Improving the performance of turbo codes via sub-optimal rate allocation

This paper looks at the problem of the implicit unequal error protection observed in the asymptotic performance of most turbo codes. In particular we propose a new formulation of the rate allocation problem associated with the distribution of a fixed quota of coded bits in the trellis sections of the upper and lower recursive systematic convolutional (RSC) codes of the turbo code structure. The main goal of this work is the design of an effective greedy approach for solving this rate allocation problem via a two phase process of puncturing and repetition coding. Sample simulation results confirm that the asymptotic performances in terms of error floor are improved confirming the potential gain of this approach.