Optimum power allocation and bit loading for systems with suboptimum receivers

Waterfilling and mercury/waterfilling power allocation policies maximize the mutual information of independent parallel Gaussian-noise channels under an average power constraint for Gaussian signals and for practical discrete constellations respectively. In this paper, the same criterion is considered for : joint power allocation and bit loading in more general scenarios where suboptimum receivers introduce a mutual information loss. The proposed approach provides the optimum solution without resorting to greedy algorithms and has nearly the same complexity as the original mercury/waterfilling.