A Decoding Algorithm for Terminated Convolutional Codes over the Blockwise Noncoherent Channel

A decoding algorithm for convolutional codes (CCs) is proposed to eliminate the need for pilot symbols in short-packet communication systems operating over a blockwise noncoherent channel, i.e., a channel that introduces a random phase rotation of the transmitted codeword block. The algorithm is tailored to M-PSK modulations and applied to zero-tail terminated CCs. It works by implementing a three-step Viterbi-based noncoherent decoder, where the performance of a first blind decoding stage is enhanced through a code-aided phase estimation (CPE) that allows for correction of previous errors. In this framework, it provides a gain of approximately 0.5 dB compared to a pilot-aided decoder, with a manageable increase in complexity. Furthermore, numerical results show a performance within a few tenths of a dB from finite-length achievability bounds with both BPSK and QPSK modulations.