I’ve been following this topic and it seems an issue dear to many. I’m trying to understand it better so I’ll just jump in and see if I am getting it.
First off the lid camera is a fisheye view of the bed so its optically distorted in a spherical pattern from center to edge. That’s a given.
It seems to correct this a factory profile is loaded that corrects most but not all distortion. On top of that the amount of correction is modulated by the thickness value you report for your material. Fudging this will sometimes result in better image accuracy but its not consistent.
The problem is the alignment of camera image to reality is affected by the lid opening and closing and the bed shifting slightly in use… so that factory file quickly goes bad.
I worked for years in the medical imaging field and we have this same issue of distortion introduced by various non linearities in optics and receptors. The answer for that and perhaps for this, is routine calibration. A calibrating phantom is put in the field of view… the image is taken and the software generates a new correction file. This is well known in imaging science.
It would seem that the accuracy of the bed image would be enhanced if a least once a week a cal program is done. First a grid phantom is placed on the bed and a perfect digital representation is displayed as an overlay. If the two align perfectly then no calibration is needed. If not then a camera image is taken and the error between the detected lines of the camera image and the calibration file’s perfect image lines would be the correction file. After cal. the camera image and the overlay should match.
Is this what Snapmark is trying to do?
They said Snapmarks were a side benefit of the ongoing development of the pass-through software. For whatever that’s worth.
I don’t think the Snapmarks actually do any calibrating, since their printed size is so crucial to getting them to work. They’re more of an alignment tool, just to line up existing artwork with something you’ve placed in the bed.
Good suggestion re: calibration, though!
It’s using the lid camera, the head camera, a digital file with fiducial markers, and a set of physical fiducial markers that are visible in the bed.
This probably isn’t the exact process that GF is using but it’s the gist of it.
The method for getting the fiducial marks on the material can vary. Some people print a design for print and cut. Some are making jigs, etc. since this was described originally as making repeat-use jigs, I’ll address it from that perspective.
A user sets up their design with Snapmarks in the file. They score the Snapmarks onto a piece of jig material and cut out or mark the area they want for the material to be processed.
When you use the Snapmarks, the user hits the Snapmarks icon. The file is checked for the presence of Snapmarks, the lid camera takes an overall image which is analyzed with some algorithm, the head camera moves over to where one Snapmark is supposed to be and takes a series of pictures to fine tune exact placement using the macro camera in the head, and then it does the same for the other Snapmark.
From these actions, it can determine the exact locations in the bed of the two Snapmarks and move the entire design accordingly adjusting the X, Y, and skew.
What it doesn’t do is correct the distortion so that the design overlay is exact. You’ll still notice a discrepancy (dependent upon how accurate that machine is), between the on-screen display and the actual placement on the material.
By combining the lid and head cameras they can definitely alleviate alignment issues in the short term and recalibrate your optical alignment. I have a feeling its on the way.
This is something that I’m hoping they will be able to pull off every time you close the lid, because it can cause the camera to be positioned slightly differently every time.
If they would have built some sort of markers into the bed, they would have been able to do this from the start with a bit more ease.
Even though this is a bit frustrating, and seemingly solvable with the tech at hand, the capabilities of a snapmark enabled Glowforge are beyond any other laser currently on the market. Pretty exciting
I really appreciate your experience and perspective. Yes. That recalibration in the field would be absolutely amazing. It should be something that is possible and necessary given the physical constraints of the machine. For example, if there ever was a need for a new crumb tray, would that change the calibration.
I would say this is in the hopper since improving accuracy has been a constant goal and it does keep improving.
As stated above, Snapmarks are different in that it is moving the design to where the snapmarka are and not a permanent calibration.
Then as @tak pointed out, there is more work possible on using the head camera to sync with the lid camera for materials thickness or surface measurement.