Camera acquisition accurancy

I’m not sure if this has been asked already but I couldn’t find any recent data and I seem to remember @dan said this was being worked on.

Is there an official statement about:

  1. The dimensional accurancy of camera acquisition
    Simply put, if i draw a square or a hole, do i get +/-1%, +/-0.5mm or what else?
    This is important if one wants to draw some simple parts by hand with a pen and ruler.
    From what I’ve see so far, this method is only good enough for artistic designs where, say, a 10% accurancy is still okay.

  2. The positioning accurancy on the bed using the camera
    Similarly, if a design is positioned in a certain point in the bed using the camera view, how accurate will it be in reality?


All I’ve seen is the same “under development” you have, and the observation that in demos not-centered stuff doesn’t position perfectly.

This is a case where the mostly-closed-cloud software toolchain may a problem, because it would be fairly trivial for people who cared to correct acquired images and then spit them back out to the laser – if only they could get to those images. (Ancient image-processing trivia: this is the reason for those grids of black dots in old pictures from space probes. The optical systems had all manner of distortion, as did the transmission, but the dots were in known “real” positions.)


Does this post by @dan assist in the answer? Knowing actual numbers of the specs will help in planning. The post is dated and you most likely read it. From recent forum discussion it’s seems to be dialed in.


I couldn’t find any public/official numbers regarding the camera resolution, but I would have to guess that the accuracy of the scan will depend heavily on the resolution of the camera.

I’ll leave the math out of this post, but I’d say we’d want 3.75 megapixels at a minimum (which would give us a pixel size of 0.008" by 0.008" AKA pixels the same size as the expected kerf width) and 15 megapixels as a realistic number to hope for (which would give us pixels of 0.004" by 0.004", AKA half a kerf).

Doesn’t the Glowforge have two cameras? I thought it had one lid camera and one on the head. It would be cool if we could use the head/macro-ish camera for extreme resolution, but that hasn’t been talked about much at all as far as I recall.


My understanding is that the camera on the head is what will enable the 3D engraving. By continuously re-focusing the camera on the surface, they would be able to determine z-depth & re-focus the laser as-needed. I don’t think the camera is meant for providing high resolution “scanning” of a drawing - I think the lid camera is responsible for that…

However, since the hardware is there, I can’t see why this couldn’t be a feature that is potentially rolled out via a firmware upgrade down the road.


The Tech Specs page has this to say about the macro camera:

Macro Camera — Able to view one square inch with resolution of 0.002” (0.05mm)

I have seen this discussed before but I can’t find the exact link. At the time, it sounded like they were shining a small beam of light at an angle towards the workpiece and using the macro camera to see where they intersected. Based on the position of the dot you can calculate the distance between the head and the work piece. They may have a different approach by now of course.

All good points guys, i had forgotten about the macro camera. At that resolution it could be used to do a high-precision scanning i suppose. The lid camera must be far from that.

I was hoping to gen an official statement from the GF team though.

1 Like

(Dan’s answer to my resolution question)
5 megapixels, if memory serves - but as any photographer will tell you, the limit is the lens, not the sensor. It’s an extremely expensive, custom lens design to view that wide, from that close, and even then it can’t resolve across the entire bed at 5 megapixel resolution

Regarding focus,

1 Like

The information here is both complete and accurate. We’ve already improved positioning since maker faire and may continue to do so, even after we’ve shipped - we have a lot of image processing work in the hopper!


Well, the information might be “complete and accurate” but it doesn’t answer the questions.
I assume it’s too soon to give the numbers.

It’s not a huge deal, for precise cuts it obviously makes sense to make the design on a PC and it works for me.
I was just curious how good it would be because I’d like my father to use the GF as well for his hobby and he’s not able to use a CAD, but pen+ruler would work for him.


Sorry, just looked at the original - I meant that what’s stated here is all we’ve shared so far. We haven’t announced our dimensional accuracy, in part because that depends on many factors (e.g. it’s slightly affected by the thickness of the material) and in part because we’re continually improving it.