Just to add this one

I think you might be on to something.

On this PRU, the place where the head gets parked after a print is normally about half an inch to the right of the stop. I also think that’s the natural home position for it.

The only time I’ve had the whacky-whacky happen is when I have accidentally moved the head all the way to the left against the stop. I think it’s just expecting to start about half an inch further over to the right. (I move it during cleaning, or if I have to stop a print for any reason.)

The whacky-whacky will continue though until the whole shebang gets reset by turning off the machine, centering the head under the lid camera, and letting it reset itself once. After that, I can turn off the machine normally, and it starts itself back up properly over on the left side of the bed.

Like I said, I don’t see it often, but I also don’t move the head by hand, even with the machine turned off. If I just don’t mess with it, it’s not usually a problem for me.

2 Likes

Ok, folks. I’m posting what I’m going to call the definitive recording of the Glowforge startup and homing procedure. Including a variety of tests which may debunk some of the theories posed earlier in this and other topics. I didn’t go into this expecting to prove anything, but I think the evidence suggests that my earlier guess is still the most plausible.

I’m very curious why rpegg’s machine apparently does an initial move down and to the left, when mine moves toward the center on both axes.

Anyway, it’s all here for the record. Pixels don’t lie:

7 Likes

Everything you showed mine does also. It doesn’t always move straight down the left side. Mine does both motion paths. The one where it moves directly toward the head camera and the one where it moves almost in a straight line down the left side but also moves toward the left stops. The one that moves toward the center never scrapes. The one I described always scrapes. Don’t know why it takes one path or the other. Might be light interference, the lid camera might see the head but be confused by cutouts in used material. It clearly has lost track of the head. Doesn’t matter to my description. When it scrapes down the left side it is hitting the stop on the gantry motor assembly which creates the sound everyone is complaining about. I can also give you a setup where it will always hit hard stops on the right side. Nothing I said conflicts with your video or your earlier statements.

BTW: you can feel the ticking noise you described in the head so doubt it is the pump.

2 Likes

That’s how they light the laser. :wink:

9BCDCE86-D25B-40FC-B5AD-4AC773C67AB7

11 Likes

ROFL! The clicking sound is part of the normal startup calibration sequence. (Rita’s mentioned it a couple of times now.) :smile:

(Although, have to admit, it would be more fun if it was a firestarter.)

3 Likes

You missed testing when people try turning on the Glowforge with material sitting in the hopper, and allowing it to try aligning.

In fact, I’d like to see a piece of light colored material sitting in there (like maple), with a box cut to the rough dimensions of the head assembly, and then see the startup procedure take place.

If you put the cut hole in the left-hand front half of the hopper, but leave the head in the back/left park position, I suspect that’s when you’ll get your head crashing into the back side and moving right. Because I think the dark hole will trick the initial calibration move into thinking it needs to move backwards/right to make the head visible for the successive centering moves.

Which brings me to another question about people reporting head crashes against the side: are people leaving material in the hopper when they try rehoming?

3 Likes

The ticking from the head at the start will be the homing of the focus stepper motor.

I wonder if there is half an inch difference in the lid camera view at the extremes between machines, so some people’s machines can’t see the head when it is parked and that is why they don’t home reliably.

@chris1’s your video looks like your machine copes with any starting condition but you also mentioned it occasionally head bangs. Can you remember what was different? Did you have material on the bed? Do you have bright lights or sunlight shining on it?

2 Likes

I have to go by memory, unfortunately, so this isn’t very reliable. But as I recall, it has happened a few (3-5) times, some of them with material on the bed (because I know I’ve removed it for the second try). I can’t say for sure that it has happened with an empty bed and no extraneous light, but I think it has. In any case, it’s not repeatable for me. Most of the time it works just fine, even with stuff on the bed and all the lights in the lab on.

Vaguely related anecdote: I have a very complicated toaster that also has a camera inside and unnecessary machine vision features. About 40% of the time, it is unable to correctly count bread slices. I’ll put in two pieces and it’ll say 6, or 3. Or it fails to detect the bread at all. I think we have a ways to go yet.

2 Likes

I’m now curious, too, since mine differs from yours as well! Mine goes down to left-center, then straight right to under the camera. I’ve gotten the headbang (down/left) a couple of times. But typically, it’s a straight shot down, then a straight shot right.

That said, I didn’t watch that action over the weekend, so it’s possible that motion has changed since the last time I watched.

Funny if they fixed it just before @chris1 made his video.

I don’t know why they can just say what goes wrong when it headbangs and why they can’t fix it easily.

1 Like

As I said I get both types of motions. For a long while only the move to center under camera was the standard move. But it changes back and forth. When I was testing yesterday and it was consistently moving down the left side, I did have material on the bed with head sized holes in it. It is quite possible that the lid camera was confused as to where the head was. Maybe the 1/2" travel to the left could have been an attempt at a 5" travel to the left if the camera thought the material was further to the right. Again this is all speculation based on one machine. If I get bored I might play with it some more . If the results change I will edit my post above.

3 Likes

OK. Looks like my thought that there were a couple standard calibration motions may be wrong. The left side down motion was the only one I saw during more than 50 tests last night but for that test I also had a piece of material with holes on the bed. The material may have been confusing the calibration routine. So lets take the camera out of the equation. I removed all material from the bed. I made sure that the lid camera could not see the back of the bed by blocking it with tape. After that I got a completely different camera motion depending on how I positioned the tape. In all cases I made sure that the head position was obstructed from view but the camera could still see a majority of the bed. So the only thing I can take away is that if the camera is unable to see the head or the machine gets confused as to where the head is during power up the calibration routine may run the head into it’s stops creating the ratcheting noise.

So the solution is the same for those that experience this. Move the head manually before startup to a place where its obvious to the lid camera. If your machine does not have the problem, don’t move the head manually.

3 Likes

I edited my speculation after further tests.

1 Like

Looks like there are two phases, one looking for the head as a black object and then looking for the logo when it is under the head. It is easy to see how it can get confused if it can’t see the logo at the extremes of view. A black box over a black crumb tray is not the easiest thing to find optically. I wonder if the switch to a black crumb tray has made it less reliable for new machines.

I also wonder if the camera view at the far extremities varies a lot between machines.

3 Likes

Given the differences in camera accuracy and focus quality it seems possible. But don’t really know. Have certainly been wrong with a lot more evidence.

As far as the crumb tray goes, I’m not sure it matters in the long run. My silver crumb tray is now black. And I have taken it apart and cleaned it back to silver twice.

2 Likes

Great video – but @dan_berry has a good point. everytime mine has had the problem, material was sitting in the bed.

2 Likes

So there aren’t limit switches on the gantry so it “knows” when to stop? the XY table/arm on my 400,000 dollar Cell Sorter, has limit switches on it so that when it powers up after a reset it sweeps from one end of travel to the other.

The first rule of Glowforge Club is we don’t talk about limit switches.

6 Likes

The very first bed comparison is probably made between it and a stored photo of the head in the back left corner. From there it would make a speculative first move to swing the head/logo under the camera, and continue doing successive repositioning until it’s within centered tolerances.

There’s ample evidence that fringe areas where there’s significant dewarping of qr codes is unreliable, so I see no reason to expect accurately sensing the Glowforge logo at far ends would be any more accurate.

I’m wagering that almost all the reported head crashes are the result of leaving material in the bed when powering up, and that the changeover to a black crumb tray was just as much about reducing optical alignment errors as it was about appealing aesthetics.

3 Likes

LOL, What a great movie that was. :slight_smile:

1 Like