[10GBASE-CX4] Test Patterns
Reviewing Annex 48A, I realized that all the current test patterns defined there send almost identical data on each XAUI/CX4 lane (except for the ||Tn||, ||S||, preamble, & CRC sets, and one byte in CJPAT. This does not stress crosstalk very much (only random disparity effects cause the lanes to differ). For the 10G serial formts, this is irrelevant, and for XAUI crosstalk is generally pretty negligible, but for CX4 it could be a significant effect. In particular, the RX3 pins are adjacent to the TX0 pins, and a crosstalk of 5% (-26dB) is only a little above the possible signal level (-16.5 db, was -19.6 dB). The margin could be further reduced by the ~3 dB difference between a 1100 1100 pattern being sent (with full pre-emphasis on one bit) against a HF pattern being received on the adjacent channel.
I would like to suggest that we raise this as a topic, maybe at the upcoming Interim meeting.
Peter Bradshaw
P.S. our mail system is having problems; if you do not get an expected reply, try sending e-mail to
pbradshaw@bitblitzcom.com