Just an idea sparked in my head today, that I also assume @bar and @anna have tested. Create a setup where anchor points are same height as arms.

To check math on assumptions, after calibration is complete, couldn’t you disconnect all the belts on the Maslow but one and retract the machine towards the connected belt all the way to to the one anchor that is attached and verify the math to the one anchor point? Then the variance could help you extrapolate better math on the assumptions?

Let me know if that doesn’t make sense. (Makes sense to me)Then you could test all 4 corners this way.

I don’t fathom this being the calibration process, but if the Maslow thinks the anchor point is at a different length away, atleast you can know by how much and perhaps divide out how many teeth and get a better measurement spacing between teeth?

Just an idea sparked in my head today, that I also assume @bar and @anna have tested. Create a setup where anchor points are same height as arms.

To check math on assumptions, after calibration is complete, couldn’t you
disconnect all the belts on the Maslow but one and retract the machine towards
the connected belt all the way to to the one anchor that is attached and
verify the math to the one anchor point? Then the variance could help you
extrapolate better math on the assumptions?

when fully retracted, the anchors are over the sled, so would not reach the
anchors.

but I don’t understand what you think you would be learning if you could do
this. Are you questioning the distance between the center of the anchor hole and
the center of the bit? that’s pretty easy to validate by just measuring the
plastic parts (as you will be plastic-to-plastic at that point with not belts
deployed at all)

Good point about not reaching the anchor. But this could be fixed with some sort of 100mm spacer or something.

I was just trying to think how you could test what the Maslow thinks the distance it is away from a certain anchor, to what it actually is in reality. And if there is a discrepancy, perhaps the assumptions (2mm tooth distance) could be more fine tuned.

Could be a dumb idea… But I said it out loud in case it sparks a good idea.

Good point about not reaching the anchor. But this could be fixed with some sort of 100mm spacer or something.

I was just trying to think how you could test what the Maslow thinks the distance it is away from a certain anchor, to what it actually is in reality. And if there is a discrepancy, perhaps the assumptions (2mm tooth distance) could be more fine tuned.

the problem is ‘how do you measure the distance accurately enough’. If the belt
is stretching, that could give you bad numbers.

but as long as you are measuring and moving based on teh belt, I don’t think it
matters for getting the machine moving. the parts may not be the right size when
measured by a tape measure, but the reason for doing the calibration with the
belts rather than with manual measurements (see the thread of maslow 4 manual
calibration for how you could do it manually) is that any errors during
calibration will be the same errors you see during cutting as the measurements
are all being done with movement of the encoder against the belt.

Could be a dumb idea… But I said it out loud in case it sparks a good idea.

it doesn’t hurt to ask. If nothing else it keeps us honest with out
explinations.

The tricky thing is that all of the internal math relies on the same assumptions so they tend to cancel out. If we retract one belt to get to the anchor point I think we will always arrive exactly the anchor point because we’re using the same value for the distance between the belt teeth both to move and to measure.

So if the Maslow thinks it is 250mm from the TL anchor. Then you run a test…disconnect all anchor points but TL and retract TL belt 250mm, it will always be exactly back at the anchor? I was thinking maybe it would be 1mm or 2mm off, and that discrepancy might be able to help you fine tune the assumptions (belt teeth and arm to router bit assumptions)

The real tricky thing is that all of maslow’s internal understanding of the world is basically based on belt teeth not real measurements so if we say

So what this really means is that maslow thinks it’s 125 belt teeth from the anchor and if we tell it to move to the anchor it will move 125 belt teeth exactly but that may not actually be exactly 250mm in the real world

In that case, the most important numbers are the teeth count and their corresponding length over a span. Would there be some way to spool out a long length while only tracking the number of teeth spooled then measure the total length spooled out and compare it against number of teeth for a per tooth measurement (basically another take on James’ idea above)? With large enough numbers, a small discrepancy should become much more transparent. I suppose, if someone here doesn’t mind losing their sanity, they could manually count the teeth over a long segment and do a measurement.

In that case, the most important numbers are the teeth count and their
corresponding length over a span. Would there be some way to spool out a long
length while only tracking the number of teeth spooled then measure the total
length spooled out and compare it against number of teeth for a per tooth
measurement (basically another take on James’ idea above)? With large enough
numbers, a small discrepancy should become much more transparent. I suppose,
if someone here doesn’t mind losing their sanity, they could manually count
the teeth over a long segment and do a measurement.

We can be very confident in the number of teeth (not just the number of teeth,
but fractions of teeth), that’s not worth manually counting. We aren’t going to
‘skip teeth’ (the way stepper motors can ‘skip steps’

Where we can’t be confident is that belts stretch and sag so a tooth worth of
movement may not be exactly 2mm. But one thing we found with the earlier
versions of the maslow is that trying to get precise measurements over long
distances is HARD. The best tape measures available (which are significantly
more expensive than most) are rated at ±0.3mm + 0.1mm per meter of length (your
run-of-the-mill tape measures are double to triple that error), and having
someone measure over a 3m length to an accuracy of 1-2mm is nor reliable.

That’s why the maslow 4 has the calibration step. It’s trying to calculate where
the anchors are, using the belts to measure distances.

@bar “The first assumption that we make is that the teeth on the belts are spaced exactly 2 mm apart. Nothing in the real world is perfect so while the belt datasheet says they are 2 mm apart, we’ve found in the past that they are actually closer to 1.9988 mm apart and we’re working to see if we can refine that number even further. “

——> I wonder if belts are getting stretched but the stretch is worse in certain segments ie tooth spacing is dynamic along the length of the belt