Simulating Gaussian Crosstalk for DSL Testing

The presence of crosstalk is a major detriment to the performance of Digital Subscriber Line (DSL) transceivers. It is therefore important to test performance under a wide variety of crosstalk scenarios. This thesis presents the development of a crosstalk noise generation system as a tool for DSL performance testing. Where many commercially available solutions are limited to specific testing needs, the system presented here is universal and can simulate virtually any noise scenario conceivable for DSL deployment.
Several requirements for the accurate frequency domain modeling of DSL crosstalk are presented. The time domain properties of crosstalk were also studied with particular attention paid to the amplitude distribution. The importance of a Gaussian amplitude distribution and the limiting effect of a finite-length pseudo-random crosstalk sequence on the "tail" of the Gaussian distribution were investigated. A method of forcing finite-length crosstalk sequences to have a "near-Gaussian" amplitude distribution is proposed.