Remora for RP2040

Been trying to find the cause of the following errors for a couple of days now and still no luck.  Have reduced the velocity, accels and scale dramatically but still the following errors persist.

Commanded position and feedback track fairly closely, but following error is accumulating.

 Here is a better capture of what we're seeing. 

Increasing the pgain for each axis to 100 definitely helps.

However, something strange with "random" steps in the position command. Are you seeing this as well?
 

Yes. Last night before calling it quits for the day, I got the debug output going and I can see that the following errors are correlated with RX failures (RX reset, so no data received).

It really looks to me like somewhere in the pipeline there is some packet loss.

Maybe not, as a RX reset is normal behaviour when e-stoped. For the controller board to disconnect by action of the communications monitoring module, multiple successive packets would need to be lost.

github.com/scottalford75/Remora-RP2040-W.../RemoraComms.cpp#L23

I found that pgain of 500 and ff1gain of 1 made a big difference.  I also found some things in the firmware that seemed to help and I have posted a PR with my changes.

Thanks for the work on this, greatly appreciated! I've merged your changes and will do some tests as well. How's the PicoBob hardware working out?

The changes have made a huge difference! pgain of 100 seems to be adequate, so I'll update the config to include that as default. ff1gain is set to 1 by default. Higher pgains actually created following errors at high speed by overshooting.

That sounds good. The PicoBOB DLX hardware is ready to go IMO. Everything works and seems to be in the correct place I will be posting the design very shortly to Github. If you want to reach out to me on Discord or something I am sure I can get you one sooner.

That said. Pausing the threads for the data copying seems to make the step timing jitter pretty bad. I'm going to try implementing a ping-pong buffer and use pointers in the threads so that they are only interrupted for as long as it takes to do an assignment and see if that helps.