SIMULATION AND IMPLEMENTATION OF BLOCK TURBO CODES

Abstract

Turbo-Decoding revolutionized error correction techniques after they were first proposed in 1993. A lot of work has been done on Convolutional Turbo Codes (CTCs) since then, however Block Turbo Codes (BTCs) have been partially neglected. At high code rates, BTCs achieve near-capacity levels. This thesis involves investigating a BTC decoding algorithm first introduced by Ramesh Pyndiah. Pyndiah’s algorithm uses soft-input-softoutput (SISO) decoders that are based on the Chase soft-decoding algorithm which is a Maximum Likelihood (ML) technique. Various encoder schemes for BTCs have also been discussed. Comparisons have been done using different numbers of competitor vectors for the Chase algorithm while varying the length of the linear block code involved. Fixed-point analysis has also been conducted for the considered BTC decoding mechanism. The second part of the thesis involves Hardware Descriptive Language (HDL) implementation of Pyndiah’s BTC algorithm. The algorithm has been broken down into various sub-modules which are designed using Algorithmic State Machines (ASMs). The SISO decoder considered inherently uses an algebraic hard-input-hard-output decoder which has been implemented via the Reformulated inversionless Berlekamp Massey (RiBM) algorithm. The RiBM algorithm is an efficient hardware design used to decode Reed-Solomon (RS) and Bose-Chauduri-Hocquenghem (BCH) codes which are used to encode data in our BTC