Capacity and coding for the block-independent noncoherent AWGN channel

Communication over the noncoherent additive white Gaussian noise (AWGN) channel is considered, where the transmitted signal undergoes a phase rotation, unknown to the transmitter and the receiver. The effects of phase dynamics are explicitly taken into account by considering a block-independent model for the phase process: the unknown phase is constant for a block of N complex symbols and independent from block to block. In the first part of the paper, the capacity-achieving input distribution is characterized. In particular, it is shown that the maximizing density has circular symmetry, is discrete in amplitude with infinite number of mass points, and always has a mass point at zero. Furthermore, asymptotic expressions and bounds for the capacity are derived. Based on these results, the capacity is evaluated through numerical optimizations for unconstrained and modulation-constrained input distributions. In the second part of this paper, inspired by the capacity results, two classes of coding and modulation schemes are proposed for fast and moderate phase dynamics. In the case of fast phase dynamics (i.e., small N), optimized modulation alphabets are designed having exponential complexity with N at the demodulator. In the case of moderate phase dynamics (i.e., moderate values of N), specially designed modulation alphabets are utilized that have linear complexity with N. These alphabets are used together with optimized irregular low-density parity-check (LDPC) codes. Simulation results show that these codes can achieve close-to-capacity performance with moderate complexity, and outperform the best known codes so far.