Statistical behavior of crosstalk power sum with dominant components

The literature of digital transmission on wire-pair cables generally considers the probability distributions of both pair-to-pair crosstalk loss and its power sum to be normal on a dB scale. This paper presents extensive measured data of crosstalk among connectors and wire pairs on the backplane and associated stub cable of 466-type apparatus cases of the existing T1 system. The measured probability distribution of crosstalk power sum “bends” toward more severe crosstalk levels in the lower tail region, which is important for T1 system engineering. This bend is because of the effects of a few dominant components (i.e., within-slot or within-harness crosstalk) in the power sum. The simple normal model is too optimistic by 4 dB in estimating apparatus-case-crosstalk power sum at 0.1 percentile level This paper shows that both the Monte Carlo and the lower bound methods for power sum calculations predict this bend in close agreement with the measured data. Although apparatus-case-crosstalk power sum is worse than previously assumed, the performance of T1 system has been adequately protected by the extra margin in the previous engineering rules to cover unknowns.