it would need to be a 4’x8’ roll, it’s hard to get such things printed
accurately.
The need isn’t for a linear strip, but for a grid, you need to figure out where
you are in 2d space.
at one point the local kinkos/fedex type places had black and white large format
printers that you could get things printed on for something like $0.50 sq ft,
but they have phased them out in favor of the color ones at $10/sq ft.
David Lang
The 4x8 vinyl banners that I and @johnboiles got from Vistaprint were about $70… but they weren’t “perfect”… really good, but not perfect. We had to do include compensation factors in the optical calibration routine.
I suspect printing is less accurate than laser cutting in that regard.
@dlang agree that a single linear strip of tractor holes is not suitable. But with five overlapped sheets centered in the work area, you will have six rows of holes that are both horizontally and vertically aligned. Although the vertical interval is not so great (nine inches for imperial paper), it might be enough intervals to perform reasonable calibration. If finer vertical calibration is required, you could add a rotated sheet strategically in whatever zone we need it most, for a followup calibration after processing the horizontal strips. To aid in counting tractor holes, use a bingo stamp to mark every fifth hole or whatever. Heck, if you want the machine to index its own location, use a sequence of bingo stamp colors, but that’d get tedious to prepare (-:
@JWoody18 I agree printing has too many issues for accurate repeatability. Often both the substrate and the printing mechanism are moving, and different materials, machines, and software are in use. A flat silkscreen or metallic mask could be better, but boutique solutions (and again the rolling of large format paper) make it hard to fulfill globally over the years. Laser cuts are only as good as the machine that did it. My self-designed 1.3^2m blue diode laser cutter is really janky, I wouldn’t trust it to do the job. Also, many laser dots are usually not circles, they are ovals or lines, and have to be taken into consideration if you want to maintain that sub-mm accuracy that is requested. This is why I keep returning to large repeating factory-stamped patterns, it’s really likely to have good repeatability if applied ad nauseum by a single rolling metal wheel.
Is this confirmed fact or opinion and why is full 4x 8 coverage required?
Jeff
Calibration issues (error) tend to be more pronounced around the edges, so having a pattern that covers the entire area helps. You could calibrate only in the middle portion and extrapolate from there, but the results won’t be as good as incorporating actual measurements from the edges.
full coverage is needed because the error is different in different places on
the sheet.
So if you are going to handle error by having a table of what the error is, you
need to have the table cover the entire sheet. If you are going to compute the
error based on measurements in only a few places, you have to have an accurate
model of what the error is everywhere (this is the approach we take with the
basic calibration, optical calibration is taking the approach of a table of
errors)
David Lang
Ok, but this is slightly different than full coverage. The more sample points the better, of course, but balanced with something shippable or reproduceable. The idea of using a smaller number of strips with a way to register them to each other to ensure square and a series of reference marks for the optical sensor to read could work, could it not? And registration points could be in multiple locations to allow for small work areas as well.
Which brings me back to my post above about using smaller pieces of laser cut mylar that are full length horizontal and vertical (I don’t agree with David that these can’t be sourced, but TBD).
So in the attached screenshot I’m showing six pieces, three horizontal and three vertical. I only showed two registration marks for the example (each piece would require them, of course) and I didn’t show any of the calibration points to be read but they can be anywhere on the strips in theory.
Assuming the sourcing can be resolved, does this meet requirements?
-Jeff
the errors are in curves (different shaped curves in different directions from different causes adding together), so just getting the midpoint and end of the resulting error may not be good enough (remember, testing trumps theory and I’m talking theory)
the problem comes when you then need to connect several together to cover the
entire machine.
I’m thinking that it may be possible to do something like map the mat area, then
pause the machine and move the mat so you still half cover the old area, go back
over that old area to orient the machine as to how the new mat position relates
to the old one, and repeat.
but using a 18x24 mat to cover 48x96 bed, half a mat at a time seems like it’s
going to be slow.
David Lang
What if we printed on Tyvek? I see many home builders with custom branded house wrap, so we might be able to order a QR code pattern. Cheap, dimensionally stable, and you could probably fold for shipping and then iron it out.
Has anyone looked at this?
That reminded me of a machine that I work on occasionally at a customer site. A robot travels down a long set of rails and uses an optical encoder (camera) to read QR codes, along the length of the rails, to know where it is. Obviously industrial tech like that is super expensive but the QR tape isn’t too bad (I had it priced before but don’t remember the number).
Here is a link to a 6m length of tape.
If you are interested in the other cool , expensive positioning tech, "fraction-of-a-millimeter position feedback " check this out (I’m not a P+F rep 
):
P+F is sort of like Festool or Porsche.
This is even better than what I was suggesting, and now it’s reminding me of the domino tape that came with my Shaper Origin. Can we place it precisely enough for calibration, though?
we’ve talked about using mice before, but we don’t trust them not to loose their
position slightly as they move around.
David Lang
remember, we are looking for sub-mm accuracy, this is why I am not sure if it’s
possible to have any pieces that need to be aligned and get them accurate
enough.
David Lang
That’s why I suggested returning home often to recalibrate the mouse. If you don’t, it certainly will gain too much error (based upon the research I did)
That’s a really interesting idea and it’s relatively light to ship too. Would certainly give broader coverage.
-Jeff
