IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 46, NO. 12, DECEMBER 1998 1693Iterative Multiuser Detection for CDMA with FEC:Near-Single-User PerformanceMark C. Reed, Student Member, IEEE , Christian B. Schlegel, Senior Member, IEEE,Paul D. Alexander,Member, IEEE, and John A. Asenstorfer, Member, IEEEAbstract—This paper introduces an iterative multiuser receiverfor direct sequence code-division multiple access (DS-CDMA)with forward error control (FEC) coding. The receiver is derivedfrom the maximum a posteriori (MAP) criterion for the jointreceived signal, but uses only single-user decoders. Iterationsof the system are used to improve performance, with dramaticeffects. Single-user turbo code decoders are utilized as the FECsystem and a complexity study is presented. Simulation resultsshow that the performance approaches single-user performanceeven for moderate signal-to-noise ratios.Index Terms—Code-division multiaccess, decoding, multiuserchannels, random codes, turbo codes.I. INTRODUCTIONWITH THE standardization of direct spread code-division multiple access (DS-CDMA) for mobilecommunications [1], a number of vendors have introducedtheir products onto the world market. This has raised a lotof interest on the potential capabilities and capacity of thismultiple-access technology [2]–[7]. In this paper we studythe uplink, or the base-station (BS) receiver. In the designof these systems most are currently symbol-synchronous orquasisymbol-synchronous so that orthogonal codes can beutilized. When orthogonal codes are used the BS linear filterreceivers perform well in detecting the signal sent by takingadvantage of this orthogonality, which gives performanceequal to single-user performance.In a true mobile wireless system, synchronization is difficultto maintain and needs tight closed-loop timing control betweenthe BS and the mobile station (MS). If this timing controlis not maintained then the orthogonal properties are lost andPaper approved by R. Kohno, the EDitor for Spread Spectrum Theory andApplications of the IEEE Communications Society. Manuscript received April28, 1997; revised December 17, 1997 and April 28, 1998. This work wassupported in part by Telstra Australia under Contract 700136 and by theCommonwealth of Australia under International S & T Grant 56. This paperwas presented in part at the International Symposium on Personal, Indoor andMobile Radio Communications, Helsinki, Finland, September 1997.M. C. Reed was with the Mobile Communications Research Center,University of South Australia, Mawson Lakes SA 5095, Australia. He isnow with Ascom Systec AG, Technology Unit, Gewerbepark, CH-5506Maegenwil, Switzerland.C. B. Schlegel is with the Department of Eletrical Engineering, Universityof Utah, Salt Lake City, UT 84112 USA (e-mail: [email protected]).P. D. Alexander is with the Centre for Wireless Communications, NationalUniversity of Singapore, Singapore 117674 (e-mail: [email protected]).J. A. Asenstorfer is with the Mobile Communications Research Center,Institute for Telecommunications Research, University of South Australia,Mawson Lakes SA 5095, Australia (e-mail: [email protected]).Publisher Item Identifier S 0090-6778(98)09380-5.performance degrades severely. Multipath effects, common inmobile radio channels, also destroy this orthogonal property.If the codes are randomly selected, however, the performanceof a synchronous system is on average the same as that of anasynchronous system. Work produced by Grant et al. [8] showsthat the capacity penalty vanishes, for a large number of users,using randomly selected spreading codes, as the ratio betweennumber of users and spreading length becomes large. Jana etal. [9] have shown a slightly different result; they showedthat the upper bound of the normalized minimum distance fora trellis-coded multiuser CDMA system with nonorthogonalspreading is identical to that of the single-user case. Thismeans that asymptotically, using nonorthogonal codes, orrandom codes, single-user performance should be possible.With such a receiver the performance under asynchronous con-ditions will be the same as that under synchronous conditions.We are therefore motivated to look for new multiuser receiverstructures that use random codes to achieve near-single-userperformance.A paper by Giallorenzi et al. [7] formulated the optimalmultiuser sequence estimator for an asynchronous DS-CDMAsystem where each user employs convolutional error controlcoding. Giallorenzi et al. found that the complexity per infor-mation bit using the MLSE solution depends exponentially onthe number of users in the system and the number of states ineach user’s encoder. We propose to partition the receiver toreduce the complexity, without sacrificing performance. Thispaper therefore describes a partitioned trellis-based receiverwith separate equalization and decoding. We develop a mul-tiuser receiver (or equalizer) from the maximum-a posteriori(MAP) criterion. The MAP criterion maximizes the probabilityof a correct bit decision and, hence, minimizes the probabilityof error [10, p. 245].Recently, a new coding method, called turbo codes, wasintroduced [11]. This technique achieves reliable transmissionwhile operating close to the Shannon limit. Turbo codescombine the concept of soft-in/soft-out decoding, iterativedecoding, nonuniform random interleaving, and parallel con-catenated convolutional codes (PCCC). Further to this, pub-lished results by Benedetto and Montorsi [12] discuss serialconcatenated convolutional codes (SCCC).Several authors have proposed using turbo codes for DS-CDMA systems [4], [5]. These papers discuss system im-plementations but show no performance results. The authorswould like to note that since the submission of this paperseveral independent publications have shown similar results0090–6778/98$10.00  1998 IEEE1694 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 46, NO. 12, DECEMBER 1998(albeit without the use of turbo codes for the channel code),namely, work by Moher [14] and Tark¨oy [15].In this paper we concatenate the MAP-based multiuserreceiver and soft-in/soft-out single-user trellis decoders toproduce a type of serial concatenated convolutional code.The proposed structure differs from successive interferencecancellation [6], because no signal cancellation takes place.Our system can be viewed as a soft parallel interferencecanceler or more correctly as an SCCC. A paper by Douillartet al. [13] discusses the use of an SCCC to