I did two simulations on the same channel with different bit patterns. The channel is very short, about 2.4 inches, connecting Xilinx GTX Serdes to O/E transceivers.
The resulted Eye diagrams are shown below. I do not quite understand the differences between the two:
1) The eye height difference: +/-195 mV in 8B/10B vs +/-320mV in PRBS15.
The eye height with PRBS15 bit pattern is ~50% more than that of 8B10B bit pattern. However, look closer at the the Eye diagram, the top/bottom of the Eye is flat for more than 0.3UI, which means each edge of each bit has enough time to reach its steady level, so this much reduced eye height with 8B/10B bit pattern is unlikely to be caused by ISI. Anybody understand the cause?
2) Why is the stripe texture at top/bottom in Eye diagram with 8B/10B bit pattern? What is the cause? It is hard to beleive that you would measure such a Eye diagram on a real scope.
My naive thinking is that people use 8B/10B encoding in order to improve Eye diagram. Whereas, following two simulations seems to tell me the opposite. Something must be wrong.
Eye diagram 8B10B bit pattern (10 million bits with worse-case patterns)
Eye diagram with PRBS15 bit pattern (10 million bits with worst-case pattern)