Power allocation policies for convolutional and turbo coded systems over fading channels

Abstract

We study adaptive power allocation (PA) policies for improving the performance of convolutional and turbo codes on fading channels. The transmitter has an average power constraint. The fading process can be continuous (e.g., Rayleigh distribution). Perfect channel state information at the transmitter (CSIT) and the receiver (CSIR) are assumed. For convolutional codes, we consider block (slow) fading and fast fading environments separately and propose new PA policies that reduce the BER. We do a comparative study of the proposed PA policies with commonly used policies, e.g. water filling, (truncated) channel inversion and an optimal policy proposed by J.F. Hayes (1968) for an uncoded system. For all the cases studied, we show that the proposed policies substantially outperform commonly used policies. Among existing policies, only Hayes' gives performance improvement over constant PA. We show that interleaving with PA can improve significantly the performance of coded systems on block fading channels. We also make the important observation that the improvements in BER obtained with PA increase with SNR, which is in sharp contrast to the negligible gain in channel capacity obtained with PA (Goldsmith, A.J. and Varaiya, P., 1997). Since direct optimization for turbo codes is difficult, we use the policies derived for convolutional codes on the constituent convolutional codes of turbo codes and show that significant performance improvements can be obtained.