Interleavers for Turbo codes

Abstract

Since their introduction in 1993, in a seminal paper by Berroud, Glavieux, and Thitimajshima, Turbo code have changed the way we look at high-performance error-control coding. After some initial scepticism from the coding community, Turbo coding gained rapid acceptance as the original results were replicated by a number of independent researchers. A flexible software Turbo codec and simulation environment is designed and tested. Using this tool, the Turbo code interleaver design problem is considered for large block sizes, where the effect of trellis termination is less marked. The performance of various interleavers with a similar block size are compared, including an implementation of the uniform interleaver. An optimised interleaver design technique based on simulated annealing is proposed - the results obtained show that the error performance may be significantly improved without increasing the delay. Finally, interleaver design for small Turbo codes is considered. In this case it is shown that while correct termination improves the performance for an average interleaver, its effect on Turbo codes with optimised interleavers is negligible. Using our simulated annealing design technique it is easier to include restrictions which make the interleaver correctly-terminating or odd-even. While the S-random algorithm serves well for specifying interleaver spread, we believe that our algorithm is better suited for more sophisticated design criteria.