OLIVER PRETZEL Imperial College, London Error-Correcting Codes and Finite Fields CLARENDON PRESS - OXFORD 192Teuie, & be ced Cree a Oxord Unnerety Press, Walon Strat, Oxford OX2 6DP ‘Oxford New York Toronto elkt Bombay Caluta Madras Karachi Petaling aya Singapore Hong Kong Tokyo ‘Nat Dar er Salaam Cape Town ‘Melbourne Auckond and associated companies ‘Bern Tbadon sfondi tade mark of Oxford University Press Published inthe United States by Oxford Untesty Press, New York (© Ober Pret 1992 Alt rights rserced. No part of ths publction may be reproduced, stored a rereval Sam, o anand in ony form or by any meas, without the pio prmsson I wring of ford Unteray Press’ Within he UK, exceptions ae allowed in respect of any fr doling forthe purpose of research or rete sud, ov elicim or vedi, aspeited under the Copyright, Designs and Patents Ae, 1988, oF i theca of reprograpte reproduction I ecordance with he toms of licence isd by the Copyright Licey Agen). Engures eonceraing ‘reproduction cute that tars and over cours shouldbe Sent ‘the Rights Department, Oxford Univerty Presa the abres bore ‘This book is a abet to the condition tha shall ot, by way of trade or oterie, be ln resol, hired out, or others icalated ‘without the pubhers pir consent any form of binding or cover other tha that which ts published and without a sia condion ‘iclading tis condion being imposed on the Subsequent purchaser ‘A catalogue reorder this book is valble from th Bish Library Libary of Congress Cataloging in Publication Data Pretzel ler Errorcorecting codes and ite lds|Olver Pretzel (Osfrd applied maton and cimping scene Series) Incles biographical rerences, 1. Broncorstng codes (Infomation theory) I. Tle. 1. Series (QAves P74 1992 08 S4edet 924088, ISBN 019 8596782 ‘Typeset by Integrl Typeseing,Goreston, Norfolk NRSI 6RO ‘Printed i Great Britain by ‘Booker (ath Led, Midsomer Norton, Ason To Christ! and RaimundPreface _ ‘This book arose out of a series of courses given to students of mathematics and electrical engineering at Imperial College. The theory of error-correcting block codes combines mathematical elegance and practical utility to an ‘unusual degree, Thus, the intention of the courses was twofold. On the one hhand I wished to introduce the mathematicians to some attractive practical problems and to address these as an essential part of the development of a ‘mathematical theory. On the other I hoped to persuade engincers of the power and elegance of modern mathematics and to give them confidence in using it. ‘There are many excellent texts on coding theory, notably that by MacWilliams and Sloane, but T found that they were either too advanced for my purposes or stopped short of providing all the mathematical tools required to implement a coding system (like the excellent introductions by Till or Pless) I therefore wrote my own set of lecture notes, which form the basis of Parts 1-3 of the book. These start with a standard elementary introduction to coding theory, then develop the theory of finite fields (which is an essential tool) and in Part 3, exploit it to construct and decode BCH and Reed-Solomon codes. 1 abhor tome-like textbooks that skim over a vast array of topics saying only trivalities about all of them. So this book does require its reader to think. My experience has been that although electrical engineers go through ‘a kind of culture shock as the material on finite fields is presented, they emerge confident that they can apply them in the many areas of their tliscipline in which they appear. Similarly mathematicians used to abstract ‘generalities and existence theorems find the concerns of coding theory unfamiliar but gain a deeper understanding of the mathematical theory by seeing it at work, "The standard courses at Imperial College covered most of the material in arts 1-3. The additional sections (the Extras), were only mentioned, or left ‘out of the course entirely. However, in writing the book, I could not resist the temptation to add a further part on Goppa codes (both classical and ‘gcometrical) including the decoder of Skorobogatov and Viddut. This part ‘vas tried out in a postgraduate course at the University of London. During that course it became evident that the major difficulty in presenting geometric Goppa codes is to find a simple presentation for the geometry of algebraic ‘curves. Chapters 21-23 are my attempt to do this. I hope that once a reader has worked through the first three parts ofthe book, these chapters will noti Preface present excessive difficulties. In treating Goppa's codes, Ihave tried to exhibit {hem as natural generalizations of BCH codes, and included proofs that BCH codes are a special class of both classial and geometric Goppa codes in the exereses. A parts contain exercises. These range from routine calculations to extensions of the theory. Routine calculations are extremely important for the understanding of the subject, and all chapters contain extensive examples to guide the reader. He or she should work through these carefully and then do the corresponding exercises to gain confidence. The more theoretical exercises are to some extent optional, They certainly contain material that will deepen the reader's understanding of the codes, but on the other hand, they are not essential, How many ofthese should be attempted depends on the reader’ purpose in studying the book and on his or her ‘mathematical saturity’. Attempt atleast afew, and if you enjoy them try more. ‘The book also contains two short appendices, one on linear algebra and the other on polynomials. The first is provided mainly because engineers frequently have not seen the rank and nullity theorem explicitly and also because linear algebra is usually taught only for real and complex numbers, ‘whereas coding theory uses finite fields. The second is a quick refresher on the properties of polynomials. Each ofthe four parts ofthe book is divided into chapters. The chapters are split into sections or paragraphs, numbered consecutively. Theorems, propositions, and definitions are referred to by their section numbers. I the reference isto Section 7in Chapter 3 itis given as 37. Some chapters have “extra" sections at the end that can be omitted at first reading. The start of these seotions is marked by the heading “Extras” External references are siven by authors’ names and the date; there are separate lists for textbooks and papers in journals. My thanks are duc to my students and colleagues who attended the lectures. frequently learned as much from them as they did from the courses. What qualities the book may have are due to them in no small measure. 1 rust particulasly thank Benjamin Baumslag who read early drafts in detail and made many suggestions for improvement. Without his enthusiasm and insistence I would never have completed the book. I would also like to thank the copy editor for his careful reading of the text. He introduced many improvements. Naturally, its deficiencies are my own. Imperial College, London March 1992 Contents eee PART 1 BASIC CODING THEORY 1 Introduction 3 Exrors of transmission, Examples from natural language. Channel models ‘The binary symmetric channel. Three simple codes (a parity check code, & triple repetition code, and a triple parity check code). 2. Block codes, weight, and distance 1B Block codes. Block length, message block length, and rate. Definition of ‘Hamming weight and distance. Minimum distance, error detection, and error correction. Block and message success probabilities. Calculation of error detection/correction probabilities for the examples of Chapter 1. Discussion ff Shannon's theorem (without proof). * 2 Definition of linear codes and fields, Dimension and rate, The generator ‘matrix. Standard form generator matrices and systematic encoding. Message ‘and check bits. The check matrix. Uniqueness of standard form generator and check matrices. 4. Error processing for linear codes 4 Decoding by cosets (standard array). Coset leaders and syndromes, Code ‘can correct single errors if and only if check matrix has distinct non-zero ‘columns. Conditions for multiple error correction. 5 Hamming codes and the binary Golay codes 6 Definition of the sequence of binary Hamming codes Ham(k) by their check ‘matrices, Suocess probabilities for Hamming codes. Long Hamming codes are very efficient, but poor at correcting errors. Perfect codes. Construction ‘of the binary Golay codes by Turyn's method. Appendix LA Linear algebra 9 ‘The laws of arithmetic: rings, domains, and fields. Elementary vector space theory, Bases and dimension. Elementary matrix theory. Row operations, rank, and nullity, Vandermonde mat