Performance of Extrinsic Information Scaling Mechanisms with Duo-Binary Turbo Codes

Abstract

Non-binary Turbo codes have gained significant research interest following the adoption of duo-binary Turbo codes in the Digital Video Broadcasting with Return Channel via Satellite (DVB-RCS) standard. In order to meet the Quality of Service (QoS) requirements of applications in communication systems, error control codes with enhanced error performance and low decoding complexity are highly desired. In this paper, the performance analysis of duo-binary Turbo codes using Sign Difference Ratio (SDR) and Regression analysis based extrinsic information scaling mechanisms has been performed. Different techniques for computing the scaling coefficients using the symbol and bit level Log-Likelihood Ratios (LLRs) have been presented and analyzed. Results show that the best gain in error performance is obtained using the mean of scaling coefficients computed from bit-level LLRs with both SDR and Regression based algorithms. The schemes give an average gain of 0.1 dB as compared to the conventional scheme.