Request: send out a weekly update email when calibration is actually working

I’ve spent a few days trying various things, and can’t get it to calibrate. It took at least a day of fiddling just to get retract all and extend all to work reliably. The extra slack when calibrating means I have to baby sit it constantly because it seems like it wants to get the belts tangled any time I’m not on top of it.

And then I’ll get fitness that ranges from .10 to 1.10 without any idea why one run worked well and another didn’t. But regardless of how good the fitness is, the behavior when trying to move it around is the same, it spits out a ton of errors, and there’s often at least one belt with slack.

I’m obviously doing something wrong, but I have no idea what.

Instead of everyone trying to get their individual machine running with the software in the current state, we should be focusing on figuring out the critical problems at each stage of the calibration process. We should come up with a clear testing plan for each step, figure out what needs to change, make the fixes, etc.

And then once calibration actually works reliably, it would be great to get an email saying “try and run calibration now, it should finally work.”

1 Like

I agree with this 100% and getting feedback on what isn’t working is the key there. Mine is largely working great at this point so we’re at the phase where getting feedback on what isn’t working is the key to continuing to improve it. There’s a big step from “Works on my desk” to “works everywhere in the world”.

I think that every week it will get a little better, but I don’t know that there is going to be a single moment when it goes from not working to working.

The issues that are, or were, stopping me from being able to calibrate are:

  • Retract all/extend all not working. I think this is a physical issue with the gears being very tight. After running them in and out a number of times they’ll retract all at 1800. A couple times I got warnings that it had exceeded 4000, so there might be room to run it even higher than 1800 initially, at least until the gears have broken in.
  • Figuring out what size calibration area could potentially work. The new Google sheet was helpful for this.
  • Too much slack when calibrating. Can we just cut this setting in half? I’m happy to do testing to tune this part of the calibration process, but I don’t think we should have people be calibrating without explicit instructions to watch every move carefully, and give guidance on exactly how and when to take up slack. I’ve had at least three belts get tangled when I wasn’t paying attention to every single move during calibration.
  • Fitness is all over the place. I’d do two calibration runs back to back with the same setup and get very different fitnesses. I assume this is something I’m doing wrong (could it have to do with taking to slack manually?) but I have no idea what I should do something different.
  • It’s not obvious that calibration has actually worked. It seems like even with a good calibration it’ll throw lots of errors? It would be great if there was a first test we should do to see if it was moving correctly before we did the first cut.

I think getting the belts to retract reliably is a critical issue because nothing can happen without that. And getting rid of the excessive slack during calibration is the other one that will likely trip up a lot of people.