Practical layered rateless codes for the gaussian channel: power allocation and implementation

Practical rateless codes for the additive white Gaussian noise channel based on the layer-dither-repeat construction of Erez, et al. [5] are developed and analyzed. We use an iteratively decoded base code for each layer and a maximal-ratio combining receiver, which provides suitable log-likelihood ratios to the decoder. For this construction we derive an algorithm for computing the optimal power allocation among the layers that takes into account the base code gap to capacity. We further derive the initial and limiting gaps to capacity of the rateless code. Finally, we verify the predicted performance characteristics via simulations using a turbo base code.