Calibration process ends with
[MSG:INFO: Center point deviation: TL: 0.161 TR: 0.092 BL: -2.687 BR: -93.730]
[MSG:ERR: Unable to move safely, stopping calibration]
And after more tests
[MSG:INFO: Center point deviation: TL: 0.008 TR: 0.675 BL: -18.410 BR: -89.216]
BR has always around 100 less offset
[MSG:INFO: Bottom Right pulled tight with offset -1970.544]
[MSG:INFO: Top Left pulled tight with offset -2074.512]
[MSG:INFO: Top Right pulled tight with offset -2083.455]
[MSG:INFO: Bottom Left pulled tight with offset -2039.083]
After multiple extend/retract I got:
[MSG:INFO: Top Left pulled tight with offset 0.333]
[MSG:INFO: Top Right pulled tight with offset 0.075]
[MSG:INFO: Bottom Right pulled tight with offset 0.086]
[MSG:INFO: Bottom Left pulled tight with offset -0.011]
Then Calibration stops at:
[MSG:INFO: Center point deviation: TL: 0.102 TR: 0.151 BL: -4.364 BR: -92.335]
I entered approximative widht and height and it is calibrating now… and stopped in BR telling:
WARNING FITNESS TOO LOW. DO NOT USE THESE CALIBRATION VALUES! Maslow-serial(3).log (15.8 KB)
this new size is different from the first:
I use it in my basement: stable 17°C and 80% humidity. Can the reinforced toothed belts suffer tension failures in this location? The frame is an OSB panel fixed to a concrete floor.
That makes sense, rerunning calibration will build on past information so it learned from its failure the last time through.
What the machine does is to adjust the two upper belts to the length which put the sled in the center of the sheet and then pull the two lower belts tight to check that they are the expected length.
It’s basically a sanity check to make sure that the frame is roughly the same size as the math thinks it is before starting calibration
What the machine does is to adjust the two upper belts to the length which put the sled in the center of the sheet and then pull the two lower belts tight to check that they are the expected length.
It¢s basically a sanity check to make sure that the frame is roughly the same size as the math thinks it is before starting calibration
so this is expecting that all four belts would end up being the same length
(within some margin of error) when at the center.
But it sounds as if this is firing at other times (when you aren’t at the
center, like in this case where it fired after many calibration points)
is this testing that every time it pulls the belts tight, the result is
plausible (within a margin of error) for the current frame size? Is this
designed to catch when the belt isn’t straight (caught on something)?
I’m wondering if this is worth it or causes more grief than it solves. It seems
like it requires a ‘good enough’ estimate of the frame size when we may be
better off making this a warning instead of an error and not worry if the
current frame size estimate is good or not.
I think it’s pretty important, without that we were getting a lot of broken belts and calibration gone wrong in weird ways. It also only happens at the first step in the calibration process.
The margin of error (or the threshold for the test passing) is that the belts are within 115mm of what they expect to be so it’s pretty forgiving and just catches cases where something is really wrong.
The one that happens during calibration is this one. This one is triggered when the math is telling the machine that to move to the left the left belts need to get longer (which is obviously not true)
Something that I would love to work on is to make is to make is to that we can attempt to automatically recover from both of these issues instead of just failing and giving up.
We should be able to recompute a new guess for the frame size and give the user an option to try again with that size rather than making them turn the machine off and back on and enter a new set of measurements.
I think it’s pretty important, without that we were getting a lot of broken belts and calibration gone wrong in weird ways. It also only happens at the first step in the calibration process.
The margin of error (or the threshold for the test passing) is that the belts are within 115mm of what they expect to be so it’s pretty forgiving and just catches cases where something is really wrong.
The one that happens during calibration is this one. This one is triggered when the math is telling the machine that to move to the left the left belts need to get longer (which is obviously not true)
Something that I would love to work on is to make is to make is to that we can attempt to automatically recover from both of these issues instead of just failing and giving up.
We should be able to recompute a new guess for the frame size and give the user an option to try again with that size rather than making them turn the machine off and back on and enter a new set of measurements.
I would say that we should allow the user to move the sled to center it.
(ideally this could be done automatically if the frame is truely rectangular,
but if the frame is a parallelagram of other shape, it may require manual
intervention to get into a good position)
115mm seems like a lot at first blush, but that’s only 4.5 inches, over a large
frame (especially with anchors to walls/floor) that’s not that much error, so at
the very least there should be a way to say “yes, this is really the center,
accept it”
It’s not so much that we’re trying to land in the center as it is that the actual frame dimensions match with the input numbers.
For example if we were to tell the machine that the frame is 1000x1000mm but in reality it is 500x500mm the machine needs to do a sanity check to make sure that things are reasonably close to accurate before attempting to run calibration.
Right now what happens is that the machine sees that something is wrong and gives up, but what we could have happen is for the machine to see that the frame seems to be smaller than the entered dimensions and we could offer to shrink the frame size and try again.
It’s not so much that we’re trying to land in the center as it is that the actual frame dimensions match with the input numbers.
For example if we were to tell the machine that the frame is 1000x1000mm but in reality it is 500x500mm the machine needs to do a sanity check to make sure that things are reasonably close to accurate before attempting to run calibration.
Right now what happens is that the machine sees that something is wrong and gives up, but what we could have happen is for the machine to see that the frame seems to be smaller than the entered dimensions and we could offer to shrink the frame size and try again.
what I’m saying is that if one anchor is out of position enough, the same alert
will trigger, and the fix isn’t going to be to shrink the frame size, it’s going
to be to accept the error and continue.
move one of your anchors about 6 inches and you will be in a situation where you
aren’t going to be able to give it frame dimensions that will satisfy it.
in terms of the 1000x1000 vs 500x500, other than the initial positioning, and
possibly a sanity check of the grid size, does anything in the calibration
actually care about what size it thinks the frame is? or is it all a case of
working from the belt lengths?
I think that a better thing to do would be that if one of the lower belts
becomes significantly shorter than the upper belts while tightening, the upper
belts should retract to roughly balance out. that will get you close to the
center with reasonable belt lengths. (and keep in mind that with a distorted
frame, tightening the bottom right belt may make the top right belt loose, so
after you think you are done, you should pull on all belts to be sure)
