| Receiver techniques and joint transmitter-receiver techniques are studied for various wireless channel models. We first analyze the performance of a reduced-rank equalizer, constrained in a Krylov subspace, for frequency-selective fading channels. Asymptotic performance is characterized, as the filter length(s) approach infinity, for both linear and decision-feedback equalizers (DFEs). A reduced-rank equalizer can achieve MMSE performance with relatively low rank(s), and significantly outperform a full-rank equalizer with limited training. A rank-recursive algorithm for computing the equalizer with rank selection rules is developed. We apply this to an adaptive reduced-rank turbo DFE, which can effectively exploit available diversity.; We also consider a multiple-input/multiple-output channel with an adaptive linear receiver. For a fixed packet length, we show that the training length, which maximizes the capacity, grows as the square root of the packet length.; We then consider joint transmitter-receiver optimization. For a CDMA system, we propose and analyze an iterative algorithm to adapt the signature with a reduced-rank receiver. Both single-user adaptation in the presence of fixed interference, and group adaptation, in which multiple users update their signatures, are considered. With little training, the algorithm offers a dramatic improvement in performance relative to full-rank adaptation. To reduce feedback requirements, reduced-rank signature adaptation is studied with a reduced-rank receiver. In many cases the reduced-rank signature performs better than full-rank signature adaptation.; Finally, we consider a multi-carrier system, in which the channel is known to the receiver, and limited channel information can be relayed back to the transmitter. We characterize the asymptotic growth in capacity as the number of sub-channels tends to infinity, which depends on the amount of feedback available. We analyze the asymptotic capacity achieved by a uniform power allocation over a subset of sub-channels with quantized rate control. We show that the asymptotic growth in capacity for this power allocation scheme is the same as for water-filling. Correlated sub-channels are also considered, which can reduce the feedback required. The results are then extended to both the uplink and downlink of a single cell with multiple users. |
