator of equation (54) yields values between 3.91 and 4.74 as
varies from 0 to 0.9. We therefore use an approximation for Ts which can be used for all values of
. The approximation is taken to be

4 Ts =
!
Evaluation of Tr

(55)

There is no precise analytical relationship between rise time and damping ratio or natural frequency. However, by experimentation a relationship can be found as shown in Slide 97.

4 EVALUATION OF SYSTEM RESPONSE

109

Damping ratio versus normalised rise time

Rise time x Natural Frequency, T
r

Slide 97

Damping Normalized ratio rise time 0.1 1.104 0.2 1.203 0.3 1.321 0.4 1.463 0.5 1.638 0.6 1.854 0.7 2.126 0.8 2.467 0.9 2.883

Damping Ratio,

For 0:866

0:5 an approximation to rise time can be taken to be Tr  1:8 !

n
(56)

This suggests that rise-time depends on !n , but as we shall see, this is rather a crude approximation and
must also be taken into account in practice. Nonetheless, we shall use the approximation in equation (56) from time to time. Example 4.6 Given the transfer function

100 Gs = s2 + 15s + 100

nd Tp , OS , Ts and Tr . Solution: From the model transfer function !n = 10 and
= 0:75. Substituting these values into the formulae for Tp , OS and Ts gives Tp = 0:475 seconds, OS = 2.838, and Ts = 0:533 seconds. From the graph in Slide 97 we nd that for
= 0:75, !n Tr  2:3 seconds. Dividing by !n gives Tr = 0:23 seconds. All without taking inverse-laplace transforms of the step response of Gs and measuring the response!

Recap
In this section we have developed formulae for the second-order system performance parameters

4 EVALUATION OF SYSTEM RESPONSE

Percentage overshoot (OS ) (52), (53) Time-to-peak (Tp ) (48) Settling time (Ts ) (55) Rise-time Tr (Fig. 97 and Eq. (56))

110

in terms of the generalized second-order parameters

and !n . Although the denitions of settling-time and rise time are the same as used for rst-order systems, percentage overshoot only applies to second-order systems which are underdamped or undamped. The same parameters may be used to categorise the performance of higher-order systems provided that they exhibit a response which is similar in shape to that of a second-order system. However, in that case, there is no direct relationship between the response parameters and the system poles. In the next lecture we shall conclude the development of performance characteristics by nding out how the location of the system poles can be used to specify a particular kind of performance.

4 EVALUATION OF SYSTEM RESPONSE

Lecture 9: Evaluation of System Response (5)

111

Preamble
In the last lecture we developed formulae for the second-order system performance parameters Percentage overshoot (OS ), Time-to-peak (Tp ), Settling time (Ts ), and Rise-time Tr , in terms of the generalized second-order parameters
and !n . In this lecture we shall conclude the development of performance charateristics by nding out how the location of the system poles can be used to specify a particular kind of performance.

4.6 Relating Response Specications to Pole Locations in the sPlane

We now have the relationships between the specications for peak-time, rise-time, settling-time and percent overshoot to the natural frequency and damping ratio of an underdamped second-order system. We now need to relate these to the locations of the poles on the s-plane. Once we have done this we will be in a position to specify the pole locations that yield a particular response. That is we shall be able to solve the so-called synthesis problem. The pole plot for an underdamped second-order system is shown again in Slide 98. It is easy to show that the radial distance from the origin to the pole is !n and the angle subtended by the radial line (measured from the negative real axis) is = cos,1
.

Pole-plot for an underdamped second-order system

s plane

j
+ j n 1 2

Slide 98

n
n

j n 1 2

Previously, when considering pole locations, we used the symbol !d (damped natural frequency) for the imaginary part of the pole and d (exponential decay

4 EVALUATION OF SYSTEM RESPONSE

frequency) for the real part of the pole. From the pole locations it is clear that

112

d =
!n
and Thus, in terms of pole locations

!d = !n 1 ,
2
d !n 1 ,
2 4 Ts =
! = 4 (58) n d Equation (57) shows that Tp is inversely proportional to the size of the imaginary part of the pole. Since horizontal lines on the s-plane are lines of constant imaginary part, then they are lines of constant peak-time. Similarly, equation (58) shows that Ts is inversely proportional to the size of the real part of the pole. Thus, vertical lines on the s-plane, which are lines of constant real part, are lines of constant settling-time. Further, since
= cos , radial lines are lines of constant
. Since percent overshoot is only a function of
, radial lines are thus lines of constant overshoot. Finally, the approximate equation !n = 1:8=Tr implies that curves of constant natural frequency (semicircles of radius = !n ) correspond to responses with constant rise- time. Again, Tr is inversely proportional to !n .

Tp =

p

 =!

(57)

Putting all this together we can show curves of constant peak time, settling time, rise time and percentage overshoot (Slide 99). 4.6.1 Effect of moving poles along design curves To illustrate the effect of the design parameters in greater detail the following slides show the effect of moving the poles along these curves. In Slide 100, the step responses for a system whose poles are moved keeping d constant are shown. As the poles move away from the real axis, the frequency !d increases, but the exponential envelope given by e, d t remains the same. So the settling time remains virtually the same although the percentage overshoot increases because the damping reduces. In Slide 101, the effect of moving the poles to the left with constant !d is shown. Now, the damped natural frequency remains the same so that Tp is constant. However, as the poles move further to the left, damping is increased so that percentage overshoot is reduced and the oscillations are damped out quicker.

4 EVALUATION OF SYSTEM RESPONSE

113

Curves of constant design parameters

%OS2

Tp2
Tp1
Slide 99

%OS1

Tr2 Tr1

Ts2

Ts1

Poles move with constant real part

j
c(t)
envelope the same motion of poles

s plane

Slide 100

4 EVALUATION OF SYSTEM RESPONSE

114

Poles move with constant imaginary part

j
c(t)
motion of poles

s plane

Slide 101

frequency same

Poles move with constant damping ratio

j
c(t)
Slide 102
overshoot same

s plane
motion of poles

4 EVALUATION OF SYSTEM RESPONSE

115

In Slide 102, the percentage overshoot remains the same as the poles move away from the origin along the radial line. The system becomes faster as the natural frequency is increased. Finally, in Slide 103 we see that the approximation Tr = 1:8=!n is not very precise. As the poles move together along the curve of constant !n the rise time varies quite by quite a large amount. This is because the damping ratio is changing also. Nevertheless, the relationship between speed of response and natural frequency for a given value of
can be obtained from design curves like that shown in the last lecture.

Poles move with constant natural frequency

j
c(t)
Slide 103
motion of poles

s plane

rise-time differs

4 EVALUATION OF SYSTEM RESPONSE

116

C = 0. 2 rad / Nm

R
(t )

q( t )
Figure 17: Rotational Mechanical System Example 4.7 The poles for a second-order systems are located at determine the system response. Solution: From trigonometry
= cos p frequency is given by !n = 32 + 72

s = ,3  j 7,

= costan,1 7=3 = 0:394. The natural = 7:616. The peak time is  Tp = ! =  = 0:449 seconds: d 7

The percent overshoot is

p OS = e,
= 1,
2   100 = 26:018

Ts = 4 = 4 = 1:333 seconds: d 3 For the rise time we see that !n Tr  1:44 when
= 0:394. Thus Tr  1:44=7:616 = 0:18 seconds.
Example 4.8 For the rotational mechanical system shown in mobility diagram form in Figure 17, determine the values of the bearing resistance R and inertia J required if the response of the angular position  to a step change in torque input of 1 N m is to have a 20% overshoot and a settling time of 2 seconds. Solution: Taking s to be the output we have

The approximate settling time is

1 Qs = Js2s + Rss + C s 1 Gs = s = Js2 + Rs + 1=C Qs 1=J = s2 + R=Js + 1=CJ

so the system transfer function is

4 EVALUATION OF SYSTEM RESPONSE

117

= q , ln0:2 = 0:46: 2 + ln2 0:2 For Ts = 2 seconds

!n = 4=2 = 2. Hence !n = 4:35 rad/s. s2 + 4s + 18:93 = s2 + R=Js + 1=CJ
Now, since C Nm/(rad/s).

For a 20 overshoot

The denominator of the system transfer function must then satisfy

= 0:2 5=J = 18:93 ! J = 0:26 kg m2. R=J = 4 ! R = 1:06

Example 4.9 Section objective example. We now put our knowledge to use to examine the performance of the open-loop antenna azimuth control system which can be used as a velocity control system. The block diagram is shown in Slide 104.

Open-Loop Velocity Control System

Slide 104
Vp ( s )
power amp Motor and load gearbox

100 s + 100

Va ( s )

2. 083 s + 1. 71

m ( s )

1 10

o ( s )

For this system: 1. Predict, by inspection the form of the open-loop angular velocity response of the load following a step change in the voltage input to the power amplier. 2. Find the damping ratio and natural frequency of the open-loop system. 3. Derive the complete analytical expression for the open-loop angular velocity response of the load to step-voltage input to the power amplier, using transfer functions. 4. Simulate, using Matlab, the step response of the open-loop system.

4 EVALUATION OF SYSTEM RESPONSE

Solution:

118

1. The transfer function relating angular velocity of the load to the voltage input to the power amplier is

20: Gs = s + 10083+ 1:71 s

The step response will be of the form

ct = A + Be,100t + Ce,1:71t

2. The damping ratio and natural frequency can be found by expanding the denominator of the transfer function

20 83 Gs = s2 + 101::71s + 171

and comparing it to s2 p response predicted above.

2 + 2
!n s + !n . Thus !n = 171 rad/s and
= 101:71=2 171 = 3:89. Thus the system is overdamped, conrming the
3.

o s = ss + 100s + 1:71

Expanding into partial fractions gives:

20:83

o s =

0:122 + 2:12  10,3 , 0:124 s s + 100 s + 1:71

Transforming to time domain gives

!o t = 0:122 + 2:12  10,3e,100t , 0:124e,1:71t

4. To get the step response in Matlab (shown in Fig. 18) >> num = 20.83; >> den = conv([1 100],[1 1.71]); % polynomial % multiplication step(num,den) title(Chapter Objective Problem) 41 Find the step response for each of the sytems with transfer function: 1. 2.

Gs = 5=s + 5 Gs = 20=s + 20

Also nd the time constant, rise time and settling time in each case.

4 EVALUATION OF SYSTEM RESPONSE

Chapter Objective Problem 0.14

119

0.12

0.1

Amplitude

0.08

0.06

0.04

0.02

0.5

1.5 Time (secs)

2.5

Figure 18: Satellite Azimuth Control System: Open-Loop Speed Response

vi ( t )

vo ( t )

42 A simple low-pass lter can be constructed from the RC circuit shown below. Show that the transfer function of such a lter is

s 1 Gs = Vos = sCR + 1 ; V

i
and determine the maximum frequency square wave that can be passed with a change in amplitude (vi to vo ) of less than 2%. 43 For each of the transfer functions shown below, nd the location of the poles and zeros, plot them on the s-plane, and then write down the expression for the general form of the step response without solving the inverse Laplace transform. State the nature of each response. 1. 2. 3. 4. 5.

2 Gs = s+2 5 Gs = s+3s+6 10s Gs = s+10+7 s+20

Gs = s2+620+144 s Gs = s+2 s+ 9

4 EVALUATION OF SYSTEM RESPONSE

6.

120

s+5 Gs = s+102

44 Determine the exact response for each of the systems in Problem 43 using Laplace transform techniques. 45 For each of the systems in Problem 43 nd the natural frequency and damping ratio, and hence conrm the type of response predicted. 46 A system has a damping ratio of 0.5, a natural frequency of 100 rad/s, and a dc gain of 1. Find its response to a unit step input. 47 For each of the second-order systems below, nd
, !n , Ts , Tp , Tr , and OS . 1. 2. 3.

Gs = 120=s2 + 12s + 120 Gs = 0:01=s2 + 0:002s + 0:01 Gs = 109=s2 + 6280s + 109 

48 For each of the second-order system specications below, nd the location of the second-order pair of poles. 1. 2.

OS = 10, Ts = 0:5 second. OS = 15, Tp = 0:25 second.

49 For a general second-order system, nd the location of the poles if the percentage overshoot is 30% and the settling time is 0.05 seconds. 410 Find J and C in the rotational mechanical system shown below if the system is to yield a 30 overshoot and a settling time of 4 seconds.

C
1 Nm / (rad / s)

(t )

q( t )

All problems except problem 41 adapted from Chapter 4 of Nise.

4 EVALUATION OF SYSTEM RESPONSE

121

Summary of the Evaluation of System Response Sequence

This section of the course was devoted to the evaluation of the transient response from the system model. It introduced the useful concept of poles and zeros. Step responses for rst and second order systems were analysed. The generalised parameter for the rst order system was found to be the time constant and the performance specications rise-time Tr and settling time Ts were related to . The generalised parameters for second-order systems were found to be damping ratio
and natural frequency !n ; Four different types of response are possible for stable second-order systems: overdamped when

1, critically damped when

= 1, underdamped when 0
1, undamped when
= 0.

The time response specications percentage overshoot; %OS, peak time Tp ; settling time Ts and rise-time Tr can be formulated in terms of
and !n . Curves of constant time response specications can be drawn in the s-plane and used to design a system which must have a given response. We also completed the objectives: Given the antenna azimuth-angle control system developed in the section on modelling Predict, by inspection, the form of the open-loop angular velocity response of the load to a step-voltage input to the power amplier; Find the damping ratio and natural frequency of the open-loop system; Derive the complete analytical expression for the open-loop response using transfer functions. In the next lecture we will return to consider the control system design problem and see how feedback enables us to satisfy some of the performance specications.

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

Lecture 10: Analysis and Design of Feedback Systems

122

Preamble
In the last section of the course we dened ways to evaluate the transient response of rst and second-order systems from knowledge of their transfer function. We now apply this knowledge to feedback control systems. In particular we shall introduce: Transfer functions for general feedback systems; The unity-gain feedback system, a so-called canonical form; The effect of gain on the system response of a system with the same form of transfer function as the antenna azimuth position control system; Introduce the root locus.

Analysis and Design of Feedback Systems

The feedback form of a control system topology is illustrated below in Fig. 19. A simplied model is shown in Figure 20.
Input transducer Controller
+
-

Plant

Rs
Input

G1 ( s)

E ( s)
Actuating signal

G2 ( s )

G3 ( s)

C( s)
Output

H2 (s )
Feedback

H1 ( s )
Output transducer

Figure 19: Feedback form of a control system topology Now, for the simplied system

E s C s C s 1 , GsH s C s Gc s = C s Rs

= = = = =

Rs , C sH s GsE s GsRs , GsH sC s GsRs Gs 1 + GsH s

(59)

The block diagram of this reduced closed-loop control system is shown in Fig. 21.

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

Plant + controller

123

Rs
Input

+
-

E (s )
Actuating signal

G( s )

C( s )
Output

H(s)
Feedback

Figure 20: Simplied feedback control system topology

Rs

G ( s) 1 + G ( s ) H (s )

C( s )

Figure 21: Reduced feedback control system

5.1 Interpretation of the generalised closed-loop transfer function

The components of equation (59) are interpreted as follows: The transfer function GsH s is called the loop transfer function.

1+GsH s = 0 is called the closed-loop characteristic equation (CLCE). As we shall see, the CLCE is a very important equation in feedback control systems analysis and design.

Gc s is called the closed-loop transfer function.

5.2 Unity-gain feedback
The unity-gain feedback canonical form1 is shown in Fig. 22.

Go s is called the open-loop transfer fuction. In comparison with the previous model, H s = 1, hence, from equation (59) we have: G Gc s = 1 + o ss (60) Go 
This form of system will be used almost exclusively in the remainder of this course.
1 canonical

authoritative, standard, accepted

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

124

Rs
Input

+
-

E ( s)
Error signal

Go ( s )

C( s )
Output

Figure 22: Unity-gain feedback canonical form

5.3 Closed-loop transient performance

Consider the example servomechanism shown in Figure 23.
Rs
+
-

K s( s + a)

C( s )

Figure 23: An example system: a position control system or servomechanism For this system:

G  K Gc s = 1 + o ss = s2 + as + K : Go
We see that this is a second order transfer function which will have various forms of damping depending on the value of K . As K is varied, the closed-loop poles move through three types of behaviour from overdamped, to critically-damped and on to underdamped response. At K = 0, the poles are the same as for the open-loop, that is p1;2 (labelled s1 in Fig. 24). For

Go s = ssK a +

= 0; ,a

0 K

a2=4

the poles are real and located at

(labelled s2 in Fig. 24).

2, p1;2 = , a  a 2 4K  2

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

125

As K increases, the poles move towards each other along the real axis and the response is overdamped (although the rise- and settling-times reduce), until both poles come together at p1;2 = ,a=2, when K = a2 =4, and the response is critically damped (s3 in Fig. 24). As K is further increased, the poles become complex with real part

d = ,a=2
and imaginary part

!d = 2 which increases in size as K increases (s4 in Fig. 24). The real-part remains constant whilst the damping ratio is reduced. Thus, OS inscreases whilst
the settling time remains constant. These results are summarised in Table 4, and the main classes of pole locations, labelled as above, are illustrated in Figure 24. Table 4: Pole locations for the example system as a function of K

p4K , a2

K 0 0 K a2 =4 a2=4 a2=4 K

p1 ,a p ,a=2 , a2 , 4K =2 ,a=2 p ,a=2 + j 4K , a2=2

p2
0

label

p ,a=2 + a2 , 4K =2 ,a=2 p ,a=2 , j 4K , a2=2

s1 s2 s3 s4

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

126

s plane

s4

s1
a

s2

s3
a 2

s2

s1

s4

Figure 24: Pole locations for example system

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

127

Example 5.1 Find Tp , OS and Ts for the feedback control system illustrated in Figure 25 below.
Rs
+
-

25 s ( s + 5)

C( s )

Figure 25:

Solution:

Hence !n

= 25 = 5, 2
!n = 5 and
= 0:5. Therefore Tp = p 2 = 0:726 sec: !n 1 ,

!2 25 Gc s = s2 + 5s + 25 = s2 + 2
!n s + !2
n

OS = e,
= 1,
2  100 = 16:303: 4 Ts =
! = 1:6 sec:
Example 5.2 For the system shown in Figure 26 below, design the gain K needed to give a OS of 10.
Rs
+
-

K s ( s + 5)

C( s )

Figure 26: Solution:

2
!n = 5, !n = K , thus

K Gc s = s2 + 5s + K
5
= p : 2 K

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

For a 10% overshoot

128

= q , ln10=100 = 0:591: 2 + ln210=100 K = 2

 5 2

= 17:892:

Note: for this system the settling time is Ts = 4=

!n  = 4=2:5 = 1:6 seconds. We cannot design for a settling time less that this since the real part of the poles is xed and is not adjustable by means of the gain K . We would need to add extra components to achieve a settling time less than 2 seconds.

5.4 The Root-Locus: A Preview

Consider the system of Figure 23 again. This has closed-loop transfer function

K Gcs = s2 + as + K :
We calculated the locations of the poles of Gc s for variations of K and plotted their locations in the s-plane (see Table 4 and Figure 24 above). We could plot two continuous curves through these points to indicate the movement of the poles as a continuous function of K (see Figure 27).
s plane

s1
a

s1
a 2

Figure 27: Motion of the closed-loop poles of the control system with open-loop transfer function Go s = K=ss + a These curves describe the locus of the closed-loop poles as K increases. This root locus can be sketched or generated for any system provided the open-loop transfer function is known. In Matlab, it is generated as shown in Slide 105 and the result is that shown in Slide 5.4.

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

129

Root Locus
For the system with open-loop transfer function Go s = 1=ss + 5 and unity-gain feedback: >> num_Go = [1]; >> den_Go = [1, 5, 0] % Go(s) = 1/(s2 + 5s)! >> rlocus(num_Go, den_Go)

Slide 105

Root Locus (2)

10 8 6 4

Imag Axis

Slide 106

2 0 -2 -4 -6 -8 -10 -10 -8 -6 -4 -2 0 Real Axis 2 4 6 8 10

x x

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

130

First Drill Problem Transient Response of Feedback Control Systems

Aim: Reduce a feedback control system to a single transfer function that relates the output to the input in order to analyse and design the closed-loop transient response. Problem: Given the block diagram for the antenna azimuth position control system with velocity feedback compensation shown in Figure 281 : 1. Find the equivalent unity-gain feedback canonical form by block diagram reduction and hence write down the open-loop transfer function Go s, the closed-loop transfer function Gc s and the closed-loop characteristic equation (CLCE). 2. Simplify the system by replacing the power amplier by its DC gain (=1) and disconnecting the tacho. Evaluate the closed-loop peak time Tp , percent overshoot OS and settling time Ts when the gain K = 800. 3. Derive the expression for the closed-loop step response of the system of part 2. 4. For the simplied system of part 2, nd the value of overshoot.

that yields a 5%

5. With the tacho connected, but the power-amplier still replaced by its DC gain, show that it is possible to design a closed-loop response in which the damping ratio and the settling time may be controlled by suitable choice of the parameters K and KT . Find suitable values for the parameters K and KT that will give a 10% overshoot and a settling time of 1.5 seconds. What is the corresponding rise time?

i s

1 Please

note the error in the gure. The input signal, shown as

s in the diagram should read

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

131

pot

pre-amp

power amp
+
-

motor & load

i (s )

Vi ( s )
+
-

Vp ( s )

100 s + 100

Va ( s )

0. 2083 s( s + 1. 71)

o (s)

Vo (s )

V ( s )
1
pot

sKT
tacho

Figure 28: Antenna azimuth position control system Answers 1.

2. 3. 4. 5.

:63K Go s = ss + 1:71s6+ 100 + 20:83K  ; T 6: Gc s = s3 + 101:7s2 + 1:7163K :83K s + 6:63K ; + 20 T CLCE ! s3 + 101:7s2 + 1:71 + 20:83KT s + 6:63K = 0: Tp = 0:435 seconds; OS = 69:1; Ts = 4:7 seconds. o t = 1 , e,0:855t cos 7:23t + 0:118 sin 7:23t: K = 23:2. K = 172:64; KT = 10:99; Tr  0:63 seconds.

Note: In the following, unless otherwise specied, assume that the closed-loop is obtained by the application of unity-gain feedback. 51 A control system has a controller and plant transfer function Gs and a feedback transfer function H s. For the closed-loop system, determine the characteristic equation and the closed-loop zeros in terms of the poles and zeros of Gs and H s for the cases where H s = 1 (unity-gain feedback canonical form) and H s = NH s=DH s. 52 An open-loop system has poles at s = 0; ,1 and ,6 and a zero at s = ,2. If the open-loop gain K = 18 nd the closed-loop step response and compare it to the open-loop step response. Comment on the result. (Hint: use Matlab to determine the zeros of the closed-loop characteristic equation.)

5 ANALYSIS AND DESIGN OF FEEDBACK SYSTEMS

53 A servomechanism has an open-loop transfer funtion

132

Go s = ss K 20 : +

Determine the value of K for which the closed-loop system is 1. critically damped, and 2. ideally damped. 54 A control system is to satisfy the following performance criteria 1. rise time: Tr 2.

 0:5 seconds. settling time: Ts  2 seconds.

OS 10.

3. Percent overshoot in the range 1

Sketch the region of the s-plane in which the dominant poles of the closedloop system must lie to meet these performance constraints and hence determine the closed-loop transfer function of a second-order system that satises the requirements. (Note: you should assume that Tr = 1:8=!n ).

Recap
In this lecture we have applied our knowledge of second-order system response to the analysis of closed-loop feedback control systems. We have seen that the openloop gain K in the transfer function effects the performance, and that by choice of suitable values for this parameter we can design closed-loop systems to have certain specied behaviour. However, there are limitations because gain typically can only be used to design one parameter, e.g. overshoot, rise-time, settling-time, and rarely can it be used to design two or more parameters simultaneously. We also observed that if we regard the open-loop gain as a continuously varying parameter we can create a locus of movement for the closed-loop poles of the feedback control system which we call the root locus. We will return to this in due course. Early in the course we stated that there are three important aspects to be considered in the design of control systems: stability, steady-state performance, and transient performance. We have spent quite some time on the third of these. In the next few lectures we turn our intention to the other, even more important, characteristics of stability and steady-state behaviour. We begin in the next lecture with a look at stability.

6 STABILITY
Lecture 11: Stability (1)

133

Preamble
Three requirements enter into the design of a feedback control system: transient response stability steady-state errors. So far, we have considered transient response. We now discuss stability. In this lecture we will: Dene stability for linear time-invariant systems and see that a system whose poles all lie to the left of the imaginary axis in the s-plane are stable; Dene instability and see that a system will be unstable if it has any poles to the right of the imaginary axis in the s-plane or it has poles of a certain structure on the imaginary axis; Dene marginal stability and see that a system is marginally stable if it has poles on the imaginary axis; Develop a test for stability which does not need the location of the system poles to be found.

Stability

Stability is the most important system specication. If a system is unstable, transient performance and steady-state errors are moot points. An unstable system cannot be designed for a specic transient response or a steady-state error requirement.

6.1 What is stability?

There are many denitions for stability depending upon the kind of system or ones point of view. In this section, we limit ourselves to a consideration of the stability of linear time- invariant systems (LTIs). Recall that the response of an LTI system is given by

ct = cforced t + cnatural t:

An LTI is stable if the natural response approaches zero as time approaches innity:

cnatural tjt!1 = 0 Only the forced response remains as t ! 1.

ctjt!1 = cforced t

6 STABILITY

134

An unstable system has a natural response that grows without bound, so that:

cnatural tjt!1 = 1
and therefore

ctjt!1 = 1

A marginally stable system has a natural response that neither grows nor decays as t ! 1 but either oscillates or remains at a constant value. Physically, an unstable system whose natural response grows without bound can cause damage to the system, adjacent property or human life. In practice many systems are designed with limit stops to prevent runaway. From the time-response point of view, instability is indicated by transients that get bigger and consequently by a total response that does not reach a steady state.

6.2 How do we determine if a system is stable?

6.2.1 Stability Recall from our study of system poles and zeros that poles to the left of the imaginary axis in the s-planea region called the left-half plane (LHP)yield responses that are either decaying exponentials or damped sinusoids. These natural responses decay to zero as time approaches innity. Thus: Closed-loop stability: a closed loop control system is stable if all the closed-loop poles are located in the left half plane. Example 6.1 Determine the stability of the closed-loop control system shown in Fig. 29.

R( s )

+ -

E( s)

3 s( s + 1)( s + 2 )

C( s )

Figure 29: Solution: The closed-loop transfer function is

Gc s = s3 + 3s23+ 2s + 3 : s3 + 3s2 + 2s + 3 = 0

The poles are the zeros (roots) of the closed-loop characteristic equation (CLCE)

That is:

s = ,2:672; ,0:164  j 1:047:

The pole-locations and the resulting response are illustrated in Slide 107.

6 STABILITY

135

Stable response
s plane
LHP

c(t)

Slide 107

6 STABILITY
6.2.2 Instability

136

A system which has all its poles in the LHP is stable: all the poles will be negative real or complex with negative real parts. On the other hand, poles in the righthalf plane are positive or complex with positive real parts. They produce responses which are increasing exponentials or increasing sinusoids. These grow without bound and hence yield unstable responses. Thus if there are poles in the right-half plane the system is unstable. Example 6.2 Determine the stability of the closed-loop control system shown in Fig. 30.

R( s )

+ -

E( s)

7 s( s + 1)( s + 2 )

C( s )

Figure 30: Solution: This time the closed-loop transfer function is

Gc s = s3 + 3s27+ 2s + 7
but now the poles are:

s = ,3:087; +0:0434  j 1:505:

Thus, two poles are in the right-half plane (RHP) and the resulting response is unstable (Slide 108). If there are double or triple poles on the imaginary axis the response will be of the form Atn cos!t +  n = 1; 2; : : :. Such responses also grow without bound since clearly tn ! 1 when t ! 1. Thus: Instability of a closed-loop control system: A system is unstable if its closed-loop transfer function has at least one pole in the right-half plane and/or poles of multiplicity 1 on the imaginary axis.

6 STABILITY

137

Unstable response
s plane
LHP

c(t)

Slide 108

6 STABILITY
6.2.3 Marginal stability

138

Finally, if a system has a single pair of poles on the imaginary axis, or a single pole at the origin, then we say that the system is marginally stable. It will have a natural response containing an undamped oscillation or a constant value as t ! 1.

6.3 Testing for Stability

It would seem to be a simple matter to test for stability. All we need to know is that all the closed-loop poles are in the LHP. But, in fact, it is not so simple as it at rst appears to be. Example 6.3 Consider the closed-loop system shown in Fig. 31. Is the system stable?

R( s )

+ -

E( s)

10 ( s + 2) s( s + 4 )( s + 6)( s + 8 )( s +10 )

C( s )

Figure 31: Solution: The closed-loop transfer function is:

10s + 2 Gc s = s5 + 28s4 + 284s3 + 1232s2 + 1390s + 20 :

Although we know the location of the open-loop poles, this doesnt help us nd the the location of the closed-loop poles. Neither is there an analytical way of nding the roots of the CLCE. We could use a root nding algorithm such as >> roots([1, 28, 284, 1232, 1930, 20]) in Matlab (maybe your calculator has a similar function). However, not everyone always has access to such facilities so we need another way to determine the sorts of roots that a system characteristic equation has. Our requirements are simplied because we only need to test the sign of the real part of the poles rather than nding their location. Namely, we need to answer the question are there any roots of the CLCE which have positive real part? and, more pertinant perhaps is there a simple method of checking that all the roots of a polynomial have negative real part without resorting to root-nding algorithms? The answer to the last question is yes. It is called the Hurwitz test and we present it now.

6.4 The Hurwitz Criterion

Consider a closed-loop characteristic polynomial in factored form:

P s =

n Y

i=1

s , pi 

6 STABILITY

139

where pi is a zero of P s (that is a root of the CLCE). If all poles are in the LHP, then all factors will be of the form s + pi  (because each pi will have a negative real part). The coefcients of the expanded polynomial will therefore only have positive sign. This is true even if some factors pi are complex

, + j!
because complex factors always appear as conjugate pairs so that

s + + j!s + , j! = s2 + 2 s +

2 + !2

which also has positive coefcients. No coefcents can be missing since this would imply cancellation between positive and negative roots, or roots on the imaginary axis, which we do not allow. Thus a necessary condition for a system to be stable is for its CLCE to have all coefcients of every term sn ; sn,1 ; : : : ; s0 present and all positive1. To be more formal: The Hurwitz Criterion: The Hurwitz criterion states that a system is unstable if there are any negative or missing coefcients of the closedloop characteristic equation. The test of stability, called the Hurwitz Test is very simple to apply: negative;

s3 + 27s2 , 26s + 24 is unstable because the coefcient of the s term is s3 + 27s2 + 26s is unstable because the s0 term is missing; but s3 + 27s2 + 26s + 24 may be stable.

The problem with the Hurwitz criterion is that a system that passes the Hurwitz test is not guaranteed to be stable, as in the nal example above. Hence we call the Hurwitz criterion a necessary but not sufcient criterion for stability. The Hurwitz Test provides a useful means of rejecting some closed-loop systems, but we need a stronger test to be able to say for denite that a system that passes the Hurwitz test is actually stable. This more stringent test is based on the Routh-Hurwitz criterion and is called the Routh Test. It is based on the so-called Routh Array which is constructed from the coefcients of the characteristic polynomial and it, and the test based on it, are described in the next lecture.

Recap
In this lecture we have dened stability for linear time-invariant (LTI) systems. An LTI system is stable if the natural response cnatural t decays to zero as t ! 1. In terms of system transfer functions, a stable system will only have poles in the s-plane which lie to the left of the imaginary axis. This region of the s-plane is called the left-half-plane (LHP).
1 or

all negative which implies multiplication of P

s by ,1

6 STABILITY

140

An LTI system will be unstable if, for any reason, the natural response cnatural t does not decay to zero as t ! 1. The conditions under which this occurs are: when the system has at least one pole to the right of the imaginary axis in the s-plane (the right-half-plane), or the system has poles of multiplicity

1 on the imaginary axis. An LTI system is marginally stable if it has poles of multiplicity = 1 on the
imaginary axis. Such a system will have a natural response that is oscillatory (undamped) or results in a constant value. In order to test for stability we need to examine the sign of the poles of the characteristic equation. Any system that only has poles in the left-half-plane must produce a characteristic polynomial that has positive coefcients. The Hurwitz test makes use of this property so that any system whose characteristic polynomial has any negative or missing coefcients must be unstable. However, the Hurwitz criterion is only a necessary condition, since polynmials that pass the Hurwitz test can still have roots in the right-half-plane. We therefore need a necessary and sufcient test. Such a test is based on the Routh-Hurwitz criterion and is called the Routh Test. It is based on the so-called Routh Array which is constructed from the coefcients of the characteristic polynomial and it, and the test based on it, are described in the next lecture.

6 STABILITY
Lecture 12: Stability (2)

141

Preamble
In the last lecture we dened stability for linear time-invariant (LTI) systems and showed how knowledge of the location of the system poles on the s-plane could tell us if a system was stable or not. However, in order to know the location of the poles, we need to nd the roots of the closed-loop characteristic equation, and this may not be possible unless we have access to a computer or sophisticated calculator. It turned out, however, that in order to judge a systems stability we dont need to know the actual location of the poles, just their signthat is whether the poles are in the right-half or left-half plane. The Hurwitz criterion can be used to indicate that a characteristic polynomial with negative or missing coefcients is unstable. However, a polynomial that passes the Hurwitz test may still have unstable roots. In this lecture we demonstrate the Routh-Hurwitz Crterion which can be used to test the stability of polynomials that pass the Hurwitz test. The Routh-Hurwitz Criterion is called a necessary and sufcient test of stability because a polynomial that satises the criterion is guaranteed to stable. The criterion can also tell us how many poles are in the right-half plane or on the imaginary axis. Having obtained a test for stability that is valid for all cases we can use it in design to ensure that, for example, the system gain is never sufcient to cause instability. Using a stability test in design in this way is called design for relative stability.

6.5 The Routh-Hurwitz Stability Criterion

To use the Routh test we rst need to construct a Routh array. We present the construction of a typical Routh array by means of an example, then present the Routh-Hurwitz Criterion. 6.5.1 The Routh array Consider the system shown in Fig. 32. The closed-loop characteristic equation is:

a4s4 + a3 s3 + a2s2 + a1 s + a0 = 0:

(61)

R( s )

N( s) 4 3 a4 s + a3 s + a2 s2 + a1 s + a0

C( s)

Figure 32: General closed-loop transfer function for a fourth-order system

6 STABILITY

142

s in the closed-loop characteristic polynomial. In this case there are 5 rows, one for each term s4 ; s3 ; : : : ; s0 but in general there will be n + 1 rows for a system with a characteristic polynomial whose highest power of s is n. There is a relationship between the rows and the powers of s which we will make use of later, so to make
the relationship explicit, it is useful to label the rows of the Routh array as shown below.

The Routh array is simply a rectangular matrix with one row for each power of

s4 : s3 : s2 : s1 : s0 :

0 B B B B @

::: ::: ::: ::: :::

1 C C C C A

The next operation is to initialise the Routh array by lling the rst two rows with the coefcients of the characteristic polynomial as follows: 1. For row s4 the rst element is a4 the coefcient of s4 . The next element is a2 the coefcient of s2 and the nal element is a0 the coefcient of s0. Thus the rst row is:

s4 : a4 a2 a0 :

2. For the second row s3 in this case and sn,1 in general the elements are simply those coefcients skipped in the second row written down in order. Thus:

Note that in fact what we are doing is skipping the odd coefcients of the characteristic polynomial when constructing the row for s4 . In general the rule is: if n is even then the sn row consists of the coefcients of the even powers of s. If n is odd, then the sn row consists of the coefcients of the odd powers of s.

s3 : a3 a1 0

(where the nal zero is added to make the two rows equal in length). At the end of initialisation, the Routh array for a fourth-order system will be as shown in Table 5 (where we now use a table for clarity in what follows). Table 5: Starting layout for Routh array

s4 s3 s2 s1 s0

a4 a2 a0 a3 a1 0

The next stage in the process is to ll in the remaining rows of the table. This is done by operating on the two initial rows to create row 3, applying the same operations on rows 2 and 3 to generate row 4, and so on until all rows are complete. The operations that are applied to do this are now described.

6 STABILITY

143

1. Starting at the left-most end of the array create a 2  2 matrix from the rst two rows such that the rst column consists of the rst element of row 1 and 2 and the second column contains the second elements of row 1 and 2. Lets call this matrix 1 . That is, for Table 5 we have:

a a R1 = 4 2
a3 a1

The rst element of the third row of the Routh array, well call it b1 , is then given by

b1 = , detR1 =R11; 2:

In other words:

, a4 a2 a3 a1 b1 = a3 = ,a4a1 , a2a3
a3 a2a3 , a4 a1 = a3
2. The second element of row 3, b2 , is calculated in much the same way. 2 is obtained by leaving column 1 as it is and replacing column 2 with the elements of column 3. Since column 1 is the same as in 1 , the divisor 2 1; 2 is also unchanged so that:

b2 =

a3 = a0 a3 , a4  0 a3

, a4 a00 a3

3. This process is continued until the determinant is zero, after which zero is put into all remaining columns of row 3. 4. The whole process is repeated for the remaining rows except that the top left element of the matrix moves down one row at the start of each new row. That is, row 4 is created from rows 2 and 3 (using the odd elements of P s and the newly calculated coefcients b1 , b2 , etc.), and row 5 is constructed from rows 3 and 4. At the end of the process the array looks like that shown in Table 6. Note: you may nd it easier to remember that , det  is given by the product of the off diagonal terms , the product of the diagonal terms. Once the Routh array is complete we are ready to perform the Routh test.

6 STABILITY

144

Table 6: Completed Routh array

s4 s3 s2 s1 s0

a4 a3

a2 a1
3 3

a0 0
3

, b1 = a2 a3 a a4a1 , c1 = a1 b1 b a3b2
1 1

, b2 = a0a3 , a4  0 = a0 b3 = 0  a3 a a4  0 = 0 a
c2 = 0  b1 , a3  0 = 0 c3 = 0  b1 , a3  0 = 0 b b
1 1 1 1

, , d1 = b2  c1c, b1  0 = b2 d2 = 0  c1 c b1  0 = 0 d3 = 0  c1 c b1  0 = 0

6.5.2 The Routh-Hurwitz Test The Routh-Hurwitz criterion enables us to test a system whose closed- loop characteristic equation passes the Hurwitz criterion. The Routh-Hurwitz Criterion: The number of roots of the characteristic polynomial that are in the right-half plane is equal to the number of sign changes in the rst column of the Routh Array. If there are no sign changes, the system is stable. Example 6.4 Test the stability of the closed-loop system shown in Fig. 33.

R( s )

+ -

E( s)

1000 ( s + 2 )(s + 3 )(s + 5)

C( s )

R( s )

1000 3 2 s + 10s + 31s + 1030

C( s )

Figure 33: Solution: Since all the coefcients of the closed-loop characteristic equation s3 + 10s2 + 31s + 1030 are present, the system passes the Hurwitz test. So we must construct the Routh array in order to test the stability further. Here is the opening Routh array:

6 STABILITY

145

s3 s2 s1 s0

1 31 0 10 1030 0

The Routh array is unchanged by the multiplication of any row by a constant. This can be used to simplify the Routh array between stages. For example, in the second row above can be simplied by multiplying each term by 1=10, thus:

s3 s2 s1 s0 s3 s2 s1 s0
1 1

1 31 0 1 103 0

We now construct the rest of the table using the rules already described:

31 103 01,10 =0 1

31  1 , 1  103 = ,72 1

,72  103 , 1  0 = 103 ,72  0 , 1  0 = 0 ,72 ,72

For clarity, we can rewrite the array:

0 1 31 0 1 B 1 103 0 C B ,72 0 0 C @ A
103 0 0

and now it is clear that column 1 of the Routh array is:

0 11 B 1C B ,72 C @ A
103

and it has two sign changes (from 1 to ,72 and from system is unstable with two poles in the right-half plane.

,72 to 103).

Hence the

6.6 Special Cases

Two special cases can occur when constructing a Routh array: a zero may appear in the rst column of the array; a complete row can become zero.

6 STABILITY
6.6.1 A Zero in the rst column

146

If the rst element of a row is zero, division by zero would be required to form the next row. To avoid this, a small number (epsilon) is used to replace the zero in the rst column. The value is allowed to approach zero from either the positive or negative side, after which the signs of the entries in the rst column can be determined. Example 6.5 Consider the control system with closed-loop transfer function

Gc s = s5 + 2s4 + 3s310 6s2 + 5s + 3 : +

Construct the Routh array and interpret the stability of the closed-loop system. Solution: The characteristic polynomial is P s = s5 + 2s4 + 3s3 + 6s2 + 5s + 3 so the Routh array will be:

s5 s4 s3 s2 s1 s0
Label

0! 6 ,7 42 , 49 , 6 2 12 , 14 3
First column

1 2

3 5 6 3 7=2 0 3 0 0 0 0 0

Considering just the sign changes in column 1:

s5 s4 s3 s2 s1 s0

1 2

! 0+
+ + +

! 0,
+ +

,
+ + +

6 ,7 42 , 49 , 6 2 12 , 14 3

,
+ +

If is chosen positive there are two sign changes. If is chosen negative there are also two sign changes. Hence the system has two poles in the right-half plane and it doesnt matter whether we chose to approach zero from the positive or the negative side. This is always the case!

6 STABILITY
6.6.2 An Entire row is zero

147

This occurs when there is an even polynomial that is a factor of the original polynomial. Example 6.6 Construct the Routh array for the system with closed-loop transfer function:

Gc s = s5 + 7s4 + 6s3 1042s2 + 8s + 56 : + s5 s4 s3 s2 s1 s0

7 ! 1 42 ! 6 56 ! 8 0 0 0 1 6 8

Solution: The Routh array for the example is:

We cannot procede any further because the third row is zero. In order to procede we have to take the so-called auxillary polynomial Qs formed from the row preceeding the zero row:

Qs = 7s4 + 6s2 + 8:

The auxillary polynomial is even (only even powers of s are present) and it is also a factor of the original characteristic polynomial (as you may easily verify for yourself). To progress further, we have to differentiate Qs with respect to s:

dQs = 4s3 + 12s + 0 ds

and replace the zero row with a row formed from the coefcients of the derivative:

s5 s4 s3 s2 s1 s0

1 6 8 1 6 8 0 ! 4 ! 1 0 ! 12 ! 3 0

The remainder of the Routh array is constructed as usual.

s5 s4 s3 s2 s1 s0

1 1 1 3 1=3 8

6 6 3 8 0 0

8 8 0 0 0 0

There are no sign changes in the completed Routh array, hence the system is stable.

6 STABILITY
6.6.3 Interpretation of a row of zeros

148

An entire row of zeros will appear in the Routh array when a purely even polynomial is a factor of the characteristic polynomial. For example the polynomial s4 +5s2 +7 is a purely even polynomial; it only has even powers of s. Even polynomials have roots which are symmetrical about the imaginary axis. This symmetry can occur under several conditions: 1. the roots are real and symmetric about the imaginary axis; 2. the roots are imaginary and symmetric about the real axis, or 3. the roots are quadrantal. These cases are illustrated in Fig. 34. each case, or any combination of these cases, will generate an even polynomial.

j
B C C A

C B

A: Real and symmetrical about the origin B: Imaginary and symmetrical about the origin C: Quadrantal and symmetrical about the origin
Figure 34: Root positions to generate even polynomials: Patterns A, B or C (or any combination) It is the even polynomial that causes a row of zeros to appear in the Routh array. Thus the row of zeros tells us that there are roots that are symmetric about the origin. Some of these roots could be on the imaginary axis (symmetry type B). On the other hand, if we do not have a zero row, we cannot possibly have roots on the j! axis. Another characteristic of the Routh array for the case in question is that the row previous to the row of zeros contains the even polynomial that is a factor of the original polynomial. Thus in the previous example, the polynomial s4 + 6s2 + 8 is a factor of the original polynomial. Finally, the Routh test from the row containing

6 STABILITY

149

the even polynomial down to the last row of the Routh array tests only the poles in the even polynomial. A further example should clarify all this. Example 6.7 For the system with closed-loop transfer function

nd how many poles are in the right-half plane, the left-half plane and on the j! axis. Solution: Construct the Routh array shown below.

Gc s = s8 + s7 + 12s6 + 22s5 + 3920 + 59s3 + 48s2 + 38s + 20 s4

s8 s7 s6 s5 s4 s3 s2 s1 s0

20 0 ,10 ! ,1 ,20 ! ,2 10 ! 1 20 ! 2 0 20 ! 1 60 ! 3 40 ! 2 0 0 1 3 2 0 0 0 0 0 0 0

1 1

12 22

39 59

48 38

For convenience the s6 row has been multiplied by 1=10 and the s4 row by 1=20. There is a complete row of zeros at the s3 row. Moving back to the s4 row, we extract the even polynomial

Qs = s4 + 3s2 + 2
and take its derivative:

pleted:

dQs = 4s3 + 6s + 0: ds The zero row is then replaced by 4; 6; 0 = 2; 3; 0 and the Routh array is com-

s8 s7 s6 s5 s4 s3 s2 s1 s0

1 12 39 48 20 1 22 59 38 0 ,1 ,2 1 2 0 1 3 2 0 0 1 3 2 0 0 2 3 0 0 0 3=2 ! 3 2 ! 4 0 0 0 1=3 0 0 0 0 4 0 0 0 0

6 STABILITY
Interpretation

150

Since all the entries from the even polynomial from the s4 to the s0 entry are a test of the even polynomial Qs, we can draw our rst conclusions about this polynomial. There are no sign changes from the s4 to the s0 row, so there are no poles in the right-half-plane (this rules out poles that are symmetric about the real axis or are quadrantal). But since there must be some symmetric poles, there must be 4 poles on the j! -axis. The remaining roots are evaluated from the remaining rows of the Routh array. There are two sign changes hence there are two RHP poles. The remaining poles must be in the LHP. To summarise: Even (4th-order) 0 RHP 0 LHP 4 j! Rest (4th-order) 2 RHP 2 LHP 0 j! Total (8th- order) 2 RHP 2 LHP 4 j!

Drill Problems
Each of the following transfer functions is the open-loop transfer function for a control system with unity-gain feedback. In each case, construct the Routh array for the closed-loop characteristic-polynomial and comment on the stability, and where appropriate, on the location of the closed-loop poles in the s-plane. 1.

2000 Go s = ss3 + 6s2 + 11s + 6

2.

Go s = s2s4 + 3s3 +12s2 + 3s + 2

3.

128 Go s = ss7 + 3s6 + 10s5 + 24s3 + 96s2 + 128s + 192

6.7 Use of the Routh-Hurwitz Criterion for Control Systems Design
To conclude this section, we examine the use of the Routh-Hurwitz criterion in the design of feedback control systems. Example 6.8 For the closed-loop system illustrated in Fig. 35, nd the range of values of the gain parameter K for which the closed-loop system is stable. Solution: The closed-loop transfer function is

Gc s = s3 + 18s2K 77s + K : +

The Routh array is:

6 STABILITY

151

R( s )

+ -

E( s)

K s( s + 7)(s + 11)

C( s )

Figure 35:

s3 s2 s1 s0

1 18 1386 , K 18 K

77 K 0 0

If K is assumed to be positive then we can use the Routh-Hurwitz criterion to set limits on the design values of K allowed for stable operation. There will be no sign changes if K 0 or if 1386 , K 0, that is if K 1386. If K 1386 there will be two sign changes so there will be two poles in the RHP and the system will be unstable. If K = 1386 then the s1 row will be zero. The previous row is Qs = 18s2 + 1386 and the derivative is 36s so the new Routh array will be:

s3 s2 s1 s0

77 K ! 1386 0 ! 36 0 1386 0

1 18

For the even polynomial Qs there are no sign changes from s1 to s0 so there must be two imaginary roots and the system is marginally stable. Example 6.9 For the antenna azimuth control problem illustrated in Fig. 36 the closed-loop transfer function is

Find the range of pre-amplier gains K for which the closed-loop system is stable. Solution: 0

Gc s = s3 + 101:71s62:63K s + 6:63K : + 171

K 2623:29.

6.8 Relative Stability

The ability to test a control system for stability against some parameter is very useful, but for design purposes we are not so much interested in absolute stability as designing a control system so that it is far removed from instability.