New applications of the simple bound: random codes, serial turbo codes and punctured codes

This paper deals with the application of the "simple" bound technique to a new family of serially concatenated turbo codes and to random codes with the goal of a fair comparison of the maximum likelihood (ML) performance in AWGN channels. Reported results confirm that outer code improvement causes great advantages both for the waterfall and the floor. In particular, it is shown that, for a given rate and block size, better outer encoders make concatenated codes very close to random codes down to the floor. Moreover, the BER floor is shown to be related to the weight distribution. As this becomes more and more binomial (random-code-like), the BER and FER curves improve. Finally, punctured communications are taken into account and a new upper bound of the ML performance is introduced; in particular, a new strategy, called "uniform puncturing", is considered with remarkable results for random codes and, also, for the considered turbo-like codes.