Another take on ringing

[foehnsturm 969]

Could a few people share short (ideally < 1hr) real life prints where they experienced serious ringing issues with the UM3 and the settings they used? Or the other way round, prints that suffered from the massive reduction of accelleration and jerk (rounded corners etc.)

I'm testing a UM3 mod and the usual ringing tests  show some  improvement. But real life prints might be a whole different story.

[SandervG 1,521]

Did you already get some models to test your mod on? ?

 

Can you say a little bit more about what you're thinking off? I'm curious! 

[geert_2 557]

I took a quick look, and in the models I have laying around here, a simple plain cube or rectangular bar shows ringing best (thus worst, with the most defects): 10mm x 10mm x 10mm, printed 100% filled. Or 15mm x 15mm. See the white dummy bar in the image below. This is where cooling is on the edge of being sufficient (and clearly insufficient in the cones).

Things that suffer from slowing down in corners are mainly tiny text: caps height 3.5mm, line-width 0.5mm, character-width ca. 2mm. Text raised 0.2 to 0.5mm above the surface. Here the ends and corners of the characters get visibly thicker. You can find test character sets on my page:

https://www.uantwerpen.be/nl/personeel/geert-keteleer/manuals/

The zip-file includes the design files (RSDOC, DesignSpark Mechanical), and STL-files with raised, recessed, and hollowed-out test texts. The hollowed-out are for transparant materials only, of course.

 

All printed on an UM2 (non plus) in PLA, 0.1mm layers. So I don't know if this applies to an UM3 too,but you could try.

 

 

 

 

[foehnsturm 969]

Here is what I found so far: 

• The resonance (ringing) frequency of my UM3 is around 26 hz;
• corner bulging and ringing is affected by wall speed and jerk;
• you can mostly ignore acceleration, as (at least with usual settings) it happens in a different time scale.

As commonly known, the bulging happens due to overpressure in the hotend. This will inevitably (pressure advance algotrithms could probably help here) build up when printing with more than very modest speed (like 20-30mm/s). "Slow" deceleration doesn't help to release the pressure, because our "slow" like 500mm/s2 or even 100mm/s2 is way too fast. Deceleration still happens in a timescale of less than 0.1 sec, which is one or two magnitudes too fast to show any helpful effect on the extrusion system.  

 

So, if you want to print with more than very low speed, you're like going downhill in a car with very weak breaks. If you want to go fast straight you have also to corner fast: I got the same perfect, almost bulging-free pre-cornering surface with standard 20/5 speed/jerk as with 50/30. With a speed-jerk difference above 20 the bulging starts, irrespective of acceleration (tried from 100 mm/s2 to 2000mm/s2).

 

But fast direction changes will produce ringing after the corner. The printhead and the belts form kind of a resonating mass-spring system. The amount of "bad" energy stored there, again monstly depends on the speed difference between the two adjacent paths, which is defined by the jerk setting. However, there's is some positve effect (less stored ernergy)  when decelerating very slowly but this is ruined by the increased bulging.

 

The ringing may possibly be reduced by some kind of hardware measures for low frequency damping. E.g. somehow decouple the belt vibrations from the printhead.  Another approach, which could show some positive effect, is to measure the ringing frequency (print speed / distance between the ripples) and find a specific deceleration rate as explained in the Duet3D forum.   

 

Left (standard): wall speed 20/30mm, acceleration 500/1000, jerk 5/10  |  30 min

Right: wall speed 40/50, acceleration 526/789, jerk 20/20 | 22 min

 

 

 

Edited September 24, 2018 by foehnsturm

[foehnsturm 969]

Well, at and least in theory, you could counteract against the ringing with smarter slicing.

If you know the resonance frequency / wavelength, the slicer could alter the path in a way that e.g. at at corner it stops the moving axis like 0.1mm earlier to compensate for the belt stretch and then add this 0.1mm to the axis position again after half of the resonance wavelength to compensate the back swing / contraction of the belt.

Edited September 23, 2018 by foehnsturm

[foehnsturm 969]

Printed at 50mm/s, 30mm/s jerk; bottom half: 20mm square, top half: same square with tiny bulging / ringing corrections. Could a slicer or even a post-processing plugin apply those corrections to the path in a generic way? I think yes!

Edited September 24, 2018 by foehnsturm

[neotko 1,417]

1 hour ago, foehnsturm said:

Printed at 50mm/s, 30mm/s jerk; bottom half: 20mm square, top half: same square with tiny bulging / ringing corrections. Could a slicer or even a post-processing plugin apply those corrections to the path in a generic way? I think yes!

 

If I get your idea, basically you want the slicer to do motion dampening acceleration by already assuming the ranges on which the printer works? 

 

Something like this? Ofc the slicer should take into consideration the motion of tge board an compensate accordingly?

 

Sounds like a great idea. Anyhow I see a real issue of the base of how the very old motion planner executes the extrusion. For perimeters (and 90%) of the thingiverses should indeed work. 

 

The slicer should then add an AI layer to distinguish between the big outside areas and the small details, to avoid getting under/over extrusion due extruder planner. Also, old atmegas would suffer BIG time from continuous changes on speeds to mimic a motion dampening.

 

 

[geert_2 557]

Maybe you could print some sort of dampened zigzag line for testing? Thus a sort of dampened triangle wave?

 

Let's say the zigzag amplitude is 10mm at the start, and then very gradually reduces 0.1mm or so?

 

If there would be resonance effects, they might show up?