Traitement du signal pour les communications mobiles multi-utilisateur en accès multiples par répartition de codes utilisant la méthode directe

The last decade of twentieth century witnessed explosive growth in wireless cellular mobile and fixed systems, and the continuing interest in this area suggests that most future communication systems will be untethered networks. With the advent of digital wireless cellular systems like GSM and IS-95, the concept of reliable global coverage in nomadic communications has for the first time seemed possible. However, these systems were representatives of the second generation of wireless communications and the target market was voice. Even though the quality still leaves much to be desired, voice communication in cellular radio is already considered to be a second generation (2G) issue - solved, and thus a deployment issue, and of little interest in terms of future growth. In today's wireless communication forums, experts therefore dare to speak of networks capable of handling high-speed data over mobile and wireless channels for multimedia applications. GSM is hoping to smoothly evolve a part of itself into EDGE, a new version, capable of handling higher (approaching 384 kb/s) data rates at the cost of mobility/coverage. GPRS, another variant, is the packet-switched service of GSM already at the brink of seeing its market debut. It is said to be capable of handling rates approaching 144 kb/s under low system loading conditions. Third generation systems like the Universal Mobile Telecommunication System (UMTS) or its North American counterpart, the CDMA-2000 are both aiming at rates approaching 2 megabits per second for cellular systems. It seems, however, that third generation (3G) systems will be based, to start with, upon a second generation backbone and the two will co-exist for long years before the natural death of the latter. Lessons learned from 2G systems and attempts to develop high-rate versions of them, however, suggest that the network must be a hybrid structure. This is motivated by the fact that all rates and all applications are never required by all customers in all situations and communication scenarios. Consequently, a major core of any network will need to support voice and relatively low-rate data at all mobile speeds and locations, ensuring global reliable connectivity, with advanced services provided wherever and whenever the need arises. The third generation of wireless networks basically aims at the following target performances:
_ Full coverage and mobility for 144 kb/s at first, and 384 kb/s later
_ Limited coverage and mobility for 2 Mb/s
_ High spectrum efficiency compared to 2G systems
_ Higher flexibility to incorporate new services
_ Backward compatibility to 2G systems
In order to satisfy some of these highly demanding requirements, a high performance physical layer
needs to be designed, incorporating sophisticated signal processing techniques to mitigate the distortion caused by radio propagation phenomena. For DS-CDMA systems which constitute the core multiple-access technology in future wireless systems, the traditionally used receiver technique is the RAKE receiver. This receiver is an anti-multipath device, but is known to operate in environments where the delay spread of the propagation channel is short relative to the symbol duration. Furthermore, strict power control needs to be exercised in order to keep the interference level down for the proper functioning of this receiver. If these conditions are not satisfied, when for example, transmission rates require the symbol duration to be short, thus resulting in a small processing gain, or when power cannot be controlled efficiently, more sophisticated interference cancelation algorithms need to be designed to either replace or supplement the RAKE receiver. This is the driving force behind the major portion of the work carried out in thesis where advanced receiver algorithms for multipath radio channels are proposed as alternatives to the traditionally used RAKE receiver.
A related and even more critical issue is that of parameter estimation which implies identification of channel parameters or its impulse response. Naturally, this estimate needs to be relatively accurate in order to build any reasonable receiver. This thesis also deals with channel identification issues and techniques through various methods. The usual method for channel identification is the use of known training data. Bandwidth efficiency lost to training data or pilot channels is another of the undesirable phenomenon in a system; to get around this problem, we propose blind methods (without the use of training information) for channel identification and also explore hybrid semi-blind channel identification and receiver algorithms. A crucial observation pertaining to advanced receivers is that the interference canceling capability for a given receiver comes about due to diversity techniques, which refers to the reception of the signal through several independent channels. These channels can be created by employing one of the well-known methods e.g., fractional oversampling or several reception antennas. This issue is discussed in detail in this thesis and spatio-temporal interference cancelation schemes are presented for both the forward and reverse link problems. Emphasis is also laid on the of exploitation of side information in the problem, like training information, transmitter filter characteristics, structure of the channel, and the knowledge of spreading sequences of DS-CDMA users, in order to derive improved and low-complexity receivers. From the same motivation, the uplink and downlink problems are treated separately, since although the latter can be made to look like the former and handled in the same fashion, appreciable gains can be achieved by considering it from a different angle while exploiting the very particular structure of the forward link. A new dimension where interference can be cancelled, and which has attracted much interest in recent years is the space dimension. Joint spatio-temporal signal processing techniques, also known in the literature as smart antenna processing, offer a significant advantage over pure beamforming strategy for forward link transmission. This is another area addressed in this thesis. We treat the problem of performing optimum spatio-temporal processing while using antenna arrays at the base-station for multiuser downlink transmission. The two transmission modes discussed are the Time-Division Duplex (TDD) and the Frequency-Division Duplex (FDD), in which varying degrees of information about the downlink channel is available at the base-station from the uplink channel estimate. It is this information that is exploited to design spatio-temporal filters at the base-station to attempt to separate
users in space/time and to improve downlink performance while reducing mobile station complexity. The TDD and FDD problems are discussed separately and solutions are proposed for both. The effect of scrambling on the structure of the problem are also discussed and solutions for this case are also presented.