Clustering Strategies for Binder Identification Using Phantom Measurements

Abstract

This paper proposes an automatic method for the identification of twisted pairs sharing the same binder, based on the analysis of phantom circuit measurements. This type of circuit is often used for improving data transmission rates in communication systems, but in this work, phantoming is used to reveal if a 4-wire loop composed by two twisted pairs are close enough and well balanced in order to be considered in the same binder. The identification is done via application of two pattern recognition techniques, K-means and Gaussian Mixture Model, on S 11 parameter obtained from the phantom-mode measurement of two twisted pairs. This paper also presents an automatic method to labeling the clusters and a method to estimate the length of the two twisted pairs that share the same binder, using Time-Domain Reflectometry analysis. Laboratory results confirm the accuracy of the proposed methods.