Signal processing for ultra fast next generation metallic access network

Abstract

With the escalating demand for high speed, high reliability, low latency, low cost and ubiquitous connectivity, the telecommunications industry is entering a new era where the ultimate optimality of the current wireline-wireless access network has to be achieved.
Regarding the current wireline network paradigm, dominated by the copper-based digital subscriber lines (DSL) technology, multi-Gigabit data rate is the ambitious design objective at the customer end for the forthcoming ITU-T G.mgfast standard. In order to prepare for the new challenges in the era of total network convergence, both the wireline and the wireless community must be able to think beyond their respective conventions and learn from each other if necessary. Overall, the current DSL-based wireline network architecture is prone to the mutual interference resulting in far-end crosstalk (FEXT).

The recently expanded 424/848 MHz spectrum of the ambitious G.mgfast project introduces far higher FEXT than that over the current 212/30 MHz G.fast/VDSL2 band.

Additionally, the coexistence of multiple standards will also cause 'alien' FEXT. In the face of these new challenges, most state-of-the-art techniques, such as linear transmit precoding or Reed-Solomon coding, fail to satisfy the escalating performance requirements. In this thesis, we conceive a general architecture for next-generation DSL networks based on recent technological advances in key areas of access networks. More specifically, we conceive a dynamic spectrum management (DSM) aided signal processing framework relying on vectored transmission, spectrum balancing and error control.

Specifically, the excessive FEXT cannot be combated by low-complexity linear transmit precoding for downstream vectoring. However, as shown in the field of wireless communications, using lattice reduction as a signal space remapping technique significantly improves the performance of traditional multi-user detectors (MUD) and of the respective multi-user precoders (MUP). These promising techniques have largely remained unexploited in commercial wireless communications, due to their complexity in the face of the rapidly fluctuating wireless channels. However, they are eminently suitable for quasi-static copper channels. Hence we design powerful lattice-reduction-aided MUP (LRMUP) as well as an optimal sphere encoder and its low-complexity variants. Furthermore, we propose the concept of vectoring mapping regions (VMR) as the transmitterside dual counterpart of the classic decision regions conceived for receiver-side signal detection. Based on the VMR, we propose a near-optimal multi-level DSM algorithm relying on the family of LRMUPs, which strikes a more favourable sum rate vs. fairness trade-off than the state-of-the-art non-linear Tomlinson-Harashima precoder (THP).

On the other hand, impulsive noise (IN) contamination constitutes another challenge in next-generation DSL networks. In particular, the copper loops used for transmitting DSL data may become contaminated by other high-power undesired wireless signals such as IN and radio-frequency interference (RFI). As a counter-measure, we amalgamate forward error correction (FEC) with sophisticated automatic repeat request (ARQ) techniques for conceiving hybrid-ARQ (HARQ) schemes. Given that low density parity check (LDPC) codes are likely to be used in G.mgfast, we propose a novel low-complexity HARQ protocol-based on the class of New Radio (NR) LDPC codes conceived for 5G cellular wireless systems, while exploiting the available a posteriori knowledge concerning the IN events at the receiver, in order to minimize the SNR degradation due to IN. Specifically, based on our proposed Tanner graph characterization as well as on the corresponding extrinsic information transfer (EXIT) chart, we propose an EXIT-chart-aided deferred iteration (DI) decoder switching technique, as well as a pair of MI-thresholding-aided curtailed iteration (CI) early stopping strategies, in order to minimize the potentially futile or redundant decoding operations.

More information

Identifiers

ORCID for Lajos Hanzo: orcid.org/0000-0002-2636-5214

Catalogue record

Export record