Inconsistent Extrusion - Githubissues

TheBrigandier commented 6 years ago

Update: @vertigo235 set up a Discord server for this issue, if you'd like to discuss/work with people on the problem. Here's the link: https://discord.gg/hYUjSnW

This issue has been closed after the overwhelming amount of replies, tests, etc. If you are needing guidance, please refer to http://prusaowners.com/wiki/index.php?title=How_to_make_prints_better_on_the_mk3 and Prusa Support.

dielessen commented 6 years ago

Btw Im on 220v. I Dont think Its a voltage thing

MoshiBin commented 6 years ago

FYI, a new release candidate just went out and "extruder linearity correction" is listed as a feature. We might be able to tweak these settings more easily now that they're on an official RC (easier to get testers?) https://github.com/prusa3d/Prusa-Firmware/releases/tag/v3.2.0-RC1

TheBrigandier commented 6 years ago

@moshibin many of us have already been playing with that feature for awhile with our own compiled firmwares. It affects the consistent moire/banding artifact, but not this inconsistent extrusion. I have used the linearity correction to very nearly get rid of the moire (it works), but it has no noticeable effect on this issue. I can confirm the commits since these tests haven't touched anything in that department since our tests as well.

Has anyone tried a cube with absolutely no filament "tug" from above from the spool? Say a test cube while holding the filament loosely above the extruder? I am curious if the extruder is tilting up slightly before the spool holder gives.

TheBrigandier commented 6 years ago

https://shop.prusa3d.com/forum/user-mods-octoprint-enclosures-nozzles--f65/mk3-filament-spool-bowden-tube-feeder-t15550.html#p77647 Bigdogbro1 has a Bowden tube rigged up in a way that it wouldn't have upward pull. Asked him if he can print us a cube.

Matts-Hub commented 6 years ago

@ff8jake I've tried exactly that. I cut some loose filament and just had it resting on the spool holder. Unfortunately no difference

TheDuggem commented 6 years ago

Have extruder temperature fluctuations been eliminated as a possible source? I saw a video on YouTube by CNC Kitchen talking about PID tuning showing that he was getting a +/- 5 degrees C oscillation around the set temperature before running a PID calibration. https://youtu.be/4NlPzp05VwY?t=1m33s

TheBrigandier commented 6 years ago

@TheDuggem Yes, I am seeing this with no temp variance. Attempted a PID tune just to make sure of this. Also attempted printing extra hot to affect this. No change.

dielessen commented 6 years ago

I tried also about all options possible. Leaving only some options :

Hardware design . Motherboard / stepper drivers Stepper motor Firmware specific prusa settings.

3d-gussner commented 6 years ago

I still wait for my MK3 but is there any chance that the MK3 speed setting get the printer to resonate at specific speeds and gcode: Checkout CNC Kitchen Prusa i3 MK2 - Practical Vibration Analysis video https://www.youtube.com/watch?v=Ws1JfHl3Y0o I know the chance with a calibration cube is small but you never know.

TheBrigandier commented 6 years ago

@3d-gussner We tried this as well in the Prusa forum thread I believe. I know I have tried slowing it down to 50%, and I have tried keeping it as fast as 400% with high heat. Very high speeds made a slight improvement (I suspect more consistent pressure on the nozzle?), but at the speeds to get that effect I was up to ungodly amounts of ringing.

Regarding the size of the print, I also did a 180x180 single wall cube awhile back (wanted something to get an Xbox a little higher so the IR sensor was visible in our cabinet lol). Same effect all the way around. Considering the varied amounts of prints showing this (including those with curved varied perimeters), I don't think resonance is to blame; however, that video before did lead me to finding that the MK3 has horrible vibration when moving in Y, particularly so in movement towards the front. If you do a calibration cube at high enough speed to see ringing, you'll notice your ringing is worse on the side the print bed is moving forward on. :)

alfskaar commented 6 years ago

@ff8jake

" If you do a calibration cube at high enough speed to see ringing, you'll notice your ringing is worse on the side the print bed is moving forward on. :)"

Very interesting that have to be magnets or binding.

Magnets slowing down the print taking up the slop/damping (mass damping) off the belt.

Raise the bed.

Read on pleas as I see your post after I write all under this line. I am starting to think that this is a combination off things.

Can someone disconnect the belt from the Y axis and try to squish / move the bed with one finger to rule out any binding or any magnets trying to grip on to something ?

You need to do the test with both the steel sheet on and off.

The weight off the bed make it harder to feel sow as a last step you should remove the heated bed to and only try to move the "Y-carriage"

If there is any binding I will remove the single U-bolt and replace it with the old method zip tiers, if it still bind I will then remove the two on the other side to and replace with zip tiers.

Not having 3 U-bolt under there give you less metal but, if it bind may be more important as I m not sure how close the magnet are.

Next I will consider to use longer spacers and longer screws most off the kit according to the manual have 12mm I will double the length.

depending on length on screws you can print some spaces but they may not last long but if the heated bed is off it will be a start.

Then you need to calibrate it, losing some Z is a small price to pay if this will help.

It is also possible that the magnet is not gripping on to something but slowing things down (inconsistency), that is why I will like to raise the bed.

And maybe some grease in the bearings is needed to.

TheBrigandier commented 6 years ago

@josefprusa @XPila @PavelSindler Guys, this ticket has been open over a week and has over 60 comments, can we at least get a "Yeah we see it" from you guys?

Also, devilhunter on the forums has built two machines from scratch with COMPLETELY different hardware except for the Einsy, magnetic bed, and fans. https://shop.prusa3d.com/forum/general-discussion-announcements-and-releases-f61/i-made-custom-scratch-built-mk3-clones-and-yes-you-t15658.html#p77755 - His machines have this exact same issue, so I think it's likely safe to say this is a software issue or an Einsy issue.

TheBrigandier commented 6 years ago

@ericclinedinst

Great. "Not much we can say about it, when we don't know what's causing it. But we can continue to ship them out to customers and collect their money, so we have that going for us!"

We've yet to find an MK3 that doesn't experience this issue. Stop jerking our chain already @josefprusa, you guys are seeing this on yours just as much as we are on ours. You should be ashamed of your business practices.

GurliGebis commented 6 years ago

@josefprusa @XPila @PavelSindler Guys, this ticket has been open over a week and has over 60 comments, can we at least get a "Yeah we see it" from you guys?

Most likely because as soon as they do that, they have to spend hours upon hours (at 1am at night it seems) to reply to facebook posts guessing what might be the cause of it, instead of working on fixing it.

And since some people seems to be very very impatient as soon as they don't get an instant reply, I understand their silence. (I also understand the frustration, since I myself has been quiet impatient with PR in the past, so I know the feeling)

MoshiBin commented 6 years ago

Lack of communication only seems to sow more distrust. A concise "We're on it" or "It's on our list" does not warrant hours upon hours of replies; just acknowledge this issue so we'll know it will be fixed eventually.

Ultimately we did get this acknowledgement, it's just unfortunate that it had to devolve into that facebook argument.

GurliGebis commented 6 years ago

History has just shown, that once things has been acknowledged, people keep asking for updates. Just look at the Facebook post above, Eric is already asking what they expect might be the issue, even though it is pretty clear they don't know yet.

cjo20 commented 6 years ago

@ericclinedinst can we try to keep this thread for technical discussion rather than harassing staff? The SNR is rapidly deteriorating at the moment.

GurliGebis commented 6 years ago

@ericclinedinst the only thing you are accomplishing is to annoy PR.

Can you explain how asking what Joseph think is the cause is helping?

If they knew, and they told, then instantly everyone would start demanding an instant fix for it. I hope they find a solution soon - trying to get different Youtubers to cover it won't make it any faster, just waist more time.

misan commented 6 years ago

From what I have done and read I reckon the problem (which I consider a mild one) could be connected to:

Bondtech gears: either some of them are defective, or the mount creates a defective working mode, or they pick up debris in a way that does not self-clean
TCM stepper driver: either the drive logic creates a not so linear motion with micro-stepping, torque ripple could easily explain what we are seeing here.
E3D hotend: some turbulence in the melt zone could cause uneven flow

That said, I think that both E3D and Bondtech are well-known products, with a proven record, so that makes it quite unlikely to be the source of the problem.

Increasing micro-stepping may induce its own problems as explained here https://www.micromo.com/media/wysiwyg/Technical-library/Stepper/6_Microstepping%20WP.pdf that might be coupled with certain extruder steppers ... ?? so not all users experience the problem ... ??

Let's see if some test firmware is made available to rule out this hypothesis.

TheBrigandier commented 6 years ago

@misan As it was pointed out earlier, we're seeing this on an MK3 built from the ground up (with completely different hardware, save for the Einsy, magnetic heat bed, and fans). That knocks out 1 and 3 on your list likely. TMC stepper driver is possible, maybe, and we're already seeing linearity correction added to firmware to help with the predictable moire pattern (not related to this, that we can tell). Could still be the Einsy or TMC drivers (or firmware).

Also in regards to torque or skipping steps, I have dropped this down from 32 microsteps to 8. No change, so I took it a step further and disabled the TMC microstep interpolation (looks at your steps and automagically interpolates up to 256 steps), again no change other than the printer got a hell of a lot louder.

alfskaar commented 6 years ago

https://youtu.be/bZGJ9M62L0M Have a look at 06:20 don't know if it can be related.

MoshiBin commented 6 years ago

@misan @ff8jake Additionally regarding the E3D V6 hotend - The MK3 has a special heatbreak marked by a red section in the middle. In the Prusa store it is marked as the MK3/MMU heatbreak. I have replaced my entire hotend with a new one from E3D, with a regular V6 heatbreak (no red section). I can confirm that it did not affect the issue we're seeing in any way - older prints with the MK3 heatbreak have these lines, newer prints with the default heatbreak also have these lines.

jonbet83 commented 6 years ago

Today I replaced the stepper motor and hot end for new, it had no effect on the problem.

I also replaced the x axis bearings for SKF, while that hasnt had an effect on the problem it’s has made that axis total silent.

rackley096794 commented 6 years ago

Here is mine. Note on what SHOULD be a flat surface there are clear ebbs and flows of the filament.

Also notice there is a clear pattern where it extrudes more, then almost comes to a stop in other areas.

Note my prints from ~2 days ago didn't have this problem. So it's clear there is something intermittent about this, probably firmware.

img_20180415_134806

devilhunter84 commented 6 years ago

@ff8jake Progress, perhaps? chris.s39 and me perhaps found something.

See here: https://shop.prusa3d.com/forum/general-discussion-announcements-and-releases-f61/i-made-custom-scratch-built-mk3-clones-and-yes-you-t15658-s10.html#p77928 and the conversation downwards.

Quoting chris.s39: In stock Marlin when you use software auto bed level correction you can choose a height to fade the correction out by (so the z doesn't keep moving throughout the entire print: M420 Z0.2 or something like that) I have no idea whether the mesh bed corrections are faded out after a while in Prusa's firmware or not. If they aren't, then yes the z would keep moving along the mesh for the entire print. If they are faded out, there is a specific point in the print where the z will not move along the corrections anymore. http://marlinfw.org/docs/gcode/M420.html

And quoting myself: Then we have found our culprit/one of our culprits for the extrusion issues. I'm currently 10cm high into a print, and the couplers still wobble from side to side.

The MK2 feverishly locks down the Z motors with a high holding current, so that the motors get hotter when holding then they would moving. And guess what, the old MK2 firmwares does not have this problem.

The MK2.5 firmware, same old board, introduces this problem for the older machines. So a software dev most likely half baked this feature into the firmware versions above 3.1.x somewhere.

I would start here, if nobody has any leads where to look.

c-stewart commented 6 years ago

Hello, I'm chris.s39 on the prusa forums. It would make sense to me that if for some reason the mesh bed level corrections were not being faded out, and the z motors were moving up and down slightly during normal layer printing, that could possibly create artifacts like we're seeing. It would be interesting to see if someone could get stock Marlin configured for a mk3...

c-stewart commented 6 years ago

However, wasn't someone able to run a print without bed leveling, and it didn't help? I think that would negate this theory

austaquio32 commented 6 years ago

fne work good night, after testing and dismantling many times I decided to put the v6 direct to bowden and the impressions to a perimeter were the same, the problem of extrusion is clear, the body of the bondtech is not well designed. I pass photos to you. bowden no bowden work bowden

TheBrigandier commented 6 years ago

@devilhunter84 @c-stewart Alright, you made me dig my MK3 out of the box and plug it back in. :)

Took 3.2.0 RC1, commented out MESH_BED_LEVELING and went through all the firmware errors that popped up because of it until I finally got a working firmware. I can confirm my original 50mm single wall cube no longer performs the G80 it was doing before. Unfortunately, no change. :(

c-stewart commented 6 years ago

Darn. Thanks for all of your continued effort @ff8jake

devilhunter84 commented 6 years ago

@ff8jake Please try it with the tape to the leadscrew method, to see if its still wobbling from side to side if you already have the MK3 out.

TheBrigandier commented 6 years ago

@devilhunter84 What do you mean tape to leadscrew, as in some indicator of movement of the screw itself? At the bottom of the lead screw there are dust covers, and I watched each one for about four or five layers. Zero turn during print on both steppers. Is that what you are meaning?

devilhunter84 commented 6 years ago

Yes, that. Tape works better bc its longer to See movement. Also print something larger like a 80x80mm rectangle (no top/bottom layers, perhaps some grid infill

TheBrigandier commented 6 years ago

@devilhunter84 found some tape, couple inches away from the lead screw, no movement except when layer change happens. Got another test going now with super old b137 firmware, to see if this was a firmware regression. Will report back in a few.

Edit: No change.

itsaryuk commented 6 years ago

JFYI. I tried to decouple lead screw from X gantry using this method:

And it changes nothing, extrusion inconsistency is not related to lead screw wobbling.

TheBrigandier commented 6 years ago

@devilhunter84 Well, I give up again. Things I tried tonight:

Disabled MESH_BED_LEVELING, no change. Confirmed no movement in Z steppers during prints except during layer changes,
Tried Slic3r vs S3D again, no difference,
Tried bumping E motor current up (to same setting as Z), no difference,
Tried old (b137) firmware vs new, no difference,
Tried locking down bondtech gears in case the springs were allowing it to vary travel distance, no difference,
Tried setting very high jerk settings on the off chance maybe some wonky acceleration was affecting things, no difference (other than some more ringing, yay),
Tried looking at other printers online to provoke jealousy and maybe get better performance through damaged confidence, no difference (printer appears to be proud of how shitty it prints).

The only thing I can say about tonight's testing is when I view the bondtech via the little viewing window, it looks like it isn't turning uniformly. I disabled everything I could to make each layer identical, no lin advance, no retraction, no wipe, aligned seams, literally every layer in the gcode has identical moves and extrudes. While watching it print, shining light completely from the side and looking completely from the side, to prevent shadows from messing with my eyes, it seems like sometimes the bondtech is turning fast, sometimes it pauses and catches up, sometimes it's consistently slow. The variance is hard to pick up on sometimes, but I feel pretty confident my eyes aren't fooling me here.

Can anyone else check this, or maybe someone has some video equipment they could use to examine this accurately?

EDIT: Better yet, does anyone have an oscilloscope and experience enough to use it? I bet that would be very telling.

Panayiotis-git commented 6 years ago

Off topic. @jonbet83 The SKF bearings are not more silent than PRUSA bearings. If you not hear the balls rotating, then they do not touch the smooth rod and the contact point is the rubber seal at the two ends of the bearing. If you preload them (press the metallic plates at the outside of the perimeter) then you will hear the balls making contact with the rod and start rolling as you move the bearing.

jonbet83 commented 6 years ago

@Panayiotis-git Thanks for the heads up, you can just hear the balls rotating. but your right I was concerned about the movement of the races but the carriage gives them a nice nip, I also got the version without DL Oil seals so the seals aren't in any contact with the rod.

@ff8jake also more on topic, I managed to get marlin 1.1.x bugfix running on my MK3 this morning and the extrusion issue was still present. I'm not sure what to make of that, there's a picture on the forum.

TheBrigandier commented 6 years ago

@jonbet83 Ah didn't realize that was you. Saw that, good find. Wonder if anyone has an alternative board they could try with this, something else running Marlin but not the Einsy?

jonbet83 commented 6 years ago

@ff8jake maybe a mini Rambo would be good, I don't have one to hand, I did think i could buy one but its £100 test.

or there's a duet but again I don't have one either. I only have a RAMPS with DVR8825 but i'm not sure how easy that would be to hook up. I guess buy doing that it proves all the hardware (Motors, extruders, heaters and thermistors) are all O.K. and all things then point to the EINSY board??

Maybe 24v is having some effect? I have no idea, the MK2 is 12V.

People have really worked hard trying to diagnose this, its getting to the point where now its looking like the EINSY isn't it?

My other printer group I'm in, ANET A8 (AM8 Steel frame conversion) has users who use the TMC2130 Divers on a RAMPS and i've seen some lovely prints coming from them.

I'm guessing we all don't have boards from the same production run and some boards have the daughter boards, are they all effected the same?

prusa3d / Prusa-Firmware

Inconsistent Extrusion #602