Methods of summing crosstalk from mixed sources .II. Performance results

For pt.I see ibid., vol.50, p.453-61, (Mar. 2002). In a companion paper, the problem of summing crosstalk from mixed sources has been deeply analyzed. The problem of determining a good model for the summation of mixed crosstalk is becoming more important now that many digital subscriber line (DSL) services are being deployed in the field. The straightforward application of the usual 1% worst-case model for only one kind of disturber to all the individual worst-case contributions is overly pessimistic. The most well-known methods of summing mixed crosstalk, the FSAN and the Mean PSD methods, are less pessimistic. Also, the two new summation methods presented previously, the Minkowski-bound and the Holder-bound methods, are less pessimistic and offer the appealing feature of choosing the level of pessimism or optimism through a parameter. However, the accuracy of these methods has never been quantified in the past. As more and more types of DSLs are deployed by multiple service providers, it is imperative to gain confidence in the methods of modeling worst-case mixed crosstalk and to understand their limitations. In this paper, the problem of defining an accurate statistical modeling of the pair-to-pair crosstalk power coupling losses is addressed first. This has been accomplished by exploiting a large measurement campaign performed on real cables. The results of a large simulation campaign that compares the accuracy of known and newly proposed methods of summing mixed crosstalk are then presented. In addition, it is proposed here to use the signal-to-noise ratio (SNR) as the metric for determining the actual system performance in the presence of mixed crosstalk. A Monte Carlo simulation is run that assigns randomly each disturber to a pair-to-pair NEXT coupling which matches the statistics of measured cables. The Monte Carlo method accurately determines the I % worst-case impact on the SNR of DSL systems affected by mixed crosstalk. ISDN, HDSL, HDSL2, T1 and ADSL are each simulated with many different numbers and types of mixed crosstalk.