My new pulse generator has more trigger jitter than I'd like, so I'm investigating the front end. There's a fast comparator that gets the external trigger, makes differential PECL out, a DPDT switch for edge inversion, and that goes into the diff clock input of an FPGA. The layout is mediocre: see the red traces.
LT Spice doesn't (yet?) include the Analog Devices comparators in its library, just the slow Linear parts, so I fudged the PECL driver from discretes.
I think that Spice is pretty good for analyzing PCBs as long as the traces aren't long enough to be very lossy. This is a sim of the original design
and one with some possible kluges
Both are probably realistic, or at least useful. Simulating a board with longer traces would probably require using the lossy transmission-line model and tweaking that to match a TDR measurement, which would be a chore.
The dilemma now is that I don't know if the Xilinx FPGA will do that
100 ohm differential termination. The documentation is not clear. Apparently what one does is ask the compiler to please do it and see if it's willing. We'll try that.