# Discussion Regarding Triangular Calibration

#121

This is theoretically possible, but introduces a lot of complexity. It also will have an impact on user experience, as the amount of drag would depend on the exact material being cut. Meaning, you might need to recalibrate any time you move to a different material, or possibly even thickness.

I think there are improvements yet to be had in the static model. The easy solution I see to addressing this would first be to minimize friction between the sled and the work surface. That won’t fix everything, particularly acceleration-induced dynamics as you mention, but it’s a start at least.

#122

Would it be possible to look at an “average” situation (wood sled cutting plywood) and add those values? Would something here be better than nothing?

#123

Sleds usually don’t change after calibration. Could there be some kind of material selection when starting a cut that would alter a few values on the back end? Just a few common materials. Understood that this is getting more complicated.

#124

So to reply to my own reply, here’s an option: Draw a line plumb straight down from the center… use a level of a plumb bob… something. Mark where the tape measure between Hole 2 and 3 crosses that line and measure the distance from the center. In this example, Hole 0 is at 0,0 and therefore Mark 5 is at 0,-5.02. Knowing that, you can precisely determine coordinates for Hole 2 and Hole 3. Then, you can determine coordinates for Hole 4 using Hole 0 and Hole 3 and for Hole 1 using Hole 0 and Hole 2. You can quantify the errors in your hand measurement by calculating using different combination of measurements and determining how different your coordinates turn out.

#125

Yep

Edit: I love the brainstorming, don’t get me wrong. I just think we need to get static model done first.

#126

Would adding holes on the vertical centerline, maybe at the same Y coordinates as the corner holes show the rotation? Using a level between those should indicate any twist.

#127

Well, we don’t need to test for rotation, we just need to be able to solve the equations and the only way to do that is to have two well known coordinates. Mark 5 becomes that well known coordinate by making it directly below Hole 0. Therefore, the distance measurement becomes Mark 5’s Y coordinate (it’s X coordinate is 0). With the above diagram, and thanks to https://math.stackexchange.com/questions/2156851/calculate-the-coordinates-of-the-third-vertex-of-triangle-given-the-other-two-an I was able to calculate the coordinates for all four holes:

H1x:-9.1250666127, H1y:-4.74991585228
H2x:-9.53121855175, H2y:5.25003953974
H3x:8.03741860947, H3y:3.71249156983
H4x:7.78735350455, H4y:-4.8377524415

#128

no, we do not. So far, it has not seemed to be a significant factor, but it
could be one we are missing.

David Lang

#129

the idea of the slots was that you could hook a tape measure into one and then
measure to the far side of the other (eliminating bit diameter as a variable)

David Lang

#130

if you have the same distance from each motor (i.e. including chain slop error),
and feed out the same from both motors, you can draw/cut a vertical line down
the center of your coordinate system.

If this doesn’t happen to line up with gravity, it’s still the vertical line at
X=0

David Lang

#131

No, there is zero accounting for acceleration in the current code.

David Lang

#132

It’s not a factor at very low speeds, and since we move, drill, move any error
from the movement is going to settle out by the time we drill down into the
material.

It’s something we need to deal with eventually, but right now we are not getting
into the correct position, and we need to be able to do that with slow/static
positioning before we worry about dynamic effects.

David Lang

#133

Exactly, until we can accurately do static positioning, trying to account for
dynamic effects is hopeless, we can’t know what is static error and what is
dynamic error.

David Lang

#134

Just a quick question. Can you describe the order of magnitude of the “sag” in chain length. For instance, what length correction is there for 1) the nearest corner to the motor, 2) the center of the workpiece (4’x8’), and 3) to the opposite low corner? Are the answers like .005, .01, .05", or something like .01. .5,1.5"?? I have no idea of the answer.

#135

The long chain is calculated to be about 2 mm longer in the bottom corners if that helps.

#136

Thank you

#137

It’s more like 10mm longer

https://www.spaceagecontrol.com/calccabl.htm?F=3.2&a=10&q=0.09&g=32.18503937&Submit+Button=Calculate

now, the horizontal distance is only 9 ft not 10, but there is the weight of 10’
of cable

at the bottom corner, the force on the chain is only ~3.2 pounds

so this results in ~0.03 ft of error in the chain length, or about 10mm of
difference in length.

This also gives the formula for calculating the sag.

now, in our case, we don’t know all the constants, so instead of trying to
calculate them all, we can lump them into one value and figure that value out
experimentally.

David Lang

#138

This is something that I along with @blurfl, @bar, and others experienced during testing of the calibration routine. We suspected it was related to frame flex near the top center of the work area. This was before the updated frame was developed however.

It’s assumed that the X=0 midpoint is directly between cut 1 and cut 2, as cut 1 and cut 2 use identical coordinates around X=0. This assumption does limitations, particularly around acceleration-induced errors in relation to chain sag for example, but so far I think works alright.

100% agree. I’m still catching up on the progress with chain tolerance. That may or may not be related, but regardless, if the measurements for this known value are varying a decent amount, it will be difficult to make any part of the calibration routine fully accurate.

I like this idea. I don’t think I’ve ever played with this concept. Not sure if anyone else has, but it shouldn’t be hard to adapt this into the code.

#139

Now that’s a much larger number, and therefor potentially the most significant contributor to the calibration calculations issues.

Is there a simple way to verify the sag calculation standalone?

For a test:

In the under chain mount method for the 12’ top beams, could one position the sled to the opposite lower corner, mark it’s location, and then hold a straight edge under the entire length of chain from the sprocket departure point ( tangent) to the sled ring connection point. Probably need a 100” or so straight edge, but that should remove the sag and the sled should move accordingly the “sag” amount to verify the calculation.

#140

Now that’s a much larger number, and therefor potentially the most significant contributor to the calibration calculations issues.

Is there a simple way to verify the sag calculation standalone?

early on when we were first noticing the chain sag, we saw very substantial
differences from adding weight to the sled (we now know that that gets us in
trouble at the top of the work area), but we did see very noticable differences
is position at the same chain length with different weight sleds.

For a test:

In the under chain mount method for the 12’ top beams, could one position the sled to the opposite lower corner, mark it’s location, and then hold a straight edge under the entire length of chain from the sprocket departure point ( tangent) to the sled ring connection point. Probably need a 100” or so straight edge, but that should remove the sag and the sled should move accordingly the “sag” amount to verify the calculation.

the 12’ top beam will have substantially less sag as is has much higher tension
(even with the added length of chain)

10’ of distance with 3.2 pounds of force requires 10.03 ft of chain (~10mm)

11’ of distance with 7.5 pounds of force requires 11.007 ft of chain (~2mm)

yet another reason to go with the 12’ top beam

David Lang