Spatially adaptive linear precoding for systems beyond G.Fast

The next generation of Digital Subscriber Line (DSL) technologies seek to exploit frequencies beyond 106 MHz and up to 500 MHz where Far-End Crosstalk (FEXT) becomes excessive; for these frequencies effective crosstalk cancellation, also known as vectoring, is essential. This paper proposes a novel spatially adaptive linear precoder and explores its performance, alongside classical Zero-Forcing Precoding (ZFP) and Tomlinson-Harashima Precoding (THP) methods, under two distinct scenarios. In this paper, the objective function is set to maximise the aggregate capacity of the copper MIMO channel. All precoding methods benefit from a coordinated bitloading to maximise the spectra within a defined power spectrum density (PSD) mask. Results suggest that THP's superiority is substantial in the fixed cable length scenario whilst the proposed Spatially Adaptive Generalised ZFP (SA-GZFP) outperforms the latter marginally in the dendritic one. The paper further suggests that a spatially adaptive nonlinear precoder may have the potential of combining the gains of THP and Maximal Ratio Transmission (MRT) for greater system performance.