5q
`
`SONYExhibit 1009 - 0001
`
`BSDS
`
`SONY Exhibit 1009 - 0001
`
`

`

`European Transactions on
`
`Telecommunications
`
`_
`
`EDITOR IN CHIEF
`Prof. EZ10 BIGLIERE
`Pobtecmoo di Tora, Dipartimente di Pietra
`emanl, meten@ ponte i
`
`MANAGING EDITOR
`Prof.ACHILLE PATTAVINA
`Pohtecnice di Milano. Dip. di Elettronica ¢ informazione
`comath pagtas mawelet pedi. it
`
`EDITORS
`
`Prof. ULRICH KILLAT
`Communication Networks
`Teche, Unawvenatat MienbureHishorg
`e-mail klar o-harberg be
`
`Prof. JORG EBERSPACHER
`Mobile Networks
`Tech, Universi: Munchen, LKN
`eomaik cherpeccher@ec: tom de
`
`Prof. UMBERTO MENGALE
`Communication Theery
`Laverna di Pisa, Dip Ingegneria dell informasione
`emai: mengal 2 etunpeat
`
`Prof. SERGE VAUDENAY
`Cryptographysecurity
`BPP O8¢°4 ASEC
`comail: wrgss astdenay epfl ch
`
`Prof. BEXIO RIMOLDI
`Information Processing
`EPPL, Mobile Communications Lab
`e-mail: baie nmoldi@epfi.ch
`
`Prof. JOHN M. SENIOR
`Optical Communications
`Vnner of Hertfordshire Faculty of Eng & Inform. Scienves
`ermail
`) tise@therts ac uk
`
`Prof. MAURICE BELLANGER
`Signal Processing
`CNAM, Dept, Elecrronisgas:
`email bellang @cnam_fr
`
`Prof. JOACHIM HAGENAUER
`Telecommunication Transmissions
`Tech Uneerwtit Minchen, LNT
`emmal: joachimhagenaicr @er tunde
`
`
`
`ETT 2000 SUBSCRIPTION | Tints|Ee
`
`Aims and Scope
`CONDITIONS
`/
`Lire
`
`European Transactions on Telecommunications (ETT)
`originates from the well known magazines «Alta Frequen-
`za» International Journal (AEL, Italy) and «ntzArchiv»
`(VDE, Germany) and aims for two ambitious goals:
`
`to be the focus ia Europe of outstanding contributions
`from researchers and engineers working in the field of
`information technology:
`
`*
`
`*
`
`to concentrate papers on the various applications of
`telecommunications:
`
`ETT is published by
`
`Subscription fee
`
`500.000
`
`258,23
`
`Member of EURELsocieties
`
`— 400.000 206,59
`
`Memberof ETT signatory societies
`
`© 150.000
`
`77,47
`
`Postage by airmail
`
`50.000
`
`25,83
`
`~ Communication Networks
`
`~ Communication Theory
`~ Cryptography/Security
`~ Information Processing
`~ Mobile Networks
`
`~ Optical Communications
`~ Signal Processing
`~ Telecommunication Transmissions
`
`In order to improve the quality and the numberofthe pa-
`pers, a staff of eight Telecommunication Editors has been
`settled. Their main tasks are:
`
`~ fo encourage paper submission
`~ to propose special issues and facuses
`~ te invite tutorials and critical reviews
`
`ETTwelcomes open contributions from every country in
`the world. It publishes:
`
`- Invited papers (tutorial orcritical reviews}
`regular papers (original scientific contributions }
`- letters (short contributions and scientific correspon-
`deneces)
`
`t
`
`E
`
`AEI
`
`AEE |
`
`
`
`F INEL |
`
`SF
`
`Assooiaione Elettrotecnica od Elettronica fraliana
`
`in cooperation with
`
`Asociaciin Electrotdenica Espartola
`
`Association des Ingeniours Electriciens sorts de
`Vinsntur Monietiore (Universite de Litge)
`
`Aatteellisten sahkoinsindorijarjestijen neuvottelukunta
`The Consulting Commutiee of the Professional
`Flectroengineers” Organizations in Finland
`
`NL
`
`Koninkiyk Instituut van Ingemeurs Afd. elektrotechnick
`
`KIVI
`
`ODE
`
`OVE,
`
`;
`
`SER
`
`P
`
`A
`
`Ss
`
` CH
`
`SRBE
`KBVE
`
`VDE -
`
`B
`
`D
`
`Ordem Dos Eagenheiros
`
`Osterremhischer Verband ftir Elekuotochnik
`
`Sveriges Licktro-och Dataingenjérers Rikstorening
`
`Schweaverscher Flekirotechniseher Verein
`Association Suisse des Electricrens
`
`Societe Royale Belge des Electnciens
`Koninklijke Beigische Vereniging der Elektrotechnici
`
`Verband Deutscher Hlektromechatker
`
`with the suppedt of the
`European Commission
`
`
`
`<itemcemeenn
`
`SONY Exhibit 1009 - 0002
`
`
`
`SONY Exhibit 1009 - 0002
`
`

`

`
`
`European Transactions
`on Telecommunications
`
`EURELPublication
`
`Vol. 1, No. 6, November-December 2000
`
`memnanncnene
`
`
`
`Published by AEI (Italy) in cooperation with:
`AEE (Spain), AIM (Belgium), FINEL
`(Finland), KIVI
`(Netherlands). ODE
`(Portugal), OVE (Austria), SER (Sweden),
`SEV/ASE (Switzerland), SRBE/KBVE
`(Belgium), VDE (Germany). With the sup-
`port of the European Commissien (EC)
`
`EDITOR IN CHIEF
`E. Bighert
`
`MANAGING EDITOR
`A. Pattavina
`
`BOARD OF DIRECTORS
`Chairman: E, Biglieri (AED - Members: JM.
`Ortiz Gonzales (AEE), AIM (to be appointed),
`G. Delf Osso (EC), | Aarai
`(FINEL),
`L.G.M.M. Niemegeers (KIVI), A. Carvalho
`Fernandes (ODE), OVE(to be appointed), LLL.
`Massey (SEV/ASE), SER(to be appointed), P.
`Delogne (SRBE/KBVE), D. Kre8 (VDE).
`
`SCIENTIFIC BOARD
`Chairman: E. Bigtiert () - Members: J.C.
`Arnbak (NLi, M. Bellanger
`(PF), L.
`Calandrino (1), ¥, Cappellini (1D, F. Carassa
`(>, A. Carvatho Fernandes (Pi, A. Danthine
`(Bi. M. Decina (1), P. Delogne (B), C. Egidi
`cD, F. Fedi (f). G. Franceschetti (1), E. Gatti
`(fh, A. Gilardini (D, V. Grandis (CEC), J.
`Hagenauer (D}, 1. Hartimo (SF), H.L.
`Hartmann (D), J.H.M. Henaff (CNET), PJ.
`Kuhn (D}) R. Lehnert (D), W. Lin (T), EL.
`Massey (CH), H. Meyr (D), V.A. Monaco(1),
`C. Massotto (1), J.-A. Nossek (Di, H.
`Ohnsorge (D), R. Paul (D), M. Pent (1), T.R.
`Rowbotham (UK), W. Rupprecht (D)}, A.
`Sangiovanni- Vincentelli (USA), A. Schroth
`(D}, K.U. Stein (D), F.L.AM. Stumpers
`(NL, G. Tartara (D, P. Usienghi (USA), G.
`Vannucchi (D, R. Van Overstraeten (B), P.
`Vary (Di. B.L.A. Waumans (NL),
`J
`Weinrichter (A), G. Wiest (D).
`
`Special Issue on Multi-Carrier Spead-Spectrum
`Guest Editors: K. Fazel, §. Kaiser
`
`K. Fazel, S. Kaiser
`Guest Editorial: Multi-Carrier Spread Spectrum oosccccscecccceccsceeccessereeeresesenee 2S
`G.B. Giannakis, A. Stamoulis. Z. Wang, P_A. Anghel
`Loud-AduptiveMUMISI-Resilient Generalized Multi-Carrier CDMA
`with Linear and DF Receivers oo... ccccccccscecseseceseveneeveccucsssenentnanstavaesensaesensacseaseraneres 327
`
`§.P.W. Jarot, M. Nakagawa
`Each Carrier Transmission Power Control with Antenna Carrier
`Diversity for OFDM/DS-CDMASystem oc. ccoccccccseessssecssvssesseessesrssvessessoseneissesecessee 539
`J.-¥. Baudais, J.-F. Hélard, J. Citerne
`An Improved Linear MMSE Detection Technique for Multi-Carrier
`CDMASystems: Comparison and Combination with Interference
`Cancellation Schemes 0. .cccccseeseuccsseeetrsisseceesenserssesssensvsesienetvataetcnusnessarcanarevaneaseees 547
`
`M. Reinhardt, J. Egle, J. Lindner
`Transformation Methods, Coding and Equalization for Time- and
`Prequency-Selective Channels ..cccccccccccncescccccouseessenssnseseseneneaecevasenecsenescescereneueanaeents 555
`L. Ping
`Randomized OFDM for Multi-cell Environments .o.ccccccceccccsesessessescsserccrencneesssseeesnees 567
`J. Kiihne, A. Nahler. G.P. Fettweis
`Comparison ofTime and Frequency Domain Processing of a Multi-
`Carrier Spredd-SpeCtraSVGOne sssscsssncassoassasacescneas sees evecne svecavassm ci asst snsveanotecessias 579
`P. Robertson, S. Kaiser
`:
`Analysis of Doppler Spread Perturbations in OFDM(A) SYStGMIS occas
`S. Zazo, F. Bader, }.M, Pdez-Borrallo
`Spread Spectrum OFDM Modem for HF Voice-Bund Link over
`Feadinng Chearerie8 occ ccc cscecesesceveuesesscsnvesuscaesonecsoneucassuseseustoveusnecagnavseecasagesiseuseaneaneeseras 593
`K. Matheus, K.-D. Kammeyer, U. Tuisel
`Flexible Implementation of Multicarrier Systems with Polyphase
`FELEPDAAKS ooo cceeccccccccs ccc cosesveneseesescuacesessuensaseuesesseeyssseasedequsatnesasassiesacueressaseneneaseaenvess 599
`
`585
`
`S. Kaiser, W.A. Krzymien, K. Fazel
`SS-MC-MA Systems with Pilot Symbol Aided Channel Estimation
`
`in the Asvachronous Uplink ........
`
`
`EDITORS
`U,
`Communication Networks:
`Killat,
`U, Mengali,
`Communication
`Theory:
` §. Vaudenay.
`CryptographyéSecurity:
`Information Processing: B. Rimoldi, Mobile
`Networks: J. Eberspacher, Qpticul Communi-
`cations: JM. Senior, Signal Processing: M.
`Rellanger. Telecommunication Trasmissions:
`J. Hagenauer.
`
`
`
`
`Books on Telecommunications
`
`
`
`
`
` Annual Index 2000.....
`
`To ETT Readers and Subscribers oo... anaaseiyne nyt
`
`
`
`aanSeonNreMARESSDERRSNAAASMOAN
`
`
`
`eatenPRCAaRatt
`
`
`
` “espanatainemeneabmeaotasoaiasesmeanonnennthSloentinamanabinMonstentmaemonensinPe
`
`
`
`
`. 20121 Milano
`Proprietaria © Associazione Elettroteenica ed Elettronica Italiana - Direzione Redazione Amministrazione - Piazzale R. Morandi 2
`Tel. 02/77790.1, Telefax 02/798817 - E-mail: riviste aei@aei.it
`Autorizz. Tribunale di Milano 29 agosto 1984, N. 394 del Registro
`Direttore responsabile: G. Lucchini
`© AET 1990. I diritti di mproduzione anche parziale sonoriservati.
`European Transactions on Telecommunications é pubblicata col concorso del Consiglio Nazionale delle Ricerche ¢ della Fondazione Ugo Bordoni
`Spedizione in abbonamentopostale art. 2 comma 20/C legge 662/96 - Filtale di Milano.
`Stampa: Arti Grafiche Stefano Pinelli Sct - Via R. Fameti, 8 - 20129 Milano
`® Associati all"USPI- Unione Stampa Perivdica kalian
`
`SONYExhibit 1009 - 0003
`
`SONY Exhibit 1009 - 0003
`
`

`

`1 Special Issue I
`
`Load-Adaptive MUI/ISI-Resilient Generalized
`Multi-Carrier CDMA with Linear and DF Receivers
`
`GEORGIOS B. GIANNAKIS, ANASTASIOS STAMOULIS, ZHENGDAO WANG, PAUL A. ANGHEL
`Dept. of Electrical and Computer Engineering, 200 Union Street S.E., University of Minnesota, Minneapolis, MN 55455, USA
`{ georgios, starnoti Ii, zhengdno, anghel} @ ece. umn. edu
`
`Abstract. A plethora of single-carrier and multi-carrier (MC) CDMA systems have 'been proposed recently to
`mitigate intersymbol interference (ISI) and eliminate multiuser interference (MUI). We present a unifying all-digital
`Generalized Multicanier CDMA framework which enables us to describe existing CDMA schemes and to highlight
`thorny problems associated with them. To improve the bit error rate (BER) performance of existing schemes, we design
`block FIR transmitters and decision feedback (DF) receivers based on an inner-code/outer-code principle, which guaran-
`tees MUVISI-elimination regardless of the frequency-selective physical channel. The flexibility of our framework allows
`further BER enhancements by taking into account the load in the system (number of active users), while blind channel
`estimation results in bandwidth savings. Simulations illustrate the superiority of our framework over competing MC
`CDMA alternatives especially in the presence of uplink multipath channels.
`
`1 INTRODUCTION
`
`The ultimate goal in wireless CDMA systems is to sup-
`port as many mobile users and as high data rates as pos-
`sible given bandwidth and power constraints. Both the
`maximum number of mobile users that the system can sup-
`port and the maximum achievable throughput depend crit-
`ically upon the bit error rate (BER) performance at the
`physical layer. The BER performance is directly related to
`the ability of recovering the transmitted symbols at the re-
`ceiver. Symbol recovery is mainly impeded by two factors:
`first, multiuser interference (MUI) and second, intersym-
`bol interference (ISI) caused by frequency selective chan-
`nels. A plethora of single-carrier and multi-carrier (MC)
`CDMA systems have been proposed recently to mitigate
`or eliminate these two factors. However, as we explain
`later on, no existing scheme guarantees symbol recovery
`in the uplink with uncoded transmissions and without im-
`posing constraints on the multipath channel nulls. This is
`exactly where we start from to develop a transmission-
`reception framework capable of delivering superior BER
`performance. While making efficient use of the available
`bandwidth, our scheme guarantees symbol recovery re-
`gardless of the possibly unknown FIR channel by achieving
`deterministic MU1 elimination and IS1 suppression.
`In the downlink, orthogonal division multiple ac-
`cess (OFDMA) is capable of complete MU1 elimination,
`provided that a sufficiently long cyclic prefix is used
`at the transmitter. To combat frequency-selective fad-
`
`ing, OFDMA employs frequency hopping andor channel
`coding[l3, pages 213-215,2201. As a result, robustness to
`multipath comes at the price of increased complexity at the
`receiver and reduced bandwidth efficiency. On the other
`hand, the ability of direct sequence (DS) CDMA to exploit
`multipath diversity (via RAKE reception) is well docu-
`mented. Capitalizing on both OFDMA's resilience to MU1
`and DS-CDMA's robustness against time-dispersiveness,
`multicarrier (MC) CDMA systems have been proposed to
`suppress both MU1 and IS1 [I, 2, 4, 25, 9, 11, 201. In the
`uplink, the use of channel coding, in the form of repetition
`or convolutional codes, results in further BER improve-
`ments at the expense of complexity and bandwidth over-
`expansion [lo, 141.
`However, without channel coding, symbol interleaving
`or assumptions on the channel nulls, there is no existing
`scheme guaranteeing symbol recovery with FIR linear or
`decision-feedback (DF) receivers.
`In order to study the
`fundamental shortcomings of existing schemes, herein we
`develop a generalized multi-carrier (GMC) CDMA frame-
`work, capable of modeling existing MC-CDMA schemes.
`Revealing thorny limitations of existing schemes is
`but one of the provisions of our all-digital GMC-CDMA
`framework. Most importantly, our GMC-CDMA frame-
`work offers guidelines for the design of user codes in or-
`der to provide symbol recovery guarantees. To this ex-
`tent, we design block FIR filterbank transmitters and linear
`or DF receivers based on an inner-codelouter-code princi-
`ple, which guarantees MUMSI-resilience regardless of the
`
`Vol. I I , No. 6, Novenibcr-December 7-000
`
`527
`
`SONY Exhibit 1009 - 0004
`
`

`

`G.B. Giannakis, A. Stamoulis, Z. Wang, P.A. Anghel
`
`Figure I : Discrete-time equivalent baseband transceiver model of the mth user.
`
`frequency-selective physical channel. Unlike the MLSE re-
`ceivers of [ 101. our multiuser DFreceivers have linearcom-
`plexity; they can be easily implemented either using filter-
`banks (in hardware) or matrix operations on a DSP chip (in
`software). Moreover, the flexibility of our code assignment
`procedure allows further BER enhancements by taking into
`account the load (i.e., the number of active users) in the
`system. Unlike existing schemes, where the code assign-
`ment does not adapt to the number of active users in the
`system, we show that without demanding extra transmitted
`power, active users in a cell can take advantage of the de-
`parture or silence of other (perhaps roaming) users by prop-
`erly re-adjusting their codes. Furthermore, in wireless mul-
`timedia networks the load-adaptability of our framework
`can certainly benefit best-effort or constant-bit rate sessions
`by possibly exploiting the fluctuations of variable-bit rate
`sessions. Finally, dispensing with bandwidth-consuming
`training sequences increases the effective data transmission
`rate; this is made possible by our novel blind channel es-
`timation algorithm which is tailor-made and suited to our
`inner-code/outer-code design.
`Our
`load-adaptive MUI/ISI-resilient GMC-CDMA
`framework follows the principles of the so-called AMOUR
`system of [7, 24, 221, whereas the development of our
`multiuser DF receivers borrows from [ 191. However,
`f7, 24, 221 assume a fully loaded system, and address
`blind channel estimation for a specific class' of user codes.
`Herein we shed light on the load-adaptive capabilities of
`our framework, and we also study the role of different pre-
`coders and their impact on BER. Moreover, we show that,
`for the same amount of induced redundancy, our inner-code
`design exhibits better BER performance than BCH codes
`with the same decoding delays. On the other hand, [19]
`refers to a single-user block transmission system.
`In Section 2 we develop our GMC-CDMA framework,
`which allows us to describe existing MC-CDMA systems
`in Section 3 and identify their inherent limitations. Thus,
`we are motivated to pursue our MUI/ISI-resilient inner-
`codelouter-code design in Section 4, and in Section 5 we
`study how the framework is modified in the case of under-
`loaded systems. In Section 6 we show how symbol recov-
`ery can be accomplished using multiuser DF receivers and
`in Section 7, we derive our blind estimation algorithm. Fi-
`nally, in Section 8 we illustrate the BER superiority of our
`designs over existing MC-CDMA schemes using extensive
`simulations, and give pointers for future research in Sec-
`
`528
`
`tion 9.
`
`2 SYSTEM MODEL
`
`First, we present a high level view and then we explain
`the knots and bolts of our system model. Fig. 1 depicts
`the all-digital baseband equivalent transmitter and receiver
`model for the mth user. User m uses an assigned code ma-
`trix C, = F,,O,
`to transmit blocks s m ( i ) of size K.
`Through the code C, , the Ii x 1 vector s,
`( i ) is mapped
`(i) which is transmitted
`to a P x 1 vector 11, (i) = C , s,
`over the channel. The channel, which is assumed to have fi-
`nite impulse response (FIR) during the transmission of the
`block, is represented by the Toeplitz (convolution) matrix
`H,. The received block x ( i ) contains multiuser interfer-
`ence (MUI) and additive noise ~ ( i ) . The receiver removes
`MU1 using the filterbank described by the matrix G,, and
`retrieves the transmitted block either using the filterbank
`rm or a DF receiver described by the matrices W, and
`B, (see Figure 2).
`
`W,
`
`Decision
`
`Figure 2: GMC-CDMA DF receiver of the rnth user.
`
`In more detail, signals, codes, and channels of the up-
`link CDMA are represented by samples of their complex
`envelopes taken at the chip rate. The data symbol sequence
`of the rn-th user is denoted by s, (k) and through serial-
`to-parallel converters is grouped into blocks of I< symbols
`s,,, (i) := [Sm (ilr') . . . sm (iI< + 11' - 1)IT. Using the or-
`der P - 1 FIR precoding filterbank {Cm,k(n)}f:',
`the
`is mapped to a block u , ( i ) of P > K chips:
`block s,,(i)
`urn(i) = C,s,,(i),wherethe(n,k)thentryoftheP x A'
`matrix C , is cn1,k(n). The mapping of I\'
`input symbols
`to P output chips is conceptually performed in two phases:
`first, the inner code is applied to facilitate IS1 suppression,
`and, second, the outer code is applied to accomplish MU1
`elimination. The inner code is represented by the J x K
`
`SONY Exhibit 1009 - 0005
`
`

`

`Load- Adaptive MUVISI-Resilient Generalized Mu1 ti-Canier CDMA with Linear and Decision-Feedback Receivers
`
`We can deduce from (2), that the precoder C , and the re-
`ceiver G,
`should be designed in such a way that MU1 is
`made as small as possible. In Section 4 we provide a joint
`design procedure for the receiver G, and the outer code
`F, which leads to MU1 elimination regardless of the phys-
`ical channel. We will also see how the proper choice of the
`inner code 0, guarantees the IS1 removal and symbol re-
`covery from the MUI-free block y, (n). But first, we mo-
`tivate the importance of our design procedure by looking at
`existing’multiuser multicarrier CDMA systems (see 161 for
`detailed derivations of all-digital equivalent models).
`
`3 EXIS.TING MC-CDMA SYSTEMS
`
`In this section we explore how our GMC-CDMA
`framework enables us to identify fundamental shortcom-
`ings of existing schemes. First we see how various CDMA
`systems can be modeled using the results of Section 2, and
`then we show why existing schemes cannot guarantee sym-
`bol recovery regardless of the physical channel.
`
`3.1 SC-DS-CDMA
`
`matrix Om, whereas the outer code is represented by the
`P x J matrix F, ( P 2 J 2 1;). The intuition behind the
`factoring of the code matrix C , into 0, and F, is that
`MC-CDMA systems introduce redundancy to combat IS1
`and MUI; ISI-related redundancy can be built-in using O m ,
`whereas MUI-related redundancy can be introduced using
`F,. Though C , is not capable of modeling all possible
`forms of channel coding, error-control codes are not pre-
`cluded from our GMC-CDMA system. A possible chan-
`nel encoder should precede the block-spreading operation
`that C,
`implements, and either operate before the S/P con-
`verter (e.g., as in the case of a convolutional code) or after
`the S/P converter (e.g., as in the case of a Reed-Solomon
`code).
`The coded chip sequence u,(n) passes through the
`discrete-time equivalent baseband channel h, (n), which
`is assumed to be of order 5 L (a common assump-
`tion in quasi-synchronous CDMA systems [7]). The im-
`pulse response h , (n) models multipath, transmit-receive
`filters, and the mth user’s asynchronism in the form
`of delay factors [7]. The chip-sampled sequence is:
`~ ( n ) = Cm=O Xrn(n) + q(n), where: ~ , , ( n ) =
`M-1
`h m ( j ) u m ( n - j ) , and q(n) is the additive noise. To
`avoid channel-induced inter-block interference (IBI), we
`design our transmitted blocks u,(n)
`to have L trailing
`zeros (TZ), which act as “guard bits”: in our matrix for-
`mulation, the lower L x J submatrix of F m is set to an
`all-zero L x J matrix 0~~ J. As detailed in [24, 221, the
`trailing zeros at the transmitter can be replaced by a cyclic
`The simplest of the CDMA transmission systems uses
`prefix (CP) that similar to OFDM must be discarded at the
`a single carrier (SC) and each symbol is spread by a code
`. . . c p ( p - 1)IT of processing gain P. It
`receiver to eliminate IBI. With TZ transmissions, the re-
`c,,O = [c,(o)
`ceivedPx lvectorx(i) := [z(iP) . . . ~ ( i P + P - l ) ] ~ i n
`follows that a SC-DS-CDMA system can be modeled in
`AGNq(i) := [ q ( i P ) . , .q(iP+ P - 1)IT can beexpressed
`thus, the P x K
`our framework by setting C, := c , , ~ ;
`as:
`in (1) reduces to a P x 1 vector, c , , ~ ,
`code matrix C,
`x(i)= C H m C m s m ( i ) + ~ ( i )
`the entries of which represent the chips of a Gold, Walsh-
`Hadamard, or pseudo-noise (PN) fl sequences. The fun-
`damental shortcoming of SC-DS-CDMA is that there may
`exist channels for which the user symbols are not identifi-
`able from the received signal. For example, suppose that
`we have a SC-DS-CDMA system with 2 users: user “1”
`with I< = 1, P = 4, uses the Walsh-Hadamard code
`c1,o = [-1 , -1 1 1IT. and transmits the data over the
`channel h l ( n ) = d(n) - S(n - 1); user “2” uses the code
`c a , ~ = 11 , -1 , -1 1IT and transmits over the channel
`h?(n) = d(n) + S(n - 1). If user “2” is silent and user
`“1” transmits sl(n) = -S(n) (which corresponds to “-1”
`in BPSK). the received signal z(n) = S(n) - 2d(n - 2) +
`S(n - 4) is identical to the case where user “1” is silent and
`user ‘2” transmits s?(n) = 6(n) (which is “1” in BPSK).
`In other words, if both users transmit at the same time, it is
`impossible to recover at the receiver the transmitted sym-
`bols, even though the user codes are orthogonal. Therefore,
`spreading at the symbol level fails to guarantee symbol re-
`covery: we will show in Section 4 that spreading (i.e., pre-
`coding) at the block of symbols level will make symbol re-
`covery possible regardless of the underlying FIR channel.
`
`M-1
`
`I
`
`(1)
`
`m=O
`where H,
`is a P x P Toeplitz (convolution) matrix with
`(i, j ) entry h, (i - j).
`At the receiver end, recovery of the transmitted block
`i)elimination of MUI, and
`sm(i) entails two actions:
`@elimination of multipath induced IS1 (within a block).
`These actions are implemented by filterbanks performing
`block processing that amounts to multiplying blocks by the
`matrices G,, W,, and B,. The first stage of the receiver
`J 1
`for user m consists of J parallel filters (gm,j ( P ) ) ~ , ~ , each
`of length P . The taps of the filters are given by the en-
`tries of the J x P matrix G, with ( j , p ) entry [G,]j,p =
`gm,,(p). The matrix G, maps the block x ( i ) to an MUI-
`free block ym ( i ) :
`ym(i) = G m H m C m s m ( i ) + Gmq(i)
`
`h I - 1
`
`-
`
`MU1
`
`Vol. I I . No. 6, November-December 2000
`
`529
`
`SONY Exhibit 1009 - 0006
`
`

`

`none of the aforementioned models guarantees MUYISI-
`free multirate transmissions and user symbol recovery in
`the presence of (possibly unknown) multipath (uplink or
`downlink) channels without bandwidth expansion.
`
`4 MUI/ISI ELIMINATING CODES
`
`G.B. Giannakis, A. Stamoulis, 2. Wang, PA. Anghel
`
`3.2 MC-CDMA
`
`By combining spreading with multicarrier modulation,
`multi-carrier (MC) CDMA schemes constitute a natural ex-
`tension of SC-CDMA systems (see e.g., [S] and [9]). The
`basic idea is that with IFFT processing at the transmit-
`ter, and FFT processing at the receiver, frequency-selective
`multipath channels are converted to flat fading channels;
`all users share the same subcarriers and MU1 suppression
`is achieved by their linearly independent chip sequences
`cp,0. Our framework encompasses MC-CDMA systems
`by setting h' := l,Vp, and C, := cp = F C , , ~ , where
`c , , ~ is chosen as in SC-DS-CDMA, while the P x P ma-
`trix F with (PI, p 3 ) entry [FIpl = exp(jZirplp?/P),
`0 5 p l , pa 5 P - 1, implements the Inverse Fast Fourier
`Transform (IFIT) of each user's chip sequence cp,0 (the
`matrix F is common to all the users).
`
`3.3 MC-DS-CDMA
`
`Whereas in MC-CDMA systems information about a
`specific source symbol is transmitted on every subcarrier,
`in MC-DS-CDMA each subcarrier carries only a subset
`of the source symbols. Specifically, the source symbols
`s p ( n ) of every user are first S/P-converted into iV, sub-
`streams { s p , k (n)};;;'.
`Then, each substream is spread
`with a P-long DS code c p , 0 , which has spectral support
`[-a, a], ( a < T ) (the DS code is user-specific). The spread
`substreams s,,k(n)cp,0 are then modulated on subcarri-
`ers { f k } k & with subcarrier spacing Afi 5 a [9, 161.
`N - 1
`The primary motivation behind MC-DS-CDMA schemes
`is their increased bandwidth efficiency (compared to MC-
`CDMA), which, however, comes at the price of BER per-
`formance degradation (because of inter-carrier interference
`and channel nulls potentially hitting a particular subcar-
`rier). Our filterbank model in (1) describes an MC-DS-
`CDMA system by choosing the P x K code matrix in (I)
`. . . Dp(f~,)cp,~]v where Dp(fi)
`as C p = [ D p ( f i ) ~ p , ~ ,
`is a P x P diagonal matrix with (n, n)th element equal to
`i = 0, . . . , N, - 1. Unfortunately, MC-DS-
`~ p ( j 2 ~ f i n ) ,
`CDMA suffers from the same symbol recovery problems
`as DS-CDMA. This is because each symbol substream
`s,,k (n) can be thought of as a separate DS-CDMA system
`which modulates the subcarrier at frequency jk.
`
`In this section we present our algorithm for the design
`of an MUVISI-resilient CDMA system capable of allevi-
`ating fading effects and of providing symbol recovery re-
`gardless of the (possibly unknown) FIR channels. First,
`we will give an intuitive description of how we go about:
`i)eliminating MU1 in the frequency (Z-) domain, and ii)
`guaranteeing symbol recovery regardless of the physical
`channel. Our approach lies in the fundamental mechanism
`of redundant spreading at the block of symbols level; in-
`terestingly enough, this mechanism is the natural evolution
`and combination of already existing techniques.
`Our origin is the observation that if users transmit over
`different frequencies (transmitting orthogonal signals in the
`FDMA sense), the orthogonality between the users' sig-
`nals will be preserved at the receiver. However, should a
`user use only one subcarrier (as in OFDMA), it would not
`be possible to recover the transmitted symbols at the re-
`ceiver, if the user's channel happened to have a null at this
`particular subcarrier. Even if the subcarrier is close to a
`channel zero, the user's signal will be greatly attenuated,
`which inevitably Ieads to poor performance. A natural so-
`lution which guarantees identifiability of the user symbols
`is given by a multi-carrier approach, where each user's data
`are transmitted over more than one frequencies (subcarri-
`ers). Such a guarantee can be given as long as the channel
`has finite number (say at most L ) zeros: then for symbol
`recovery, it suffices for every user to transmit each sym-
`bol on L + 1 frequencies. unfortunately, like MC-CDMA,
`the price paid is bandwidth over-expansion: we need L
`times larger bandwidth than the OFDMA system. This is
`exactly where the novelty of our spreading at the block-
`of-symbols level stems from: to overcome this bandwidth
`over-expansion, we transmit a group of 1; symbols using
`A' + L frequencies with 1; >> L ; as a direct result, the
`bandwidth expansion factor (1; + L ) / K can be brought ar-
`bitrarily close to 1. Under such a transmission scheme, the
`challenge is in designing the code so that the A' symbols
`can be recovered from any of the K subcarriers. By doing
`this, symbol recovery will also be guaranteed, as at most L
`of the K + L subcarriers can be nullified by the channel.
`A special case of MC-DS-CDMA are Multi-tone (MT)
`To see how the aforementioned line of thought can
`CDMA systems: the subcarrier frequencies are now cho-
`be cast rigorously in the mathematical framework of Sec-
`tion 2. we assume that N = N ( K + L ) subcarriers are
`sen as fi = i / P [20]. It is straightforward to see how MT-
`CDMA systems can be modeled using our GMC-CDMA
`available to users. We look at the practically appealing
`0
`framework (see [6] for details).
`case where the N subcarriers are distributed uniformly in
`an allocated frequency band, and we refer to each of the
`1 = 0,1, . . . , N - 1 subcarriers using the FFT frequen-
`cies exp(j'Ld/N). We consider the partition of the set
`
`3.4 MT-CDMA
`
`As we have shown, the GMC-CDMA model can be
`used to describe several existing CDMA schemes. But
`
`530
`
`SONY Exhibit 1009 - 0007
`
`

`

`Load- Adaptive MUI/lSI-Resilient Generalized Multi-Carrier CDMA with Linear and Decision-Feedback Receivers
`
`F = {exp(jSd/iV) , I = 0.. . (N - 1)) into M disjoint
`subsets 3,; each subset contains the subcarriers which are
`allocated to user m. Then the outer code matrix F, can be
`built as:
`
`(3)
`
`is a diagonal matrix comprised of the frequency response of
`the mth users’ channel at the respective allocated subcarri-
`ers. The transmitted symbols s, (i) can be retrieved from
`y,(i) using a linear receiver I?,,,.
`From (5) we observe
`that the outer code rendered the multi-user system equiv-
`alent to M single users. As a result, the linear receiver is
`given by [15]: Fz = (DH, 0,)’ for a zero-forcing linear
`receiver. In Section 6 we will see that a DF receiver results
`in improved BER performance.
`At the price of increased implementation overhead,
`general frequencies on the 2-plane (instead of frequencies
`on the unit circle used here and in [24]) can be used to
`carry the users’ symbols, which is basically the approach
`taken in [7]. At this point, we underline that though the
`design of the outer code is reminiscent of FDMA, there are
`distinct features which make the GMC-CDMA approach
`superior. First, our solution is based on an all-digital im-
`plementation which dispenses with FDMA-induced analog
`implementation problems such as subchannel leakage and
`inflexibility in subcarrier allocation (in our scheme, sub-
`carrier allocation amounts to properly setting the entries of
`Second, our novel design of the inner code guaran-
`CP,).
`tees symbol recovery even when CSI is not available at the
`transmitter as long as there is an upper bound on the chan-
`nel order. In cellular networks the channel is time-varying
`but an upper bound on the channel order is available (based
`on previous measurements). Hence, unlike existing work
`on multi-user DMT (see, e.g., [3J) which requires CSI at
`the transmitter, our work has a significant practical appeal
`in wireless environments. The latter include cellular and ad
`hoc (Bluetooth[8]-like) networking if frequency-hopping is
`introduced via Q,
`along the lines of [26].
`We remark that our code design is amenable to opti-
`mization when CSI is available at the transmitter. At the
`user level, following a “water-filling” principle, the inner
`code 0, could be optimized to maximize the mutual in-
`formation rate. The optimization of the outer code is at the
`system level, where the allocation of subcarriers could take
`into account the deep fades of the users‘ channels (i.e., a
`subcarrier close to a user’s channel null should not be allo-
`cated to this specific user). Both optimizations delineate fu-
`ture research avenues along the lines of our previous work
`on single-user block transmission systems.
`We summarize the inner-code/outer-code design proce-
`dure in the following algorithm:
`
`Global Design Decisions
`0 Given allocated frequency band, decide on the number of
`subcarriers N (a power of two).
`0 Given an upper bound on the channel order, and the num-
`ber of users select I< and hf.
`Partition the set 3 of available subcarriers to 11.1 disjoint
`sets 3,.
`
`The N x N matrix F has entries [F]l,k = exp(-j2dk/N),
`0 5 1 5 N - 1 , O _< k 5 N - 1, whereas the subcanier-
`selector N x J matrix CP, has entries [CP,]l,k = 1, if
`exp(j2n[/N) E Frn, and 0 otherwise.
`From an implementation point of view, the outer code
`F,
`takes the IFFT of the users’ data (which is computa-
`tionally not expensive) and zero-pads them to obviate IBI.
`The matrix CP, performs the subca