Frequency synchronization techniques for coordinated multibase MIMO wireless communication systems

This thesis is a study into the synchronization of the downlink of a coordinated multibase/multiuser (CMBMU) system. The goal of CMBMU systems is to form a system-wide MIMO network using multiple co-channel basestations (BSs) and mobile subscribers (MSs). Compared to existing cellular systems, CMBMU benefits include increased micro- and macrodiversity. The effects of channel and system impairments on the sum-rate capacity of a linear downlink CMBMU beamforming technique are examined in order to establish which impairments are most problematic. The presence of multiple carrier frequency offsets (CFOs) within the system is identified as the largest problem. It is demonstrated that the effect of local oscillator induced CFO on the sum-rate capacity is more significant than Doppler-induced CFO. Then, multi-CFO estimation techniques are explored. A MS-side estimator that exploits the convexity of the maximum likelihood (ML) metric is derived. Since convexity of the ML metric is dependent on the length of the training sequence and the severity of the CFO, segmentation methods that guarantee convexity of the ML metric are formulated. The performance of both the convex estimators and the segmentation methods are shown to asymptotically meet the Cramer-Rao lower bound for multi-CFO estimates. As it is impossible for the MS to correct for the BS CFOs in the CMBMU downlink, an estimation and correction technique that requires co-operation between the BSs and MSs is presented. The initial MS-side estimates, which are fed back to the BSs, are shown to be biased. The proposed technique overcomes this bias, with the end result being that the entire system is brought to a common carrier frequency. During multi-CFO estimation, additional channel impairments such as timing offset, frequency selective channels and time selective channels can complicate matters. Straight-forward modifications to the ML metric are made that compensate for timing offset and frequency selective channels. A new ML metric based on a parameterized channel estimate that uses the autocorrelation function of the channel gains is used to compensate for time selective channels. Improvements to the CFO MSE of nearly two orders of magnitude are achievable over using the original channel estimator.