—
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`INTERNATIONA
`
`PUBLICATION
`
`CONTENTS
`pages 1-96
`
`ANTENNAS
`Accurate modelling of anti-
`resonant dipole antennas
`using the method of moments
`D.H. Werner and
`R.J. Allard (USA)
`Dual-polarised uniplanar conical-
`beam antennasfor HIPERLAN
`E.M.Ibrahim, N.J. McEwan,
`R.A. Abd-Alhameed and
`PS. Excell (United Kingdom)
`
`CIRCUIT THEORY & DESIGN
`Analogue CMOS high-frequency
`continuous wavelet transform
`circuit
`E.W. Justh and F.J. Kub (USA)
`Apparent power transducer for
`three-phase three-wire system
`S. Kusui and M. Kogane
`(Japan)
`Efficient and fast iterative reweighted
`least-squares nonrecursivefilters
`Yue-Dar Jou, Chaur-Heh Hsieh
`and Chung-Ming Kuo( Taiwan)
`Input switch configuration
`suitableforrail-to-rail
`operation of switched
`opamp circuits
`M. Dessouky and
`A. Kaiser (France)
`Unified model of PWM switch
`including inductor in DCM
`Sung-Soo Hong (Korea)
`
`COMMUNICATIONS & SIGNAL
`PROCESSING
`Adaptive multiwavelet prefilter
`Yang Xinxing and
`Jiao Licheng (China)
`Decision feedback equalisation
`of coded I-O OPSKin mobile
`radio environments
`A. Adinoyi, S. Al-Semari
`and A, Zerguine (Saudi Arabia)
`Detection algorithm andinitial
`laboratory results using
`V-BLAST space-time
`communication architecture
`G.D. Golden, C.J. Foschini,
`R.A. Valenzuela and
`PW. Wolniansky (USA)
`
`page
`
`10
`
`11
`
`13
`
`14
`
`7th January 1999 Vol, 35 No.1
`
`
`
`
`
`Efficient complexity reduction
`techniquein trellis decoding
`algorithm
`Sooyoung Kim Shin
`and SooIn Lee (Korea)
`Extended complex RBF and
`its application to M-QAM
`in presence of co-channel
`interference
`Ki Yong Lee and
`Souhwan Jung (Korea)
`Fair queueing algorithm with
`rate independent delay for
`ATM networks
`S. Ho, S. Chan and
`K.T. Ko (Hong Kong)
`Integrated space-time equaliser
`for DS/CDMA receiver with
`unequalreduced lengths
`V.D. Pham andT.B. Vu
`(Australia)
`Investigation of sensor failure
`with respect to ambiguities
`in linear arrays
`Fs
`V. Lefkaditis and A. Manikas
`(United Kingdom)
`Learning algorithms for minimum
`cost, delay bounded multicast
`routing in dynamic environments
`J. Reeve, P. Mars and
`T. Hodgkinson (United Kingdom)
`Multiple target tracking using
`constrained MAPdata association
`Hong Jeong and
`Jeong-HoPark (Korea)
`Passbandflattening and
`broadening techniques for
`high spectralefficiency
`wavelength demultiplexers
`E.G. Churin and
`P. Bayvel (United Kingdom)
`Performance of COMA/PRMA
`protocol for Nakagami-m
`frequency selective fading
`channel
`R.P.F. Hoefel and
`C. de Almeida (Brazil)
`Thit/s switching scheme
`for ATM/WDM networks
`J. Nir, |. Elhanany
`and D. Sadot(/srae/
`
`20
`
`24
`
`27
`
`30
`
`«ia
`
`(continued on back cover)
`
`Apple 1003
`
`Apple 1003
`
`

`

`CONTENTS
`(continued from front cover)
`
`ELECTROMAGNETIC WAVES
`Electromagnetic penetration into
`2D multiple slotted rectangular
`cavity: TE-wave
`H.H. Park and H.J. Eom (Korea)
`
`IMAGE PROCESSING
`Encoding edge blocks by partial
`blocks of codevectors in vector
`quantisation
`Hui-Hsun Huang, Cheng-Wen Ko
`and Chien-Ping Wu( Taiwan)
`Techniquefor accurate correspondence
`estimation in object borders and
`occluded image regions
`E. Izquierdo M. (United Kingdom)
`
`page
`
`31
`
`32
`
`34
`
`
`
`
`
`
`
`
`35°.
`
`37
`
`
`
`
`
`
`page
`
`51
`
`Near room-temperature continuous-
`waveoperation of electrically
`pumped 1.5511m vertical cavity
`lasers with InGaAsP/InP bottom
`mirror
`S. Rapp, F. Salomonsson,
`J. Bentell (Sweden),
`|. Sagnes, H. Moussa,
`C. Mériadec,R. Raj (France),
`K. Streubel and
`M. Hammar(Sweden)
`Record high characteristic
`temperature (7, = 122K) of 1.55\1m
`strain-compensated AlGalnAs/
`AlGalnAs MQW lasers with
`AlAs/AllnAs multiquantum barrier
`N. Ohnoki, G. Okazaki,
`F. Koyama and K.Iga (Japan)
`Red light generation by sum frequency
`mixing of Er/Yb fibre amplifier
`output in OPM LiNbO,
`D.L. Hart, L. Goldberg
`and W.K. Burns (USA)
`
`INFORMATION THEORY
`Analysis of turbo codes with
`asymmetric modulation
`Young Min Choi and
`Pil Joong Lee (Korea)
`Improved group signature
`MICROWAVE GUIDES &
`schemebased on discrete
`COMPONENTS
`logarithm problem
`Lumped DC-50GHz amplifier
`Yuh-Min Tseng and
`Jinn-Ke Jan (Taiwan)
`using InP/InGaAs HBTs
`A. Huber, D. Huber,
`Low density parity check codes with
`C. Bergamaschi, T. Morf
`semi-random parity check matrix
`and H. Jackel (Switzerland)
`Li Ping, W.K. Leung (Hong Kong)
`RF tunable attenuator and modulator
`and Nam Phamdo (USA)
`using high Tc superconductingfilter
`Non-binary convolutional
`Lu Jian, Tan Chin Yaw, C.K. Ong
`codesforturbo coding
`and Chew Siou Teck (Singapore)
`GiBerrou cand M. Jézéquel (France)
`
`NEURAL NETWORKS
`» INTEGRATED OPTOELECTRONICS
`Compact building blocks for
`
`46GHz bandwidth monolithic
`artificial neural networks
`InP/InGaAs pin/SHBTphotoreceiver
`M. Meléndez-Rodriguez and
`D. Huber, M.Bitter, T. Morf,
`J. Silva-Martinez (Mexico)
`_C. Bergamaschi, H. Melchior
`and H. Jackel (Switzerland)
`
`55
`
`56
`
`57
`
`59
`
`61
`
`41
`
`f
`LASERS
`1.5m InGaAlAs-strained MQW ridge-
`waveguide laser diodes with hot-
`carrier injection suppression structure
`
`H. Fukano, Y. Noguchi
`~~and S, Kondo (Japan)
`9.5W CW output power from high
`+ de
`_ brightness 980nm InGaAs/AlGaAs
`laser arrays
`F.J. Wilson, J.J. Lewandowski,
`B.K. Nayar, D.J. Robbins,
`PJ. Williams, N. Carr and
`F.O. Robson(United Kingdom)
`Investigation of data transmission
`characteristics of polarisation-
`controlled 850nm GaAs-based
`VCSELs grown on (311)B substrates
`H. Uenohara,K. Tateno,
`T. Kagawa, Y. Ohiso,
`H. Tsuda, T. Kurokawa
`and C. Amano (Japan)
`Low current and highly reliable
`operation at 80°C of 650nm
`5mW LDs for DVD applications
`M. Ohya,H.Fujii,
`K. Doi and K. Endo (Japan)
`Modelocked distributed
`Braggreflector laser
`H. Fan, N.K. Dutta, U. Koren,
`C.H. Chen and A.B.Piccirilli (USA)
`
`OPTICAL COMMUNICATIONS
`40Gbit/s single channel dispersion
`managed pulse propagation in
`standard fibre over 509km
`S.B. Alleston, P. Harper,
`1.S. Penketh, |. Bennion and
`N.J. Doran (United Kingdom)
`All-optical 2R regeneration based
`on interferometric structure
`incorporating semiconductor
`optical amplifiers
`D. Wolfson, P.B. Hansen,
`A. Kioch and K.E. Stubkjaer
`(Denmark)
`Demonstration of time interweaved
`photonic four-channel WOM
`sampler for hybrid analogue-
`digital converter
`J.U. Kang and R.D. Esman (USA)
`Design of short dispersion decreasing
`fibre for enhanced compression of
`higher-order soliton pulses around
`1550nm
`M.D.Pelusi, Y. Matsui
`and A. Suzuki (Japan)
`Experimental measurement of
`group velocity dispersion in
`photonic crystal fibre
`M.J. Gander, R. McBride,
`J.D.C. Jones, D. Mogilevtsev,
`T.A. Birks, J.C. Knight and
`PSt.J. Russell (United Kingdom)
`
`(continued on inside back cover)
`
`

`

`ELECTRONICS
`LEACERS
`
`
`
`
`THE INSTITUTION OF ELECTRICAL ENGINEERS
`
`7 JANUARY 1999
`ELLEAK 35(1)
`
`VOLUME35
`1-96
`
`NUMBER1
`ISSN 0013-5194
`
`ELECTRONICS LETTERS (ISSN 0013-5194) is published every other week except for one issue in December(total of 25 issues). 1999 annual
`subscription price £630.00. Single copy £26.00. Available also in a concurrent microfiche edition at the same subscription rate as for the printed
`edition. Air mail service direct to subscribers an additional £62.00.
`
`All subscription inquiries and orders should be sent to IEE Publication Sales, PO Box 96, Stevenage SG1 2SD, United Kingdom.
`Contributions should be in accordance with the Guide to Authors, which is usually to be found on the inside back coverorthe last page of each issue,
`and should be addressedasindicatedin the Guide.
`
`All other correspondence,including advertisements, should be sent to the IEE Publishing Department.
`The advantages of rapid publication given by Electronics Letters are nullified if subscribers are subject to excessive delays in the receipt of the
`journal. The Executive Editor would be grateful to learn of anydifficulties subscribers may experience.
`Electronics Letters Online
`
`
`
`Electronics Letters is available online through the IEE Online Journals service on the IEE’s World-Wide Webserver. Articles are stored in PDF format
`and may be viewed and printed using the Adobe® Acrobat® Reader. Features include: browsable table of contents pages for each journalissue,
`cross-journal searching of bibliographic records (inc. authors, titles, abstracts), and the ability to print articles locally. The issues are added in
`advance of print publication. For furtherinformation, please look at http://ioj.iee.org.uk/ or contact the Marketing Departmentat Stevenage.
`Copyright and copying
`This publication is copyright under the Berne Convention and the Universal Copyright Convention. All rights reserved. Apart from any copying under
`the UK Copyright, Designs and Patents Act 1988, Part 1, Section 38, whereby a single copy of an article may be supplied, under certain conditions,
`for the purposesof research orprivate study, byalibrary of a class prescribed by The Copyright (Librarians and Archivists) (Copying of Copyright
`Material) Regulations 1989: SI 1989/1212, no part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by
`any meanswithoutthe prior permission of the copyright owners. Permission is, however, not required to copy abstracts ofindividual contributions on
`condition that a full reference to the source is shown.
`
`Single copies may be madefor the purposeof research or private study.
`Authorisation to photocopyitems for internal or personal use, or for the internal or personal use of specific clients, is granted by theInstitution of
`Electrical Engineers forlibraries and other users registered with the Copyright Clearance Center Transactional Reporting Service, provided that the
`base fee of $10.00 per copyis paid directly to the Copyright Clearance Center, Inc., 21 Congress Street, Salem, MA 01970, USA. 00135194/97
`$10.00. This authorisation does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional
`purposes, for creating new collective worksorfor resale.
`Multiple copying of the contentof this publication without permissionis alwaysillegal.
`The IEE is not as a body responsiblefor the opinions expressed by individual authors in Electronics Letters.
`
`Secretary of the IEE:
`J. C. Williams, OBE, PhD, FEng, FIEE
`
`© 1999: The Institution of Electrical
`Engineers
`
`Editorialstaff:
`Gill Wheeler (Managing Editor)
`Joanne Davies
`Rachel Smyly
`
`Typeset in England by
`
`Pindarplc
`Ryedale Building
`First Floor
`60 Piccadilly
`York YO1 SWU
`
`Printed in UK by
`A. McLay & Co. Limited
`Longwood Drive
`Forest Farm
`Cardiff CF4 7ZB
`
`
`
`The IEE is a memberof the Association of Learned & Professional Society Publishers.
`
`The Institution of Electrical Engineers
`Savoy Place
`London WC2R OBL
`United Kingdom
`
`Telephone
`
`Facsimile
`
`WWW
`
`+44 (0)1 71 240 1871
`
`+44 (0)1 71 240 7735
`
`http://www.iee.org.uk
`
`+44 (0) 1438 313311
`825578 IEESTV G
`
`Publishing Department:
`|IEE
`Michael Faraday House
`Six Hills Way
`Stevenage SG1 2AY
`United Kingdom
`Telephone
`Telex
`Facsimile:
`+44 (0) 1438 742792
`Sales
`+44 (0) 1438 742849
`Editorial
`+44 (0) 1438 318361
`Advertising
`+44 (0) 1438 742840
`Marketing
`E-mail:
`
`HONORARYEDITORS
`Sir Eric Ash, CBE, FEng, FRS
`Prof. Peter Clarricoats, CBE, FEng, FRS
`Prof. Chris Toumazou
`
`

`

`
`
`the proposed
`
`Security analysis: Some possible attacks against
`schemeare presented below.
`Attack I; Although the group authority has the knowledge of(r,.
`s, k,), the group signature cannot be forged without the secret key
`x, of U.. It is impossible to obtain x, from y, without being able to
`solve the discrete logarithm problem. Moreover, because the gen-
`erator o, = g’*' mod p,k,€ Z, anda e Z,, both o, and g have
`the same order g. Therefore, forging (R, S) is as difficult as break-
`ing ElGamal’s scheme [3]. Since the group authority cannot forge
`the group signature, forgery by an adversary is even more diffi-
`cult. Thus, the impersonation attack can not be successful.
`Attack 2: The signer U, can be identified if we can obtain v, from
`the signature {R, S, h(n), A, B,C, D, E}. Since the receiver does
`not knowthe (r,. s,, k,) of the group authority, he cannot check the
`equation D’. yy» E = D* mod p. Obtaining (v, 5, &,) from given
`(4, B, C, D, FE} depends on the discrete logarithm.
`Attack 3: The group authority may publish the information (r;, s;.
`y) for the message m's signature to enable a verifier to check the
`identity of U,, This does not damage the anonymity of Us previ-
`ous group signatures because the information (r,, s,, ¥,)
`is only
`provided for the specific group signature {R, S, h(i), A, B,C, Dy
`FE}. Fordifferent messages, U, will have chosen different random
`integers a and / to generate group signatures. If an adversary
`wants to obtain a, 6 and (r,, s,) from given {A, B, C.D, E} this is
`as difficult as solving the discrete logarithm.
`
`Discussion: The improved scheme preserves the main merits inher-
`ent in most of the Lee-Chang scheme. In the case of a later dis-
`pute, the group authority may publish the information (1; 57. 1)
`to enable a verifier to check the identity of the signer, although
`this does not damage the anonymity of the other previous signa-
`tures of the signer. Meanwhile.
`the group authority need not
`renew any key of the signer. The reason is that the information
`(ry, 8p. ¥,) is only provided for the specific group signature {| R, S,
`iim), A, B. ©. D, E}. Compared to the original scheme,
`the
`improved scheme requires some additional cost
`in terms of com-
`putational time and the size of the group signature. For generating
`a group signature. the signer U, may precompute several different
`‘a, A, B,C, D. FE} using (r,. s,) to reduce the real-trme computa-
`tional time.
`
`Conclusions: We have proposed an improved group signature
`scheme based on the discrete logarithm. In our improvement, a
`group signature can be opened to reveal the identity of the signer.
`the anonymity of the other previous signatures signed by this
`group member are not damaged. Meanwhile, the group authority
`also need not renewthe keys of the signer. We have demonstrated
`somepossible attacks against the proposed scheme. Underthe dif-
`ficulty of computing the discrete logarithm problem, we have
`shownthat the improved schemeis secure against these attacks.
`
`© IEE 1999
`Electronics Letters Online No: 1999007]
`
`28 October 1998
`
`Low density parity check codes with semi-
`random parity check matrix
`
`Li Ping, W.K. Leung and Nam Phamdo
`
`A semi-random approach to lowdensity parity check code design
`is shown to achieve essentially the same performance as an
`existing method, but with considerably reduced complexity.
`
`Introduction; Recently, there has been revived interest in the low
`density parity check (LDPC) codes originally introduced in 1962
`by Gallager [1]. It has been shown that such codes can achieve
`very good performances (within 1.5dB oftheoretical limits) with
`modest decoding complexity[2].
`An LDPCcode is defined from a randomly generated parity
`check matrix H [2]. For the purpose of encoding,it is necessaryto
`transfer H into H,,,, the equivalent systematic form of H, which
`can be accomplished by Gaussian elimination. For a rate R = k/n
`(kK = information length, n = coded length), the size of H is (nk) x
`n. When nvis large, Gaussian elimination can be costly in terms of
`both memoryand the operations involved. Besides, a considerable
`amount of memoryis required to store H,,,
`in the encoder, which
`is not necessarily sparse even though H is usually designed so.
`In this Letter, we report a modified approach to LDPC code
`design. We adopt a semi-randomtechnique, i.e. only part of H is
`generated randomly, and the remaining part is deterministic. The
`new method can achieve essentially the same performance as the
`standard LDPC encoding method withsignificantly reduced com-
`plexity.
`
`Proposed approach: For simplicity we will only consider binary
`codes. Decompose the codeword ¢ as ¢ = [p, d]', where p and d
`contain the parity and information bits, respectively. Accordingly,
`we decompose 7 into H = [H’, H“]. Then
`
`a?a) (P) =0
`
`d
`
`(1)
`
`In the proposed method, H’is constructed in some deterministic
`form. Empirically, we found the following a good choice (recall
`that He must be a square matrix [3]):
`]
`
`0
`
`He = a
`0
`11
`
`(2)
`
`Weadoptthe following rules to create H¢. Let ¢ be a preset integer
`constrained by (/) ¢ divides n-k and(ii) n-& divides kr. Partition H#
`(which has »-k rows) into ¢ equal sub-blocks as
`
`H? =
`
`Hal
`:
`Ho’
`
`(3)
`
`Yuh-Min Tseng and Jinn-Ke Jan (/nstitute of Applied Mathematics,
`National Chung Hsing University, Taichung, Tatwan 402, Republic of
`China)
`
`Jinn-Ke Jan: corresponding author
`E-mail: jkjan@amath.nchu.edu.tw
`
`¢ = 1, 2 -- -t, we randomlycreate exactly
`In each sub-block H*,
`one element | per columnand kt/(n-k) 1s per row. The partition in
`eqn. 3 is to best increase the recurrence distance of each bit in the
`encoding chain (see below) and, intuitively, reduces the correlation
`during the decoding process. The resultant H’ has a column
`weight of ¢ and a row weight of k1/(m-&) (the weight of a vectoris
`the number of Is amongits elements).
`References
`Based on eqns. | and 2,p= {p,} can easily be calculated from a
`
`given d =
`{d,} as
`signatures’,
`Proc,
`
`1 CHAUM.D,—and HEYST, F:: ‘Group
`
`}
`7
`Pr >it),
`aid p= Piast S_ Aid,
`(med 2)
`EUROCRYPT9I, 1992, pp. 257-265
`(4)
`2) CHEN.L., and PEDERSEN, T.P.: ‘New group signature schemes’, Proc.
`EUROCRYPT 94, 1995, pp. 171-181
`ELGAMAL.T.:
`‘A public key crypto system and a signature scheme
`based on discrete logarithms’, JEEE Trans.
`Inf. Theory, 1985, 31,
`(4). pp. 469-472
`LEE. Wes. and CHANG. ¢.c:: “Efficient group signature scheme based
`on the discrete logarithm’. JEE Proc. Comput. Digit. Tech., 1998,
`145. (1), pp. 15-18
`5 NYBERG. K.. and RUEPPEL. R.A “Message recovery for signature
`schemes based on the discrete logarithm problem’. Designs, Codes
`and Cryptography, 1996, 7. pp. 61-81
`
`3
`
`4
`
`the above
`Compared with the standard LDPC code design [2],
`method has several advantages. First, the encoding process in eqn.
`4 is much simpler than a full Gaussian elimination. Secondly. a
`random H® can be singular, which causes additional programming
`complexity in realising a specified rate. On the other hand, H’ in
`egn. 2 is always non-singular so the new method canrealise any
`given rate directly and precisely. Thirdly,
`it requires verylittle
`memoryto store H# in the encoder if H* is sparse (this can be
`ensured using small 1).
`
`38
`
`ELECTRONICS LETTERS
`
`7thJanuary 1999 Vol.35 No.1
`
`

`

`
`
`
`
`decoders passing the result of their work to each other, at each
`iteralive step.
`
`x
`systematic
`part
`
`data
`
`
`interleaving |
`1 redundancy
`
`|||
`
`
`
`go
`(955/1
`1 Two-dimension turbo code with generators 15, 13
`
`0.6
`
`1.0
`
`1.4
`E,/No.dB
`
`1.8
`
`2.2
`
`oir
`
`| Performances of LDPC codes generated by semi-random pariti
`Fig.
`check matrixes with k = 30000
`
`@ R= 13
`BR-
`AR-2? Lahot
`
`Fig.
`
`| contains the simulated performances of
`Simulation study; Fig.
`the proposed encoding method for various rates (1/3, 1/2. 2/3)
`using f = 4. The decoding algorithmfollows that in [2]. The results
`are essenually the same as those obtained using fully random H.
`
`It has been shown that a semi-random approach to
`Conclusion:
`LDPC code design can achieve essentially the same performance
`as the existing method with considerably reduced complexity.
`
`© IEE 1999
`Electronics Letters Online No; 19990065
`
`23 November 1998
`
`Li Ping and W.K. Leung (Department of Electronic Engineering, City
`University of Hong Kong, Hong Kong)
`E-mail: eeliping@cityu.edu.hk
`Nam Phamdo (Department of Electrical and Computer Engineering,
`State University of New York at Stony Brook, Stony Brook, N¥ 11794-
`2350, USA)
`
`References
`
`IT
`
`GALLAGER, R.G.: “Lowdensity parity check codes’, JRE Trans. Inf.
`Theory, 1962, 1T-8, pp. 21 28
`2) Mack AY, DJC.. and NEAL, R.M.: “Near Shannonlimit performance of
`low density parity check codes’, Electron. Lett., 1997, 33, (6), pp.
`457 458
`PROAKIS. 1G.: “Digital communications’ (McGraw-Hill, 1995)
`PETERSON, W.W., and WELDON. E.., Jr: “Error-correcting codes’ (MIT
`Press, Cambridge, Massachusetts. 1972) 2nd edn.
`
`3)
`4
`
`Binary codes versus quaternary codes: Fig. 2a represents a block of
`size A encoded bythe code of Fig.
`|. This block is seen as a two-
`dimensional vA x V& block and for simplicity we consider that the
`interleaver is a regular one:
`the sequence is encoded first by C,,
`following the horizontal or linewise dimension, and secondly by
`C., following the vertical or columnwise dimension. The dashes on
`beth dimensions symbolise the path error packets at the output of
`the two decoders, at a particular step of the iterative process.
`These packets do not contain only erroneous decisions but they
`indicate where a wrong path has been chosen either by the
`decoder of C, or by the decoder of C,. This corresponds to a cer-
`tain path error density per dimension, which is the same in both
`dimensions if the component codes are identical.
`
`
`
`vki2
`
`Fig. 2 Path error packets in turbo decoding
`a Binary codes
`b Quaternary codes
`
`82]
`
`
`systematic part
`
`>
`
`|
`
`Non-binary convolutional codes for turbo
`coding
`
`C. Berrou and M., Jézéquel
`
`The authors consider the use of non-binary convolutional codes in
`turbo coding.
`It
`is
`shown that quaternary codes can be
`advantageous.
`both
`from performance
`and
`complexity
`standpoints, but that higher-order codes may not bring further
`improvement.
`
`Introduction: Turbo codes are error correcting codes withat least
`two dimensions (i.e. each datum is encoded at least
`twice). The
`decoding of turbo codes is based on aniterative procedure using
`the concept of extrinsic information. Fig.
`1 gives an example of a
`two-dimensional turbo code built from aparallel concatenation of
`two identical recursive systematic convolutional (RSC) codes with
`generators
`15.
`13 (octal notation). The global
`(non-iterative)
`decoding of such a code is too complex to be envisaged because of
`the very large numberofstates induced by the interleaver. Aniter-
`ative procedure is therefore used,
`the two codes being decoded
`alternately in
`their own dimensions and the two associated
`
`Fig. 3 S-state quaternary recursive systematic convolutional
`code with generators 15, 13
`
`( RSC)
`
`redundancy
`
`The performance of turbo decoding is strongly dependent on
`the path error density per dimension. Obviously, the more numer-
`ous and the longer the horizontal and vertical dashes in the square
`box are, the harder the convergence to the correct codeword is.
`Nowfor each component code, replace the binary code of Fig.
`|
`by the quaternary code of Fig. 3. The data are thus encoded and
`interleaved in couples. The size of the block is 4/2 couples and the
`square box nowhas the dimensions Vk/2 x Vk/2 (Fig. 25). When a
`decoder selects a path in the decodingtrellis, the same amount of
`information is used in the cases of both binary and quaternary
`codes, therefore with half the number oftransitions in the case of
`a quaternary code, giving path error packets which are half the
`
`ELECTRONICS LETTERS
`
`7th January 1999 Vol.35 No.
`
`17
`
`39
`
`