Frame size and squareness of the cutouts posted above.
Recalibrated, and adjusted for the corner mount Z lift (~9mm), ended up with 0.80 fitness, which is my best result yet. .log file posted below.
Two cocktails and two calibrations later, I realized the easiest way to check X and Y accuracy is to measure the edge of the sled to the edge of the workpiece, and then jog the machine that exact amount and see if it lines up. As expected, the X axis was 4mm short on the right, and the left of center, the Y axis was spot on. This was on the spoil board at zMin after a fresh calibration. So likely as accurate as it can get. Excited about this test, since it’s so easy and quick to check. It is a little short of what I experienced on the cuts, so something might still be missing there. (And FWIW, I jogged the same distance back and it landed perfectly at home.)
Posting some photos below of my whole rig, just in case y’all spot anything off, and the X and Y axis jog tests.
If you jog to the left or right edge as shown in the pictures then then rise or lower in Y only does the measurement stay consistent along the whole edge or have any deviation?
Just right, left, down, and up from home. Suppose I can try far right, then down and see what happens. But generally the y axis has been pretty accurate.
I am curious if going down one of the edges and measuring along the way will be consistent or have any change. For example start in top left corner and work down the bottom left.
Seems pretty straight and square. Guessing it’s more of a scale issue on the x axis.
The maslow hardware doesn’t have an X or Y axis, they are simulated for ease of
programming
instead what it has is 4 axis that fight each other, and as a result, instead of
getting a nice linear error, we tend to get errors in curves instead (which is
why I’m looking for errors that change in the corners compared to the centers of
the tops/sides)
I’m not entirely sure how the diagonal jog works, it looks like 45*, but mathematically doesn’t make sense if you jog the path of a right triangle. Guessing it calculates the hypotenuse of a right triangle with the A and B side being your jog length.
Just got back home and tried jogging down then right. Basically 0 offset at the bottom and about 5mm on the right. Seems consistent with the “Y” axis distortion I’m seeing.
Let me know what else I can test or where we can go from here to try and sort this out. I believe I posted my calibration data earlier in this thread.
@dlang Jogged the machine and ran a grid measuring how far off from where it should be. Results are in the thin black lines below. This time I got to about 2" from the edges since that’s about the clearance I ran with the cabinet. The skew in the X axis corresponds to the amount it was off with the last project. Also, I noticed that when the machine is at the far right side of the board, the left belts are quite loose, wonder if that’s where the issues are stemming from. Posted a video below of that as well.
Please let me know what we can do next to try and improve the accuracy here.
belts being loose is a key thing, check that your Z offsets are correct, save
your existing maslow.yaml and try the calibration again to see if that improves
things.
We are also seeing that having the calibration tension higher than it needs to
be is causing grief (the belts get stretched as they are measured)
Is the error symmetrical (top to bottom and left to right?
looking at the calculator above, the default grid size has some data points that are far enough outside the green area to be questionable, so I would also shrink the grid used for the calibration to 1500X1000
The belts are only loose at the far edge of the board, otherwise they are tight and look as they should.
Never changed the calibration tension, do you think I should lower it from the factory setting?
Only checked the BR quadrant, I can check the others, but it takes a while to do. Maybe I’ll do less points this time.
I can shrink the grid and recalibrate, but that was the stock calibration size I believe. I run the calibration with support skirts so the machine doesn’t fall or tilt off the edges, but I don’t think it gets that close to doing so any way.
Can you elaborate on what that graphic is telling us? I’m not too familiar with that tool (looks super cool though!)
Can you elaborate on what that graphic is telling us? I’m not too familiar with that tool (looks super cool though!)
that page is looking for where the arms hit the uprights (either opposite
(corner) or adjacent (edges)). Where they do not hit is the green area. The
bands outside of that are 1 degree of bend between the arm and the belt per band
(there is a little chart showing how much error each degree of bend would result
in for what would otherwise be a 1m distance to the anchor)
just a little outside the green doesn’t make much difference when cutting, but I
think that if you calibrate outside the green you have more problems. (it would
be good to get a calibration that could identify points that have more error and
drop them, but we don’t have that right now)
note that there is disagreement about how significant it is to move outside the
green area, I am on the pananoid side.
the grid is 100mm squares with 4 corners at the center.
Never changed the calibration tension, do you think I should lower it from the factory setting?
I can shrink the grid and recalibrate, but that was the stock calibration size I believe. I run the calibration with support skirts so the machine doesn’t fall or tilt off the edges, but I don’t think it gets that close to doing so any way.
this is an area that we are just starting to look at, but indications are that
right now this will help.
you need to calibration force high enough that the belts pull tight at every
point, but any beyond that may cause more grief
The belts are only loose at the far edge of the board, otherwise they are tight and look as they should.
Only checked the BR quadrant, I can check the others, but it takes a while to do. Maybe I’ll do less points this time.
I would be especially interested in seeing if there are errors that show up
outside the green area of the graph
Okay, I’ll reduce the calibration area, calibrate, then reduce the tension and calibrate to see how much each changes the calibration result. What tension would you recommend? Also, should I return the tension back to stock for when it’s operating?
Let me know what errors to look for or how to test them. I had some 15mm position errors when running the machine previously, but it kept on trucking. I posted them above, it was cutting the doors so the machine was about 2" from the TR corner.
Okay, I’ll reduce the calibration area, calibrate, then reduce the tension and
calibrate to see how much each changes the calibration result. What tension
would you recommend?
I don’t know with a vertical frame.
Also, should I return the tension back to stock for when it’s operating?
there is no need, during operation it will apply up to 4000 and works to keep
the belts at the desired lengths. During calibration it’s pulling and trying to
figure out if the belts are tight or not, so it needs a limit.
Let me know what errors to look for or how to test them.
at each point, you want the belts to pull tight (not be loose).
and you need the right Z offsets and make sure the machine is all the way down
when you run the calibration.
I had some 15mm position errors when running the machine previously, but it
kept on trucking. I posted them above, it was cutting the doors so the machine
was about 2" from the TR corner.
the 15mm position errors should be the result of staticn electricity on the
direction pin (the JST boards solve this, or you can solder a jumper to the pins
to fix the older boards)
@bar Multiple times. Here’s the errors from earlier in this thread. This was maybe the third time it happened while cutting this project and on the JSTs, the other instances were pretty similar. Despited exceeding the threshold, the machine kept on trucking. Belts and everything else looked just fine to me — I was watching it pretty much the whole time. If it helps, the doors are the rectangular panels with 3 pocket holes located in the top right and the bottom middle position (reposted the image w/ measurements so you can see where they are located.)
I don’t think that this is related to the measurements issue, but that is something to keep an eye on. I haven’t seen that with the JST-XH cables. It might be worth unplugging that top right cable and plugging it back in just to make sure everything is seated fully and making good contact