Hi. My GF is generally working really well and I love it. However, I find that there are a few things that hint at a greater problem.
The main gantry (I think that’s the right word) doesn’t rest as far back as I think it should. The unit I have is a replacement after my first one died and I remember that the first one slid all the way back when it returned to it’s resting position. This one sits an inch or so forward. This causes me to need to slide my material forward in order to get the best use out of it.
When I cut a large rectangle and then flip it upside down to engrave the back, it doesn’t fit, even if the material was perfectly flat. It seems to have cut slightly skewed. I posted about this a while ago for reference: Not Square (skewed).
Circles are not perfectly round. If I cut a circle the inside is not able to spin around freely, implying that it’s oblong. Even with the kerf of the laser, it gets stuck pretty soundly when turning it.
To better show the problems I made a couple of horribly produced videos with very bad sound. (Sorry, but you may need to turn up your volume to hear it.). It shows the gantry in the first one and specifically the circle issue well in the second. The rectangle problem does not come across, but I think that’s because the shapes I cut were too long and narrow. The other post is clear on that, though.
Anyway, I’d love to know if this is something that I can fix on my end. The precision of a laser cutter is what really attracted me to Glowforge and I would hate to have to live with it being off like this if I don’t have to.
It’s where it’s supposed to be. Neither the PRU that I tested or the unit that I currently have has a gantry that goes all the way to the back, and they produce excellent results.
As far as some earlier rectangles being skewed, it might look like the material is flat, but even a mm of warp in it is going to affect how it cuts. If you’re not using the honeycomb pins, you might have warp in there that is affecting the shape of the circles and squares that you are trying to cut. If you pin the material down and it’s still skewing, then you might have a problem.
The videos that you showed did show a slight amount of skew in your circles, but without inspecting the file, it’s hard to know whether you actually created perfect circles. It’s very easy to drag one out that is not exactly perfect, and if you just duplicated the original, they would all show the same skew problem. The rectangles in the same file did not show any skew issues, leading me to believe that the problem was with the circles in the design.
You can look at the width and height measurements on the circle. If they are not the exact same number, you found the answer.
The Glowforge has been incredibly precise. Every time that I have had a problem with something fitting back into a hole that was cut for it, it was a problem with my file. So that’s the first thing I always look for.
Anyway, it’s something to check. Maybe try a different circle or one created in some other program.
Could also be that the gantry is crooked, which makes everything come out crooked. Support will follow up with instructions on how to attempt to fix it.
Yes the GF currently has no way to square the gantry automatically. It seems to rely on the wheels keeping it square but they don’t because it is long and heavy.
You need to manually square it with the power off using a set square against one of the Y rails. Hopefully it will then stay reasonably square but it may jump a bit when the motors energise.
Thanks @Jules! As always, I appreciate your quick and thoughtful responses.
This very well might be the case. I just didn’t remember the first one having the head in resting position being actually over the honeycomb. I admit that I may be remembering incorrectly, though.
I can assure you that the materials were very flat. I check them before putting them in the machine, use some serious magnets to keep them in place, and then tap test them around to check for anything sticking out. Any bend would not have been enough to justify this type of skewing.
The files are definitely clean and exact. I’m a graphic designer by trade for many a year, so this is one area that I’m confident in. My guess is that the rectangles didn’t show skew because of the normal give inherent in MDF and would have had the same issue had a cut a wider sample.
I’m very happy to hear this. I just hope mine is not the exception. I’ve done lots of work on it and generally it’s amazing. For most projects, though, if something is just slightly off then it’s not noticeable. When exact precision is needed, though, I keep running into things like this.
My fingers are crossed for you. There are so few user-serviceable adjustments, I predict you’ll end up with a replacement. The only things I’ve seen suggested here are to turn off the power and push the gantry all the way back and even on both sides, and to ensure the rails are clean of any debris. If that’s been done and it skips steps or goes out of square on its own, I don’t think they’re going to send you in with a screwdriver.
I’m not sure on that, when it does its calibration the head first moves under the camera to get its world position, then the head moves to the far left and the camera takes a couple pics and it sounds like the steppers move sometimes, I always assumed it was looking at the y gantry and squaring it by moving only one of the steppers. But this is pure speculation…
Oh, you’ve got to try the pins! I used everything from magnets that were so strong I couldn’t lift them off, to Gorilla tape, to practically nailing it down…the only thing so far that has removed both warping and twisting were those pins. I won’t cut without them now. They even keep partially cut up sheets as flat as the topography around here.
I don’t think it does because a) people have solved this problem by manually squaring the gantry, b) support even recommend that sometimes, and c) the motion plan only has one step and direction bit for Y, so to move them independently (which the electronics is capable of) would need some other mechanism.
Also anything the camera does is has not proved accurate enough to detect a small skew in the gantry when each end of it could be 1/4" off.
Not saying that it does it, but seems like one would be able to pick up rotation in the bug. The calibration seems to be right on/repeatable every time for me (in that I can power down and back on, and use old jigs and have perfect runs).
I think if the gantry happens to start square it stays square but it doesn’t actively compensate, so if it starts skewed it stays skewed.
To some extent the wheels will try to pull it square but you can’t rely on that. I have a Shapoko 2 that uses the same axis mechanism and that naturally runs a bit skewed due to the plastic wheels not being perfect. It needs two limit switches and independent Y axis homing to pull it straight. After that both Y axes can step together.
If you are feeling brave you can skew your own axis and see if calibration fixes it by moving the ends independently somehow.
If the source code they released is the same as that which is currently generating the outputs for the steppers (and all indications are that it is), then the Glowforge is not currently capable of driving the Y axis steppers independently.
The code (specifically lines 32-38 for direction, and lines 55-59 for the step) sets both Y stepper motors based only on a single Y step/dir instruction.
Also, the current puls file format they send to the device doesn’t support independent Y-axis direction/steps.
This isn’t to say it can’t be done. It just can’t be done with the existing firmware.
I extracted the logs to investigate the laser arm positioning, and I can confirm that your Glowforge is calibrating properly.
I’m sorry your prints aren’t turning out the way you expect.I have a method you can use to reset your laser arm. All you’ll need are your own two hands and large square layout tool with a foot. You’ll find the instructions here.
Yes if they can ever make the lid camera accurate to 10 thou the machine will meet its specification for accuracy but at this late stage I doubt they will ever get it that accurate. Not having limit switches (and two on Y) is probably their biggest design mistake.
