Capacity of a dispersive channel and algorithms of power and bit loading

Some algorithm, such as HH algorithm, are very slow for applications like ADSL, where a large number of bits will be contained in each symbol and transmitted over a large number of subchannels.

While this algorithm may be slightly suboptimal relative to the Hughes-Hartogs algorithm, it will typically converge much faster than Hughes-Hartogs with a running proportional to O(N). Chow et al. proposed an algorithm that is not guaranteed to find the optimal solution, but has the advantage of being very fast and simple.

Their results show that the performance loss due to suboptimality is usually negligible. The Chow-Cio Binghams algorithm is suboptimal and approximated, it use the hypotheses of convexity, constant granularity and constant gap approximation.

Would be suboptimal even if the constant gap hypothesis were satis ed exactly. How it works? Estimates which subchannels to turn off, and distributes power almost equally among remaining subchannels. The key assumption is that performance is mostly determined by the choice of subchannels to be turned off.