Visit https://ebooknice.

com to download the full version and

explore more ebooks

(Ebook) Digital control and state variable

methods : conventional and intelligent control
systems by M. Gopal ISBN 9780070668805,
0070668809

_____ Click the link below to download _____

https://ebooknice.com/product/digital-control-and-
state-variable-methods-conventional-and-intelligent-
control-systems-10945070

Explore and download more ebooks at ebooknice.com

Here are some suggested products you might be interested in.
Click the link to download

(Ebook) Digital control and state variable methods : conventional and

intelligent control systems by M. Gopal ISBN 9780071333276, 0071333274

https://ebooknice.com/product/digital-control-and-state-variable-
methods-conventional-and-intelligent-control-systems-10945072

(Ebook) Digital Control and State Variable Methods by M. Gopal ISBN

9780070483026, 0070483027

https://ebooknice.com/product/digital-control-and-state-variable-
methods-5186136

(Ebook) Control Systems: Principles and Design by M. Gopal ISBN

9780070668799, 0070668795

https://ebooknice.com/product/control-systems-principles-and-
design-23300446

(Ebook) Intelligent Systems: Modeling, Optimization, and Control

(Automation and Control Engineering) by Yung C. Shin, Chengying Xu
ISBN 9781420051766, 1420051768

https://ebooknice.com/product/intelligent-systems-modeling-
optimization-and-control-automation-and-control-engineering-2021666
(Ebook) Product Engineering: Tools and Methods Based on Virtual
Reality (Intelligent Systems, Control and Automation: Science and
Engineering) by Doru Talaba, Angelos Amditis ISBN 9781402081996,
1402081995
https://ebooknice.com/product/product-engineering-tools-and-methods-
based-on-virtual-reality-intelligent-systems-control-and-automation-
science-and-engineering-1742748

(Ebook) Intelligent Control, Filtering and Model Reduction Analysis

for Fuzzy-Model-Based Systems: 385 (Studies in Systems, Decision and
Control, 385) by Xiaojie Su, Yao Wen, Yue Yang, Peng Shi ISBN
9783030812133, 3030812138
https://ebooknice.com/product/intelligent-control-filtering-and-model-
reduction-analysis-for-fuzzy-model-based-systems-385-studies-in-
systems-decision-and-control-385-34085372

(Ebook) Variable Gain Control and Its Applications in Energy

Conversion (Automation and Control Engineering) by Chenghui Zhang, Le
Chang, Cheng Fu ISBN 9781032491844, 1032491841

https://ebooknice.com/product/variable-gain-control-and-its-
applications-in-energy-conversion-automation-and-control-
engineering-49163108

(Ebook) State-Space Approaches for Modelling and Control in Financial

Engineering: Systems theory and machine learning methods by Gerasimos
G. Rigatos (auth.) ISBN 9783319528656, 9783319528663, 3319528653,
3319528661
https://ebooknice.com/product/state-space-approaches-for-modelling-
and-control-in-financial-engineering-systems-theory-and-machine-
learning-methods-5882724

(Ebook) Intelligent Systems in Cybernetics and Automation Control

Theory by Radek Silhavy, Petr Silhavy, Zdenka Prokopova ISBN
9783030001841, 3030001849

https://ebooknice.com/product/intelligent-systems-in-cybernetics-and-
automation-control-theory-7319892
Digital control and state variable methods conventional
and intelligent control systems 3 ed. Edition M. Gopal
Digital Instant Download
Author(s): M. Gopal
ISBN(s): 9780070668805, 0070668809
Edition: 3 ed.
File Details: PDF, 56.28 MB
Year: 2009
Language: english
THIRD EDITION
About the Author

M Gopal is presently Professor, Department of Electrical Engineering, Indian

Institute of Technology, Delhi. His teaching and research stints span over three
decades at prestigious institutes.
Dr Gopal is the author/co-author of six books. His Video Course on Control
Engineering is being transmitted periodically through EKLAVYA Technology
Channel of the IITs. He is author of an interactive web compatible multimedia
course on Control Engineering, available on the site: [email protected]
Dr Gopal has a large number of research publications to his credit. His current
research interests are in the areas of Machine Learning, Soft-Computing
Technologies, and Intelligent Control.
 THIRD EDITION

M GOPAL
Professor
Department of Electrical Engineering
Indian Institute of Technology
Delhi

Tata McGraw-Hill Publishing Company Limited

NEW DELHI
McGraw-Hill Offices
New Delhi New York St Louis San Francisco Auckland Bogotá Caracas
Kuala Lumpur Lisbon London Madrid Mexico City Milan Montreal
San Juan Santiago Singapore Sydney Tokyo Toronto
Published by the Tata McGraw-Hill Publishing Company Limited,
7 West Patel Nagar, New Delhi 110 008.

Copyright © 2009, by the Tata McGraw-Hill Publishing Company Limited.

No part of this publication may be reproduced or distributed in any form or by any means, electronic, mechanical,
photocopying, recording, or otherwise or stored in a database or retrieval system without the prior written permission of
the publishers. The program listings (if any) may be entered, stored and executed in a computer system, but they may
not be reproduced for publication.

This edition can be exported from India only by the publishers,

Tata McGraw-Hill Publishing Company Limited

ISBN-13: 978-0-07-0668805
ISBN-10: 0-07-0668809

Managing Director: Ajay Shukla

General Manager : Publishing—SEM & Tech Ed: Vibha Mahajan
Sponsoring Editor: Shukti Mukherjee
Editorial Executive: Suman Sen
Executive—Editorial Services: Sohini Mukherjee
Junior Manager—Production: Anjali Razdan

General Manager : Marketing—Higher Education & School: Michael J Cruz

Product Manager : SEM & Tech Ed : Biju Ganesan

Controller—Production: Rajender P Ghansela

Asst General Manager—Production: B L Dogra

Information contained in this work has been obtained by Tata McGraw-Hill, from sources believed to be reliable.
However, neither Tata McGraw-Hill nor its authors guarantee the accuracy or completeness of any information
published herein, and neither Tata McGraw-Hill nor its authors shall be responsible for any errors, omissions, or
damages arising out of use of this information. This work is published with the understanding that Tata McGraw-Hill
and its authors are supplying information but are not attempting to render professional services. If such services are
required, the assistance of an appropriate professional should be sought.

Typeset at Script Makers, 18, DDA Market, A-1B Block, Paschim Vihar,
New Delhi 110063 and printed at Unique Color Carton, 15/72, Punjabi Bagh, New Delhi-110 026.

Cover: Rashtriya Printers

RAXYYRYFDZZDZ
 Dedicated
with all my love to my
son Ashwani
and
daughter Anshu
 Contents

Preface XI
Part I Digital Control: Principles and Design in Transform-Domain 1
1. Introduction 3
1.1 Control System Terminology 3
1.2 Computer-Based Control: History and Trends 9
1.3 Control Theory: History and Trends 12
1.4 An Overview of the Classical Approach to Analog Controller Design 14
1.5 Scope and Organization of the Book 19
2. Signal Processing in Digital Control 21
2.1 Why Use Digital Control 21
2.2 Configuration of the Basic Digital Control Scheme 23
2.3 Principles of Signal Conversion 24
2.4 Basic Discrete-Time Signals 30
2.5 Time-Domain Models for Discrete-Time Systems 32
2.6 The z-Transform 41
2.7 Transfer Function Models 52
2.8 Frequency Response 58
2.9 Stability on the z-Plane and the Jury Stability Criterion 61
2.10 Sample-and-Hold Systems 69
2.11 Sampled Spectra and Aliasing 72
2.12 Reconstruction of Analog Signals 77
2.13 Practical Aspects of the Choice of Sampling Rate 80
2.14 Principles of Discretization 82
Review Examples 100
Problems 107
3. Models of Digital Control Devices and Systems 113
3.1 Introduction 113
3.2 z-Domain Description of Sampled Continuous-Time Plants 115
3.3 z-Domain Description of Systems with Dead-Time 122
3.4 Implementation of Digital Controllers 126
3.5 Tunable PID Controllers 133
3.6 Digital Temperature Control System 146
3.7 Digital Position Control System 151
3.8 Stepping Motors and their Control 157
3.9 Programmable Logic Controllers 163
Review Examples 180
Problems 183
 viii Contents
4. Design of Digital Control Algorithms 193
4.1 Introduction 193
4.2 z-Plane Specifications of Control System Design 196
4.3 Digital Compensator Design using Frequency Response Plots 212
4.4 Digital Compensator Design using Root Locus Plots 224
4.5 z-Plane Synthesis 237
Review Examples 242
Problems 246

Part II State Variable Methods in Automatic Control: Continuous-Time and

Sampled-Data Systems 253
5. Control System Analysis Using State Variable Methods 255
5.1 Introduction 255
5.2 Vectors and Matrices 256
5.3 State Variable Representation 266
5.4 Conversion of State Variable Models to Transfer Functions 276
5.5 Conversion of Transfer Functions to Canonical State Variable Models 282
5.6 Eigenvalues and Eigenvectors 293
5.7 Solution of State Equations 303
5.8 Concepts of Controllability and Observability 314
5.9 Equivalence between Transfer Function and State Variable Representations 325
5.10 Multivariable Systems 329
Review Examples 334
Problems 341
6. State Variable Analysis of Digital Control Systems 351
6.1 Introduction 351
6.2 State Descriptions of Digital Processors 352
6.3 State Description of Sampled Continuous-Time Plants 359
6.4 State Description of Systems with Dead-Time 364
6.5 Solution of State Difference Equations 367
6.6 Controllability and Observability 372
6.7 Multivariable Systems 376
Review Examples 379
Problems 386
7. Pole-Placement Design and State Observers 393
7.1 Introduction 393
7.2 Stability Improvement by State Feedback 394
7.3 Necessary and Sufficient Conditions for Arbitrary Pole-Placement 397
7.4 State Regulator Design 400
7.5 Design of State Observers 404
7.6 Compensator Design by the Separation Principle 413
7.7 Servo Design: Introduction of the Reference Input by Feedforward Control 418
7.8 State Feedback with Integral Control 420
 Contents ix
7.9 Digital Control Systems with State Feedback 422
7.10 Deadbeat Control by State Feedback and Deadbeat Observers 432
Review Examples 434
Problems 440
8. Linear Quadratic Optimal Control Through Lyapunov Synthesis 451
8.1 Introduction 451
8.2 The Concept of Lyapunov Stability 452
8.3 Lyapunov Functions for Linear Systems 454
8.4 Parameter Optimization and Optimal Control Problems 458
8.5 Quadratic Performance Index 461
8.6 Control Configurations 467
8.7 Optimal State Regulator 470
8.8 Optimal Digital Control Systems 479
8.9 Constrained State Feedback Control 484
Review Examples 488
Problems 495

Part III Nonlinear Control Systems: Conventional and Intelligent 501

9. Nonlinear Systems Analysis 505
9.1 Introduction 505
9.2 Some Common Nonlinear System Behaviours 507
9.3 Common Nonlinearities in Control Systems 508
9.4 Describing Function Fundamentals 511
9.5 Describing Functions of Common Nonlinearities 514
9.6 Stability Analysis by the Describing Function Method 519
9.7 Concepts of Phase Plane Analysis 527
9.8 Construction of Phase Portraits 529
9.9 System Analysis on the Phase Plane 535
9.10 Simple Variable Structure Systems 542
9.11 Lyapunov Stability Definitions 546
9.12 Lyapunov Stability Theorems 549
9.13 Lyapunov Functions for Nonlinear Systems 556
Review Examples 561
Problems 567
10. Nonlinear Control Structures 574
10.1 Introduction 574
10.2 Feedback Linearization 575
10.3 Model Reference Adaptive Control 581
10.4 System Identification and Generalized Predictive Control in Self-Tuning Mode 587
10.5 Sliding Mode Control 600
10.6 Reinforcement Learning Control 607
Problems 615
 x Contents
11. Neural Networks 620
11.1 Introduction 620
11.2 Neuron Models 624
11.3 Network Architectures 633
11.4 Learning in Neural Networks 640
11.5 Training the Multilayer Neural Network—Backpropagation Tuning 649
11.6 Function Approximation with Neural Networks 654
11.7 System Identification with Neural Networks 655
11.8 Control with Neural Networks 660
11.9 Support Vector Machines 665
Review Examples 679
Problems 684
12. Fuzzy Logic Models 688
12.1 Introduction 688
12.2 Fuzzy Quantification of Knowledge 695
12.3 Fuzzy Inference 706
12.4 Designing a Fuzzy Logic Controller (Mamdani Architecture) 709
12.5 Data Based Fuzzy Modeling (Sugeno Architecture) 719
12.6 System Identification and Control with Neuro-Fuzzy Systems 723
12.7 Genetic Algorithms 726
Review Examples 737
Problems 740
References 748
Answers to Problems 755
Appendix A: MATLAB Aided Control System Design: Conventional Control
URL: http://www.mhhe.com/gopal/dc3e
Appendix B: MATLAB Aided Control System Design: Intelligent Control
URL: http://www.mhhe.com/gopal/dc3e

Index 775
 Preface

The dramatic development of computer technology has radically changed the boundaries of practical control
system design options. It is now possible to employ very complicated, high-order digital controllers, and to
carry out the extensive calculations required for their design. These advances in implementation and design
capability can be achieved at low cost due to the widespread availability of inexpensive, powerful digital
computers and related devices. There is every indication that a high rate of growth in the capability and
application of digital computers will continue far into the future.
Fortunately, control theory has also developed substantially over the past 45 years. The classical design
methods have been greatly enhanced by the availability of low-cost computers for system analysis and
simulation. The graphical tools of classical design like root locus plots, Nyquist plots, Bode plots, and
Nichols chart can now be more easily used with computer graphics. Coupled with hardware developments
such as microprocessors and electro-optic measurement schemes, the classical control theory today provides
useful design tools to practising control engineers.
The modern control theory (which can’t be termed modern any longer) refers to the state-space based
methods. Modern control methods initially enjoyed a great deal of success in academic circles, but did not
perform very well in many areas of application. Modern control provided a lot of insight into system struc-
ture and properties; nevertheless, it masked other important feedback properties that could be studied, and
manipulated, using classical control. During the past three decades, a series of methods, which are a combi-
nation of modern state-space methods and classical frequency-domain methods, have emerged. These tech-
niques are commonly known as robust control.
The rapid development in digital computers and microelectronics has brought about drastic changes in
the approach to analysis, design, and implementation of control systems. The flourishing of digital control
has just begun for most industries and there is much to be gained by exploiting the full potential of new
technology.
Implementation of nonlinear robust control schemes—model reference adaptive control, self-tuning con-
trol, variable structure sliding mode control—has now become a relatively simple task. In the process of
understanding and emulating salient features of biological control functions, a new field called ‘intelligent
control’, has emerged.
Fuzzy logic, artificial neural networks, and genetic algorithms have grown into three distinct disciplines
with the aim of designing “intelligent” systems for scientific and engineering applications. The theory of
fuzzy logic provides a mathematical morphology to emulate certain perceptual and linguistic attributes,
associated with human cognition. It aims at modelling the inexact modes of reasoning and thought processes,
that play an essential role in the remarkable human ability to make rational decisions in an environment of
uncertainty and imprecision. Artificial neural networks attempt to emulate the architecture and information
representation schemes of the human brain. Genetic algorithms provide an adaptive, robust, parallel and
randomized searching technique where a population of solutions evolves over a sequence of generations, to
a globally optimal solution.
While the development of individual tools—fuzzy logic, artificial neural networks, genetic algorithms—
was in progress, a group of researchers felt the need of integrating them in order to enjoy the merits of
 xii Preface

different biologically inspired techniques into one system. The result is the development of several hybrid
paradigms, like neuro-fuzzy, neuro-genetic, fuzzy-genetic, and neuro-fuzzy-genetic. These hybrid paradigms
are suitable for solving complex real-world problems for which only one tool may not be adequate.
We are passing through a phase of rosy outlook for control technology: there has been a tendency to
portray neural network and/or a fuzzy-logic based controller as magical; a sort of black box that does
amazing things. Enthusiastic researchers have dreamed big dreams and made sweeping predictions on the
potential of intelligent control systems with neural processor/fuzzy logic chips in the control loop. The
growth of the new field during the past few years has been so explosive that the traditional model-based
control theory and the traditional op amp/computer-based implementations of controllers are under threat.
New design methods are required to prove themselves in actual practice, before they can displace well-
accepted techniques. Intelligent control technology is slowly gaining wider acceptance among academics
and industry. Even if there has been some over-enthusiastic description of the field, the scientific commu-
nity, and industry, are converging to the fact that there is something fundamentally significant about this
technology. Also, preparations are underway, to accord a warm welcome to integrated control technology–
integration of intelligent control theory with the traditional model-based theory, and integration of the VLSI
microprocessors with neuro-fuzzy processors to implement the controller.
This book includes a tutorial introduction to the knowledge-based tools for control system design. Rigor-
ous characterization of theoretical properties of intelligent control methodology will not be our aim in our
tutorial presentation; rather we focus on the development of systematic engineering procedures, which will
guide the design of the controller for a specific problem.
The arrangement of topics in the text, is a little different from the conventional one. Typically, a book on
digital control systems starts with transform-domain design and then carries over to state space. These books
give a detailed account of state variable methods for discrete-time systems. Since the state variable methods
for discrete-time systems run quite parallel to those for continuous-time systems, full-blown repetition is not
appreciated by the readers conversant with state variable methods for continuous-time systems. And for
readers with no background of this type, a natural way of introducing state variable methods, is to give the
treatment for continuous-time systems, followed by a brief parallel presentation for discrete-time systems.
To meet this objective, we have divided the treatment of transform-domain design and state-space design
in two parts of the book. Part I deals with digital control principles and design in transform domain, assum-
ing that the reader has had an introductory course in control engineering concentrating on the basic princi-
ples of feedback control and covering various classical analog methods of control system design. The mate-
rial presented in this part of the book is closely related to the material already familiar, but towards the end
a direction to wider horizons is indicated. Part II of the book deals with state variable methods in automatic
control. State variable analysis and design methods are usually not covered in an introductory course. It is
assumed that the reader is not exposed to the so-called modern control theory. Our approach is to first
discuss state variable methods for continuous-time systems, and then give a compact presentation of the
methods for discrete-time systems using the analogy with the continuous-time systems.
For the purpose of organizing different courses for students with different backgrounds, the sequencing
of chapters, and dependence of each chapter on previous chapters, has been properly designed in the text.
The typical engineering curriculum at the second-degree level includes core courses on ‘digital control
systems’, and ‘linear system theory’. Parts I and II of the book have been designed to fully meet the require-
ments of the two courses. In Part III of the book, a reasonable detailed account of nonlinear control schemes,
both the conventional and the intelligent, is given. The requirements of elective courses on ‘nonlinear
 Preface xiii
control systems’, and ‘intelligent control’, will be partially or fully (depending on the depth of coverage of
the courses) served by Part III of the book.
The typical engineering curriculum at the first-degree level includes a core course on feedback control
systems, with one or two elective courses on the subject. The book meets the requirements of elective courses
at the first-degree level.

WEBSITE URL
The Online Learning Center for the book:
http://www.mhhe.com/gopal/dc3e
· provides the students with the source codes of the MATLAB problems given in Appendix A and
Appendix B of the book; and
· provides the faculty with the Solution Manual for the exercise problems given at the end of all chapters
in the book. This part of the site is password protected, and will be made available by the Publisher to
the faculty on request.

M Gopal
[email protected]
 Part I
DIGITAL CONTROL: PRINCIPLES AND DESIGN IN
TRANSFORM-DOMAIN

Automatic control systems play a vital role in the (technological) progress of human civilization. These
control systems range from the very simple to the fairly complex in nature. Automatic washing machines,
refrigerators, and ovens are examples of some of the simpler systems used in the home. Aircraft auto-
matic pilots, robots used in manufacturing, and electric power generation and distribution systems repre-
sent complex control systems. Even such problems as inventory control, and socio-economic systems
control, may be approached from the theory of feedback control.
Our world is one of continuous-time variables type. Quantities like flow, temperature, voltage,
position, and velocity are not discrete-time variables but continuous-time ones. If we look back at the
development of automatic control, we find that mass-produced analog (electronic) controllers have been
available since about the 1940s. A first-level introduction to control engineering, provided in the
companion book ‘Control Systems: Principles and Design’, deals with basics of control, and covers
sufficient material to enable us to design analog (op amp based) controllers for many simple control loops
found in industry.
From the 1980s onward, we find microprocessor digital technology beginning to take over. Today,
most complex industrial processes are under computer control. A microprocessor determines the input to
manipulate the physical system, or plant; and this requires facilities to apply this input to the physical
world. In addition, the control strategy typically relies on measured values of the plant behaviour; and this
requires a mechanism to make these measured values available to the computing resources. The plant can
be viewed as changing continuously with time. The controller, however, has a discrete clock that governs
its behaviour and so its values change only at discrete points in time. To obtain deterministic behaviour
and ensure data integrity, the sensor must include a mechanism to sample continuous data at discrete
points in time, while the actuators need to produce a continuous value between the time points with
discrete-time data.
Computer interfacing for data acquisition, consists of analog-to-digital (A/D) conversion of the input
(to controller) analog signals. Prior to the conversion, the analog signal has to be conditioned to meet the
input requirements of the A/D converter. Signal conditioning consists of amplification (for sensors
generating very low power signals), filtering (to limit the amount of noise on the signal), and isolation
(to protect the sensors from interacting with one another and/or to protect the signals from possible
damaging inputs). Conversion of a digital signal to an analog signal (D/A) at the output (of controller), is
to be carried out to send this signal to an actuator which requires an analog signal. The signal has to be
amplified by a transistor or solid state relay or power amplifier. Most manufacturers of electronic instru-
mentation devices are producing signal conditioners as modules.
The immersion of the computing power into the physical world has changed the scene of control
system design. A comprehensive theory of digital “sampled” control has been developed. This theory
requires a sophisticated use of new concepts such as z-transform. It is, however, quite straightforward to
translate analog design concepts into digital equivalents. After taking a guided tour through the analog
design concepts and op amp technology, the reader will find in Part I of this book sufficient material to
enable him/her to design digital controllers for many simple control loops, and interfacing the controllers
to other sub-systems in the loop; thereby building complete feedback control systems.
 The broad space of digital control applications can be roughly divided into two categories: industrial
control and embedded control. Industrial control applications are those in which control is used as part of
the process of creating or producing an end product. The control system is not a part of the actual end
product itself. Examples include the manufacture of pharmaceuticals and the refining of oil. In the case of
industrial control, the control system must be robust and reliable, since the processes typically run
continuously for days, weeks or years.
Embedded control applications are those in which the control system is a component of the end prod-
uct itself. For example, Electronic Control Units (ECUs) are found in a wide variety of products including
automobiles, airplanes, and home applications. Most of these ECUs implement different feedback con-
trol tasks. For instance, engine control, traction control, anti-lock braking, active stability control, cruise
control, and climate control. While embedded control systems must also be reliable, cost is a more sig-
nificant factor, since the components of the control system contribute to the overall cost of manufacture
of the product. In this case, much more time and effort is usually spent in the design phase of the control
system to ensure reliable performance without requiring any unnecessary excess of processing power,
memory, sensors, actuators etc., in the digital control system. Our focus in this book will be on industrial
control applications.
Perhaps more than any other factor, the development of microprocessors has been responsible for the
explosive growth of the computer industry. While early micoprocessors required many additional
components in order to perform any useful task, the increasing use of large-scale integration (LSI) or
very large-scale integration (VLSI) semiconductor fabrication techniques has led to the production of
microcomputers, where all of the required circuitry is embedded on one or a small number of integrated
circuits. A further extension of the integration is the single chip microcontroller, which adds analog and
binary I/O, timers, and counters so as to be able to carry out real-time control functions with almost no
additional hardware. Examples of such microcontrollers are Intel 8051, 8096 and Motorola MCH
68HC11. These chips were developed largely, in response to the automotive industries’ desire for
computer-controlled ignition, emission control and anti-skid systems. They are now widely used in proc-
ess industries. This digital control practice, along with the theory of sampled-data systems will be dealt
with in Chapters 2–4 of the book.
 Introduction
1
1.1 CONTROL SYSTEM TERMINOLOGY

A Control System is an interconnection of components to provide a desired function. The portion of the
system to be controlled is given various names: process, plant, and controlled system being perhaps the most
common. The portion of the system that does the controlling is the controller. Often, a control system
designer has little or no design freedom with the plant; it is usually fixed. The designer’s task is, therefore, to
develop a controller that will control the given plant acceptably. When measurements of the plant response
are available to the controller (which in turn generates signals affecting the plant), the configuration is a
feedback control system.
A digital control system uses digital hardware, usually in the form of a programmed digital computer, as
the heart of the controller. In contrast, the controller in an analog control system is composed of analog
hardware; an electronic controller made of resistors, capacitors, and operational amplifiers is a typical
example. Digital controllers normally have analog devices at their periphery to interface with the plant; it is
the internal working of the controller that distinguishes digital from analog control.
The signals used in the description of control systems are classified as continuous-time and discrete-time.
Continuous-time signals are defined for all time, whereas discrete-time signals are defined only at discrete
instants of time, usually evenly spaced steps. The signals for which both time and amplitude are discrete, are
called digital signals. Because of the complexity of dealing with quantized (discrete-amplitude) signals,
digital control system design proceeds as if computer-generated signals were not of discrete amplitude. If
necessary, further analysis is then done, to determine if a proposed level of quantization is acceptable.
Systems and system components are termed continuous-time or discrete-time according to the type of
signals they involve. They are classified as being linear if signal components in them can be superim-
posed—any linear combination of signal components, applied to a system, produces the same linear combi-
nation of corresponding output components; otherwise a system is nonlinear. A system or component is
time-invariant if its properties do not change with time—any time shift of the inputs produces an equal time
shift of every corresponding signal. If a system is not time-invariant, then it is time-varying.
A typical topology of a computer-controlled system is sketched schematically in Fig. 1.1. In most cases,
the measuring transducer (sensor) and the actuator (final control element) are analog devices, requiring,
respectively, analog-to-digital (A/D) and digital-to-analog (D/A) conversion at the computer input and
 4 Digital Control and State Variable Methods
output. There are, of course, exceptions; sensors which combine the functions of the transducer and the A/D
converter, and actuators which combine the functions of the D/A converter and the final control element are
available. In most cases, however, our sensors will provide an analog voltage output, and our final control
elements will accept an analog voltage input.
Command input Disturbance Controlled
(Desired plant behaviour) inputs output

Final
A/D Computer D/A control Plant
element

Clock

Sensor

Fig. 1.1 Basic structure of a computer-controlled system

In the control scheme of Fig. 1.1, the A/D converter performs the sampling of the sensor signal (analog
feedback signal ) and produces its binary representation. The digital computer (control algorithm) generates
a digital control signal using the information on desired and actual plant behaviour. The digital control signal
is then converted to analog control signal via the D/A converter. A real-time clock synchronizes the actions
of the A/D and D/A converters, and the shift registers. The analog control signal is applied to the plant
actuator to control the plant’s behaviour.
The overall system in Fig. 1.1 is hybrid in nature; the signals are in the sampled form (discrete-time
signals) in the computer, and in a continuous form in the plant. Such systems have traditionally been called
sampled-data systems; we will use this term as a synonym for computer control systems/digital control
systems.
The word ‘servomechanism’ (or servosystem) is used for a command-following system wherein the con-
trolled output of the system is required to follow a given command. When the desired value of the controlled
outputs is more or less fixed, and the main problem is to reject disturbance effects, the control system is
sometimes called a regulator. The command input for a regulator becomes a constant and is called set-point,
which corresponds to the desired value of the controlled output. The set-point may however be changed in
time, from one constant value to another. In a tracking system, the controlled output is required to follow, or
track, a time-varying command input.
To make these definitions more concrete, let us consider some familiar examples of control systems.

Example 1.1 Servomechanism for Steering of Antenna

One of the earliest applications of radar tracking was for anti-aircraft fire control, first with guns and later
with missiles. Today, many civilian applications exist as well, such as satellite-tracking radars, navigation-
aiding radars, etc.
The radar scene includes the radar itself, a target, and the transmitted waveform that travels to the target
and back. Information about the target’s spatial position is first obtained by measuring the changes in the
back-scattered waveform relative to the transmitted waveform. The time shift provides information about
the target’s range, the frequency shift provides information about the target’s radial velocity, and the
 Introduction 5
received voltage magnitude and phase provide infor-
nna ax
ismation about the target’s angle1[1].
Ante In a typical radar application, it is necessary to
point the radar antenna toward the target and follow its
movements. The radar sensor detects the error be-
tween the antenna axis and the target, and directs the
antenna to follow the target. The servomechanism for
Elevation Azimuth steering the antenna in response to commands from
a b radar sensor, is considered here. The antenna is
designed for two independent angular motions, one
about the vertical axis in which the azimuth angle is
Fig. 1.2 Antenna configuration varied, and the other about the horizontal axis in
which the elevation angle is varied (Fig. 1.2). The
servomechanism for steering the antenna is described by two controlled variables—azimuth angle b and
elevation angle a. The desired values or commands are the azimuth angle br and the elevation angle ar of the
target. The feedback control problem involves error self-nulling, under conditions of disturbances beyond
our control (such as wind power).
The control system for steering antenna can be treated as two independent systems—the azimuth-angle
servomechanism, and the elevation-angle servomechanism. This is because the interaction effects are
usually small. The operational diagram of the azimuth-angle servomechanism is shown in Fig. 1.3.
The steering command from the radar sensor, which corresponds to target azimuth angle, is compared with
the azimuth angle of the antenna axis. The occurrence of the azimuth-angle error causes an error signal to pass
through the amplifier, which increases the angular velocity of the servomotor in a direction towards an error
reduction. In the scheme of Fig. 1.3, the measurement and processing of signals (calculation of control signal)
is digital in nature. The shaft-angle encoder combines the functions of transducer and A/D converter.

Fig. 1.3 Azimuthal servomechanism for steering of antenna

Figure 1.4 gives the functional block diagrams of the control system. A simple model of the load
(antenna) on the motor is shown in Fig. 1.4b. The moment of inertia J and the viscous friction coefficient B
are the parameters of the assumed model. Nominal load is included in the plant model for the control design.
1
The bracketed numbers coincide with the list of references given at the end of the book.
 6 Digital Control and State Variable Methods
The main disturbance inputs are the deviation of the load from the nominal estimated value as a result of
uncertainties in our estimate, effect of wind power, etc.
Load
disturbance

˜r Motor ˜ B
+ e Power
+ J
amplifier
– Load

(a) (b)
Load
disturbance

˜r + Motor ˜
e + Power
amplifier +
Load
– –

Tachogenerator
Rate
signal

(c)
Fig. 1.4 Functional block diagrams of azimuthal servomechanism

In the tracking system of Fig. 1.4a, the occurrence of error causes the motor to rotate in a direction
favouring the dissolution of error. The processing of the error signal (calculation of the control signal) is based
on the proportional control logic. Note that the components of our system cannot respond instantaneously,
since any real-world system cannot go from one energy level to another in zero time. Thus, in any real-world
system, there is some kind of dynamic lagging behaviour between input and output. In the servosystem of
Fig. 1.4a, the control action, on occurrence of the deviation of the controlled output from the desired value
(the occurrence of error), will be delayed by the cumulative dynamic lags of the shaft-angle encoder, digital
computer and digital-to-analog converter, power amplifier, and the servomotor with load. Eventually, how-
ever, the trend of the controlled variable deviation from the desired value, will be reversed by the action of
the amplifier output on the rotation of the motor, returning the controlled variable towards the desired value.
Now, if a strong correction (high amplifier gain) is applied (which is desirable from the point of view of
control system performance, e.g., strong correction improves the speed of response), the controlled variable
overshoots the desired value (the ‘run-out’ of the motor towards an error with the opposite rotation), causing
a reversal in the algebraic sign of the system error. Unfortunately, because of system dynamic lags, a reversal
of correction does not occur immediately, and the amplifier output (acting on ‘old’ information) is now
actually driving the controlled variable in the direction it was already heading, instead of opposing its excur-
sions, thus leading to a larger deviation. Eventually, the reversed error does cause a reversed correction, but
the controlled variable overshoots the desired value in the opposite direction and the correction is again in
the wrong direction. The controlled variable is thus driven, alternatively, in opposite directions before it
settles on to an equilibrium condition. This oscillatory state is unacceptable as the behaviour of antenna-
 Introduction 7
steering servomechanism. The considerable amplifier gain, which is necessary if high accuracies are to be
obtained, aggravates the described unfavourable phenomenon.
The occurrence of these oscillatory effects can be controlled by the application of special compensation
feedback. When a signal proportional to motor’s angular velocity (called the rate signal) is subtracted from
the error signal (Fig. 1.4c), the braking process starts sooner before the error reaches a zero value.
The ‘loop within a loop’ (velocity feedback system embedded within a position feedback system)
configuration utilized in this application, is a classical scheme called minor-loop feedback scheme.

Example 1.2 Variable Speed dc Drive

Many industrial applications require variable speed drives. For example, variable speed drives are used for
pumping duty to vary the flow rate or the pumping pressure, rolling mills, harbour cranes, rail traction, etc.
[2–4].
The variable speed dc drive is the most versatile drive available. The Silicon Controlled Rectifiers (SCR)
are almost universally used to control the speed of dc motors, because of considerable benefits that accrue
from the compact static controllers supplied directly from the ac mains.
Basically, all the dc systems involving the SCR controllers are similar but, with different configurations
of the devices, different characteristics may be obtained from the controller. Figure 1.5 shows a dc motor
driven by a full-wave rectified supply. Armature current of the dc motor is controlled by an SCR which is, in
turn, controlled by the pulses applied by the SCR trigger control circuit. The SCR controller thus combines
the functions of a D/A converter and a final control element.

Load

Speed Actual
reference + – speed

Tachogenerator

A /D Motor
Full Over-current
bridge protection
ac Digital
rectifier
computer
circuit
SCR SCR
trigger control
circuit

Fig. 1.5 Variable speed dc drive

Firing angle of the SCR controls the average armature current which, in turn, controls the speed of the dc
motor. The average armature current (speed) increases as the trigger circuit reduces the delay angle of firing of
the SCR, and the average armature current (speed) reduces as the delay angle of firing of the SCR is increased.
In the regulator system of Fig. 1.5, the reference voltage which corresponds to the desired speed of the dc
motor, is compared with the output voltage of tachogenerator, corresponding to the actual speed of the
motor. The occurrence of the error in speed, causes an error signal to pass through the trigger circuit, which
 8 Digital Control and State Variable Methods
controls the firing angle of the SCR in a direction towards an error reduction. When the processing of the
error signal (calculation of the control signal) is based on the proportional control logic, a steady-state error
between the actual speed and the desired speed exists. The occurrence of steady-state error can be elimi-
nated by generating the control signal with two components: one component proportional to the error signal,
and the other proportional to the integral of the error signal.

Example 1.3 Liquid-level Control System

This example describes the hardware features of the design of a PC-based liquid-level control system. The
plant of our control system is a cylindrical tank. Liquid is pumped into the tank from the sump (Fig. 1.6). The
inflow to the tank can be controlled by adjusting valve V1. The outflow from the tank goes back into the sump.
Valve V1 of our plant is a rotary valve; a stepping motor has been used to control the valve. The stepping
motor controller card, interfaced to the PC, converts the digital control signals into a series of pulses which
are fed to the stepping motor using a driver circuit. Three signals are generated from the digital control
signal at each sampling instant, namely, number of steps, speed of rotation, and direction of rotation. The
stepping-motor driver circuit converts this information into a single pulse train, which is fed to the stepping
motor. The valve characteristics between the number of steps of the stepping motor and the outflow from the
valve, are nonlinear.
The probe used for measurement of liquid level, consists of two concentric cylinders connected to a
bridge circuit, to provide an analog voltage. The liquid partially occupies the space between the cylinders,
with air in the remaining part. This device acts like two capacitors in parallel, one with dielectric constant of
air ( ~- 1) and the other with that of the liquid. Thus, the variation of the liquid level causes variation of the
electrical capacity, measured between the cylinders. The change in the capacitance causes a change in the
bridge output voltage which is fed to the PC through an amplifier circuit. The characteristics of the sensor
between the level and the voltage are approximately linear.

Step motor
controller card
Expansion
slots of PC

Driver Step
circuit motor
V1 V2

Pump Sump

A /D
conversion card Bridge and
amplifier circuit

Fig. 1.6 Liquid-level control system

In the liquid-level control system of Fig. 1.6, the command signal (which corresponds to the desired level
of the liquid in the cylinder) is fed through the keyboard; the actual level signal is received through the A/D
 Introduction 9
conversion card. The digital computer compares the two signals at each sampling instant, and generates a
control signal which is the sum of two components: one proportional to the error signal, and the other,
proportional to the integral of the error signal.

1.2 COMPUTER-BASED CONTROL: HISTORY AND TRENDS

Digital computers were first applied to the industrial process control in the late 1950s. The machines were
generally large-scale ‘main frames’ and were used in a so-called supervisory control mode; the individual
temperature, pressure, flow and the like, feedback loops were locally controlled by electronic or pneumatic
analog controllers. The main function of the computer was to gather information on how the overall process
was operating, feed this into a technical-economic model of the process (programmed into computer
memory), and then, periodically, send signals to the set-points of all the analog controllers, so that each
individual loop operated in such a way as to optimize the overall operation.
In 1962, a drastic departure from this approach was made by Imperial Chemical Industries in England—
a digital computer was installed, which measured 224 variables and manipulated 129 valves directly. The
name Direct Digital Control (DDC) was coined to emphasize that the computer controlled the process
directly. In DDC systems, analog controllers were no longer used. The central computer served as a single,
time-shared controller for all the individual feedback loops. Conventional control laws were still used for
each loop, but the digital versions of control laws for each loop resided in the software in the central compu-
ter. Though digital computers were very expensive, one expected DDC systems to have economic advantage
for processes with many (50 or more) loops. Unfortunately, this did not often materialize. As failures in the
central computer of a DDC system shut down the entire system, it was necessary to provide a ‘fail-safe’
back-up system, which usually turned out to be a complete system of individual loop analog controllers, thus
negating the expected hardware savings.
There was a substantial development of digital computer technology in the 1960s. By the early 1970s,
smaller, faster, more reliable, and cheaper computers became available. The term minicomputers was coined
for the new computers that emerged. DEC PDP11 is by far, the best-known example. There were, however,
many related machines from other vendors.
The minicomputer was still a fairly large system. Even as performance continued to increase and prices to
decrease, the price of a minicomputer main frame in 1975, was still about $10,000. Computer control was still
out of reach for a large number of control problems. However, with the development of microcomputer, the
price of a card computer, with the performance of a 1975 minicomputer, dropped to $500 in 1980. Another
consequence was that digital computing power in 1980 came in quanta as small as $50. This meant that
computer control could now be considered as an alternative, no matter how small the application [54-57].
Microcomputers have already made a great impact on the process control field. They are replacing analog
hardware even as single-loop controllers. Small DDC systems have been made using microcomputers.
Operator communication has vastly improved with the introduction of colour video-graphics displays.
The variety of commercially available industrial controllers ranges from single-loop controllers through
multiloop single computer systems to multiloop distributed computers. Although the range of equipment
available is large, there are a number of identifiable trends which are apparent.
Single-loop microprocessor-based controllers, though descendants of single-loop analog controllers, have
greater degree of flexibility. Control actions which are permitted, include on/off control, proportional action,
integral action, derivative action and the lag effect. Many controllers have self-tuning option. During the self-
tune sequence, the controller introduces a number of step commands, within the tolerances allowed by the
operator, in order to characterize the system response. From this response, values for proportional gain, reset
time, and rate time are developed. This feature of online tuning in industrial controllers is interesting, and
permits the concept of the computer automatically adjusting to changing process conditions [11-12].
 10 Digital Control and State Variable Methods
Multiloop single computer systems have variability in available interface and software design. Both the
single-loop and multiloop controllers may be used in stand-alone mode, or may be interfaced to a host
computer for distributed operation. The reducing costs and increasing power of computing systems, has
tended to make distributed computing systems for larger installations, far more cost effective than those built
around one large computer. However, the smaller installation may be best catered for by a single multiloop
controller, or even a few single-loop devices.
Control of large and complex processes using distributed computer control systems (DCCS), is facili-
tated by adopting a multilevel or hierarchical view point of control strategy. The multilevel approach sub-
divides the system into a hierarchy of simpler control design problems. On the lowest level of control (direct
process control level), the following tasks are handled: acquisition of process data, i.e., collection of instan-
taneous values of individual process variables, and status messages of plant control facilities (valves, pumps,
motors, etc.) needed for efficient direct digital control; processing of collected data; plant hardware moni-
toring, system check and diagnosis; closed-loop control and logic control functions, based on directives
from the next ‘higher’ level.
Supervisory level copes with the problems of determination of optimal plant work conditions, and genera-
tion of relevant instructions to be transferred to the next ‘lower’ level. Adaptive control, optimal control, plant
performance monitoring, plant coordination and failure detections are the functions performed at this level.
Production scheduling and control level is responsible for production dispatching, inventory control,
production supervision, production re-scheduling, production reporting, etc.
Plant(s) management level, the ‘highest’ hierarchical level of the plant automation system, is in charge of
the wide spectrum of engineering, economic, commercial, personnel, and other functions.
It is, of course, not to be expected that in all available distributed computer control systems, all four
hierarchical levels are already implemented. For automation of small-scale plants, any DCCS having at least
two hierarchical levels, can be used. One system level can be used as a direct process control level, and the
second one as a combined plant supervisory, and production scheduling and control level. Production
planning and other enterprise-level activities, can be managed by the separate mainframe computer or the
computer centre. For instance, in a LAN (Local Area Network)-based system structure, shown in Fig. 1.7a,
the ‘higher’ automation levels are implemented by simply attaching the additional ‘higher’ level computers
to the LAN of the system [89].

Process Computer Higher

Management Centre hierarchical
Computer levels

LAN

Supervisory and Lower

direct process control hierarchical
levels
Control Plant
signals inputs

PLANT

Fig. 1.7 a Hierarchical levels in Computer Integrated Process Systems (CIPS)

 Introduction 11
For complex process plant monitoring, SCADA (Supervisory Control And Data Acquisition) systems are
available. The basic functions carried out by a SCADA system are:
· Data acquisition and communication · Events and alarms reporting
· Data processing · Partial process control
The full process control functions are delegated to the special control units, connected to SCADA system,
and are capable of handling the emergency shut down situations.
The separation of SCADA and DCCS is slowly vanishing and the SCADA systems are being
brought within the field of DCCS; the hierarchical, distributed, flexible and extremely powerful Computer
Integrated Process Systems (CIPS), is now a technical reality.
The other main and early application area of digital methods was machine tool numerical control, which
developed at about the same time as computer control in process industries. Earlier, numerically controlled
(NC) machines used ‘hard-wired’ digital techniques. As the microcomputer price and performance im-
proved, it became feasible to replace the hard-wired functions with their software-implemented equivalents,
using a microcomputer as a built-in component of the machine tool. This approach has been called Compu-
terized Numerical Control (CNC) [20]. Industrial robots were developed simultaneously with CNC systems.
A quiet revolution is going on in the manufacturing world that is changing the look of the factory. Com-
puters are controlling and monitoring the manufacturing processes [21-22]. The high degree of automation
that, until recently, was reserved for mass production only, is also applied now to small batches. This
requires a change from hard automation in the production line, to a flexible manufacturing system (FMS),
which can be more readily rearranged to handle new market requirements.
Flexible manufacturing systems, combined with automatic assembly and product inspection on one hand,
and CAD/CAM systems on the other, are the basic components of the modern Computer Integrated Manu-
facturing System (CIMS). In a CIMS, the production flow, from the conceptual design to the finished prod-
uct, is entirely under computer control and management.
Figure 1.7b illustrates hierarchical structure of CIMS. The lowest level of this structure contains stand-
alone computer control systems of manufacturing processes and industrial robots. The computer control of
processes includes all types of CNC machine tools, welding, electrochemical machining, electrical discharge
machining, and a high-power laser, as well as the adaptive control of these processes.

Supervisory computer

Transfer Automatic
CAD/CAM FMS system inspection

Raw Material
Products
DNC Manufacturing Assembly
system cell station

Computer Control Industrial

of Processes Robots

Fig. 1.7 b Hierarchical levels in Computer Integrated Manufacturing Systems (CIMS)

 12 Digital Control and State Variable Methods
When a battery of NC or CNC machine tools is placed under the control of a single computer, the result
is a system known as direct numerical control (DNC).
The operation of several CNC machines and industrial robots, can be co-ordinated by systems called
manufacturing cells. The computer of the cell is interfaced with the computer of the robot and CNC
machines. It receives “completion of job” signals from the machines and issues instructions to the robot to
load and unload the machines, and change their tools. The software includes strategies permitting the
handling of machine breakdown, tool breakage, and other special situations.
The operation of many manufacturing cells can be coordinated by flexible manufacturing system (FMS).
The FMS accepts incoming workpieces and processes them under computer control, into finished parts.
The parts produced by the FMS must be assembled into the final product. They are routed on a transfer
system to assembly stations. In each station, a robot will assemble parts, either into a sub-assembly or (for
simple units), into the final product. The sub-assemblies will be further assembled by robots located in other
stations. The final product will be tested by an automatic inspection system.
The FMS uses CAD/CAM systems to integrate the design and manufacturing of parts. At the highest
hierarchical level, there will be a supervisory computer, which coordinates participation of computers in all
phases of a manufacturing enterprise: the design of the product, the planning of its manufacture, the auto-
matic production of parts, automatic assembly, automatic testing, and, of course, computer-controlled flow
of materials and parts through the plant.
In a LAN-based system, the ‘higher’ automation levels (production planning and other enterprise-level
activities), can be implemented by simply attaching the additional ‘higher’ level computers to the LAN of
the system.
One of the most ingenious devices ever devised to advance the field of industrial automation, is the
Programmable Logic Controller (PLC). The PLC, a microprocessor-based general purpose device, pro-
vides a ‘menu’ of basic operations that can be configured by programming to create logic control system for
any application [23–25]. So versatile are these devices, that they are employed in the automation of almost
every type of industry. CIPS and CIMS provide interfaces to PLCs for handling high-speed logic (and other)
control functions. Thousands of these devices go unrecognized in process plants and factory environments—
quietly monitoring security, manipulating valves, and controlling machines and automatic production lines.
Thus, we see that the recent appearance of powerful and inexpensive microcomputers, has made digital
control practical for a wide variety of applications. In fact, now every process is a candidate for digital
control. The flourishing of digital control is just beginning for most industries, and there is much to be
gained by exploiting the full potential of new technology. There is every indication that a high rate of growth
in the capability and application of digital computers, will continue far into the future.

1.3 CONTROL THEORY: HISTORY AND TRENDS

The development of control system analysis and design can be divided into three eras. In the first era,
we have the classical control theory, which deals with techniques developed during the 1940s and 1950s.
Classical control methods—Routh-Hurwitz, Root Locus, Nyquist, Bode, Nichols—have in common the use
of transfer functions in the complex frequency (Laplace variable s) domain, and the emphasis on the graphi-
cal techniques. Since computers were not available at that time, a great deal of emphasis was placed on
developing methods that were amenable to manual computation and graphics. A major limitation of the
classical control methods was the use of single-input, single-output (SISO) control configurations. Also, the
use of the transfer function and frequency domain limited one to linear time-invariant systems. Important
results of this era will be discussed in Part I of this book.
In the second era, we have modern control (which is not so modern any longer), which refers to state-
space based methods developed in the late 1950s and early 1960s. In modern control, system models are
 Introduction 13
directly written in the time domain. Analysis and design are also carried out in the time domain. It should be
noted that before Laplace transforms and transfer functions became popular in the 1920s, engineers were
studying systems in the time domain. Therefore, the resurgence of time-domain analysis was not unusual, but
it was triggered by the development of computers and advances in numerical analysis. As computers were
available, it was no longer necessary to develop analysis and design methods that were strictly manual.
Multivariable (multi-input, multi-output (MIMO)) control configurations could be analysed and designed.
An engineer could use computers to numerically solve or simulate large systems that were nonlinear and/or
time-varying. Important results of this era—Lyapunov stability criterion, pole-placement by state feedback,
state observers, optimal control—will be discussed in Part II of this book.
Modern control methods initially enjoyed a great deal of success in academic circles, but they did not
perform very well in many areas of application. Modern control provided a lot of insight into system struc-
ture and properties, but it masked other important feedback properties that could be studied and manipulated
using the classical control theory. A basic requirement in control engineering is to design control systems
that will work properly when the plant model is uncertain. This issue is tackled in the classical control theory
using gain and phase margins. Most modern control design methods, however, inherently require a precise
model of the plant. In the years since these methods were developed, there have been few significant
implementations and most of them have been in a single application area—the aerospace industry. The
classical control theory, on the other hand, is going strong. It provides an efficient framework for the design
of feedback controls in all areas of application. The classical design methods have been greatly enhanced by
the availability of low-cost computers for system analysis and simulation. The graphical tools of classical
design can now be more easily used with computer graphics for SISO as well as MIMO systems.
During the past three decades, the control theory has experienced a rapid expansion, as a result of the
challenges of the stringent requirements posed by modern systems such as: flight vehicles, weapon control
systems, robots, and chemical processes; and the availability of low-cost computing power. A body of
methods emerged during this third era of control-theory development, which tried to provide answers to the
problems of plant uncertainty. These techniques, commonly known as robust control, are a combination of
modern state-space and classical frequency-domain techniques. For a thorough understanding of these new
methods, we need to have adequate knowledge of state-space methods, in addition to the frequency-domain
methods. This has guided the preparation of this text.
Robust control system design has been dominated by linear control techniques, which rely on the key
assumption of availability of uncertainty model. When the required operation range is large, and a reliable
uncertainty model could not be developed, a linear controller is likely to perform very poorly. Nonlinear
controllers, on the other hand, may handle the nonlinearities in large range operation, directly. Also,
nonlinearities can be intentionally introduced into the controller part of a control system, so that the model
uncertainties can be tolerated. Advances in computer technology have made the implementation of nonlinear
control schemes—feedback linearization, variable structure sliding mode control, adaptive control,
gain-scheduling—a relatively simple task.
The third era of control-theory development has also given an alternative to model-based design meth-
ods: the knowledge-based control. In this approach, we look for a control solution that exhibits intelligent
behaviour, rather than using purely mathematical methods to keep the system under control.
Model-based control techniques have many advantages. When the underlying assumptions are satisfied,
many of these methods provide good stability, robustness to model uncertainties and disturbances, and speed of
response. However, there are many practical deficiencies of these ‘crisp’ (‘hard’ or ‘inflexible’) control algo-
rithms. It is generally difficult to accurately represent a complex process by a mathematical model. If the
process model has parameters whose values are partially known, ambiguous, or vague, crisp control algorithms
that are based on such incomplete information, will not usually give satisfactory results. The environment with
 14 Digital Control and State Variable Methods
which the process interacts, may not be completely predictable and it is normally not possible for a crisp
algorithm, to accurately respond to a condition that it did not anticipate, and that it could not ‘understand’.
Intelligent control is the name introduced to describe control systems in which control strategies are
based on AI techniques. In this control approach, which is an alternative to model-based control approach,
a behavioural (and not mathematical) description of the process is used, which is based on qualitative
expressions and experience of people working with the process. Actions can be performed either as a result
of evaluating rules (reasoning), or as unconscious actions based on presented process behaviour after a
learning phase. Intelligence comes in as the capability to reason about facts and rules, and to learn about
presented behaviour. It opens up the possibility of applying the experience gathered by operators and
process engineers. Uncertainty about the knowledge can be handled alongwith ignorance about the structure
of the system.
Fuzzy logic, and neural networks are very good methods to model real processes which cannot be described
mathematically. Fuzzy logic deals with linguistic and imprecise rules based on expert’s knowledge. Neural
networks are applied in the case where we do not have any rules but several data.
The main feature of fuzzy logic control is that a control engineering knowledge base (typically in terms of a
set of rules), created using expert’s knowledge of process behaviour, is available within the controller and the
control actions are generated by applying existing process conditions to the knowledge base, making use of an
inference mechanism. The knowledge base and the inference mechanism can handle noncrisp and incomplete
information, and the knowledge itself will improve and evolve through learning and past experience.
In the neural network based control, the goal of artificial neural network is to emulate the mechanism of
human brain function and reasoning, and to achieve the same intelligence level as the human brain in learn-
ing, abstraction, generalization and making decisions under uncertainty.
In the conventional design exercises, the system is modelled analytically by a set of differential equations,
and their solution tells the controller how to adjust the system’s control activities for each type of behaviour
required. In a typical intelligent control scheme, these adjustments are handled by an intelligent controller, a
logical model of thinking processes that a person might go through in the course of manipulating the system.
This shift in focus from the process to the person involved, changes the entire approach to automatic control
problems. It provides a new design paradigm such that a controller can be designed for complex, ill-defined
processes without knowing quantitative input-output relations, which are otherwise required by conven-
tional methods.
The ever-increasing demands of the complex control systems being built today, and planned for the
future, dictate the use of novel and more powerful methods in control. The potential for intelligent control
techniques in solving many of the problems involved is great, and this research area is evolving rapidly. The
emerging viewpoint is that model-based control techniques should be augmented with intelligent control
techniques in order to enhance the performance of the control systems. The developments in intelligent
control methods should be based on firm theoretical foundations (as is the case with model-based control
methods), but this is still at its early stages. Strong theoretical results guaranteeing control system properties
such as stability are still to come, although promising results reporting progress in special cases have been
reported recently. The potential of intelligent control systems clearly needs to be further explored and both
theory and applications need to be further developed. A brief account of nonlinear control schemes both the
conventional and the intelligent, will be given in Part III of this book.

1.4 AN OVERVIEW OF THE CLASSICAL APPROACH TO ANALOG CONTROLLER

DESIGN

The tools of classical linear control system design are the Laplace transform, stability testing, root locus, and
frequency response. Laplace transformation is used to convert system descriptions in terms of integro-
 Introduction 15
differential equations to equivalent algebraic relations involving rational functions. These are conveniently
manipulated in the form of transfer functions with block diagrams and signal flow graphs [155].
The block diagram of Fig. 1.8 represents the basic structure of feedback control systems. Not all systems
can be forced into this format, but it serves as a reference for discussion.
In Fig. 1.8, the variable y(t) is the controlled variable of the system. The desired value of the controlled
variable is yr(t), the command input. yr(t) and y(t) have the same units. The feedback elements with transfer
function H(s) are system components that act on the controlled variable y(t) to produce the feedback signal
b(t). H(s) typical represents the sensor action to convert the controlled variable y(t) to an electrical sensor
output signal b(t).

+ e
w
–
+ y
yr r + e u m+
A (s ) D (s) G A (s ) G P (s )
–
b
H (s )

Fig. 1.8 Generalised operational block diagram of a feedback system

The reference input elements with transfer function A(s) convert the command signal yr(t) into a form
compatible with the feedback signal b(t). The transformed command signal is the actual physical input to the
system. This actual signal input is defined as the reference input.
The comparison device (error detector) of the system compares the reference input r(t) with the feedback
signal b(t) and generates the actuating error signal e$ (t). The signals r(t), b(t), and e$ (t) have the same units. The
controller with transfer function D(s) acts on the actuating error signal to produce the control signal u(t).
The control signal u(t) has the knowledge about the desired control action. The power level of this signal
is relatively low. The actuator elements with transfer function GA(s), are the system components that act
on the control signal u(t) and develop enough torque, pressure, heat, etc. (manipulated variable m(t)), to
influence the controlled system. GP(s) is the transfer function of the controlled system.
The disturbance w(t) represents the undesired signals that tend to affect the controlled system. The
disturbance may be introduced into the system at more than one location.
The dashed-line portion of Fig. 1.8 shows the system error e(t) = yr – y(t). Note that the actuating error
signal e$ (t) and the system error e(t) are two different variables.
The basic feedback system block diagram of Fig. 1.8 is shown in abridged form in Fig. 1.9. The output
Y(s) is influenced by the control signal U(s) and the disturbance signal W(s) as per the following relation:
Y(s) = GP(s) GA(s) U(s) + GP(s) W(s) (1.1a)
= G(s) U(s) + N(s) W(s) (1.1b)
where G(s) is the transfer function from the control signal U(s) to the output Y(s), and N(s) is the transfer
function from the disturbance input W(s) to the output Y(s). Using Eqns (1.1), we can modify the block
diagram of Fig. 1.9 to the form shown in Fig. 1.10. Note that in the block diagram model of Fig. 1.10, the
plant includes the actuator elements.
 16 Digital Control and State Variable Methods
W (s)

R (s) + E (s) U (s) + Y (s)

Y4 ( s ) +
A (s) D (s) G A (s) GP ( s )
–
B (s)
H (s)

Fig. 1.9 A general linear feedback system

W (s) N (s )

Yr (s) R (s ) + E (s) U (s) + Y (s)

+
A (s) D (s) G (s)
–
B (s) Plant
H (s)

Fig. 1.10 Equivalent representation of the block diagram of Fig. 1.9

The actuating error signal

E$ (s) = R(s) – B(s) = A(s) Yr (s) – H(s) Y(s)

The control signal
U(s) = D(s) A(s) Yr (s) – D(s) H(s) Y(s) (1.2a)

= D(s) H(s)
LM A(s ) Y (s)
r Y (s)
OP (1.2b)
N H (s) Q
Using Eqns (1.2a) and (1.2b), we can simplify Fig. 1.10 to obtain the structure shown in Fig. 1.11.

W (s) N (s )

Yr (s) + U (s) + + Y (s)

A (s )
D (s) H (s) G (s)
H (s)
–

Fig. 1.11 Simplification of the block diagram of Fig. 1.10

W (s) N (s ) A further simplification of Fig. 1.11 is

possible if H = A. In this case, which is quite
common, we can model the system as the
Yr (s) + E (s) U (s) + + Y (s)
D (s) H (s) G (s) unity feedback system shown in Fig. 1.12,
and take advantage of the fact that now the
–
actuating signal is the system error e(t).
The block diagrams in Figs 1.10–1.12 are
Fig. 1.12 Unity feedback system very useful for the purpose of system design.
 Introduction 17
However, it should be clear that these block diagrams have lost the physical significance. For example, the
block in Fig. 1.11 with transfer function A(s)/H(s), does not refer to any physical portion of the original
system. Rather, it represents the result of manipulating Eqn. (1.2a) into the form given by Eqn. (1.2b).
Thus, the reader is advised to think in terms of the equations that the block diagrams represent, rather than
attach any special significance to the block diagrams themselves. The only role played by a block diagram is
that it is a convenient means of representing the various system equations, rather than writing them
out explicitly. Block diagram manipulation is nothing more than the manipulation of a set of algebraic
transform equations.
For the analysis of a feedback system, we
W (s) N (s ) require the transfer function between the
input—either reference or disturbance—and
R (s ) + E (s) U (s) + Y (s) the output. We can use block diagram ma-
+
D (s) G (s) nipulations to eliminate all the signals except
–
the input and the output. The reduced block
B (s) diagram leads to the desired result.
H (s)
Consider the block diagram of Fig. 1.13.
The feedback system has two inputs. We
Fig. 1.13 A typical feedback system with two inputs
shall use superposition to treat each input
separately.
R (s ) + E (s) U (s) Y (s ) When disturbance input is set to zero, the
D (s) G (s)
– single-input system of Fig. 1.14 results. The
transfer function between the input R(s) and
H (s) the output Y(s) is referred to as the reference
transfer function and will be denoted by
Fig. 1.14 Block diagram without disturbance input M(s). To solve for M(s), we write the pair of
transform equations
E (s) = R(s) – H(s) Y(s); Y(s) = G(s) U(s) = G(s) D(s) E$ (s)
$
and then eliminate E$ (s) to obtain
[1 + D(s) G(s) H(s)] Y(s) = D(s) G(s) R(s)
which leads to the desired result:
Y ( s) D( s)G ( s)
M(s) = = (1.3)
R( s) W ( s ) =0 1 + D( s)G ( s) H ( s)
Similarly, we obtain the disturbance transfer function Mw(s) by setting the reference input to zero in
Fig. 1.13 yielding Fig. 1.15, and then solving for Y(s)/W(s). From the revised block diagram,
E$ (s) = – H(s) Y(s); Y(s) = G(s) D(s) E$ (s) + N(s) W(s)
from which E$ (s) can be eliminated to give
Y ( s) N ( s)
Mw(s) = = (1.4)
W ( s) R ( s) = 0
1 + D( s)G ( s) H ( s)
The response to the simultaneous application of R(s) and W(s) is given by
Y(s) = M(s) R(s) + Mw(s) W(s) (1.5)
 18 Digital Control and State Variable Methods
Figure 1.16 shows the reduced block diagram model of the given feedback system.

W (s) N (s )
W (s)

E (s ) U (s) + + Y (s)
D (s) G (s) Mw ( s )

–1
R (s ) + + Y (s)
H (s) M (s)

Fig. 1.15 Block diagram without reference input Fig. 1.16 Reduced block diagram model
for system of Fig. 1.13

The transfer functions given by Eqns (1.3) and (1.4) are referred to as closed-loop transfer functions. The
denominator of these transfer functions has the term D(s)G(s)H(s) which is the multiplication of all the
transfer functions in the feedback loop. It may be viewed as the transfer function between the variables R(s)
and B(s) if the loop is broken at the summing point. D(s)G(s)H(s) may therefore be given the name open-
loop transfer function. The roots of denominator polynomial of D(s)G(s)H(s) are the open-loop poles, and
the roots of numerator polynomial of D(s)G(s)H(s) are the open-loop zeros.
The roots of the characteristic equation
1 + D(s)G(s)H(s) = 0 (1.6)
are the closed-loop poles of the system. These poles indicate whether or not the system is bounded-input
bounded-output (BIBO) stable, according to whether or not all the poles are in the left half of the complex
plane. Stability may be tested by the Routh stability criterion.
A root locus plot consists of a pole-zero plot of the open-loop transfer function of a feedback system,
upon which is superimposed the locus of the poles of the closed-loop transfer function as some parameter is
varied. Design of the controller (compensator) D(s) can be carried out using the root locus plot. One begins
with simple compensators, increasing their complexity until the performance requirements can be met. Prin-
cipal measures of transient performance are peak overshoot, settling time, and rise time. The compensator
poles, zeros, and multiplying constant are selected to give feedback system pole locations, that result in
acceptable transient response to step inputs. At the same time, the parameters are constrained so that the
resulting system has acceptable steady-state response to important inputs, such as steps and ramps.
Frequency response characterizations of systems have long been popular because of the ease and practi-
cality of steady-state sinusoidal response measurements. These methods also apply to systems in which
rational transfer function models are not adequate, such as those involving time delays. They do not require
explicit knowledge of system transfer function models; experimentally obtained open-loop sinusoidal
response data can directly be used for stability analysis and compensator design. A stability test, the Nyquist
criterion, is available. Principal measures of transient performance are gain margin, phase margin, and
bandwidth. The design of the compensator is conveniently carried out using the Bode plot and the Nichols
chart. One begins with simple compensators, increasing their complexity until the transient and steady-state
performance requirements are met.
There are two approaches to carrying out the digital controller (compensator) design. The first approach
uses the methods discussed above to design an analog compensator, and then transform it into a digital one.
The second approach, first transforms analog plants into digital plants, and then carries out the design using
digital techniques. The first approach performs discretization after design; the second approach performs
 Introduction 19
discretization before design. The classical approach to designing a digital compensator directly using an
equivalent digital plant for a given analog plant, parallels the classical approach to analog compensator
design. The concepts and tools of the classical digital design procedures will be given in Chapters 2–4. This
background will also be useful in understanding and applying the state-variable methods to follow.

1.5 SCOPE AND ORGANIZATION OF THE BOOK

This text is concerned with digital control and state variable methods for a special class of control systems,
namely time-invariant, lumped, and deterministic systems. We do not intend to solve all the problems that
can be posed under the defined category. Coverage of digital control theory and practice is modest. Various
concepts from interdisciplinary fields of computer science and computer engineering which relate to digital
control system development—number representation, logical descriptions of algorithmic processes, compu-
ter arithmetic operations, computer system hardware and software—are beyond the scope of this book. In
fact, a course on control engineering need not include these topics because specialized courses on computer-
system architecture are normally offered in undergraduate curricula of all engineering disciplines.
It is assumed that the reader has had an introductory course in control engineering concentrating on
the basic principles of feedback control and covering various classical analog methods of control system
design. The classical digital methods of design are developed in Part I of this book, paralleling and extend-
ing considerably the similar topics in analog control.
State variable analysis and design methods are usually not covered in an introductory course. It is
assumed that the reader is not exposed to the so-called modern control theory. Our approach in Part II of this
book is to first discuss state variable methods for continuous-time systems, and then give a compact presen-
tation of the methods for discrete-time systems, using the analogy with the continuous-time case.
This text also prepares a student for the study of advanced control methods. However, detailed coverage
of these methods is beyond the scope of the book; only a brief account is given in Part III.
There are eleven chapters in the book in addition to this introductory chapter. In each chapter, analysis/
design examples are interspersed to illustrate the concepts involved. At the end of each chapter, there are a
number of review examples that take the reader to a higher level of application; some of these examples also
serve the purpose of extending the text material. The same approach is followed in unsolved problems.
Answers to problems have been given to inspire confidence in the reader.
The examples we have considered in this book are generally low-order systems. Such a selection of
examples helps in conveying the fundamental concepts of feedback control without the distraction of large
amounts of computations, inherent in high-order systems. Many of the real-life design problems are more
complex than the ones discussed in this book. High-order systems are common and, in addition, several
parameters are to be varied during design stage to investigate their effect on the system performance. Com-
puter-Aided-Design (CAD) tools are extremely useful for complex control problems. Several software
packages with computer graphics are available commercially for CAD of control systems [151–154].
Let us now go through the organization of the chapters. Chapters 2–4 deal with digital control theory and
practice. The philosophy of presentation is that the new material should be closely related to material
already familiar, and yet, by the end, a direction to wider horizons should be indicated. The approach leads
us, for example, to relate the z-transform to the Laplace transform and to describe the implications of poles
and zeros in the z-plane to those known meanings attached to poles and zeros in the s-plane. Also, in devel-
oping the design methods we relate the digital control design methods to those of continuous-time systems.
Chapter 2 introduces the sampling theorem and the phenomenon of aliasing. Methods to generate
discrete-time models which approximate continuous-time dynamics, and stability analysis of these models
are also included in this chapter.
 20 Digital Control and State Variable Methods
Chapter 3 considers the hardware (analog and digital) of the control loop with emphasis on modelling.
Models of some of the widely used digital control systems are also included.
Chapter 4 establishes a toolkit of design-oriented techniques. It puts forward alternative design methods
based on root-locus and Bode plots. Design of digital controllers using z-plane synthesis is also included in
this chapter. References for the material in Chapters 2–4 are [11–12, 23–25, 30–31, 52, 80–98].
Chapters 5–8 deal with state variable methods in automatic control. Chapter 5 is on state variable analysis.
It exposes the problems of state variable representation, diagonalization, solution, controllability, and
observability. The relationship between the transfer function and state variable models is also given.
Although it is assumed that the reader has the necessary background on vector-matrix analysis, a reasonably
detailed account of vector-matrix analysis is provided in this chapter for convenient reference.
State variable analysis concepts, developed in continuous-time format in Chapter 5, are extended to
digital control systems in Chapter 6.
The techniques of achieving desired system characteristics by pole-placement using complete state
variable feedback are developed in Chapter 7. Also included is the method of using the system output to
form estimates of the states for use in state feedback. Results are given for both continuous-time and
discrete-time systems.
Lyapunov stability analysis is introduced in Chapter 8. In addition to stability analysis, Lyapunov func-
tions are useful in solving some optimization problems. We discuss in this chapter, the solution of linear
quadratic optimal control problem through Lyapunov synthesis. Results are given for both continuous-time
and discrete-time systems. References for the material in Chapters 5–8 are [27–28, 99–119, 122–124].
Describing function and phase-plane methods, which have demonstrated great utility in analysis of
nonlinear systems, have been paid considerable attention in Chapter 9. Also included is stability analysis of
nonlinear systems using Lyapunov functions. References for the material in Chapter 9 are [125-129].
The concepts of feedback linearization, model reference adaptive control, system identification and self-
tuning control, and variable structure control are briefly introduced in Chapter 10. Also included in this
chapter, is an introduction to an emerging nonlinear-control architecture—the reinforcement learning
control. For detailed study, refer [120–121, 125–136, 147–150].
A tutorial introduction to knowledge-based tools (neural networks, support vector machines, fuzzy logic,
genetic algorithms) for control system design is given in Chapters 11 and 12. Rigorous characterization of
theoretical properties of intelligent control methodology is not our aim in our tutorial presentation; rather,
we focus on the development of systematic engineering procedures, which will guide the design of controller
for a specific problem. For detailed study, refer [137–146].
Appendices A and B provide an introduction to the MATLAB environment for computer-aided control
system design.
 Signal Processing in Digital

2 Control

2.1 WHY USE DIGITAL CONTROL

Digital control systems offer many advantages over their analog counterparts. Of course, there are possible
problems also. Let us first look at the advantages of digital control over the corresponding analog control
before we talk of the price one has to pay for the digital option.

2.1.1 Advantages Offered by Digital Control

Flexibility An important advantage offered by digital control is in the flexibility of its modifying controller
characteristics, or of adapting the controller if plant dynamics change with operating conditions. The ability to
‘redesign’ the controller by changing software (rather than hardware) is an important feature of digital control
as against analog control.
Wide Selection of Control Algorithms Implementation of advanced control techniques was earlier
constrained by the limitations of analog controllers and the high costs of digital computers. However, with
the advent of inexpensive digital computers with virtually limitless computing power, the techniques of
modern control theory may now be put to practice. For example, in multivariable control systems with more
than one input and one output, modern techniques for optimizing system performance or reducing inter-
actions between feedback loops can now be implemented.
Integrated Control of Industrial Systems Feedback control is only one of the functions of the compu-
ter. In fact, most of the information transfer between the process and the computer exploits the logical
decision-making capability of the computer. Real-time applications of information processing and decision-
making, e.g., production planning, scheduling, optimization, operations control, etc., may now be integrated
with the traditional process control functions.
To enable the computer to meet the variety of demands imposed on it, its tasks are time-shared.
Future Generation Control Systems The study of emerging applications shows that artificial intelli-
gence (AI) will affect design and application of control systems, as profoundly as the impact of the micro-
processor has been in the last two decades. It is clear that future generation control systems will have a
significant AI component; the list of applications of computer-based control will continue to expand.
 22 Digital Control and State Variable Methods

2.1.2 Implementation Problems in Digital Control

The main problems associated with implementation of digital control, are related to the effects of sampling
and quantization.
Most processes that we are called upon to control, operate in continuous time. This implies, that we are
dealing largely with an analog environment. To this environment, we need to interface the digital computer,
through which we seek to influence the process.
The interface is accomplished by a system
Discrete-time signals of the form shown in Fig. 2.1. It is a cascade of
analog-to-digital (A/D) conversion system fol-
lowed by a computer which is, in turn, followed
A/D Computer D/A by a digital-to-analog (D/A) conversion system.
The A/D conversion process involves deriving
Continuous-time signals
samples of the analog signal at discrete instants
of time separated by sampling period T sec. The
Fig. 2.1 Discrete-time processing of continuous- D/A conversion process involves reconstructing
time signals continuous-time signals from the samples given
out by the digital computer.
Quantization Effects The conversion of signals from analog into digital form and vice versa is per-
formed by electronic devices (A/D and D/A converters) of finite resolution. A device of n-bit resolution
has 2n quantization levels. Here, the analog signal gets tied to these finite number of quantization levels in
the process of conversion to digital form. Therefore, by the sheer act of conversion, a valuable part of
information about the signal, is lost.
Furthermore, any computer employed as a real-time controller must perform all the necessary calcula-
tions with limited precision, thus introducing a truncation error after each arithmetic operation has been
performed. As the computational accuracy is normally much higher than the resolution of real converters, a
further truncation must take place before the computed data are converted into the analog form. The
repeated process of approximate conversion–computation–conversion may be costly, if not disastrous, in
terms of control system performance.
The process of quantization in signal conversion systems will be discussed shortly.
Sampling Effects The selection of a sampling period is a fundamental problem in digital control systems.
Later in this chapter, we will discuss the sampling theorem which states that the sampling period T should be
chosen such that
T < p /wm
where wm is the strict bandwidth of the signal being sampled. This condition ensures that there is no loss of
information due to sampling and the continuous-time signal can be completely recovered from its samples
using an ideal low-pass filter.
There are, however, two problems associated with the use of this theorem in practical control systems:
(i) Real signals are not band-limited and hence strict bandwidth limits are not defined.
(ii) The ideal low-pass filter, needed for the distortionless reconstruction of continuous-time signals from its
samples, is not physically realizable. Practical devices, such as the D/A converter, introduce distortions.
Thus, the process of sampling and reconstruction also affects the amount of information available to the
control computer, and degrades control system performance. For example, converting a given continuous-
time control system into a digital control system, without changing the system parameters, degrades the
system stability margin.
 Signal Processing in Digital Control 23
The ill-effects of sampling can be reduced, if not eliminated completely, by sampling at a very high rate.
However, excessively fast sampling (T ® 0) may result in numerical ill-conditioning in the implementation
of recursive control algorithms (described later in this chapter).
With the availability of low-cost, high-performance digital computers and interfacing hardware, the
implementation problems in digital control do not pose a serious threat to its usefulness. The advantages of
digital control outweigh its implementation problems for most of the applications.
This book attempts to provide a modest coverage of digital control theory and practice. In the present
chapter, we focus our attention on the digital computer and its interface to signal conversion systems
(Fig. 2.1). The goal is to formulate tools of analysis necessary to understand and guide the design of pro-
grams for a computer acting as a control logic component. Needless to say, digital computers can do many
things other than control dynamic systems; our purpose is to examine their characteristics when doing the
elementary control task.

2.2 CONFIGURATION OF THE BASIC DIGITAL CONTROL SCHEME

Figure 2.2 depicts a block diagram of a digital control system showing a configuration of the basic control
scheme. The basic elements of the system are shown by the blocks.
Controlled
Disturbance
output
Digital
set-point
Final
A/D Computer D/A control Plant
element

Clock

Anti-aliasing
S/H Sensor
filter

Fig. 2.2 Configuration of the basic digital control scheme

The analog feedback signal coming from the sensor is usually of low frequency. It may often include high
frequency ‘noise.’ Such noise signals are too fast for control system to correct; low-pass filtering is often
needed to allow good control performance. The anti-aliasing filter shown in Fig. 2.2 serves this purpose. In
digital systems, the phenomenon called aliasing introduces some new aspects to the noise problems. We will
study this phenomenon later in this chapter.
The analog signal, after anti-aliasing processing, is converted into digital form by the A/D conversion
system. The conversion system usually consists of an A/D converter preceded by a sample-and-hold (S/H)
device. The A/D converter converts a voltage (or current) amplitude at its input into a binary code represent-
ing a quantized amplitude value closest to the amplitude of the input. However, the conversion is not
instantaneous. Input signal variation, during the conversion time of the A/D converter, can lead to erroneous
results. For this reason, high performance A/D conversion systems include a S/H device, which keeps the
input to the A/D converter, constant during its conversion time.
 24 Digital Control and State Variable Methods
The digital computer processes the sequence of numbers by means of an algorithm and produces a new
sequence of numbers. Since data conversions and computations take time, there will always be a delay when
a control law is implemented using a digital computer. The delay, which is called the computational delay,
degrades the control system performance. It should be minimized by the proper choice of hardware and by
the proper design of software for the control algorithm. Floating-point operations take a considerably longer
time to perform (even when carried out by arithmetic co-processor) than the fixed-point ones. We therefore
try to execute fixed-point operations whenever possible. Alternative realization schemes for a control
algorithm will be given in the next chapter.
The D/A conversion system in Fig. 2.2 converts the sequence of numbers in numerical code into a
piecewise continuous-time signal. The output of the D/A converter is fed to the plant through the actuator
(final control element) to control its dynamics.
The basic control scheme of Fig. 2.2 assumes a uniform sampling operation, i.e., only one sampling rate
exists in the system and the sampling period is constant. The real-time clock in the computer, synchronizes
all the events of A/D conversion–computation–D/A conversion.
The control scheme of Fig. 2.2 shows a single feedback loop. In a control system having multiple loops,
the largest time constant involved in one loop may be quite different from that in other loops. Hence, it may
be advisable to sample slowly in a loop involving a large time constant, while in a loop involving only small
time constants, the sampling rate must be fast. Thus, a digital control system may have different sampling
periods in different feedback paths, i.e., it may have multiple-rate sampling. Although digital control
systems with multi-rate sampling are important in practical situations, we shall concentrate on single-rate
sampling. (The reader interested in multi-rate digital control systems may refer Kuo [87]).
The overall system in Fig. 2.2 is hybrid in nature; the signals are in a sampled form (discrete-time signals/
digital signals) in the computer and in continuous-time form in the plant. Such systems have traditionally
been called sampled-data control systems. We will use this term as a synonym to computer control systems/
digital control systems.
In the present chapter, we focus our attention on digital computer and its analog interfacing. For the time
being, we delink the digital computer from the plant. The link will be re-established in the next chapter.

2.3 PRINCIPLES OF SIGNAL CONVERSION

Figure 2.3a shows an analog signal y(t)—it is defined at the continuum of times, and its amplitudes assume
a continuous range of values. Such a signal cannot be stored in digital computers. The signal, therefore, must
be converted to a form that will be accepted by digital computers. One very common method to do this is
to record sample values of this signal at equally spaced instants. If we sample the signal every 10 msec,
for example, we obtain the discrete-time signal sketched in Fig. 2.3b. The sampling interval of 10 msec
corresponds to a sampling rate of 100 samples/sec. The choice of the sampling rate is an important one,
since it determines how accurately the discrete-time signal, can represent the original signal.
In a practical situation, the sampling rate is determined by the range of frequencies present in the original
signal. Detailed analysis of uniform sampling process, and the related problem of aliasing will appear later
in this chapter.
Notice that the time axis of the discrete-time signal in Fig. 2.3b, is labelled simply ‘sample number’ and
index k has been used to denote this number (k = 0, 1, 2, ...). Corresponding to different values of sample
number k, the discrete-time signal assumes the same continuous range of values assumed by the analog
signal y(t). We can represent the sample values by a sequence of numbers ys (refer Fig. 2.3b):
ys = {1.7, 2.4, 2.8, 1.4, 0.4, ...}
 Signal Processing in Digital Control 25
y (t) y ( t)

3 3

2 2

1 1

0 10 20 30 40 0 1 2 3 4
t (msec) Sample number k
y q (k ) (a) (b)

3
k Digital word
0 10
2
1 10
2 11
1
3 01
4 00
k
0 1 2 3 4
(c) (d)
Fig. 2.3 Sampling, quantization and coding of an analog signal

In general,
ys = {y(k)}, 0 ≤ k < ∞
where y(k) denotes the kth number in the sequence.
The sequence defined above is one-sided sequence; ys = 0 for k < 0. In digital control applications, we
normally encounter one-sided sequences.
Although, strictly speaking, y(k) denotes the kth number in the sequence, the notation given above is often
unnecessarily cumbersome, and it is convenient and unambiguous to refer to y(k) itself as a sequence.
Throughout our discussion on digital control, we will assume uniform sampling, i.e., sample values of the
analog signal are extracted at equally spaced sampling instants. If the physical time, corresponding to the
sampling interval is T seconds, then the kth sample y(k), gives the value of the discrete-time signal at t = kT
seconds. We may, therefore, use y(kT) to denote a sequence wherein the independent variable is the physical time.
The signal of Fig. 2.3b is defined at discrete instants of time. The sample values are, however, tied to a
continuous range of numbers. Such a signal, in principle, can be stored in an infinite-bit machine because a
finite-bit machine can store only a finite set of numbers.
A simplified hypothetical 2-bit machine can store four
numbers given alongside:
Binary number Decimal equivalent
The signal of Fig. 2.3b can be stored in such a machine if the
00 0 sample values are quantified to four quantization levels. Figure
01 1 2.3c shows a quantized discrete-time signal for our hypotheti-
10 2 cal machine. We have assumed that any value in the interval
[0.5, 1.5) is rounded to 1, and so forth. The signals for which
11 3
both time and amplitude are discrete, are called digital signals.
 26 Digital Control and State Variable Methods
After sampling and quantization, the final step required in converting an analog signal to a form accept-
able to digital computers is coding (or encoding). The encoder maps each quantized sample value into a
digital word. Figure 2.3d gives coded digital signal, corresponding to the analog signal of Fig. 2.3a for our
hypothetical 2-bit machine.
The device that performs the sampling, quantization, and coding is an A/D converter. Figure 2.4 is a
block diagram representation of the operations performed by an A/D converter.
Continuous-time
continuous-amplitude Digital
signal words

Sampler Quantizer Encoder

Discrete-time Discrete-time
continuous-amplitude discrete-amplitude
signal signal
Fig. 2.4 Operations performed by an A/D converter
It may be noted that the quantized discrete-time signal of Fig. 2.3c and the coded signal of Fig. 2.3d carry
exactly the same information. For the purpose of analytical study of digital systems, we will use the quantized
discrete-time form for digital signals.
The number of binary digits carried by a device is its word length, and this is obviously an important
characteristic related to the resolution of the device—the smallest change in the input signal that will pro-
duce a change in the output signal. The A/D converter that generates signals of Fig. 2.3 has two binary digits
and thus four quantization levels. Any change, therefore, in the input over the interval [0.5, 1.5) produces no
change in the output. With three binary digits, 23 quantization levels can be obtained, and the resolution of
the converter would improve.
The A/D converters in common use have word lengths of 8 to 16 bits. For an A/D converter with a word
length of 8 bits, an input signal can be resolved to one part in 28, or 1 in 256. If the input signal has a range
of 10 V, the resolution is 10/256, or approximately 0.04 V. Thus, the input signal must change by at least
0.04 V, in order to produce a change in the output.
With the availability of converters with resolution ranging from 8 to 16 bits, the quantization errors do not
pose a serious threat in the computer control of industrial processes. In our treatment of the subject, we
assume quantization errors to be zero. This is equivalent to assuming infinite-bit digital devices. Thus we
treat digital signals as if they are discrete-time signals with amplitudes assuming a continuous range of
values. In other words, we make no distinction between the words ‘discrete-time’ and ‘digital.’
A typical topology of a single-loop digital control system is shown in Fig. 2.2. It has been assumed that the
measuring transducer and the actuator (final control element) are analog devices, requiring respectively A/D
and D/A conversion at the computer input and output. The
Digital Discrete-time Analog
signal signal D/A conversion is a process of producing an analog signal
words
from a digital signal and is, in some sense, the reverse of
the sampling process discussed above.
Zero-order The D/A converter performs two functions: first,
Decoder
hold generation of output samples from the binary-form
Fig. 2.5 Operations performed by a D/A digital signals produced by the machine, and second, con-
converter version of these samples to analog form. Figure 2.5 is
 Signal Processing in Digital Control 27
a block diagram representation of the operations performed by a D/A converter. The decoder maps each
digital word into a sample value of the signal in discrete-time form. It is usually not possible to drive a
load, such as a motor, with these samples. In order to deliver sufficient energy, the sample amplitude
might have to be so large that it is infeasible to be generated. Also large-amplitude signals might saturate
the system being driven.
The solution to this problem is to smooth the output samples to produce a signal in analog form. The
simplest way of converting a sample sequence into a continuous-time signal is to hold the value of the
sample until the next one arrives. The net effect is to convert a sample to a pulse of duration T—the sample
period. This function of a D/A converter is referred to as a zero-order hold (ZOH) operation. The term zero-
order refers to the zero-order polynomial used to extrapolate between the sampling times (detailed discus-
sion will appear later in this chapter). Figure 2.6 shows a typical sample sequence produced by the decoder,
and the analog signal1 resulting from the zero-order hold operation.

y (k ) y h ( t)

3 3

2 2

1 1

k t
0 1 2 3 4 0 T 2T 3T 4T
(a) (b)
Fig. 2.6 (a) Sampled sequence (b) Analog output from ZOH

2.3.1 D/A Converter Circuits

Most D/A converters use the principle shown in the three-bit form in Fig. 2.7 to convert the HI/LO digital
signals at the computer output to a single analog voltage. The circuit of Fig. 2.7 is an ‘R–2R’ ladder; the
value of R typically ranges from 2.5 to 10K ohms.
Suppose a binary number b2 b1 b0 is given. The switch (actually, electronic gates) positions in Fig. 2.7
correspond to the digital word 100, i.e., b2 = 1 and b1 = b0 = 0. The circuit can be simplified to the equivalent
form shown in Fig. 2.8a. The currents in the resistor branches are easily calculated and are indicated in the
circuit (for the high gain amplifier, the voltage at point A is practically zero [155]). The output voltage is
i2 1
V0 = 3R = Vref
2 2
If b1 = 1 and b2 = b0 = 0, then the equivalent circuit is as shown in Fig. 2.8b. The output voltage is
i 1
V0 = 3R 1 = Vref
4 4

1
In the literature, including this book, the terms ‘continuous-time signal’ and ‘analog signal’ are frequently
interchanged.
 28 Digital Control and State Variable Methods

Full scale

7/8
(ratio to full scale)
6/8
Analog output

5/8
4/8

3/8

2/8
1/8

000 001 010 011 100 101 110 111

Digital code

3R

2R R R 2R
–
A
V0
2R 2R 2R +

b0 b1 b2

– V ref
Fig. 2.7 Three-bit D/A converter

Similarly, if b0 = 1 and b2 = b1 = 0, then the equivalent circuit is as shown in Fig. 2.8c. The output
voltage is
i 1
V0 = 3R 0 = Vref
8 8
In this way, we find that when the input data is b2b1b0 (where the bi’s are either 0 or 1), then the output
voltage is
V0 = (b22–1 + b12–2 + b02–3)VFS (2.1)
where VFS = Vref = full scale output voltage.
The circuit and the defining equation for an n-bit D/A converter easily follow from Fig. 2.7 and Eqn. (2.1),
respectively.
 Signal Processing in Digital Control 29
3R
1
i
2 2
2R 2R
–
1 1 A
i i
2 2 2 2 V0
i2
2R +

– V ref
(a)
3R
1
i
2R R 2R 4 1
–
1 1 1
i i i
2 1 2 1 4 1 V0
i1 +
2R 2R 1
i
4 1

– V ref
(b)
3R
1
i
2R R R 2R 8 0
–
1 1 1 1
i i i i
2 0 2 0 4 0 8 0 V0
2R i0 2R 2R +
1 1
i i
4 0 8 0

– Vref
(c)
Fig. 2.8 Equivalent circuits of the D/A converter shown in Fig. 2.7;
(a) b0 = b1 = 0, b2 = 1; (b) b0 = 0, b1 = 1, b2 = 0; (c) b0 = 1, b1 = b2 = 0.

2.3.2 A/D Converter Circuits

Most A/D converters use the principle of successive approximation. Figure 2.9 shows the organization of an
A/D converter that uses this method. Its principal components are a D/A converter, a comparator, a succes-
sive approximation register (SAR), a clock, and control and status logic.
On receiving the (start-of-conversion) SOC command, the SAR is cleared to 0’s and its most significant
bit is set to 1. This results in a V0 value that is one-half of the full scale (refer Eqn. (2.1)). The output of the
comparator is then tested to see whether VIN is greater than or less than V0. If VIN is greater, the most
significant bit is left on; otherwise it is turned off (complemented).
 30 Digital Control and State Variable Methods

Clock

Analog + Control and status SOC

input VIN logic
– EOC
Comparator
SAR

Digital
output

V0
D /A V FS

Fig. 2.9 Organization of a successive approximation A/D converter

111 In the next step, the next most significant bit of the
111
110 SAR is turned on. At this stage, V0 will become either
110 three-quarters or one-quarter of the full scale, depend-
101
101
ing on whether VIN was, respectively, greater than or
100 less than V0 in the first step. Again, the comparator is
100 tested and if VIN is greater than the new V0, the next
011
most significant bit is left on. Otherwise it is turned off.
011
010 The process is repeated for each remaining SAR bit.
010 When the process has been carried out for each bit, the
001 SAR contains the binary number that is proportional to
001
000 VIN, and the (end-of-conversion) EOC line indicates
Fig. 2.10 Code sequence for a three-bit that comparison has been completed and digital output
successive approximation is available for transmission. Figure 2.10 gives the code
sequence for a three bit successive approximation.
Typical conversion times of commercial A/D units range from 10 nsec to 200 msec. The input VIN
in Fig. 2.9 should remain constant during the conversion time of A/D converter. For this reason, a high
performance A/D conversion system includes a S/H device which keeps the input to the A/D converter,
constant during its conversion time. The S/H operation will be described in Section 2.10.

2.4 BASIC DISCRETE-TIME SIGNALS

There are a number of basic discrete-time signals which play an important role in the analysis of signals and
systems. These signals are direct counterparts of the basic continuous-time signals.2 As we shall see, many of
the characteristics of basic discrete-time signals are directly analogous to the properties of basic continuous-
time signals. There are, however, several important differences in discrete time, and we will point these out
as we examine the properties of these signals.
2
Chapter 2 of the companion book [155].
 Signal Processing in Digital Control 31
Unit Sample Sequence The unit sample sequence contains only one nonzero element and is defined by
(Fig. 2.11a)

d (k) =
RS1 for k = 0
(2.2a)
T0 otherwise
The delayed unit sample sequence, denoted by d (k – n), has its nonzero element at sample time n (Fig. 2.11b):

d (k – n) =
RS1 for k = n
(2.2b)
T0
otherwise
One of the important aspects of the unit sample sequence is that an arbitrary sequence can be represented
as a sum of scaled, delayed unit samples. For example, the sequence r (k) in Fig. 2.11c can be expressed as
r(k) = r(0)d (k) + r(1)d (k – 1) + r(2)d (k – 2) + L
¥
= å r(n)d (k – n) (2.3)
n = 0
r(0), r(1), K , are the sample values of the sequence r(k). This representation of a discrete-time signal is
found useful in the analysis of linear systems through the principle of superposition.

d (k ) d (k – n ) r (k )

1 1

k k k
0 1 2 3 0 1 2 n 0 1 2 3
(a) (b) (c)

m (k) m (k – n ) cos ( k) m (k)

1 1

k 3 5
k k
0 1 2 0 1 n 0 1
(d) (e)

(f)
Fig. 2.11 Basic discrete-time signals

As we will see, the unit sample sequence plays the same role for discrete-time signals and systems, that
the unit impulse function does for continuous-time signals and systems. For this reason, the unit sample
sequence is often referred to as the discrete-time impulse. It is important to note that a discrete-time impulse
does not suffer from the mathematical complexity that a continuous-time impulse suffers from. Its definition
is simple and precise.
 32 Digital Control and State Variable Methods
Unit Step Sequence The unit step sequence is defined as3 (Fig. 2.11d)

m(k) =
RS1 for k ³ 0
(2.4)
T0 otherwise
The delayed unit step sequence, denoted by m(k – n), has its first nonzero element at sample time n (Fig. 2.11e):

m(k – n) =
RS1 for k ³ n
(2.5)
T0 otherwise
An arbitrary discrete-time signal r(k) switched on to a system at k = 0 is represented as r(k)m(k).
Sinusoidal Sequence A one-sided sinusoidal sequence has the general form (Fig. 2.11f )
r(k) = A cos(W k + f) m(k) (2.6)
The quantity W is called the frequency of the discrete-time sinusoid and f is called the phase. Since k is a
dimensionless integer, the dimension of W must be radians (we may specify the units of W to be radians/
sample, and units of k to be samples).
The fact that k is always an integer in Eqn. (2.6) leads to some differences between the properties of
discrete-time and continuous-time sinusoidal signals. An important difference lies in the range of values the
frequency variable can take on. We know that for the continuous-time signal r(t) = A cos w t = real {Ae jw t},
w can take on values in the range (– ∞ , ∞ ). In contrast, for the discrete-time sinusoid r(k) = A cos Wk =
real {Ae jWk}, W can take on values in the range [– p, p].
To illustrate the property of discrete-time sinusoids, consider W = p + x, where x is a small number
compared with p. Since
e jWk = e j(p + x)k = e j(2p – p + x)k = e j(– p + x)k
a frequency of (p + x) results in a sinusoid of frequency (– p + x). Suppose now, that W is increased to 2p.
Since e j2pk = e j0, the observed frequency is 0. Thus, the observed frequency is always between – p and p, and
is obtained by adding (or subtracting) multiples of 2p to W until a number in that range is obtained.
The highest frequency that can be represented by a digital signal is therefore p radians/sample interval.
The implications of this property for sequences obtained by sampling sinusoids and other signals will be
discussed in Section 2.11.

2.5 TIME-DOMAIN MODELS FOR DISCRETE-TIME SYSTEMS

A discrete-time system is defined mathematically, as a transformation, or an operator, that maps an input

sequence r(k) into an output sequence y(k). Classes of discrete-time systems are defined by placing con-
straints on the transformation. As they are relatively easy to characterize mathematically, and as they can be
designed to perform useful signal processing functions, the class of linear time-invariant systems will be
studied in this book. In the control structure of Fig. 2.2, the digital computer transforms an input sequence
into a form which is in some sense more desirable. Therefore, the discrete-time systems we consider here
are, in fact, computer programs.
As we shall see, there is a similarity in the structure of models of continuous-time and discrete-time systems.
This has resulted in the development of similar methods of analysis. For example, the simulation diagrams of
discrete-time systems are similar to those for continuous-time systems, with only the dynamic element changed
3
In discrete-time system theory, the unit step sequence is generally denoted by u(k). In control theory, u(k) is used to
represent the control signal. In this book, m(k) has been used to represent the unit step sequence while u(k) denotes
the control signal.
Exploring the Variety of Random
Documents with Different Content
Paris. Comme si la cité avait dû revêtir, dès cette époque, le
caractère cosmopolite qui la distingue aujourd'hui, on y voyait,
alignés le long de la chaussée ou espacés des deux côtés dans les
champs, des tombeaux qui emmenaient la pensée aux extrémités
les plus opposées du monde ancien. Les inscriptions y parlaient les
deux langues de la civilisation, et le voyageur s'acheminait à travers
des avenues funéraires qui faisaient passer tour à tour sous ses
yeux les monuments du paganisme romain, les édicules étranges de
Mithra, et les chastes et sobres emblèmes de la foi chrétienne. Là
dormait, au milieu de plusieurs de ses successeurs, l'évêque de
Paris, saint Prudence, et l'on veut que les chrétiens des premiers
âges y aient possédé une catacombe où ils célébraient les sacrés
mystères, et qui s'élevait sur les ruines d'un ancien sanctuaire de
Diane, la déesse des forêts[317].
[316] Voir tome I, p. 103.
[317] Lebeuf, Histoire de la ville et de tout le diocèse de Paris, nouvelle édition,
Paris, 1883, t. I, pp. 228 et suiv.; Saintyves, Vie de sainte Geneviève, pp. 101, 130,
295; Franklin, dans Paris à travers les âges, IX, p. 2
Plus d'une fois, le regard de Clovis et de Clotilde s'était arrêté sur ce
tranquille horizon, des hauteurs duquel semblait descendre jusqu'à
eux, à travers le murmure des verdoyants ombrages, la solennelle
invitation de la mort. L'idée leur sourit d'y répondre en préparant là-
haut la place de leur dernière demeure, à l'abri d'un sanctuaire qui
serait le monument durable de leur foi commune, et qui dresserait
au-dessus de toute la vallée le signe glorieux de la résurrection.
Toujours le souvenir de Clotilde a été associé à celui de Clovis dans
l'histoire de cet édifice sacré[318]; il n'est guère douteux qu'elle en ait
suggéré la première idée au roi. Un chroniqueur parisien du huitième
siècle, dont les souvenirs locaux ont souvent une grande valeur
historique, attribue formellement cette initiative à Clotilde. Il est vrai
que, d'après lui, c'était dans la pensée du couple royal une église
votive, qui devait être bâtie si le roi revenait victorieux de la guerre
d'Aquitaine[319]. Ce qui est certain, c'est que la construction n'en fut
commencée que dans les dernières années, puisqu'elle n'était pas
achevée lorsque Clovis mourut.
[318] Grégoire de Tours, ii, 43: Basilica sanctorum Apostolorum quam cum
Chrodechilde regina ipse construxerat.—Le même, iv, 1: Nam basilicam illam ipsa
construxerat.
[319] Liber historiæ, c. 17.

Le roi voulut inaugurer les travaux avec toute la solennité du rite

germanique, si nouveau et si curieux, dans son formalisme barbare,
pour la population romaine de Paris. Aussi en a-t-elle gardé le
souvenir comme d'une chose qui se voit rarement, et son
chroniqueur a-t-il cru devoir transmettre à la postérité le récit de
l'étrange cérémonie. Debout et en armes sur le terrain qu'il se
proposait d'attribuer à la nouvelle église, Clovis, de toute la force de
son bras, lança droit devant lui sa hache d'armes, cette francisque
dont le tranchant avait fendu plus d'un crâne ennemi[320]. Par cet
acte symbolique, il indiquait qu'il entendait prendre possession du
sol à la façon du guerrier victorieux, ou encore du dieu Thor lui-
même quand, lançant le redoutable marteau de sa foudre sur la
terre, il s'emparait à jamais du domaine qu'avaient touché ses traits
enflammés.
[320] Liber historiæ, l. c. Je ne sais pas pourquoi M. Krusch, Neues Archiv., xviii,
p. 42 conteste à cette tradition historique toute valeur, comme s'il s'agissait d'une
invention légendaire. En réalité, elle se borne à nous apprendre la cérémonie par
laquelle fut inaugurée la construction: or cette cérémonie, essentiellement barbare,
était usitée dans tous les cas analogues et resta jusqu'en plein moyen âge en
vigueur dans les pays germaniques, et dans ceux qui avaient subi l'influence des
Germains. Cf. J. Grimm, Deutsche Rechtsalterthümer, 2e édition, (1854) pp. 54-
68.

Bientôt l'église surgit du sol, appuyée sur une crypte qui devait
recevoir les sépultures royales, et offrant aux regards l'aspect des
primitives basiliques. Elle pouvait avoir, nous dit un historien, deux
cents pieds de long sur cinquante à soixante de large[321]. L'intérieur
en était non voûté, mais lambrissé à la manière antique; de riches
mosaïques ainsi que des peintures murales en animaient les parois.
On y avait accès, du côté occidental, par un triple portique orné,
comme l'intérieur, de mosaïques et de peintures représentant des
scènes de l'Ancien et du Nouveau Testament[322]. A côté de l'église
s'élevèrent de spacieux bâtiments conventuels pour la demeure des
chanoines réguliers qui devaient la desservir. Un vaste territoire,
longeant les jardins du palais et allant d'un côté jusqu'à la Seine et
de l'autre jusqu'à la Bièvre, forma la seconde enceinte de cette
fondation vraiment royale. La plus grande partie en était occupée par
des closeries et des vignobles à travers lesquels circulaient
d'ombreux sentiers de noyers et d'amandiers chantés au douzième
siècle, en vers agréables, par le poète Jean de Hautefeuille. Le
douaire assigné au monastère était considérable: il comprenait
Nanterre, Rosny, Vanves, Fossigny, Choisy, et la terre connue sous
le nom de fief de Sainte-Clotilde[323].
[321] Viallon, Vie de Clovis le Grand, pp. 448 et suiv.
[322] Vita sanctæ Genovefæ, xi, 53 (Kohler): Miracula sanctæ Genovefæ; cf. du
Molinet, Histoire de sainte Geneviève et de son église royale et apostolique à
Paris, manuscrit de la bibliothèque Sainte-Geneviève, livre III, chap. ii.
[323] Du Molinet, o. c., livre III, chap. iii, suivi par les autres historiens de sainte
Geneviève. Le livre de du Molinet, resté inédit, est un travail excellent, qu'il n'y
aurait plus intérêt à publier toutefois, parce que la meilleure partie en a passé
depuis dans les travaux consacrés au même sujet.
Clovis ne vécut pas assez longtemps pour voir l'achèvement de
cette fondation grandiose; c'est Clotilde qui la mit sous toit et qui en
termina les dépendances[324]. Il paraîtrait toutefois, si l'on en croit le
chroniqueur parisien auquel nous avons déjà fait des emprunts, que
le roi put encore assister à la consécration de l'église. Cet écrivain
ajoute que Clovis pria le pape de lui envoyer des reliques des saints
Pierre et Paul, parce qu'il voulait en faire les patrons du nouveau
sanctuaire, et qu'à cette occasion il fit tenir au souverain pontife de
riches cadeaux[325]. C'est probablement alors aussi qu'il lui envoya
une superbe couronne d'or, garnie de pierres précieuses, qu'on
appelait «le Règne»[326]. Plusieurs historiens du moyen âge ont
parlé de cette couronne qui mérite une mention ici, puisqu'elle fut le
premier hommage de la royauté très chrétienne à l'Église
universelle.
[324] Vita sancti Genovefæ, xi, 53 (Kohler).
[325] C'est ce que dit une note d'un des manuscrits du Liber historiæ, c. 17. Dans
tous les cas, elle se trompe tout au moins sur le nom du pape, qu'elle appelle
Hormisdas. Hormisdas ne monta sur le trône de saint Pierre qu'en 514, trois ans
après la mort de Clovis. Peut-être faut-il garder le nom d'Hormisdas, et remplacer
celui de Clovis par celui de Clotilde?
[326] Eodem tempore venit regnus cum gemmis preciosis a rege Francorum
Cloduveum christianum, donum beato Petro apostolo. Liber Pontificalis, éd.
Duchesne, t. I, p. 271. Ce passage, écrit au sixième siècle, a passé en substance
dans le Vita sancti Remigii de Hincmar, Acta Sanctorum, p. 156, F, et de là dans
Sigebert de Gembloux, Chronicon (dom Bouquet, III, p. 337). M. l'abbé Duchesne
écrit à ce sujet, o. c., p. 274: «Clovis mourut trois ans avant l'avènement du pape
Hormisdas. Il est possible que l'envoi du regnus ou couronne votive, dont il est ici
question, ait souffert quelque retard. Du reste, le nom de Clovis n'est attesté ici
que par les manuscrits de la seconde édition; l'abrégé Félicien coupe la phrase
après Francorum.» Cf. A. de Valois, I, pp. 270 et 299; dans ce dernier passage, il
fait dire à ses sources que Clovis envoya la couronne qu'il avait reçue d'Anastase.

Pendant que l'on poussait activement les travaux, le roi était

emporté par une mort précoce à l'âge de quarante-cinq ans, le 27
novembre 511[327]. Succomba-t-il aux fatigues de ses campagnes
ou aux vices d'une constitution minée par quelque mal héréditaire?
On ne saurait le dire, mais les courtes destinées de la plupart de ses
successeurs confirment une supposition que nous avons déjà faite
au sujet de ses ancêtres, et rendent plus vraisemblable la dernière
alternative. Son corps, enfermé dans un sarcophage de pierre de
forme trapézoïde, et dont des croix étaient tout l'ornement, fut
déposé dans la crypte vierge encore de la colline de Lutèce. Il
disparaissait de la scène du monde au moment où il semblait qu'il
allât goûter en paix les fruits de ses grands travaux, et consolider,
comme chef d'État, ce qu'il avait fondé comme guerrier. La
Providence, après l'avoir employé pendant trente ans à la création
d'une œuvre prédestinée, ne lui avait laissé que le temps de
préparer son tombeau.
[327] Sur l'âge de Clovis et sur l'année de sa mort, Grégoire de Tours, ii, 43. Sur le
jour, Viallon, Histoire de Clovis le Grand, p. 473, et les missels de sainte
Geneviève (manuscrit nº 1259, fol. 8, et manuscr. 90), qui donnent le 27 novembre
comme anniversaire de la depositio de Clovis. Dubos, III, p. 403, se demande s'il
ne s'agit pas plutôt du jour où les restes de ce roi furent déposés dans la crypte
après l'achèvement de l'église. La date du 27 novembre 511 est solidement
justifiée par M. Levison, Zur Geschichte des Frankenkönigs Clodowech (Bonner
Jahrbücher, t. 103, p. 47 et suivantes.)
Le moment est venu de porter un jugement d'ensemble sur cette
puissante personnalité.
Le fondateur de la monarchie franque n'est pas un de ces génies
transcendants comme Charlemagne, qui créent les événements par
la seule force de leur volonté souveraine, et qui laissent leur
empreinte indélébile sur toutes les choses auxquelles ils touchent.
C'est une nature hardie et énergique de conquérant, qui regarde les
destinées en face, et qui, sans trembler devant elles, va à leur
rencontre l'épée à la main. Il ne se préoccupe pas d'étudier la
signification prophétique des choses que l'histoire déroule devant lui;
il lui suffit de voir, d'un coup d'œil ferme et juste, la place qu'il y peut
prendre, et il se la fait large et belle. Son ambition n'est pas
insatiable; il a une idée bien nette de la limite de ce qui lui est
possible, et il ne la franchit point, quoi qu'il lui en puisse coûter. A
deux reprises, il évite prudemment d'en venir aux mains avec le seul
rival qui fût digne de lui; Théodoric put le gourmander après sa
victoire sur les Alamans, il put même arrêter dans le sud-est le cours
de ses victoires sans le décider à prendre les armes. Il faut savoir
gré au conquérant franc de cette modération, n'eût-elle même sa
source que dans un calcul. En politique, c'est une vertu encore pour
l'homme d'État d'obéir à la voix de son intérêt plutôt qu'aux chimères
de l'imagination. A la force et à la prudence, nous le voyons joindre
l'adresse. Il ne fait pas une entreprise sans se procurer des alliés, et
ceux-ci il les trouve, non seulement parmi les princes de sa famille,
mais encore parmi ses ennemis d'hier, à preuve l'arien Gondebaud,
dont il obtient l'alliance dans sa campagne contre les Visigoths
ariens. Il n'est pas moins avisé dans ses relations avec Byzance: il
accepte les honneurs qui lui sont offerts par l'empereur, il sait en
faire état aux yeux des populations gallo-romaines, mais il ne donne
rien en échange, et toute la finesse de la diplomatie impériale est
tenue en échec par sa tranquille réserve. Eut-il un idéal de
gouvernement, et cet idéal, quel fut-il? L'histoire n'a point pris la
peine de nous le dire, et nous ne le saurons peut-être jamais. C'est
pour cette raison sans doute qu'on a cru pouvoir lui préférer
Théodoric, dont la correspondance officielle parle souvent un si
magnifique langage. Mais cette supériorité n'est qu'apparente. Si
c'était le roi franc qui eût eu à sa disposition la plume de Cassiodore,
nul doute qu'on n'admirât le civilisateur dans Clovis, et que dans
Théodoric on ne vît que l'assassin d'Odoacre, le meurtrier de Boëce
et de Symmaque. De tout temps l'histoire s'est laissé faire illusion
par les lettres. Et le plus grand malheur, aux yeux de la postérité,
pour des créateurs d'État comme Clovis, c'est de n'avoir pas eu à
leur service une plume éloquente: carent quia vate sacro.
S'il s'agit d'apprécier l'homme après le souverain, nous connaissons
trop mal Clovis pour porter sur sa personne un jugement complet et
motivé. L'histoire ne nous a conservé de lui que le souvenir de
quelques faits d'armes; elle ignore tout le reste, elle ne sait rien de
sa vie privée. Cette lacune a été comblée par l'épopée, qui a enlaidi
sa physionomie en la dessinant d'après un idéal barbare, et qui a
mis un type de convention à la place du héros historique. Enfin, les
mœurs atroces des rois mérovingiens qui sont venus par la suite ont
jeté leur ombre sinistre en arrière sur la grande mémoire du
fondateur du royaume. Si bien que, l'histoire se taisant et
l'imagination ayant seule la parole, le Clovis qu'on nous a montré est
toujours le barbare d'avant le baptême. On ne voit pas en quoi il est
converti, on ne sait pas à quoi lui sert d'avoir été baptisé.
Pour retrouver la figure véritable du fondateur de la France, il faut
donc effacer de sa physionomie tous les traits dont la poésie
populaire l'a chargée à son insu. Ce travail, nous l'avons fait, et nous
avons lieu de croire qu'il est définitif. Il faut ensuite se prémunir
contre les suggestions fallacieuses de l'analogie. Invoquer la
barbarie des petits-fils pour faire croire à celle de l'aïeul, sous
prétexte que toutes les barbaries se ressemblent, c'est une erreur.
Le barbare converti, qui, touché de la grâce, est venu à Jésus-Christ
par le libre mouvement de sa volonté, ne doit pas être comparé à
celui qui a reçu le baptême dès l'enfance, mais qui ne réagit pas
contre les influences d'un milieu encore saturé de mœurs païennes.
Comme les convertis anglo-saxons, Ethelbert et Edwin, Clovis
occupe un niveau religieux fort supérieur à celui de ses
descendants. Les contemporains ne s'y sont pas trompés, à preuve
le parallèle établi entre eux et lui par Grégoire de Tours, et que nous
avons reproduit plus haut[328].
[328] Voir ci-dessus p. 190.
Et le poète inconnu qui a dépeint les visions prophétiques de la reine
Basine ne porte pas un autre jugement. Pour lui, Clovis est le lion;
ses fils sont comparés à des rhinocéros et à des léopards; ses
petits-fils ne sont plus que des ours et des loups[329]. Cette
impression eût été celle de tous les historiens, s'ils n'avaient eu
l'esprit prévenu par les légendes apocryphes. Non, il n'est pas
permis d'attribuer uniformément le même degré de barbarie à tous
les Francs. Les mœurs frénétiques des descendants de Clovis ne
suffisent pas pour accuser celui-ci, non plus que les crimes d'une
Frédégonde ne sont un argument contre la sainteté de Clotilde, de
Radegonde et de Bathilde. Ces nobles et chastes figures qui
passent, voilées et en prière, à travers un monde secoué par la
fièvre de toutes les passions, sont la preuve de la fécondité du
christianisme parmi les Francs, et protestent contre l'hypothèse
d'une barbarie qui n'aurait pas connu d'exception.
[329] Voir tome I, p. 201.
Si nous nous en tenons, pour juger Clovis, au petit nombre des faits
avérés qui composent l'histoire de son règne, il ne nous apparaîtra
pas sous un jour défavorable. Sans doute, nous le voyons, avant sa
conversion, frapper avec une vigueur cruelle un de ses guerriers qui
l'a offensé, de même qu'après son baptême il tue de sa main le
soldat qui a violé le ban du roi en pillant un homme de Saint-Martin;
mais il ne faut pas oublier qu'il usait d'un droit du pouvoir royal, et
que si, dans le premier cas, il satisfait sa soif de vengeance, dans le
second, en tuant un pillard, il préservait des milliers d'innocents.
Toutes les guerres de cette époque étaient atroces; mais les siennes
furent relativement humaines, car ses édits protégèrent des contrées
entières contre les déprédations de ses soldats, et, la lutte terminée,
il aidait l'Église à fermer les plaies en lui fournissant des ressources
pour racheter les prisonniers. Loin que nous trouvions chez lui des
actes de véritable cruauté, nous le voyons au contraire user de
clémence envers les Alamans vaincus, et renoncer à poursuivre son
avantage sur Gondebaud. Converti à la foi catholique, il se montre
tolérant envers ceux de sa nation qui sont restés païens; il les reçoit
souvent à sa table, et rien ne laisse croire qu'ils soient exclus de sa
faveur. Dans ses relations domestiques, il est accessible aux
sentiments affectueux: il pleure sa sœur Alboflède, il s'attache de
tout cœur à sa femme Clotilde, et lui laisse prendre un grand et
légitime ascendant sur sa vie. Malgré ses répugnances
personnelles, il lui permet de faire baptiser ses deux enfants, et c'est
en grande partie sous l'influence de Clotilde qu'il se convertit. Fidèle
à ses devoirs, Clovis est, ce semble, un des rares princes de sa
famille qui aient su respecter le lit conjugal. Ses mœurs sont pures;
on ne voit pas qu'il ait donné une rivale à Clotilde. Et le palais,
transformé en harem après lui, a été de son vivant le sanctuaire
d'une famille chrétienne.
Ajoutons, pour ne rien omettre du peu qu'il nous est donné de
discerner, que le premier roi des Francs est resté un vrai Germain.
Le baptême qui l'a enlevé à ses dieux n'a pas effacé en lui les traits
de son origine. Chaque fois que sa personnalité se dégage assez
des nuages de l'histoire pour frapper nos yeux, on reconnaît le fils
des races épiques d'Outre-Rhin. Comme ses ancêtres, comme son
père Childéric, dans le tombeau duquel on retrouva sa francisque, il
reste fidèle à la vieille hache de guerre des Istévons; c'est elle qu'il
abat sur la tête des soldats indisciplinés et des rois ennemis; c'est
elle encore qui, lancée d'un bras puissant, vole de ses mains pour
aller frapper le sol dont il prend possession, par un rôle marqué au
coin de la plus pure liturgie barbare. Il conserve pieusement, pour
les transmettre à ses fils et à ses descendants, les traditions de la
dynastie. Même alors qu'il est devenu le collègue honoraire des
empereurs, et qu'il a revêtu la chlamyde de pourpre et le diadème
d'or, il garde intacte la royale crinière qui ondule sur ses épaules, et
qui restera jusqu'au dernier jour le signe distinctif de tous les princes
de sa famille. Et n'est-ce pas à lui encore qu'il faut faire remonter
cette autre tradition domestique des Mérovingiens, qui ne permet
pas à un seul prénom romain d'altérer l'aspect fièrement barbare de
leur arbre généalogique?
Voilà les seuls traits que les brouillards de l'histoire nous permettent
d'entrevoir dans la physionomie du premier roi chrétien des Francs.
Ils sont bien loin de s'accorder avec l'image que nous ont tracée de
lui les chants populaires des barbares. Aussi la nation française n'a-
t-elle jamais connu ce Clovis païen et sanguinaire. Son Clovis à elle,
ç'a été, dès le temps de Grégoire de Tours, le roi catholique,
protecteur-né de tous les chrétiens opprimés, épée victorieuse au
service de l'Église et de la civilisation. «Dieu prosternait devant lui
tous ses ennemis, dit le chroniqueur, et ne cessait d'augmenter son
royaume, parce qu'il marchait le cœur droit devant lui et qu'il faisait
ce qui lui était agréable[330].»
[330] Grégoire de Tours, ii, 40.
Il était pour le peuple le type anticipé de Charlemagne, dans lequel
sa physionomie poétique est souvent allée se fondre, et la nation ne
se l'est jamais représenté autrement que comme celui qui a réalisé
le programme de saint Remi et de saint Avitus, en étendant le
royaume de Dieu. Quoi d'étonnant si l'on a voulu parfois le faire
participer au culte rendu à la mémoire de Charlemagne, et s'il a été
l'objet, à son tour, d'une espèce de canonisation populaire? On nous
dit qu'au moyen âge plusieurs églises lui étaient dédiées comme à
un saint, et le chroniqueur Aymeri de Peyrac ne craint pas de
l'invoquer sous ce titre[331]. Au dix-septième siècle, plusieurs
écrivains allèrent jusqu'à soutenir la thèse de sa sainteté avec des
arguments empruntés à l'histoire, et l'un d'eux, soit par
enthousiasme sincère, soit par esprit d'adulation, proposa même
formellement à Louis XIII de faire célébrer dans tout son royaume le
culte et la fête de saint Louys I, de même que Philippe le Bel y avait
fait célébrer le culte et la fête de Louys IX[332].
[331] Chronique d'Aymeri de Peyrac, manuscrit 4991 A de la Bibliothèque
nationale de Paris, fonds latin, fol. 104, verso.
[332] J. Savaron, De la saincteté du roi Louys, dit Clovis, Paris, 1620. Ce livre a eu
trois éditions en deux ans. Voir encore le P. Dominique de Jésus, la Monarchie
sainte, historique, chronologique et généalogique de France, etc. etc., traduite et
enrichie par le R. P. Modeste de Saint-Aimable, Clermont, 1670. Saussay, disent
les Acta Sanctorum, cite deux écrivains du seizième siècle, Jacques Almainus et
Paul Émile, qui donnent le nom de saint à Clovis.
L'histoire ne fait de Clovis ni un barbare sanguinaire avec les Francs
du sixième siècle, ni un saint avec les Français du quatorzième et du
dix-septième. Écartant l'image stylisée que lui présentent les uns et
les autres, et constatant qu'elle ne dispose pas d'assez de
renseignements pour tracer de lui un véritable portrait, elle doit
s'abstenir de porter sur lui un jugement formel et absolu. Elle peut
cependant reconnaître, dans le peu qu'elle sait de sa carrière, de
sérieux indices d'une vie morale épurée par l'Évangile, et elle doit
protester contre ceux qui le flétrissent comme un barbare brutal,
pour qui le baptême aurait été une formalité inefficace. Si l'on veut
absolument qu'il ait été un barbare, il ne faudra pas omettre de dire
que ce fut un barbare converti. C'est précisément la rencontre, dans
le même homme, du naturel indompté et de la grâce civilisatrice qui
semble avoir été le trait caractéristique de sa physionomie. Sachons
la respecter dans la pénombre où elle disparaît à nos regards, et,
jugeant ce grand ouvrier de Dieu d'après son œuvre, reconnaissons
que ni l'Église ni la France n'ont à rougir de lui.
Clovis laissait une famille jeune encore, mais en état de lui succéder
d'emblée, tous ses enfants masculins ayant atteint l'âge de la
majorité salique. Son fils aîné, Théodoric ou Thierry, né d'une
première union, avait déjà fait une campagne, et était arrivé au
moins à la vingtième année. Des trois fils de Clotilde, Clodomir,
l'aîné, pouvait avoir seize ans; les deux autres, Childebert et
Clotaire, les suivaient de près. A côté de ces princes grandissait une
fille qui portait, comme sa mère, le nom de Clotilde, et qui était
encore enfant lorsque son père mourut[333]. L'héritage paternel fut
morcelé en quatre parts, dont la plus considérable sans contredit fut
celle de Thierry Ier. Outre l'Austrasie, on lui attribua encore les
provinces dont il avait lui-même fait la conquête pendant la guerre
de 507, c'est-à-dire l'Auvergne avec le Velay, le Gévaudan, le
Rouergue et le Quercy. Les héritiers du glorieux fondateur de la
France eurent son énergie et ses qualités guerrières: ils continuèrent
son œuvre, conquirent la Bourgogne, achevèrent la soumission de
l'Aquitaine, domptèrent la Thuringe et humilièrent la Saxe. Ils
comprirent aussi l'influence sociale du christianisme, et, les
premiers, ils firent passer dans les lois civiles quelques-unes des
principales prescriptions de la loi canonique. Si l'on ne connaissait
leur vie privée, ils auraient des titres au respect de la postérité. Mais
leurs tempéraments étaient d'une frénésie et d'une brutalité qui les
ramenaient bien au delà de Clovis, dans les âges les plus sombres
de la barbarie primitive.
[333] Sur Théodechilde, voir ci-dessus, p. 186. Sur une autre prétendue fille du
nom d'Emma ou d'Emmia, qui figure dans le martyrologe d'Usard, au 4 novembre,
sous cette indication: Sancta Emmia virgo eximiæ sanctitatis filia Chlodovei regis,
voir Adrien de Valois, Rerum francicarum libri VIII, t. I, p. 32. Le prétendant
Munderic, dont Grégoire de Tours raconte l'aventure (H. F. iii, 14) semble avoir
voulu se faire passer pour un fils de Clovis, né, comme Théodoric, d'une alliance
irrégulière, peut-être de la même mère; du moins nous voyons qu'il ne veut
partager qu'avec Théodoric, et non avec les fils de Clotilde. Mais Grégoire de
Tours montre par son récit même qu'il ne croit pas à la parenté.
Clotilde cependant vieillissait dans l'espèce d'isolement moral qui lui
faisait sa supériorité sur son milieu. Sa vie, à partir de son veuvage,
fut plus que jamais une suite de bonnes œuvres. «Elle se faisait
vénérer de tous, écrit un contemporain. L'aumône remplissait ses
journées, et elle passait la nuit à veiller et à prier. Ses largesses ne
cessèrent de se répandre sur les monastères et sur les lieux saints.
La chasteté, la dignité la plus parfaite furent la marque de toute sa
vie. Insensible aux vaines préoccupations du siècle, cette femme qui
avait pour fils des rois était un modèle d'humilité. Ce n'était pas une
reine, c'était, à la lettre, une fidèle et consciencieuse servante de
Dieu[334].»
[334] Grégoire de Tours, iii, 18.
Nombreuses sont les églises qui se vantent de l'avoir eue pour
fondatrice. On cite parmi les plus célèbres Saint-Georges de
Chelles, où elle mit une petite congrégation de religieuses qui fut
plus tard augmentée par la reine Bathilde, et qui devint une des
perles monastiques de la France. A Laon et à Tours, elle éleva des
monastères qu'elle consacra, comme celui de Paris, au prince des
apôtres. A Reims et à Rouen, elle répara et agrandit des églises qui
avaient le même saint pour patron. A Andély, elle bâtit un couvent
qu'elle plaça sous l'invocation de la sainte Vierge[335]. A Auxerre,
elle éleva une superbe basilique sur le tombeau de saint
Germain[336]. La dévotion particulière qu'elle avait toujours eue pour
saint Martin de Tours l'attira de bonne heure près du tombeau de ce
saint: elle finit même par s'y établir définitivement[337], et les
Tourangeaux virent avec édification la veuve du plus grand roi du
siècle mener, à l'ombre de leur basilique, la vie humble et retirée
d'une religieuse. Il ne lui fut pas donné d'y vivre entièrement
absorbée en Dieu, et de se dérober, comme Radegonde et Bathilde,
à un monde qui n'était pas digne d'elle: il lui fallut traîner jusqu'au
dernier jour le fardeau de sa grandeur royale et les soucis d'une
maternité cruellement éprouvée. Cette barbarie à laquelle elle avait
arraché son époux, elle la voyait, indomptée et farouche, envahir sa
famille et s'épanouir dans le naturel des siens. Pendant qu'elle
prenait son essor vers le ciel, ses enfants la ramenaient malgré elle
dans l'enfer de leurs passions. Plus d'une fois, son cœur de mère et
de chrétienne saigna cruellement à la vue des excès auxquels se
livraient ces natures violentes et implacables. Elle vit son cousin, le
roi Sigismond de Bourgogne, ramené en captivité par son fils
Clodomir; elle le vit massacrer avec ses enfants, et leurs cadavres
jetés au fond d'un puits[338]. Ses larmes coulèrent plus amères
encore lorsque la seconde expédition que ses fils conduisirent en
Burgondie revint de ce pays sans Clodomir. Frappé par la main
vengeresse de Dieu, le cruel était tombé sans gloire au milieu des
ennemis, et sa tête, reconnaissable à sa longue chevelure royale,
avait été promenée sur une pique à la vue de l'armée franque
désespérée[339]. Il laissait trois fils en bas âge: Théodebald, Gunther
et Clodoald. Leur grand'mère les recueillit, pendant que leurs oncles
fondaient sur l'héritage du père et le dépeçaient entre eux.
[335] Vita sanctae Chlothildis, c. 11-13, dans M. G. H., Scriptores Rer. Meroving.,
t. I, pp. 346-347.
[336] Miracula sancti Germani Autissiodorensis, dans les Acta Sanctorum des
Bollandistes, 31 juillet, t. VII, p. 263.
[337] Chrodéchildis autem regina post mortem viri sui Turonus venit ibique ad
basilica beati Martini deserviens, cum summa pudicita atque benignitate in hoc
loco commorata est omnibus diebus vitæ suæ, raro Parisius visitans. Grégoire de
Tours, ii, 43.
[338] Grégoire de Tours, iii, 6; Passio sancti Sigismundi, c. 10; M. G. H., Script.
Rer. Merov., II, p. 338.
[339] Grégoire de Tours, iii, 6; Marius d'Avenches, année 524; Agathias, Histor., i,
3.
De nouvelles épreuves étaient réservées à la noble femme par la
triste destinée de sa fille Clotilde. Cette princesse avait été donnée
en mariage par ses frères au roi des Visigoths Amalaric, qui, se
souvenant peut-être du système d'alliances politiques pratiqué par
son grand-père, le roi d'Italie, avait jugé utile de devenir le parent de
ses puissants voisins[340]. Mais son mariage avec Clotilde était une
de ces unions contre nature, que la nature elle-même se charge de
défaire. La princesse catholique devint bientôt un objet d'aversion
pour son époux arien; la fille de Clovis ne pouvait inspirer que des
sentiments de haine au fils d'Alaric. Les passions du sectaire et les
ressentiments du fils se liguèrent contre la jeune reine, que le roi son
mari accablait des plus indignes traitements, lui faisant même jeter
de la boue et des immondices lorsqu'elle allait à l'église catholique.
En 531, Childebert, étant en Auvergne, reçut un messager qui lui
remit de la part de sa sœur un mouchoir trempé du sang qu'elle
avait versé sous les coups de son mari. Saisi de douleur et
d'indignation, Childebert se mit à la tête de son armée et alla fondre
sur la Septimanie. Amalaric fut vaincu dans une grande bataille
livrée près de Narbonne, et, peu de temps après, il périt sous les
coups des Francs à Barcelone, en essayant de gagner sa flotte. Le
roi de Paris rentra victorieux en France avec sa sœur délivrée. Mais
l'infortunée avait été brisée par tant de cruelles émotions; elle expira
en route, âgée de trente ans à peine, et son frère ne rapporta que
son cadavre à Paris. La crypte de Clovis devint la dernière demeure
de cette triste victime des mariages politiques[341]. Mais la France
devait plus tard venger cruellement sur Brunehaut les griefs de
Clotilde.
[340] Grégoire de Tours, iii, 1.
[341] Grégoire de Tours, iii, 10.
Toute la tendresse de la mère éplorée se reporta sur ses petits-fils,
les orphelins de Clodomir. Ils étaient le lien qui rattachait à la vie
cette âme sur laquelle l'infortune semblait s'acharner; ils furent, sans
le savoir, la cause de ses suprêmes douleurs. Son fils Childebert
s'émut de l'affection dont elles les entourait. Prévoyant que, quand
ils auraient grandi, elle voudrait les faire couronner, et qu'alors leurs
oncles seraient mis en demeure de leur restituer leur royaume, il
s'aboucha avec Clotaire, qu'il manda à Paris. Là, il y eut entre les
deux frères un colloque sinistre dans lequel fut décidée la mort des
malheureux enfants.
Pendant ce temps, ils faisaient courir la rumeur que leur entrevue
avait pour but d'élever leurs neveux sur le trône. Puis, ne craignant
pas de se jouer de leur mère, ils lui mandèrent de leur remettre ses
petits-fils, parce qu'ils avaient l'intention de les faire proclamer rois.
Clotilde, remplie d'allégresse, fit venir les enfants, leur donna à boire
et à manger, puis se sépara d'eux en leur disant:
«Il me semblera que je n'ai pas perdu mon fils, lorsque je vous verrai
prendre sa place.»
Les enfants partirent joyeux: ils allaient à la mort. A peine arrivés au
palais de Childebert, qui se trouvait dans la Cité, ils furent
brutalement séquestrés et séparés de leur suite. Peu de temps
après, Clotilde voyait arriver chez elle un grand seigneur clermontois
attaché à la personne du roi de Paris; c'était Arcadius, petit-fils de
Sidoine Apollinaire et arrière-petit-fils de l'empereur Avitus. Ce fier
patricien, transformé en valet de bourreau, portait d'une main des
ciseaux et de l'autre une épée nue.
«Je suis chargé par les rois, dit-il à Clotilde, de vous demander ce
qu'il faut faire de vos petits-enfants: les tondre ou les mettre à
mort?»
Folle de douleur et d'épouvante, et ne sachant ce qu'elle disait, la
malheureuse femme laissa échapper dans son délire ces paroles
irréfléchies:
«J'aime mieux les voir morts que tondus.»
Arcadius ne lui laissa pas le temps de reprendre ce propos; il courut
dire aux deux rois que Clotilde consentait à la mort de ses petits-fils.
Alors se passa dans le palais de Childebert une des scènes les plus
déchirantes dont l'histoire ait gardé le souvenir. Les malheureux
enfants de Clodomir, à l'expression de physionomie de leurs oncles,
aux armes qu'ils brandissaient, devinent le sort dont ils sont
menacés; ils courent à travers la chambre pour échapper aux royaux
assassins, mais Clotaire, empoignant l'aîné par le bras, lui plonge le
couteau dans le flanc. Pendant que l'enfant agonise, son cadet se
jette aux genoux de Childebert, qui, épouvanté du drame
monstrueux qu'il a mis en scène, s'attendrit sur l'innocente victime et
supplie son frère de l'épargner. Mais Clotaire a respiré l'odeur du
sang, il a maintenant l'ivresse du meurtre; il s'emporte, reproche à
l'autre sa lâcheté, menace de le frapper lui-même s'il prétend mettre
obstacle à l'exécution du projet commun. Alors Childebert mollit; il
repousse l'enfant qui s'était jeté dans ses bras et l'abandonne aux
mains homicides de Clotaire, qui lui fait subir le même sort qu'à son
aîné. Après quoi, par un inutile raffinement de cruauté, le gouverneur
et toute la suite des enfants royaux furent également massacrés.
L'une des victimes était âgée de dix ans, l'autre de sept. Quant à
leur frère Clodoald, qui avait cinq ans tout au plus, il fut sauvé par
quelques hommes de cœur, qui parvinrent à le dérober aux fureurs
de leurs oncles[342], et la tradition veut qu'il soit mort sous l'habit
monastique dans l'abbaye de Saint-Cloud, à laquelle il aurait laissé
son nom[343]. Ainsi la justice divine frappait encore Clodomir dans
ses enfants, et réalisait la prophétie de saint Avitus de Mici:
[342] Grégoire de Tours, iii, 18. Le récit du chroniqueur n'est pas sans obscurité.
Clotilde paraît n'avoir livré que deux enfants, puisque Arcadius lui demande
«utrum incisis crinibus eos vivere jubeas an utrumque jugulare». Il est d'ailleurs
assez difficile de croire que, si Clodoald avait été livré, il eût pu échapper aux
mains de ses oncles. Mais, si Clodoald a été sauvé avant l'extradition, tout le récit
devient invraisemblable, et on est amené à se demander si les circonstances n'en
ont pas été dramatisées par la poésie populaire.
[343] Grégoire de Tours, iii, 18: His, postpositum regnum terrenum, ad Dominum
transiit, et sibi manu propria capillos incidens, clericus factus est, bonisque
operibus insistens, presbiter ab hoc mundo migravit. Frédégaire, III, 38, écrit:
Clodoaldus ad clerecatum tundetur, dignamque vitam gerens, ad cujus sepulcrum
Dominus virtutes dignatur ostendere. Et le Liber historiæ, c. 24: Qui postea,
relictum regnum terrenum, ipse propria manu se totundit. Clericus factus est, bonis
operibus præditus, presbiter ordinatus, plenus virtutibus migravit ad Dominum,
Noviente villa Parisiace suburbana depositus requiescit. Le Vita sancti Clodoaldi
est un écrit du neuvième siècle, fait d'après Grégoire de Tours.
«Si vous vous souvenez de la loi de Dieu, et que, revenant à une
meilleure inspiration, vous épargniez Sigismond et sa famille, Dieu
sera avec vous, et vous remporterez la victoire. Si, au contraire,
vous les faites mourir, vous tomberez vous-même aux mains de vos
ennemis, vous périrez sous leurs coups, et il sera fait à vous et aux
vôtres comme vous aurez fait à Sigismond et aux siens[344].»
[344] Grégoire de Tours, iii, 6.
Clotilde surmonta sa douleur pour rendre elle-même les derniers
devoirs à ses infortunés petits-enfants. Elle les fit mettre dans des
cercueils, et, au son de lugubres psalmodies, fit porter leurs petits
corps dans l'église du mont Lutèce, où on les déposa auprès de leur
grand-père Clovis[345]. Puis, le cœur brisé, elle se hâta de regagner
sa retraite de Tours. Rarement, dit l'historien, on la revit à Paris[346].
Le séjour où elle avait passé des années si heureuses à coté de
l'époux aimé lui était devenu insupportable; il n'évoquait plus pour
elle que la sanglante vision d'un forfait dont la seule pensée
bouleversait toute son âme, puisqu'elle devait pleurer sur les
assassins autant que sur les victimes.
[345] Id., iii, 18.
[346] Id., ii, 43 Raro Parisios visitans.
Mais on eût dit que ses fils avaient juré de lui briser le cœur. Repus
de carnage, ils finirent par tourner leurs armes contre eux-mêmes.
Childebert avait décidé la mort de Clotaire; il s'unit à son neveu
Théodebert, et les deux rois donnèrent la chasse au malheureux roi
de Neustrie. Réfugié dans la forêt de la Brotonne[347], aux environs
de Caudebec, avec des forces bien inférieures à celles des deux
alliés, Clotaire n'attendait plus que la défaite et la mort. Mais il avait
une mère, et, devant la suprême détresse qui menaçait le fils
dénaturé, Clotilde oublia tout pour ne penser qu'à le sauver. Sans
pouvoir sur des âmes féroces qui semblaient se rire de ses larmes,
elle courut se jeter aux pieds du céleste ami qui recevait depuis tant
d'années la confidence de ses douleurs. Prosternée en prières
devant le tombeau de saint Martin, pendant toute la nuit qui précéda
la bataille, elle pleura et pria, suppliant le Ciel, par l'intercession du
grand confesseur, de ne pas permettre cette lutte fratricide entre ses
enfants. Et le Ciel exauça ses prières, car un ouragan épouvantable,
qui jeta le désordre dans l'armée des alliés pendant que pas une
goutte de pluie ne tombait sur celle de Clotaire, parut le signe
surnaturel de la volonté d'en haut; il désarma sur-le-champ des
barbares qui ne cédaient qu'à un Dieu irrité. La paix fut faite, et
Clotaire fut sauvé. «Nul ne peut douter, écrit l'historien, que ce ne fût
un miracle de saint Martin, accordé aux prières de la reine
Clotilde[348].»
[347] Le Liber historiæ, c. 25, est seul à mentionner cette forêt, qu'il appelle
Arelaunum; sur l'identification, voyez Longnon, p. 136.
[348] Grégoire de Tours, iii, 28.
Cette grâce, obtenue au prix de tant de larmes, fut une des
dernières consolations de la mère cruellement éprouvée. Plusieurs
années s'écoulèrent encore pour elle, vides de joies humaines et
remplies seulement, comme toute sa vie, par l'humble et assidue
pratique de toutes les vertus.
Soumise à la haute volonté qui avait appesanti avec les années le
fardeau de ses tribulations, elle l'avait porté sans murmure et en
bénissant Dieu, et maintenant, détachée de tout lien terrestre, elle se
trouvait devenue mûre pour le ciel. Elle s'éteignit enfin à Tours, le 3
juin 545[349], à l'âge de plus de soixante-dix ans, pleine de jours et
de bonnes œuvres. Un cortège imposant transporta sa dépouille
mortelle à Paris, où ses fils la déposèrent auprès de Clovis et de ses
enfants.
[349] Id., iv, 1. Le jour est donné par le Vita sanctæ Chrothildis, c. 14.

Les fidèles ne cessèrent de vénérer la mémoire de Clotilde, et de

porter leurs pieux hommages à son tombeau. Et quels hommages
furent plus mérités? Ils n'allaient pas seulement aux vertus
héroïques dont la défunte avait donné le spectacle durant sa vie; ils
s'adressaient aussi à l'épouse qui avait été l'instrument providentiel
de la conversion de Clovis. Si la France a quelque droit de se
féliciter d'être une nation catholique, elle le doit avant tout à sa
première reine chrétienne. Il est vrai, les poètes populaires, qui
entonnaient sur les places publiques des chants faits pour des
auditeurs grossiers, n'ont pas su comprendre cette suave
physionomie rencontrée par eux dans l'histoire de leurs rois. Ils ont
transformé l'épouse chrétienne en virago barbare; ils ont mis dans
son cœur tous les sentiments de leur propre barbarie; à la noble
veuve agenouillée sur des tombeaux, à la douce orante qui,
semblable aux chastes figures des catacombes, prie les bras
ouverts pour des enfants cruels, ils ont substitué la furie germanique
altérée de sang, la valkyrie soufflant la haine et la vengeance, et
armant ses parents les uns contre les autres pour des guerres
d'extermination. Leurs récits sont parvenus à se glisser dans les
écrits des premiers historiographes, et à jeter comme une ombre sur
l'auréole radieuse de la sainte. Mais l'histoire est enfin rentrée en
possession de ses droits, et elle ne permettra plus désormais à la
légende de calomnier ses noms les plus beaux.
Avant que la femme et les enfants de Clovis fussent allés le
rejoindre dans le repos du tombeau, la crypte royale avait donné
l'hospitalité à une gloire qui devait faire pâlir la leur aux yeux de la
postérité. Quand Geneviève mourut après avoir été pendant plus
d'un demi-siècle le bon génie de Paris, la reconnaissance publique
ne trouva pas d'abri plus digne de ses cendres que le souterrain où
dormait son roi. La vierge de Nanterre y fut donc déposée dans un
sarcophage; mais dès que cette royauté pacifique eut pénétré dans
le caveau, son nom et son souvenir y éclipsèrent tous les autres.
L'église Saint-Pierre du mont Lutèce ne fut plus pour les Parisiens
que l'église Sainte-Geneviève. Ce nom, le plus populaire de tous
ceux du sixième siècle, se communiqua au monastère et à la
montagne elle-même. Du haut de sa colline, Geneviève fut la
patronne céleste de Paris adolescent; de là, comme un phare
tranquille et lumineux, sa pure et touchante mémoire brilla sur la
grande ville qu'elle aimait, et sur la dynastie dont le fondateur
reposait à son ombre, comme un client fidèle. Aucune gloire
française n'est composée de rayons plus purs; aucune n'a pénétré à
une telle profondeur dans l'âme du peuple, pas même celle de
Jeanne d'Arc, cette Geneviève du quinzième siècle, sœur cadette de
la vierge de Paris. Quoi d'étonnant si, dès les premières générations
après sa mort, elle était pour la foule la seule habitante de la
basilique du mont Lutèce, tandis que le tombeau de Clovis, isolé de
la série des sépultures royales qui s'alignaient à Saint-Denis,
s'oubliait peu à peu et ne fut bientôt plus connu que des moines qui
le gardaient?
Que devinrent les sarcophages royaux de la crypte de Sainte-
Geneviève, et que devint en particulier celui de Clovis? Abandonné
aux heures du danger par les moines, qui fuyaient avec la châsse de
la sainte, il resta exposé trois fois en un siècle aux outrages des
Normands, qui vinrent piller les environs de Paris en 845, en 857 et
en 885. Fut-il violé à l'une de ces occasions, ou les cendres
échappèrent-elles à la triple profanation du sanctuaire? Nous
l'ignorons; mais les multiples tourmentes du neuvième siècle et la
sécularisation des chanoines au dixième ne durent pas augmenter à
Sainte-Geneviève la sollicitude pour un souvenir qui n'était pas
protégé contre l'oubli par l'auréole de la sainteté.
C'est seulement au douzième siècle, quand une réforme profonde et
salutaire eut rappelé les chanoines réguliers dans le cloître tombé en
décadence, qu'on se souvint enfin du trésor national que la France
avait confié à la garde des Génovéfains. L'illustre abbé Étienne de
Tournai, qui gouverna la communauté de 1176 à 1191, consacra ses
quinze années de prélature à la restauration morale et matérielle de
la maison. Le sanctuaire portait encore les traces lamentables des
profanations d'autrefois; sur les murs calcinés apparaissaient par
espaces les restes des mosaïques primitives. Étienne répara ces
ruines, orna l'église d'un nouveau plafond lambrissé, et couvrit le
tout d'une toiture de plomb[350]. Par ses soins, le tombeau de Clovis
fut transporté dans l'église supérieure à l'entrée du chœur. C'était un
monument d'élévation médiocre, sur lequel était couchée la statue
de ce roi[351]. La base en était ornée d'une inscription en vers latins,
due à la plume d'Étienne lui-même[352]. Ce mausolée subsista
pendant plusieurs siècles dans la basilique restaurée, où l'ont
encore vu les plus anciens historiens de Sainte-Geneviève. On ne
sait s'il contenait en réalité les cendres de Clovis, ou si c'était un
simple cénotaphe.
[350] Sur les travaux d'Étienne à Sainte-Geneviève, il faut lire sa propre
correspondance, lettres 176, 177, 178, 181 et 182, édition Desilve, Paris-
Valenciennes, 1893.
[351] Étienne ne parle pas de ce tombeau, mais il est décrit comme un monument
de peu d'élévation avec une statue royale couchée dessus, par Lejuge, l'Histoire
de sainte Geneviefve, patronne de Paris, 1586, fol. 174, verso, et par Dubreuil, le
Théâtre des antiquitez de Paris. 1612, p. 271, qui donne une reproduction de la
statue, p. 272.
[352] Cette inscription, faussement attribuée à saint Remi, se trouve dans un
manuscrit d'Aimoin du quatorzième siècle (Bibliothèque nationale, manuscrit 5925,
ancien fonds latin): mais les meilleurs manuscrits de cet auteur ne la contiennent
pas, et elle n'est manifestement pas de lui, quoi qu'en dise l'Histoire littéraire, t. III,
p. 161 (voir dom Bouquet, t. II, p. 538, note, et t. III, p. 44, note). Elle a donc été
composée entre le onzième et le quatorzième siècle. De plus, elle s'est réellement
trouvée sur le tombeau de Clovis, où l'a vue Robert Gaguin, Compendium super
gestis Francorum, fol. 6, verso. Nous savons en outre qu'Étienne était poète; v. sur
ce point ses propres paroles dans ses lettres 43 (au cardinal Pierre de Tusculum),
et 277 (à l'abbé de la Sauve). Nous possédons de lui l'épitaphe du roi Louis VII
(Desilve, Lettres d'Étienne de Tournai, p. 443) et celle de Maurice de Sully, évêque
de Paris. Il a composé aussi un office de saint Giraud (V. lettre 278). Tout donc le
désigne comme le véritable auteur de l'épitaphe de Clovis.

Mais une nouvelle décadence de la maison appela, au

commencement du dix-septième siècle, une nouvelle réformation. Le
cardinal de la Rochefoucauld, devenu abbé de Sainte-Geneviève,
entreprit cette grande tâche avec la même énergie et le même zèle
qu'y avait apportés Étienne de Tournai. Lui aussi voulut s'occuper du
tombeau de Clovis, qui était alors en assez mauvais état, et dont la
statue, rongée par le temps, était devenue presque entièrement
fruste[353]. Il en fit faire une autre à l'imitation des modèles du
douzième siècle, exhaussa le monument et en renouvela les
inscriptions[354]. Aux grandes fêtes, les moines venaient encenser le
tombeau[355], et tous les ans ils chantaient, le 27 novembre, pour le
repos de l'âme du roi, une messe solennelle qui contenait l'oraison
suivante:
[353] «Sur le caveau où le corps du roi Clovis, fondateur de cette abbaye, fut
inhumé, l'on voioit ci-devant le tombeau de ce roi eslevé à la hauteur de deux
pieds ou environ, au-dessus duquel estoit sa statue. Mais l'Ém. cardinal de la
Rochefoucauld, abbé de ladite abbaye, fit lever ce tombeau mangé et defformé
d'antiquité, et en faisant fouiller quelques fondements du cloistre, s'y trouvèrent
deux hautes et grandes statues de marbre blanc, de l'une desquelles il fit tailler la
statue de Clovis, qui se voit aujourd'hui couchée sur le mesme tombeau au milieu
du chœur.» Dubreuil, le Théâtre des antiquitez de Paris, édition de 1639. Millin,
Antiquités nationales, Paris, an VII, t. V, article lx, p. 85, démontre par des
arguments archéologiques que la statue exécutée par ordre du cardinal de la
Rochefoucauld est copiée sur des modèles plus anciens.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com