Auto calibrating acceleration

So I was reading this topic [forums.reprap.org] and automating calibration seemed to be a theme there, and I thought, hey we can do that for some stuff right now. Using the end stops it seems like you could write a routine that would quickly find the limits of acceleration on your machine. Just home, then accelerate out softly and back in at some higher acceleration then back to home. Check the step count out and back in. The difference should be skipped steps (lash likely cancels out) Do it again at a higher acceleration still. Keep going till you get a good idea of your max acceleration, repeat for the other axes then set acceleration to 80% of the lowest value or something.

Seems like you could then use the same basic algorithm but use a safe acceleration and vary the top speed and get your max speed the same way.

Quite possible. With acceleration, I'd be a bit conservative, as this also influences wear and print quality. Higher acceleration means higher forces, so more machine bending.

There's already M117 for checking against step losses. Not sure how many firmwares implement it.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

From my experiences developing the Fablicator, the pure acceleration limit is not what causes most skipped steps.

It's usually somewhat below that speed, and because of system resonances.

To do it properly, I think you would home an axis, center the head, and then run through a variation of amplitudes for a few seconds apiece. Ie start back and forth motion at 60mm/s with acceleration between 1 inch, then .8, .6, .5, .4... all the way to 1/2 a filament diameter. Then see if the head position is where you thought it should be. That way if there is a resonance that causes problems it will be discovered.

Just a thought.

---

www.Fablicator.com

Seems like there should be a way to calibrate lash that way too but I can't figure out how to do it with just one sensor per axis. The forward lash and the backward lash always cancel out in my head.

There should be no need to calibrate backlash. If there is backlash in the machine, it really needs to be solved mechanically.

There is just no way to properly compensate for backlash in a dynamic time dependent system.

---

www.Fablicator.com