On the coding gain of incremental redundancy over chase combining

We develop an information-theoretic model to explain and predict the coding gains of the incremental redundancy (IR) over the chase combining (CC) scheme using random coding bounding techniques. From extensive simulations, the model is found to provide accurate coding gain predictions in conjunction with either optimal or practical repeated low-rate codes. Contrary to a widely held belief that IR is superior to CC for all circumstances, we demonstrate that this is not always the case when parallel concatenated turbo codes with severe puncturing are used in IR on fading channels. Based on the theoretical model, we further propose an adaptive algorithm to remedy this anomaly.