Turbo Coded Modulation for Partially Coherent Channels

Washington State University, Pullman WA

Investigators

Abstract

Abstract: This research considers the design of bandwidth efficient coded modulation for partially coherent channels. Such channels occur in practice when coherent receivers are operated at low or moderate signal-to-noise ratios (SNRs), resulting in noisy phase estimates. This happens for example on wireless channels, where multipath fading causes fluctuations in received SNR. This research leverages a recent breakthrough known as turbo coding, wherein the outputs of two or more encoders at the transmitter are iteratively decoded at the receiver, resulting in reliable communication at SNRs within 1 dB of channel capacity. This research integrates phase-tracking error into all aspects of turbo coded modulation design, thereby reducing the required SNR for practical coherent communication systems. The new design techniques stemming from this research will therefore interest not only communication and coding theorists, but also engineers designing practical communication systems. The goal of this research is to develop theoretically based, practical design methods for parallel and serially concatenated turbo coded modulation over partially coherent (PC) channels, including PC additive white Gaussian noise (AWGN) channels, and PC fading channels with and without channel state information (CSI), with and without correlated fading, and with closed-loop and open-loop phase estimation. The investigators study the information capacity of PC channels, and determine capacity-achieving input probability density functions (PDFs) and capacity-optimal constellations for equiprobable and non-equiprobable signaling under average and peak power constraints. Distance metrics based on pairwise symbol error probabilities will be derived for PC fading channels, and iterative decoder performance prediction tools based on union bounds and convergence-region analysis will be developed. The investigators will then study set-partitioning, bits-to-symbol mapping, and code-search procedures for both linear and non-linear recursive convolutional encoders used as constituent codes in concatenated trellis-coded modulation schemes. Finally, shaping code design techniques for turbo coded modulation on power-constrained PC channels will be developed, based on the capacity-achieving PDFs of these channels, which are known to be non-Gaussian. These codes will shape the high-dimensional space of encoder output symbol sequences so that the induced two-dimensional PDF will closely approximate the capacity-optimal input PDF.