Today's digital subscriber line (DSL) deployments are often conservatively engineered to function in a statistically worst-case environment. Crosstalk is treated as unknown and uncontrollable random noise, even though it is man-made. Other impairments are often treated by simply adding margin to crosstalk. While this simplistic practice currently suffices, it often provisions unnecessarily low bit rates. However, the loop plant can be indeed optimized to provide significantly higher bit rates than currently provided, offering higher speeds without installing new remote terminals or repeaters. Recent developments in Dynamic Spectrum Management (DSM) may allow a two to three fold improvement in DSL bit rates, improved reliability, and increased range for DSL services. This process involves several engineering tasks, the two most important of which are:


In the present talk, we will describe sophisticated techniques that allow solving the problems of loop make-up and crosstalk identification, a problem seldom addressed in the literature.

In our approach, identification of a loop make-up is achieved via single-ended measurements based on the use of an enhanced Time Domain Reflectometer (TDR). As it will be shown, the TDR trace obtained by probing a loop with a pulse consists of an unknown number of closely spaced echoes, some overlapping some not, some spurious some not, that exhibit unknown amplitudes, unknown times of arrival and unknown shapes. The problem of resolving such echoes is a problem that arises in many applications such as radar and sonar processing, geological sounding, etc., and the conventional approach is to assume the availability of an array of sensors located in the far field of the sources. However, in the case of loop make-up identification, the resolution of such echoes must be carried out via a single sensor (and not an array). This is a non-trivial problem, seldom addressed in the scientific literature.

The problem of identifying the crosstalk present on a given pair is an important case of multiuser channel estimation. We will describe a method to identify crosstalk sources by finding the maximum correlation with a “basis set” (dictionary) of representative measured coupling functions. It will also be shown that this can be considered equivalent to finding an optimal sparse representation of a vector from an over-complete set of vectors.

The algorithms described here are required to exploit fully the power of emerging DSL technologies, and to deal effectively with the increasingly complex and dynamic unregulated local loop environment. In fact, the knowledge of both the loop make-up and the noise/crosstalk environment allows us to calculate precisely the performance of any DSL type. Moreover, the techniques described here will allow operators to optimize jointly DSL transmit spectra and signals to minimize crosstalk and maximize received signals, allowing substantially higher DSL speeds than in current practice. Simulation results confirm that increases of several hundred percent in average DSL bit rates can be indeed achieved. This opens the door for new services, including symmetric enterprise services and full video service requiring minimal physical plant upgrade.