Not a complaint, as everything works really well (and I have reported this to support) – but over the last week or so, when I start my GF, as it goes through its “first act” if you will, it the carriage moves down to the middle of the bed with the head on the left side as you face it. As it moves, the head rocks on the left side as if it were being pulled over a bumpy road and makes the sound I think others have discussed here. I immediately stop it by turning it off, gently move the carriage to the middle of the bed and the head under the camera and restart it. It then starts up perfectly. I have taken to doing that placement now before I even turn it on,
When I contacted support, they advised to do exactly what I was doing and let them know if it persists. I did – telling them that if I don’t do the move carriage and head placement steps as described, the same start up action occurs. It does not seem to impair or affect the smooth operation of the GF from that point on.
I am wondering if when finishing and shutting down, the head moves just a little too far to the left.
I have the same issue. I was also told the same by support. I even gave them a video of the head homing successfully and then parking too far over. Here is their response.
Hi,
Thanks for letting us know that happened. To help your printer head avoid hitting the side, please:
Turn off your Glowforge (this is important to avoid damage to your unit)
Open the lid and, using both hands, gently move the laser arm to the center of the bed
Gently move the head under the lid camera
Turn your Glowforge back on
If your Glowforge has trouble calibrating again, please let us know.
Can’t figure out what is different with your machines and the PRU I have. Never have to move the head. Haven’t had a head crash for many months, though it did it a couple times in the beginning Which is an indicator to me that our mechanical design is similar enough. Since the head location is unknown without the lid camera it leads me to speculate a lot of possibilities but truth is we just don’t understand. I would think Support has figured it out with this many reports.
I haven’t had to do a manual positioning since the first week of with my production unit. I don’t use it every day, but it gets turned on and off frequently enough throughout the week. It is puzzling.
Mine had been doing the same. What bothers me is you can tell it’s in a pretty good bind when it does it. Been trying to get a video of it but if I pull my phone out it works fine
My hypothesis: the startup homing procedure first has to determine where the head currently is, which it seems to do with a couple of pictures and a small jog. It then tries to move it to approximately the center, in a single diagonal motion. If it incorrectly “sees” the head over on the right, it tries to move down and to the left, continuously crunching against the end of travel.
I would imagine the lid camera can’t see the corners very well, and a bit of shadow or reflection, especially if the bed isn’t empty, might throw off the algorithm enough to cause this. Most other machines have you-know-whats for this purpose, but they’re unreliable if you break them, so GF has had the courage to think different.
Perhaps feedback from the accelerometers during this process could detect that something has gone wrong and stop it.
The displayed camera image does not show the head. Might be cropped. The head does not hit the Y-rail or anything on the left side. It hits a hard stop on the gantry a full 3/4" away from anything on the left side. When it happens it just looks like something is scraping down the left side. It’s not. Just being driven against the stops while simultaneously moving forward.
The ironic thing is if it is happy to head bang on the left it could simply move left 20" and then right 10" (or whatever the exact measurements are) and be bang under the head.
Some 3D printers actually use that method to home X and Y although I wouldn’t advocate it.
Since the current is programmable it could do it gently. Just drop the current and slowly drive left. Since it parks back left it can bump off the end stops very quickly and then zoom to under the head for fine registration with the camera.
With the accelerometers in the head it could even detect the bump and stop without grinding
With modern computer vision it would be easy to locate the head logo unless the camera is dazzled by something.
So I really don’t know why after two years of development it can’t home reliably and why when there are daily reports of it they don’t roll out a fix post-haste.
Mine has done this so consistently that I’m wired on moving the head every time. I initially read the others posts about moving it below the camera, but I’ve found moving it even just a couple of inches to the right is enough to fix the issue.
To clarify, this is only ever the case when the machine has been off and I turn it back on, but I have never experienced the issue if I do more than one print in a session.
OK. I spent some time trying to figure out why so many people are having problems with the head bumping into the left stops. My machine does not normally have this issue but I was able to reproduce the problem. I power cycled my machine and moved the head to maybe 50 different positions for calibration tests. The following is speculation on my part but seems to give consistent results.
Edit: I had to edit the rest of this post because I’ve done additional tests. Hope it doesn’t confuse the topic flow
It seems that when the head scrapes down the side of the machine during calibration it is because the S/W has lost track of the head. When the S/W gets confused the head could be driven into hard stops on the gantry. (stepper motor assembly) Can only speculate as to the cause. The camera might not be able to see the head clearly or maybe cutouts in the material on the bed may confuse the camera. Might not be one reason. Either way, the solution seems to be to move the head to a position where the camera can easily recognize the head before power on and calibration.
I had originally thought that the camera alignment accuracy could be part of the problem. Now I don’t think so but can’t rule it out.
