La mise en œuvre de systèmes basés OFDMA nécessite une synchronisation de la fréquence très fine en raison de l'extrême sensibilité de l'OFDMA aux décalages en fréquence porteuse (CFO). La synchronisation en fréquence devient plus difficile dans les systèmes OFDMA en liaison montante. Nos objectifs comprennent l'étude et l'analyse des problèmes résultant de décalages de fréquence et la proposition de solutions pour lutter contre ces problèmes. Nous examinons d'abord les interférences résultant de décalages de fréquence porteuse présents dans l'oscillateur du terminal utilisateur. Nous démontrons que l'on doit prendre en compte le préfixe cyclique tout en analysant les interférences résultant du CFO. Ensuite, nous montrons qu'il existe une contradiction entre la diversité de fréquence de canal et la robustesse contre le CFO. Nous proposons un compromis sous la forme d'une taille de bloc de seuil, afin de permettre un bon compromis entre la diversité des canaux et de robustesse pour les CFO pour le cas où aucune connaissances du canal n'est disponible. Quand le canal est connu, nous proposons une allocation optimale par bloc grâce à laquelle la robustesse aux CFO et à la diversité de fréquence de canal peut être réalisée en utilisant une petite taille de bloc petit pour des CFO de petite valeur. Nous proposons également une valeur CFO critique, en dessous de laquelle l'allocation optimal par bloc est très performant.Ensuite, nous proposons des solutions pour deux problèmes importants rencontrés dans un système OFDMA en liaison montante.Premièrement, nous proposons une méthode efficace pour l'estimation conjointe des réponses impulsionnelles des canaux et fréquences porteuses basée sur l'approximation polynomiale.Notre méthode d'estimation conjointe est plus simple que les méthodes existantes, sans aucune dégradation de performance. Ensuite, nous proposons une méthode de compensation de CFO basée sur l'annulation des interférences successives (SIC). La méthode d'annulation proposée réduit la complexité de mise en oeuvre quand le nombre de porteuses est important. / One of the most prominent issues in the design and implementation of OFDMA based systems is the need for a very fine frequency synchronization due to the fact that OFDMA, like OFDM, is extremely sensitive to carrier frequency offsets (CFO). The task of frequency synchronization becomes more challenging in the uplink OFDMA-based systems where one OFDMA symbol is generated by the contribution of many different users. Our goals include the study and analysis of problems resulting from frequency mismatches provide solution to combat these problems. We first look at the interference resulting from CFOs, resulting from user terminal oscillator mismatch. We demonstrate that one must take into account the cyclic prefix while analyzing interference resulting from CFO. A new analytical expression of the ICI that takes into account the effect of CFO on the cyclic prefix is proposed. Then we focus our attention on analysis of the trade-off between channel frequency diversity and robustness against CFO and show that there exists a contradiction between the two. We propose a trade-off in the form of a Threshold blocksize, to allow a good compromise between the channel diversity and robustness for CFO for the case when no CSI is available. For system where CSI is available, we propose an optimal block carrier allocation scheme through which both robustness to CFO and channel frequency diversity can be achieved with small blocksize for small CFO. We also propose a Critical CFO value, above which the performance of the optimal block carrier allocation loses interest.Next we propose solutions for two important issues encountered in an uplink OFDMA system. First, we propose an efficient method for joint estimation of channel impulse responses and carrier frequency at the receiver based on polynomial approximation. Our proposed joint estimation method is simpler than the existing methods without any performance degradation. Next we propose a CFO compensation method based on successive interference cancellation. The proposed cancellation method reduces the implementation complexity faced in case of large DFT matrices.

Parameter estimation and signal detection are the two key components of a wireless communication system. They directly impact the bit-error-ratio (BER) performance of the system. Several criteria have been successfully applied for parameter estimation and signal detection. They include maximum likelihood (ML), maximum a-posteriori probability (MAP), least square (LS) and minimum mean square error (MMSE) etc. In the linear detection framework, linear MMSE (LMMSE) and LS are the most popular ones. Nevertheless, these criteria do not necessarily minimize the BER, which is one of the key aspect of any communication receiver design. Thus, minimization of BER is tantamount to an important design criterion for a wireless receiver, the minimum bit/symbol error ratio (MBER/MSER). We term this design criterion as the minimum-error-probability (MEP). In this thesis, parameter estimation and signal detection have been extensively studied based on the MEP framework for various unexplored scenar-ios of a wireless communication system. Thus, this thesis has two broad categories of explorations, first parameter estimation and then signal detection. Traditionally, the MEP criterion has been well studied in the context of the discrete signal detection in the last one decade, albeit we explore this framework for the continuous parameter es-timation. We first use this framework for channel estimation in a frequency flat fading single-input single-output (SISO) system and then extend this framework to the carrier frequency offset (CFO) estimation of multi-user MIMO OFDM system. We observe a reasonably good SNR improvement to the tune of 1 to 2.5 dB at a fixed BER (tentatively at 10−3). In this context, it is extended to the scenario of multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) or MIMO-OFDM with pa-rameter estimation error statistics obtained from LMMSE only and checked its effect at the equalizer design using MEP and LMMSE criteria. In the second exploration of the MEP criterion, it is explored for signal detection in the context of MIMO-relay and MIMO systems. Various low complexity solutions are proposed to alleviate the effect of high computational complexity for the MIMO-relay. We also consider various configurations of relay like cognitive, parallel and multi-hop relaying. We also propose a data trans-mission scheme with a rate of 1/Ns (Ns is the number of antennas at the transmitter) with the help of the MEP criterion to design various components. In all these cases, we obtain considerable BER improvement compared to the existing solutions.

科学研究費補助金 研究種目:基盤研究(C)(2) 課題番号:07805037 研究代表者:片山 正昭 研究期間:1995-1996年度