Statistical Behavior of Multipair Crosstalk

Abstract

The literature of twisted-pair communication cables generally considers the probability distribution of both pair-to-pair crosstalk losses and their power sums to be normal on a decibel scale. A deficiency of this model is that the assumed normal distribution for pair-to-pair crosstalk loss must be truncated to correctly predict the power sum distribution. The appropriate truncation limit is selected to fit available cable data and varies with the data selected and the binder unit size. In this paper, we present experimental data from more than 600 cables, comprising 91,875 measurements, to show that the gamma distribution (with log variate) is a more satisfactory approximation to modeling the multipair crosstalk behavior. The predicted power sum distributions agree well with the data — without imposition of truncation. The gamma model leads to a more optimistic estimate of the number of digital systems which can be deployed in a cable. The result is, therefore, a lower cost per circuit, wider application of T-carrier system, or more efficient cable utilization than the normal model would permit.