Exploiting the nature of extrinsic information in iterative decoding

Histograms of extrinsic information are used to explain the performance gain of standard turbo codes with number of iterations. The simulation results presented clearly show the Gaussian nature of the extrinsic information and they also establish a heuristic link with performance gain. Based on the histogram plots, a modified form of turbo encoding and decoding structure is proposed. The encoder is formed with three parallelly concatenated recursive systematic convolutional (RSC) encoding blocks separated by two random interleavers, subsequently a heuristic method is proposed about the way the extrinsic information is fed to the three blocks in the decoder. A new scaling of extrinsic information is proposed based on the observations made on new structure as well as performance gain in standard turbo structure.