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.
