Bit and Power Allocation for Goodput Optimization in Coded Parallel Subchannels With ARQ

This paper addresses the problem of adaptively allocating the bits and the power among a set of parallel subchannels. A frame-oriented transmission with convolutional coding, hard Viterbi decoding, and automatic repeat request (ARQ) retransmission protocol is considered. The objective is to maximize the number of information bits delivered without error to the user by unit of time, or goodput. This criterion is a tradeoff between the bit rate and bit error probability criteria. First, a mathematical expression for the goodput of the system is presented. Then, using that expression, different bit and power allocation strategies are derived and compared. It turns out that having an equal bit error rate (BER) on each subchannel is near-optimal in terms of goodput. Finally, we prove that the waterfilling solution with adequate signal to noise ratio (SNR) gap value (or equivalently adequate target BER value) is near-optimal. We analytically show how this SNR gap value depends on the convolutional code used, the frame length, the type of ARQ protocol used, the available transmit power, and the subchannel gains.