Discussion Regarding Triangular Calibration

Technical Details
201

I like the general idea, but would tweak it a bit.

we have the Z axis touch sensor, how about putting a 1/2" carriage bolt in the
chuck, hooking the sensor to it, and then manually drill some holes in known
places, put carriage bolts through them and wire them all to ground (or have a
clip you can hook to each one as the sled approaches it)

Then the machine can go to where it thinks the bolt should be, and do a spiral
search from there.

once it finds a bolt, it can then move around to find it from different
directions to figure out where the center of the bolt is. (back off until you no
longer have contact, approach slowly to get the most accurate position, cross
over the bolt head and approach it from the other direction, this then gives you
two points, move halfway between them and drop until you no longer have contact,
raise up until you make contact, keep going up until you loose contact, drop
until you make contact)

the sled should be able to ride up over the bolt heads, and then it will drop
down when they get to the center hole, so the sensing should be accurate. the
approach of getting clear and approaching slowly should work around any catching
of the two bolt heads against each other.

David Lang

202

I like the idea as well because it eliminates the camera and need for a router analog. My concern is the manual drilling of holes in known places. The accuracy of the calibration will be dependent upon the accuracy of the holes. if you’re off 0.5 mm in drilling a hole and you are off 0.5 mm in calibration. Either you need to drill the holes in precise locations or you need to be able to measure the precise locations of the holes you drilled. The latter is similar to “holey maslow” and I had difficulty determining the precise location of the holes. Maybe we can get someone to make a 4-foot by 8-foot etched PCB :slight_smile: ?

1 Like
203

I agree, but let’s work on one problem at a time.

First, let’s get this sensor approach working.

Then we can talk about the problems of drilling the holes accurately

It’s pretty easy to scribe a line parallel to the edge of the plywood. It’s not
hard to measure the actual length and width of the plywood.

This makes it pretty easy to do at least a known location in each of the 4
corners, and not that hard to make a regular pattern.

Even if we don’t use this for the ‘check every few inches and create a
correction matrix’ approach, this could work a lot better for the ‘measure a few
points to set constants’ approach that the standard calibration uses than the
current cut+measure approach (a lot less material wasted with repeated
calibrations at least :slight_smile: )

It would be a lot easier to support in a firmware calibration B-code routine

something like

Bxx x1 y1 x2 y2 …

goes to the locations specified (any number allowed), searches for a pin,
records the error matrix possibly including, possibly followed by

Byy T no-motor no-rotation

try to adjust the various constants to get a good match, consider success if you
are within T of all locations, and don’t adjust various items (motor spacing and
rotation radius in this case)

David Lang

1 Like
204

I am not sure about drilling holes accurately. With the approach I described above, all you need to do is accurately screw the bar to the plywood. Further, you only need to be accurate in one dimension since the bar can slide long-ways without impacting its position as it is measured. Get a story-stick and use the edge of the plywood for reference. Screw the bar down when it is resting against the story-stick.

205

I think he was replying to my modification where instead of a bar we use the
heads of carriage bolts.

trying to poke through the workpiece to bars behind it has a few problems

  1. you need to get close enough to get into the holes before you probe for it

  2. more Z depth is needed as you need to not just hit the top of the bar, you
    need to get down to hit the side

  3. you are trusting the edges of the plywood to be accurate, at this resolution,
    not always true. (they are generally pretty good at being straight, not bad at
    being parallel, but not always the exact same width and length)

  4. you are trusting that the bars are perfectly straight

I like the basic approach of using a touch sensor, I just think there are better
ways to make use of it.

David Lang

206

Yes, sorry.

If you had a sheet of metal and could cover it with an insulating material that had regularly spaced holes, then you could accomplish this.

207

Or if you had a sheet of metal with large holes regularly spaced, you could “activate” on loss of continuity.

208

I think the aluminum foam core boards come with a peelable protective layer. Score through that to the metal, and use a bullet-nosed pogo pin for the ‘feeler’ and you would have something like that. No idea of the cost of something like that, though.

1 Like