RE: VFA Artifacts on X+Y Straight Edges
It definitely has the feel of a resonance issue - there are some speeds where the belt ripple really stands out from the speeds directly above and below.
I wonder if anything can be inferred from those specific speeds - this is a test with my 1.5mm belt pitch setup.
RE:
Hi guys, I have also been following this thread even before getting my hands on my Core One Kit (my first 3D printer BTW). I also had severe VFA artifacts when I printed VFA tower found in printables.com. The result was disheartening. I tried adjusting XY pulleys positions to prevent them touching belts or rather belts touching them, no result. Then, I came along to the post by @shushuda link to the post, and tried out XY motor pulley swap with Mellow branded ones. The result is kind of noticeable, but this mod doesn't achieve what @shushuda described in one of his post "very low speeds where there are no VFAs yet". Here are the photos at each angles (top-printed with default prusa pulleys, bottom-mellow pulleys):
0 - degrees:
45 - degrees:
60-degrees:
30-degrees:
90-degrees:
Result: ripples at some speed are completely gone, amplitude of ripples are smaller.
Also, I weighted pulleys of both brands using cheap AliExpress scales. To compensate measurement error of scale by weighing pulleys at least ten times. Here are the results:
As you can see from screenshot Prusa pulleys are heavier than Mellow ones. Surprisingly, the best pulleys are lighter than other which are 1st and 3rd in screenshot above. But, the results are not conclusive as the scale is not accurate 😊
Note: During pick process of the best pulleys and for final full X-axis and Y axis swap I tuned belt tension to 85Hz and checked gantry was square before printing VFA test tower. Used silk black PLA(no band) for printing test towers (didn't change print settings as heatbed temp and nozzle temp was in range of recommended temps on the box of filament).
Next, I think I am going to swap smooth idler pulley touching toothed part of the GT2 belts with toothed idler pulleys with Mellow ones and then try out GT1.5 mod thanks to @chris-hill 's detailed instructions in this post.
Also, @@noizie-works-2 shared his results in his youtube video.
RE: VFA Artifacts on X+Y Straight Edges
I will try finding and installing little rubber grommets around the head of the belt tensioner screw, I can't be bothered solving the Hamiltonian for the system for the best solution but that should modify the effective spring constant of the belt system enough that the resonant frequency should change drastically.
RE: VFA Artifacts on X+Y Straight Edges
I’m itching to know what the ‘official’ plan of action is from Prusa to remove the VFAs.
RE: VFA Artifacts on X+Y Straight Edges
It definitely has the feel of a resonance issue - there are some speeds where the belt ripple really stands out from the speeds directly above and below.
I wonder if anything can be inferred from those specific speeds - this is a test with my 1.5mm belt pitch setup.
I'll have a think, one should be able to compute the effective change in spring constant of the new belt compared to the old, I don't know if that helps anything though. If you can measure the spring constant (force required for a certain change in belt length) then we could confirm the relationship.
RE: VFA Artifacts on X+Y Straight Edges
Next, I think I am going to swap smooth idler pulley touching toothed part of the GT2 belts with toothed idler pulleys with Mellow ones and then try out GT1.5 mod thanks to @chris-hill 's detailed instructions in this post.
Also, @@noizie-works-2 shared his results in his youtube video.
Thanks for the link to @noizie-works-2's video. I checked his channel earlier today, but I was apparently an hour or two too early!
Nice to see his conclusions regarding the GT1.5 conversion, and the results of his belt tension tests. I went up to 95Hz with the 1.5mm belt pitch, and saw a change in the results, but I haven't tried lower tension. It looks like he's suggesting going down as far as 72Hz with GT1.5 - I guess I need to do another VFA/speed tower!
RE: VFA Artifacts on X+Y Straight Edges
@noizie-works-2
Thanks for the video. Your conclusion, that Prusa are unlikely to want to send out 1.5mm belts and pulleys to everyone, both for cost/logistics reasons and because anyone who didn't want to do that conversion would be tempted to return the printer, seems logical. However, it would be nice to think that a) they would not void the warranties for anyone who does the conversion (open source, self-assembly-friendly etc), and b) that they might make changes to the firmware to make it easier to adopt GT1.5, i.e. to avoid the need for the 'M92 X98.44 Y98.44' custom g-code.
Have your contacts hinted that either of those things might come true?
RE: VFA Artifacts on X+Y Straight Edges
@chris-hill After having a think it's not particularly useful. Since the system consists of both motors, the gantry and the tool then that's 4 different masses all coupling with each other, that's a lot of eigenmodes that could occur and so would have too many free parameters to determine anything.
RE: VFA Artifacts on X+Y Straight Edges
@chris-hill After having a think it's not particularly useful. Since the system consists of both motors, the gantry and the tool then that's 4 different masses all coupling with each other, that's a lot of eigenmodes that could occur and so would have too many free parameters to determine anything.
I had no idea how to approach it - I figured there are too many harmonics that need to be considered.
But I have no idea what I’m talking about, and you sound like you do!
RE: VFA Artifacts on X+Y Straight Edges
there are some speeds where the belt ripple really stands out from the speeds directly above and below.
I don't think this is belt ripple at all. I have the exact same patterns at certain speeds, exact same pitch, and I use 2GT belts. Other people in this thread show the same tighter pitch pattern despite also running 2GT, and at the same speeds too.
What you have left is IMO motor resonance artefacting, not belt ripple.
As a side note, this is one of the issues I battle now after I solved my belt ripple with the pulley swap. Absolutely atrocious motor resonance artefacting, so strong you can play a tune by running your fingernail over it. Seems to clear up nicely at 55mm/s tho, but that speed is resonating so loudly I can hear it outside my flat. And this is through solid European walls!
I have a spare MK4 Y motor on hand so I will be trying a swap to see what happens. Prusa promised me a replacement printer for a different issue and then has been ghosting me for almost 2 weeks now, so I suppose I can still fiddle with what I have while I wait...
what @shushuda described in one of his post "very low speeds where there are no VFAs yet".
Just to clarify, those aren't my words and I don't agree with those words. I quoted a different person. This has not been my experience.
However, it would be nice to think that a) they would not void the warranties for anyone who does the conversion
If you bought a kit, you have warranty only for individual parts so any assembly changes like the 1.5GT mod don't void it. The user assembly and the machine as a whole unit are not warrantied in such case anyway.
Different story for assembled units.
RE:
I don't think this is belt ripple at all. I have the exact same patterns at certain speeds, exact same pitch, and I use 2GT belts. Other people in this thread show the same tighter pitch pattern despite also running 2GT, and at the same speeds too.
What you have left is IMO motor resonance artefacting, not belt ripple.
I think it is most likely belt ripple. All the motor resonances I've seen are in the really low speeds, like 10-50mm/s. GT1.5 and GT2 resonances look almost the same - .5mm difference is hard to eyeball with no reference. I suspect at certain speeds/angle moves you get resonances from the teeth over the flat idlers and the motor pulleys reinforcing or cancelling each other out.
RE:
I read a few papers on timing belts in combustion engines, it would seem the same results apply here. Tokaro et al. put strain gauges on individual teeth and find that the tooth engagement and disengagement is the major source of vibration in the system, with secondary causes being pulley eccentricity and belt non-uniformity. Interestingly they find the best way to reduce vibrations is to shrink the tooth size, this explains why 1.5 mm belts show improvements. Scurtu et al. examine the interaction between longitudinal and transverse oscillations finding that tension changes shift oscillations between the two modes, which seems to explain why these artefacts seem pop up elsewhere at different speeds when the tension is changed, including on 1.5mm belts. An interesting thing that regularly popped up is that these sorts of vibrations drastically reduce the service life of the systems, which means it's going to be in Prusa's interest to resolve this since they will have a much higher rate of replacing prematurely worn out motion systems.
Ultimately I think the length of the belt path and the increased stiffness of the core one frame have converged to create coupled longitudinal and transverse resonance issues in the belt-pulley system. If the this is the case the only software fixes I can see working is to modify the gcode or the firmware to disallow problematic angular velocities in the AB motors or some means to actively damp these oscillations in the motors. I think the latter is too computationally expensive but current sensing abilities of the drivers may be able to measure these without needing an accelerometer. Physical changes that I think would work are introducing a vibration damping idler (some papers got good results), moving to a herringbone timing belt as previously suggested in this thread, using a much thicker and stiffer belt to push resonant frequencies higher and out of the operating regime or using less stiff parts in the motion system (mathematically equivalent to a vibration damping idler). I think 1.5mm belts won't work for everyone, it'll reduce the rate of issues but some people will be unlucky enough to have certain combinations of belt Young's modulus due to belt variability, suboptimal pulley geometry and mechanical property changes due to thermal effects (eg belt stiffness might change with temperature) that they will still be haunting people intermittently.
For those interested an fairly decent review on pulley system vibration models "Dynamic modeling, simulation and experiment of power transmission belt drives: A systematic review", https://doi.org/10.1016/j.jsv.2020.115759. Note that this is done for v-belts so the effect of teeth are not considered but system resonance will be the same.
Tokaro et al., Analysis of transverse vibration in engine timing belt, https://doi.org/10.1016/S0389-4304(96)00049-5
Scurtu et al., Coupled longitudinal and transverse vibration of automotive belts under longitudinal excitations using analogue equation method, https://doi.org/10.1177/1077546311418866
RE: VFA Artifacts on X+Y Straight Edges
Physical changes that I think would work are introducing a vibration damping idler (some papers got good results), moving to a herringbone timing belt as previously suggested in this thread,
Both of these are things that I've considered. The herringbone belts and pulleys seem hard to find in the required sizes. The damping idea is interesting - I wonder how you dampen the system without affecting accuracy on corners etc. I also had the thought of using several thinner belts in parallel with their teeth out of phase in the hope of cancelling out resonances.
RE: VFA Artifacts on X+Y Straight Edges
I wonder why not all, at least CoreXY printers, have 1.5 belts?
Many have problems with VFA.
It wouldn't cost the manufacturers any more.
Or do the 1.5 have other disadvantages?
RE: VFA Artifacts on X+Y Straight Edges
@jondoe
You can't really damp it without reducing accuracy in theory, however since the oscilations only seem to be a few micron then you just need something that changes it for a few micron. Think a small elastic tube inside a very slightly larger stiff tube, it will squish until it uses up the volume then the stiffness will rise dramatically since it's now using the volumetric compressibility of the material rather than elasticity. It would need to be adjustable for the belt tension though. Practically I think any loss in accuracy that occurs compared to a theoretically infinite stiffness system plus this dampener will be offset by reductions in vibration induced inaccuracies.
RE: VFA Artifacts on X+Y Straight Edges
I read a few papers on timing belts in combustion engines, it would seem the same results apply here....
Now that's the sort of academic rigour that's needed here! Thank you for this post.
Picking up on the vibration damping idlers - would it matter where along the belt's length these were introduced? There are a few long belt runs where it might be possible to add another idler - this was suggested a few pages back by someone else, I think for the purpose of changing the effective vibrational length. Would an additional vibration damping idler do the trick, or is it better to convert one of the existing idlers? What would a vibration damping idler look like?
RE: VFA Artifacts on X+Y Straight Edges
Now that's the sort of academic rigour that's needed here! Thank you for this post.
Would an additional vibration damping idler do the trick, or is it better to convert one of the existing idlers? What would a vibration damping idler look like?
I was thinking of redesigning one of the existing idlers to include a vibration damping plastic spring in it that has an adjustable tension and adjustable constraint on the range of motion for the spring so that it can only be active for a few micron of motion. I know the composite filaments (pccf,ppa-gf or cf etc) are both good at absorbing a wide range of vibrations and don't creep anywhere near as much so they would hopefully work.
RE:
It's probably worth noting if it hasn't already been mentioned that this is an issue on most corexy machines to some degree. You can see it clearly on prints from the new Elegoo, the latest Sovols etc. Interestingly a lot of Voron builds seem to be free from it. I sold my Bambu P1s because of the belt ripple (example below); it was terrible on some angles (you can see it spread out here as it cancelled and amplified):
RE: VFA Artifacts on X+Y Straight Edges
I also had the thought of using several thinner belts in parallel with their teeth out of phase in the hope of cancelling out resonances.
I was thinking about the dual (or practically quad) belt concept and I think it wouldn't do anything useful. If it is oscillations of the sort I am increasingly confident is happening, then you wouldn't cancel out the tooth engagement and disengagement adding energy to the system, you would just make it twice as often while increasing the number of different modes the system could resonate at as I think the transverse belt resonance space would be an outer product of the two belt resonance spaces. Practically an example is you have one belt oscillating at it's fundamental, add a second and they can oscillate in phase or out of phase. Maybe by changing the locations where the belt moves onto and off of the pulley one may be able to get those forces to cancel out more.
RE: VFA Artifacts on X+Y Straight Edges
That's exactly what Prorifi3D did, though: https://prorifi3d.com/pages/technical-details