Estimation and detection of transmission line characteristics in the copper access network

Sammanfattning: The copper access-network operators face the challenge of developing and maintaining cost-effective digital subscriber line (DSL) services that are competitive to other broadband access technologies. The way forward is dictated by the demand of ever increasing data rates on the twisted-pair copper lines. To meet this demand, a relocation of the DSL transceivers in cabinets closer to the customers are often necessary combined with a joint expansion of the accompanying optical-fiber backhaul network. The equipment of the next generation copper network are therefore becoming more scattered and geographically distributed, which increases the requirements of automated line qualification with fault detection and localization. This scenario is addressed in the first five papers of this dissertation where the focus is on estimation and detection of transmission line characteristics in the copper access network. The developed methods apply model-based optimization with an emphasis on using low-order modeling and a priori information of the given problem. More specifically, in Paper I a low-order and causal cable model is derived based on the Hilbert transform. This model is successfully applied in three contributions of this dissertation. In Paper II, a class of low-complexity unbiased estimators for the frequency-dependent characteristic impedance is presented that uses one-port measurements only. The so obtained characteristic impedance paves the way for enhanced time domain reflectometry (a.k.a. TDR) on twisted-pair lines. In Paper III, the problem of estimating a nonhomogeneous and dispersive transmission line is investigated and a space-frequency optimization approach is developed for the DSL application. The accompanying analysis shows which parameters are of interest to estimate and further suggests the introduction of the concept capacitive length that overcomes the necessity of a priori knowledge of the physical line length. In Paper IV, two methods are developed for detection and localization of load coils present in so-called loaded lines. In Paper V, line topology identification is addressed with varying degree of a priori information. In doing so, a model-based optimization approach is employed that utilizes multi-objective evolutionary computation based on one/two-port measurements. A complement to transceiver relocation that potentially enhances the total data throughput in the copper access network is dynamic spectrum management (DSM). This promising multi-user transmission technique aims at maximizing the transmission rates, and/or minimizing the power consumption, by mitigating or cancelling the dominating crosstalk interference between twisted-pair lines in the same cable binder. Hence the spectral utilization is improved by optimizing the transmit signals in order to minimize the crosstalk interference. However, such techniques rely on accurate information of the (usually) unknown crosstalk channels. This issue is the main focus of Paper VI and VII of this dissertation in which Paper VI deals with estimation of the crosstalk channels between twisted-pair lines. More specifically, an unbiased estimator for the square-magnitude of the crosstalk channels is derived from which a practical procedure is developed that can be implemented with standardized DSL modems already installed in the copper access network. In Paper VII the impact such a non-ideal estimator has on the performance of DSM is analyzed and simulated. Finally, in Paper VIII a novel echo cancellation algorithm for DMT-based DSL modems is presented.