MM Bowden Auto lengthtoption  

Re: MM Bowden Auto lengthtoption

Yes, this would be an option. However, the PR has recognized that filament detection is not reliable for all filaments, and has therefore shifted towards disabling it completely.

It sounds like the current direction is to allow Bowden length calibration for each of the 5 filaments independently (which would take care of the issue).

Coordinating the two filament detectors would be another option for filaments that it detects reliably. Hopefully they'll consider that as an option (which can be disabled for filaments that don't register).

Personally, I had to turn off filament detection long ago, and pretty much left it off. It didn't give me significant advantages (I almost never run out of filament, and the 'autoload' while kind of interesting isn't that much better than just selecting 'load filament' on the display), and having it periodically stop my prints because it thought (falsely) that I was out of filament was annoying, especially on long unattended prints.

Re: MM Bowden Auto lengthtoption

I found it rarely went off, except on TPE/TPU's, and liked the autoload which I used a lot.

Turning the sensor on, AVERAGING MANY READS, setting the length and turning it back off makes a WHOLE lot of sense. I'm not sure who to yell that to, but I would like to get it heard.

Just like you couldn't do the X-Y calibration on a single read.

Re: MM Bowden Auto lengthtoption

Having the sensor on continuously during printing leads to false triggering of missing/jammed filament, which is annoying. And it's hard to make work because it has to be right 100% of the time, continuously, for the entire printing time.

Having the sensor only active when expecting new filament, and only to trigger filament loading, is much easier (IMO) because it only has to operate when the extruder is unloaded and waiting for new filament, not during printing, and it doesn't have to detect filament motion, just the arrival of new filament.

I'd be very happy if Prusa found a way to make "new filament" detection reliable even if they couldn't make the "continuous filament monitoring" work reliably. Not so much for loading filament for prints - that's easy enough to do manually - but I believe that it'd make the MMU2 filament loading more reliable.

Re: MM Bowden Auto lengthtoption

I like to see the original filament sensor used as auto loading/unloading sensor this way if it falls short to reach extruder gears or clear them on unload (sometimes it will not pull it out all the way and get stuck). it would solve a lot of issue. But and again i'n not a programmer and i'm sure they thought of it as well. wonder what would stop them from doing it.

Re: MM Bowden Auto lengthtoption

As if it matters, but yes, I continue to support using the optical sensor as much as possible. One could feed until NEAR the sensor, and move the filament back and forth until you get a repeatable signal.

Re: MM Bowden Auto lengthtoption

I know that the optical sensor isn't reliable enough for tracking filament motion continuously during printing, so it occasionally thinks there is a jam and pauses the print. Which is why I, and probably most everyone else, has it turned off, and Prusa has disabled it for the MMU2 loading.

But it occurs to me that just detecting when new filament is inserted into the extruder should be an easier task. First, it doesn't have to be right for the entire duration of a print, but only look when it knows it's loading filament (MMU2) and trigger once per filament load, which means fewer chances to be wrong. Second, it only has to detect that something appeared in front of the sensor, which I am guessing is much simpler than detecting the rate of motion of filament.

So if Prusa could solve the simpler problem of detecting new filament, and make that work consistently enough to help the MMU2 load filament, rather than the harder problem of jam detection, that'd be sufficient for auto-loading, and, even better, it'd likely make the MMU2 much more consistent because the load could be based on filament actually arriving at the extruder rather than estimated filament motion from the MMU2 down to the extruder.

Re: MM Bowden Auto lengthtoption

As if it matters, but yes, I continue to support using the optical sensor as much as possible. One could feed until NEAR the sensor, and move the filament back and forth until you get a repeatable signal.

I strongly agree. The sensor could be off almost all the time, and only on when the MMU2 is actively loading filament, which is a much easier goal than tracking filament motion continuously during printing!

Re: MM Bowden Auto lengthtoption

I'd be happy, as a start, to see it used to auto-calibrate the bowden tube lengths.

Better still, to do it when a print starts - right now I think it puts in a multiplier tied up in what filament you tell it it has. Because if during calibration I put in flex, it never gets there. If at print time, I tell it it's pet but put in TPU, I get missing layers. If I tell it it's TPU, it works fine.

Re: MM Bowden Auto lengthtoption

Unfortunately, the detector is too unreliable even if used only for filament presence detection. I have a blue pet filament from 3dsparks that doesn't register at all. I've never been able to make autoload trigger with it. Nearly evevery other filament I tried was iffy at best. The only reliable one I've found so far is black priline petg, the filament has a distinct textured exterior the seems to help the sensor.

Steph

Re: MM Bowden Auto lengthtoption

I have a few other priline PETg's I could test. While I have found stuff that's iffy, it always works if I just wiggle it a couple times. And just because it didn't trigger the auto-feed doesn't mean there was NO signal. It means the signalling didn't get past the filtering that is done to prevent false triggers.

So if you were in a mode where you EXPECTED to see a weak signal, you "turn the volume up", and use averaging to eliminate false positives.

At least, that's what I'm thinking. And I welcome any responses!

If there's a reasonable debug mode (so I can monitor stuff over serial port or have data displayed live to screen or written to SD card) it would be entirely possible to experiment and see just how likely this is to work.

Re: MM Bowden Auto lengthtoption

Good point about the filtering. Maybe it would be possible to increase the sensitivity for a very short duration when the filament is expected. Probably much simpler than getting the sensor 100% reliable during a complete multi hours print.