@mykepredko. Thx for the input!
At a guess, your scope is running at 400kSamples/second (1/(50us/s) (for the timebase) x 10 (as a guess) x 2 (for the Nyquist frequency)) which just won’t be fast enough for the 500 kbps signals you’re trying to look at.
So I used two probes. One on Can H, one on Can L with common ground.
Cyan - CAN H
Yellow - CAN L
Magenta is A-B (differential)
DS212 scope
My scope is not made for this
I guess I wasn’t trying to decode the CAN data, was just really trying to see if there were obvious anomalies. Don’t know many people with logic analyzers/high-end scopes. Long story short, it would make sense to even buy a entry-level $500 scope to diagnose this.
If I can throw in another variable - how do you know your 12V/24V power to the EBB is good and continuous? I’m wondering if you have a break in a power line that is failing very intermittently when the wires are being flexed.
Definitely not a PSU issue. I eliminated bad wiring early on in my tests. Replaced and re-crimped power and can wires. Since I can make klipper (notice I say klipper, not can interface) fall over using cangen. This is without running a print and the tool head is stationary. My deduction is that the physical layer is as good as it gets…
I have also swapped out the EBB42 with another one, to eliminate a bad board.
I’m just concerned that using cangen as a method to test is not valid. IT could simply be overwhelming the available bandwidth… There’s no good way to really measure this.
In trying to figure this out… the “B” in the log is a Bitrate switch flag.

I’m not sure why this is even there, but it seems, scratching in the firmware code, that the stm32G0B1 is set up in FD (Flexible Datarate). Not sure if that is relevant, but the packets I generate with cangen don’t have that at all. Seems like it’s only the message generated by the klipper, and only the one?
I think I’m downgrading back to old-school tomorrow. Can’t afford more downtime on my printer. So not more debugging from my side… I’m hoping someone else has a similar experience and might find the underlying issue.