| This dissertation is composed of two main parts. The first and largest part deals with joint phase estimation and turbo-decoding in a flat Rayleigh fading channel. At the region of low SNR where turbo-codes operate, and particularly if the variation of the Rayleigh channel is quite large—i.e., large Doppler—the task of channel estimation becomes quite challenging and should be done jointly with turbo-decoding for better results. To this end, a Markov model is developed for a discretized version of the channel phase (since this is a bigger problem for PSK transmission than amplitude variation) and then the Forward-Backward algorithm is used on the phase trellis implied by this Markov model to acquire the channel phase iteratively, while performing turbo-decoding. Clearly, as the iterations proceed, the reliability of the coded symbols increases, causing them to act somewhat as pilots and facilitate the phase estimation process also.; This channel estimation scheme combines well with spectrally efficient trellis turbo-codes and offers comparable performance to existing pilot averaging techniques at half the bandwidth. To assess the proximity of the performance to channel capacity, upper bounds to the capacity of idealized Markovian channel models are developed, and it is demonstrated that performance as close as 1.3 dB from these upper bounds to capacity without explicit CSI is possible. Also, this technique for iterative quantized phase estimation is extended to the case where antenna diversity is available at the receiver, and the performance improvement due to diversity is shown to be almost as much as the increase in channel capacity.; The second part of this dissertation addresses joint channel estimation and equalization for a general system with nT transmitter and nR receiver antennas, impaired by co-channel interference and ISI. A Kalman filter is used to track the frequency-selective channel, which is modeled as a first-order vector autoregressive process. The Kalman filter is aided by delayed decisions from a MIMO m.m.s.e. DFE, which equalizes and decouples the transmitted signals, based on channel estimates received from the Kalman filter. This approach works much better than conventional adaptive algorithms such as LMS and RLS, at the expense of higher complexity. Furthermore, suitable coding options for that equalization and interference cancellation scheme are briefly discussed. |
