Zero-forcing precoding using generalized inverses for G.fast DSL systems

Abstract

Previous work on precoding for G.fast systems focused on linear and non-linear precoding techniques to maximize the overall throughput. It has been shown that the existing algorithms almost achieve the sum capacity. However, the rate region does no longer feature a rectangular shape as for VDSL systems, that motivates the utilization of operating points suboptimal w.r.t. sum rate to meet the individual users demands. In these points, the gap between existing zero-forcing techniques and the corresponding weighted sum capacity is significantly larger. Therefore, we propose to improve the linear zero-forcing precoding by employing generalized inverses, for which we present methods that enable us to cope with the large system size of G.fast systems. Our simulations show that our proposed precoding scheme outperforms linear ZF solutions based on the Moore-Penrose pseudoinverse significantly in scenarios with greatly varying user demands.