If your high speed differential pairs don't work, maybe the solution is in software

[@bertsimonovich7297]

Another great video Robert and Don! The equalization explanations on single bit response was particularly helpful.

[@jimjjewett]

What is going on around the 40:00 minute mark? ( kzbin.info/www/bejne/iX6aZ3lraJypoMUsi=XtDe7qTBd9qpJ2TA&t=2400 ) As best I can tell, there is a differential pair that is jumping over a bunch of gaps, and a parallel signal from the other side of the ground plane is causing crosstalk. But why would you put those gaps in the plane in the first place? It isn't for a via, or neither track would be there. If you're pulling the plane away/using a hatched plane for greater impedance, then why not do that on the rest of the track, too? (e.g. why first go South and circle around on solid areas? And why does the other signal layer run over those cutouts, but over (under?) solid ground to the left?) And would coupling be a problem even with the ground plane, since the plane is pretty thin, and both layers would be using it for return current at the same time?

[@robertpeters9438]

So, will the receivers know when a reflections are coming by time and dynamics change for the first say 10 reflections over a threshold from training?

[@jimjjewett]

Ish. You can calculate it, but the exact amount will depend on things like whether the signal was at the same level on the previous bit, and whether random crosstalk happened to be towards the same level or the opposite level or out of phase ... and if you spend much time calculating it, you have just added a bunch of latency, and hey, if you can do lots of calculations quickly, maybe you could also sample more often and increase the bitrate instead...

[@MikhailGoncharov-tl4cr]

it's unexpectedly interesting at all

[@peiman1988]

An interesting SI conversation with Matthew McConaughey 😂