Techniques and analysis of forward link channel estimation in duplex wireless communications

This dissertation investigates novel, low cost channel estimation and compensation schemes for the base station transmitter of broadband wireless systems. In particular, a bent-pipe feedback structure is investigated to allow a base station to identify both downlink and uplink channel impulse responses. The feedback structure achieves significant downlink bandwidth saving because no downlink or even uplink training is necessary. Direct knowledge of the downlink channel in the base station is then used to design precompensation schemes against downlink channel distortion via precoding, power loading or adaptive modulation. Compared with conventional channel estimation and closed-loop feedback approaches, our proposed method significantly improves downlink throughput and bandwidth efficiency, and reduces mobile receiver complexity.;The dissertation first presents detailed feedback channel estimation algorithms for single-input-single-output (SISO) channels. Receiver oversampling enables use of a prior transceiver filter knowledge, thereby providing substantial performance improvement over baud rate sampled systems. Exact expressions for channel estimation Cramer Rao lower bound are derived and applied to analyze the impact of system design parameters. With help of receiver oversampling and bent-pipe feedback, a non-conventional fractionally spaced precoder in the base station is designed and outperforms its baud rate equalizer counterpart.;Next, the bent-pipe signal feedback concept is extended to orthogonal frequency division multiplexing (OFDM) systems. We present an integration of OFDM bit- and power-loading with feedback channel estimation, and tackle identifiability issues associated with power loaded multicarrier systems.;The more challenging multiple-input-multiple-output (MIMO) channel estimation problem can be solved using the proposed decimated signal feedback scheme. Moreover, with a simple linear ambiguity resistant precoder, unique channel matrix identification is achieved in the base station up to a scalar. This is a significant improvement over conventional blind MIMO channel identification algorithms that are typically subject to an inherent invertible matrix ambiguity.;In practice, many MIMO preceding schemes only require transmitter knowledge of the channel information in the quadratic form H HH. Therefore we develop a bandwidth efficient channel estimation algorithm that only acquires the quadratic channel information for downlink preceding. Both flat fading and frequency selective fading MIMO systems can benefit from this simpler feedback estimation scheme.