Low-Complexity Nonlinear Zero-Forcing Precoding Under Per-Line Power Constraints for Improved Downstream G.fast Active-User Peak-Rates

We consider nonlinear zero-forcing (ZF) precoding design to improve the downstream G.fast peak-rates when only a few users in the cable binder are active. In order to compute the optimal nonlinear ZF precoder under per-line power constraints (PLPCs), we present a novel low-complexity dual decomposit...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on communications 2018-06, Vol.66 (6), p.2696-2707
Main Authors: Lanneer, Wouter, Tsiaflakis, Paschalis, Maes, Jochen, Moonen, Marc
Format: Article
Language:English
Subjects:
Algorithm design and analysis
Algorithms
Complexity
Computer simulation
Crosstalk
Decomposition
DSL
dynamic spectrum management
Frequency ranges
G.fast
Lagrange multiplier
nonlinear precoding
Optimization
per-line power constraints
Precoding
Signal to noise ratio
State of the art
zero-forcing (ZF)
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We consider nonlinear zero-forcing (ZF) precoding design to improve the downstream G.fast peak-rates when only a few users in the cable binder are active. In order to compute the optimal nonlinear ZF precoder under per-line power constraints (PLPCs), we present a novel low-complexity dual decomposition algorithm, in which the key is the use of Lagrange multiplier based virtual precoders to transform the PLPCs into an easier virtual sum-power constraint (SPC), such that the SPC-optimality of the QR decomposition-based precoder may be exploited. We show a reduced computational complexity of this algorithm over the state-of-the-art SVD-block-diagonalization-based dual decomposition algorithm. We present simulations of a 10-line cable binder that demonstrate substantial peak-rate gains over standard QR decomposition-based ZF precoding in DSL, due to the increasingly stronger crosstalk channels in the G.fast frequency range (up to 212 MHz). Furthermore, we show that the proposed algorithm naturally extends to the scenario with multiple lines terminating at the customer premise equipments.
ISSN:0090-6778
1558-0857
DOI:10.1109/TCOMM.2018.2799613