| High-speed communication over metallic wire conductors encompasses a very large number of widely adopted systems, such as Digital Subscriber Lines (DSL), Ethernet over copper twisted pairs, and chip-to-chip serial links over cable or backplane traces. A prevailing characteristic of such systems is that multiple lines emanate from a central node and reach remote terminals. Also, crosstalk among the lines is typically the dominant factor constraining communication. This thesis examines how coordination at the central node among signals of different lines facilitates reliable, high-speed transmission by mitigating the effects of crosstalk.; It is shown that the matrix representations of the multi-line channel have distinct properties, with diagonal elements dominating the columns for upstream transmission and dominating the rows for downstream transmission. It is deduced that this structure results in multiuser capacity regions approximating rectangular parallelepipeds. Coordination strategies are proposed for crosstalk cancellation, combining the principle of multicarrier transmission with the structure of the Generalized Decision Feedback Equalizer (GDFE). The upstream scheme involves successive cancellation at the receivers, while the downstream scheme employs modulo precoding at the transmitters. It is found that crosstalk suppression with these techniques achieves performance that is very close to that of an equivalent crosstalk-free channel. Additionally, it is shown that the benefit of employing a Minimum-Mean-Square-Error (MMSE) criterion for crosstalk reduction instead of a Zero-Forcing (ZF) criterion is very minor.; Transmission optimization is examined with the objective of maximizing a weighted sum of the user data rates. When a coordinated system operates independently, the optimum energy allocation across the users and across frequency is shown to require independent water-fillings for each user. If neighboring non-coordinated systems must be protected from excessive crosstalk, it is shown that the energy allocation problem can be solved numerically by exploiting the convex form of the problem. In scenarios where energy allocation is combined with upstream/downstream frequency selection, a near-optimum solution is obtained by applying a numerical, low-complexity algorithm. Finally, a blind adaptive implementation of the upstream coordinated receivers is proposed, utilizing decorrelation conditions as the adaptation criteria. |
