Efficient water-filling algorithms for a Gaussian multiaccess channel with ISI

Although the optimum transmit power spectral density (PSD) of each user in a Gaussian multiaccess channel is theoretically available, efficient algorithms are needed to obtain the optimal PSD numerically. This paper presents two water-filling algorithms for the scenario where all users have the same priorities. First we propose for a two-user channel an exact algorithm with a computational complexity of O(Nlog₂N), where N is the number of subchannels. Then we develop a heuristic algorithm for K>2 users that has O(K²Nlog₂N) complexity and achieves more than 98% of the optimum data rate in all of the channels we simulated