Alignment Error Map - What is acceptable?

One problem I can immediately see is that the lowest input the material height will accept is .010”, which I’m guessing doesn’t match the paper very well.

4 Likes

Up to .25" the farther from center-bed you go. If you’re experiencing greater than that, Support will need to chime in. They’ll very likely want you to test with Proofgrade material in order to properly determine any issue. You might want to save some time and do that and UL a shot of that.

While @jbmanning5 has a point about the height, I doubt it’s going to account for all of your errors.

You can always numerically define the position by placing it accurately within the SVG. Your problem, though, will be that the tray can move. There have been a ton of discussions about precise registration using jigs, it is (to date) our best solution to the precise alignment problem. Search around and you’ll find a lot of resources.

Here’s one for jigging coasters that I wrote a while back:
https://community.glowforge.com/t/alignment-of-irregular-objects-coasters/16116?source_topic_id=18355

The matrix has a bunch of more info, including a recent addition of a jig demo video.

1 Like

The paper comes in at .004". I’m doubtful a .006" discrepancy in height could create that extreme an error. The lid not sitting perfectly flush with the rest of the top surface of the GF probably has more influence than the paper thickness.

I decided to perform this test, because placing cuts on remanent sheets of proof grade was a frustrating game of chance.

I know if I want to ensure precise alignment a jig is the way to go, but I’d like the camera to be more accurate than this when it comes to remanent cutting.

Not sure how the thickness of material works. Try it in some good flat cardboard and see. This is way off compared to what I’ve been accustomed to.

1 Like

This has been discussed before and I believe it was determined to be part of the design, not a problem.

No. But it is certainly contributing. That’s an error of 1/16th of an inch. Right?

I recorded this a little while back demonstrating the difference 1/32nd of an inch can make.

image

5 Likes

Did you input the correct thickness of the material in the Unknown Materials column before you set up the rest of the file? (For paper - I usually use about 0.01")

If you didn’t, then yes, it can be off by that much.

If you do not have the feet of the tray sitting down in the dimples correctly, it can also introduce that much error, so that’s something else to check.

2 Likes

@Jules the custom material height was set to 0.010" (as low as the GF can go, but close considering the paper is .004") Also the tray is fully seated in the dimps. I’ve slid it left and right and back and forth to make sure it is fully seated.

@jbmanning5 the discrepancy between the paper and the min allowable thickness is .006" (1/167 of an inch) not .0625" (1/16 of an inch)

The severity of the discrepancies I was seeing when using draft board is what made me run this test in the first place.

It’s late, and my brain is fried, so forgive me if this should be totally obvious, but are you using the after image of the bed to determine how accurately things were cut?

I’m using the after image to determine how accurate the camera is at predicting the placement of cuts. Ideally the cut in the after image should line up perfectly with the vector objects in the GFUI.
The bull’s eye lets me quickly measure discrepancy between the actual cut and the predicted cut. Each ring is .125" wide.

The parts being cut and their relationships to one another are spot on. It’s just the GFUI’s ability to accurately locate the parts relative to existing features on the board that is problematic.

1 Like

I could be wrong, but I’m fairly certain the post image on mine is not the same as the pre image, but I dont know if theyre all that way. On mine, the resolution changes and it all seems to shift. Which, if thats true and not me just imagining things (a total possibility) would make things seem more misaligned than they really were. You might want to try your experiment again without relying on the post image.

3 Likes

To check accuracy start with images on the paper and try to align an engraving over the top. You can make the original marks with a 2D printer or use the GF.

4 Likes

As Erin pointed out the post and pre images do not align. Do a search using the terms “post image alignment” and you will find multiple people who have pointed this out. Palmercr’s suggestion is the way to avoid any pre/post image discrepancy.

1 Like

Oh, that’s really cleverly simple. I was coming up with all kinds of ridiculous processes that involved, like, a key for scale and alignment marks and overlaying photos and all kinds of nonsense. I’m just going to try your way instead.

3 Likes

I printed out my alignment targets on paper and taped the paper to the crumb tray.


The vectors in the GFUI do not consistently align with the picture captured by the lid camera. An interesting item of note. The vectors in the GFUI that are misaligned are always to the left of the targets captured by the lid camera. Considering the lid camera picture is a distortion corrected fish-eye photo, I would suspect the misalignment to be biased left, right, up, or down depending on the targets location relative to the center perfectly aligned target.

Here is the GFUI after running the program.


Two items of note:
1.) The misalignment between the vector elements and the picture captured by the lid camera remain the same. If you stack the two images in PS or GIMP and do a difference between layers, the only significant changes you see are the cuts in the paper.

2.) The cuts in the paper are all uniformly misaligned. Relative to the printed target, every cut is roughly .120" to the right and 20 degrees above horizontal. This can be seen more clearly in a cellphone photo I took of the paper taped to the bed.

In conclusion, it appears the algorithm being used to flatten out the fish eye photo taken by the lid camera is not creating an accurate representation of the material within the GFUI. This error is then compounded by the actual cuts being shifted right and up. The .006" difference between the paper’s actual surface and the minimum height of .010" may account for some of the errors, but I highly doubt it accounts for a significant portion of the errors, especially considering my experience attempting to cut on proof grade remanents.

3 Likes

That’s what I expected, too! It’s so interesting that you’ve got those centered ones perfectly aligned, but all the misaligned ones are off in the same direction. Maybe that’s the effect of the lens correction they’re trying out as the moment?

But this is what’s weirdest to me. I wonder how much rounding up the height and/or focus height are contributing to that. It seems odd because those centered ones looked perfectly aligned, so I wonder what it’s using to determine placement. Maybe ungrouping all the targets after they aligned but before they print would give a different result? And I wonder if you can correct for the uniform offset manually (at least until algorithms or whatever get tweaked - I believe this is an ongoing thing). Like, if your machine always prints slightly down and to the left, could you line up your design in the gfui, get it exactly where you want, and then nudge it, like, 3 times right and one down, or whatever.

I don’t mean to imply you should have to test anything further, I’m just always full of theories :blush: If I get around to doing any of that myself, I’ll be sure to post it. For science.

This sounds like this…created by @cynd11 ;

A way to improve object placement

I set this up and use it quite often.

2 Likes

Fellow Glowforge’ians

I did something like this 3 weeks ago and sent my results to the GF team on Jan 1st and to be honest was a little disappointed in the reply.

After I reported they asked me to perform one more test on the Owners Ruler and send those results. The reply I got was:
Thanks for taking the time to perform the troubleshooting and provide the details. The test with the Gift of Good Measure shows that your Glowforge is performing within the range of accuracy that we expect right now. Most importantly, that means that as we make improvements to the software, I expect you to see much better alignment.

If you look at my target results they are similar to what others are seeing, and as you get closer to the edges of the bed they get even worse.

My Results.:


The second image is the final test they had me do with the object directly under the center of the lens and even that was off . I was using the PG Draft Board for my test.

Hopefully GF can get some serious updates flowing here to correct this error, as like many of you I have burned through a lot of materials with the cut not aligning up to the material placement.

Hope this helps,
Alex

1 Like

One can see the alignment will be out when the outline of the board is not shown square.

Did anybody notice an improvement in accuracy in Dec? Examples like this seem just as bad as they ever were. The announcement implies there should be a big improvement if they have started using machine specific calibration data. Perhaps the machines shift in transit or need to be on a perfectly flat surface for the calibration data to be valid.

It would be interesting to see how it varies with a small shim under a corner.