Power- and bit-loading algorithms for multitone systems

Optimal power- and bit-loading in a multitone system requires prohibitive nonlinear programming. A useful near-optimal solution has previously been proposed which imposes a uniform error probability distribution across the tones. In this paper we examine two alternative near-optimal solutions and offer a comparison with the existing solutions. The first imposes uniform power across the used tones and optimizes the bitloading and error probability distribution. The second is a default scheme in which both uniform power and error probability are enforced on all used tones, and only the bitloading is variable. These four algorithms are compared in terms of estimated throughput and complexity, using the eight standard CSA loops as testbeds. It is found that the new near-optimal algorithm achieves similar rates to the existing near-optimal scheme. More interestingly, it is found that the default scheme also provides rates which typically exceed 99% of the theoretical maximum rate at vastly reduced complexity.