Design and Implementation of Improved Decoding Algorithms for LDPC Convolutional Codes

A windowed decoder in its basic form converges rather slowly and has a large performance gap to a full-block decoder. In this work, we propose two techniques to improve the performance of windowed decoders for Low-Density Parity-Check Convolutional Codes (LDPC-CCs). The first technique: the LRL decoder, focuses on the movement direction of the window in which the window moves forward and backward across the Parity-Check Matrix (PCM). The second technique: the IPSC, focuses on the convergence criterion for the windows where the criterion is dependent on window size. We chose the LDPC-CCs specified in the standard IEEE 1901 Broadband Power Line (BPL) to evaluate our techniques. We found that a proper end-termination for the BPL’s LDPC-CCs is infeasible. We show that although the termination procedure mentioned in the standard fails to reduces the Check Node (CN) degree, the known termination bits at the decoder effectively reduce the CN degree. Simulation results show that compared to the sliding-window decoder, the LRL decoder has a decoding performance gain of about 1.6 dB while simultaneously reducing the decoding complexity by up to 40% . On the other hand, the IPSC technique proves to reduce the decoding complexity by up to 34% depending on the window sizes.