Anyone have any ideas for testing sensor less homing accuracy?
I was thinking mount a dial indicator on my MPCNC far out at 300, 300. Then home, and hit the same spot several times to see how close it is. I figure doing both axes at the same time will magnify any errors?
Should I just try one axis with the dial indicator, heck just draw a few large square like we do for dual endstop calibration?
Isn’t there a builtin test for that in Marlin? Something where it measures the steps before reaching the switch and backs off the same amount.
The bounce?
I found the info, the sensor less part is only accurate to between 1 and 4 full steps. We need more accuracy to be useful. I had hopes of setting up the new board using this for simplicity but, nope.
Doesn’t “sensorless homing” really just push against a (reliable, repeatable) hard stop until the machine knows it has missed the step? If the hard stops are tweakable, wouldn’t that allow for the needed adjustments to get squared up?
Edit to expand my thought - Let me know if I’m missing something in my thought process. I would think that as long as the zero point (or home offset) wasn’t set until after the steps were missed, the zero position ought to be reliable. No steps (or micro steps) should be missed after the homing sequence was completed. I don’t have hardware capable of lost step detection, so haven’t played with this myself so this is a thought experiment, based on watching some of the Prusa “lost steps detection” demo videos. They seem to say “I notice we lost a step, better re-home that axis so we’re working from a known good datum.” They don’t seem to care how many steps they miss, they re-home and assume they’re good from then on.
If I remember when I go back in I will check my history to find the article I read. There is the accuracy of the actual resistance drop and the speed it takes to get it that accurate. Turns out if all the machines are not identical (prusas) it seems tuning them to accurately see the bump is not easy either.
There’s probably a reason so many machines that really need them rely on good old mechanical switches. Sometimes it good to remember “newer” is not necessarily congruent with “better.”
I understand the goal is to achieve end stop functionality without the additional part count, which would be nice, but IMHO the switches aren’t that hard to add. My personal situation with the v3.00 cnc shield and switch wires running in a single shielded cable along side the motor wires is probably a good example of noisiest possible condition and I was able to work through it in a few days with the grbl web site noise abatement references.
I was under the impression they were reading the increased current as the stepper tries to push against the edge, which I would think doesn’t involve any actual missed steps. It could be the current doesn’t climb fast enough on a softer physical stop to respond in a quick enough manner. So you end up compressing something by a step or two before it decides it’s at the end.
You do need a hard edge to get it to trigger. I have noticed on mine that if it’s close, it can miss it and grind away sometimes. But if it’s flying across from the other end of the printer, it figures it out pretty quickly.
I think if sensorless homing were first, we would all be asking for the new and improved microswitches. But the real reason I have them is… I hate wiring.