Reduced dimensional power optimization using class AB and G line drivers in DSL

It has been shown recently that power-efficient digital subscriber line (DSL) access networks can be designed via an optimization problem with an objective that assigns a value to user's data rate while assessing a penalty proportional to the line driver power consumption. An optimization algorithm was proposed that combines elements of linear programming and trust region optimization, having a complexity that scales linearly with the number of users and the number of used tones. Motivated by the expansion of DSL technology from hundreds to thousands of tones, and by operator's preference to implement power spectral profiles that can be specified with a small number of parameters, we propose in this paper a reduced dimensional formulation of the optimization problem. The new formulation reduces complexity with little loss in performance, by exploiting the smoothness of the problem data as a function of frequency. Using the new formulation, we compare the efficiency achievable using two classes of line drivers: class AB and class G.