Adaptive GDFE based detection and I/Q mismatch compensation for MIMO systems

Abstract

This thesis investigates signal detection and compensation of I/Q mismatch and DC offset for multi-input multi-output (MIMO) systems. First, an adaptive receiver for MIMO systems is derived based on the generalized decision feedback equalizer (GDFE) in an attempt to reduce the computational complexity of the vertical Bell Labs layered space time (V-BLAST) processing in time-varying channels. The proposed algorithm rearranges the detection ordering and adaptively adjusts the filter tap weights of the GDFE using the recursive least squares (RLS) algorithm. Since the proposed adaptive technique does not need explicit channel tracking and estimation of noise variance, it requires less computation than the V-BLAST detector. In particular, the proposed algorithm can be further simplified by using the fact that the detection orderings are fairly similar for successive transmission data. Computer simulations show that the proposed receiver performs comparable to the V-BLAST algorithm with channel tracking. In the second part, the proposed adaptive GDFE is applied to MIMO systems with layered space-time block codes (STBCs). By dividing the transmit antennas into groups, the MIMO system with the STBCs provides a tradeoff between spatial multiplexing and transmit diversity. It is shown that the adaptive GDFE can be used for the detection of the layered STBCs through some modification. In the modified GDFE algorithm, the orders of weight vectors are increased by the size of an antenna group, and the detection orderings are rearranged among antenna groups. In time-varying channels, the proposed adaptive receiver needs less computation than the V-BLAST processor with channel tracking, and performs comparable to the latter. Finally, compensation of I/Q mismatch and DC offset is considered for MIMO systems with low intermediate frequency (IF) or direct conversion architectures. In the proposed scheme, I/Q mismatch and DC offset parameters are estimated in a least squares (LS) ...