RE: VFA Artifacts on X+Y Straight Edges
The main problem I have is knowing what the right spring and damper rates are, and how to set those more methodically than just turning the screw a bit and waiting another 80 minutes for another VFA tower.
Yeah, that's the hard part. I have been spending a little time looking at how to use the accelerometer to find the worst VFA frequencies and come up with some way optimise it algorithmically. An untuned damping system is just as likely to be a resonator. Unfortunately the only option I can see is modified firmware.
RE: VFA Artifacts on X+Y Straight Edges
The main problem I have is knowing what the right spring and damper rates are, and how to set those more methodically than just turning the screw a bit and waiting another 80 minutes for another VFA tower.
Yeah, that's the hard part. I have been spending a little time looking at how to use the accelerometer to find the worst VFA frequencies and come up with some way optimise it algorithmically. An untuned damping system is just as likely to be a resonator. Unfortunately the only option I can see is modified firmware.
The interesting thing about the accelerometer is, I think it’s more important if you build the machine yourself that if you had Prusa build it for you.
I am positive that some of the VFA issues are more related to people assembling the printers themselves as opposed to them coming directly from the factory, already built.
obviously, if you’re modifying the printer yourself then yeah running that is probably a really good idea. Also, I suspect if you move the printer from one location to another, it might be wise to do that if you’ve rerun the accelerometer test previously. How much of a difference it makes, I really have no idea. I know that I didn’t see a huge difference and that the firmware made a much bigger difference…. And of course, using the appropriate print speeds. It’s also definitely the case that if I use a very small slice setting, e.g. .10 or .15, I really don’t see any VFA.
RE: VFA Artifacts on X+Y Straight Edges
The interesting thing about the accelerometer is, I think it’s more important if you build the machine yourself that if you had Prusa build it for you.
I am positive that some of the VFA issues are more related to people assembling the printers themselves as opposed to them coming directly from the factory, already built.
What makes you think that?
When it comes to homing issues, I have seen various posts which describe how errors during the build have caused them: Skewed gantry, reversed motor pulley, cable obstructing Y travel... When it comes to VFAs, I do not recall such examples of simple blunders causing them. And we have this present thread where technically very competent users are chasing the VFA issue without unearthing any simple build or calibration step as the solution.
On the other hand, we have seen a lot of evidence here on the forum that Prusa's in-house builds are far from perfect. Jamming Z drives, rattling XY mechanisms, absurdly high belt tensions... These printers are built by people who work on the clock and can't afford the level of diligence and attention to detail which many kit builders put in.
Hence, I see no reason to assume that Prusa-built printers show less VFAs than those built from a kit.
RE: VFA Artifacts on X+Y Straight Edges
Yeah I fully agree.
Also you've got the added factor of kit machines not being thrown/kicked/dropped on and off multiple delivery vehicles in their final built state.
That and the reduced price were major factors in me opting for a kit.
RE: VFA Artifacts on X+Y Straight Edges
My conclusion from the observations so far is that there are multiple possible sources of VFAs, including belt tension, rubbing belts and bad pulleys. If none of those things apply, it's probably the motors. Early review units used either the old LDO steppers or some of the best new ones, for everyone else it's a lottery now. There were extreme cases of motors so bad that you could feel them being defective when turned by hand, those obviously have to be replaced.
For the "good enough" motors that still show VFAs, it's a difficult story. I believe the situation could be improved via firmware, but the problem is that this requires the accelerometer. So Prusa has to spend money to ship accelerometers (which should have been part of the kit from the beginning!) to everybody affected. This probably won't happen, since I suspect that an integrated accelerometer will be part of Core One S.
The only cheap solution for Prusa is coming up with a solution like modified idlers that lessen the effect of motor vibrations on the print head. Although it doesn't fix the cause but rather the symptoms, it's probably the solutions we will get, together with the promised checklist to find issues in the core XY-assembly.
RE: VFA Artifacts on X+Y Straight Edges
For the "good enough" motors that still show VFAs, it's a difficult story. I believe the situation could be improved via firmware, but the problem is that this requires the accelerometer. So Prusa has to spend money to ship accelerometers (which should have been part of the kit from the beginning!) to everybody affected. This probably won't happen, since I suspect that an integrated accelerometer will be part of Core One S.
The only cheap solution for Prusa is coming up with a solution like modified idlers that lessen the effect of motor vibrations on the print head. Although it doesn't fix the cause but rather the symptoms, it's probably the solutions we will get, together with the promised checklist to find issues in the core XY-assembly.
How about a mobile phone app that records audible noises and reports them back to the printer via WiFi -- to be used in the iterative phase stepping calibration, as an alternative to the accelerometer? 😛
This would avoid making and shipping physical gadgets. And it might actually work if one puts the phone inside the printer, in physical contact with a suitably chosen surface that vibrates along. But the variation between microphones in different phone types will probably make it impractical.
RE:
For the "good enough" motors that still show VFAs, it's a difficult story. I believe the situation could be improved via firmware, but the problem is that this requires the accelerometer. So Prusa has to spend money to ship accelerometers (which should have been part of the kit from the beginning!) to everybody affected. This probably won't happen, since I suspect that an integrated accelerometer will be part of Core One S.
The only cheap solution for Prusa is coming up with a solution like modified idlers that lessen the effect of motor vibrations on the print head. Although it doesn't fix the cause but rather the symptoms, it's probably the solutions we will get, together with the promised checklist to find issues in the core XY-assembly.
How about a mobile phone app that records audible noises and reports them back to the printer via WiFi -- to be used in the iterative phase stepping calibration, as an alternative to the accelerometer? 😛
This would avoid making and shipping physical gadgets. And it might actually work if one puts the phone inside the printer, in physical contact with a suitably chosen surface that vibrates along. But the variation between microphones in different phone types will probably make it impractical.
I wouldn't hold my breath on Prusa being able to develop that, when they've struggled with being able to give us a simple belt tuning implementation.
RE: VFA Artifacts on X+Y Straight Edges
Every phone has an accelerometer, which is probably more sensitive than the microphone. So this could actually work. BUT. Prusa hasn't been able to implement consistent phase stepping calibration using their own accelerometer, so this will surely not happen. Remember: For phase stepping on the XL they put out a large blog post, while on Core One it's a feature for "motor enthusiasts" or whatever they wrote.
For the "good enough" motors that still show VFAs, it's a difficult story. I believe the situation could be improved via firmware, but the problem is that this requires the accelerometer. So Prusa has to spend money to ship accelerometers (which should have been part of the kit from the beginning!) to everybody affected. This probably won't happen, since I suspect that an integrated accelerometer will be part of Core One S.
The only cheap solution for Prusa is coming up with a solution like modified idlers that lessen the effect of motor vibrations on the print head. Although it doesn't fix the cause but rather the symptoms, it's probably the solutions we will get, together with the promised checklist to find issues in the core XY-assembly.
How about a mobile phone app that records audible noises and reports them back to the printer via WiFi -- to be used in the iterative phase stepping calibration, as an alternative to the accelerometer? 😛
This would avoid making and shipping physical gadgets. And it might actually work if one puts the phone inside the printer, in physical contact with a suitably chosen surface that vibrates along. But the variation between microphones in different phone types will probably make it impractical.
RE:
Sorry to pile on @gbmaryland, but I know some people with pre built core ones that have all these issues. The difference is that a lot of the people buying assembled printers don't have the time, understanding or care enough. They generally aren't interested or passionate enough to come onto the forums and discuss it. Having serviced or helped people with dozens of Prusas, most of them purchased assembled, I am quite confident that I assemble printers to a higher standard than Prusa. They almost always come with more than one little issue, a part that is a bit wonky, a screw a over/under-tightened, missing thermal paste (historically), a cable pinched etc.
I buy kits and assemble them for myself and others because I know proper assembly and QA of components saves a lot more time and money in the long run. Assembly issues have been getting worse over time in my experience. Maybe I've just gotten better at spotting them though.
I think VFA probability and kit/assembled are not correlated.
On the subject of using a microphone @jurgen-7. I don't think it would be practically achievable. You've got at least 400 control points for each motor to individually, multiply that by the micro-stepping, with the only input for the objective function for all of them together being if the amplitude of the sound increased or decreased. You might be able find speed up with a time-angular position based convolution to weight the control points optimisation based on where the motors are on the recording however since most of the sounds come from machine resonances from experience this doesn't reduce the search space much. I can tell you immediately one solution it is likely to find that satisfies the null vector is multiplying the stepper current tables by a vector full of zeroes. No movement means no noise. Ultimately noise amplitude is only weakly related to motor position.
RE: VFA Artifacts on X+Y Straight Edges
Thanks everyone for the comments on the "mobile phone pickup" suggestion! It wasn't meant all that seriously, as indicated by the "tongue" smiley -- more intended as a comment on nbu's observation that Prusa is not likely to deploy a solution which costs them money for every printer to be fixed.
RE: VFA Artifacts on X+Y Straight Edges
Carrying a thought here from the tensioner thread - belt damping might be easier to achieve/tune as another pulley applying sideways pressure along the length of each belt (like an automatic tensioner likely would have to work). Just another idea at this point.
RE: VFA Artifacts on X+Y Straight Edges
@jurgen-7 I think it's good to suggest things that may seem crazy/silly and still consider them. I was told by one of the managers who worked at CERN building the internet (Tim Berners-Lee was in his team) told me that one of the most useful practices would have a bottle of wine to give to the craziest idea people had in brainstorming meetings. He said that when you're stuck on a problem, out there thinking is often needed. We all use what they built now, so can't argue with results.
RE: VFA Artifacts on X+Y Straight Edges
Carrying a thought here from the tensioner thread - belt damping might be easier to achieve/tune as another pulley applying sideways pressure along the length of each belt (like an automatic tensioner likely would have to work). Just another idea at this point.
Along that thought I have been toying with the idea to simply add more support points on the longer free stretches of the belts, a simple pulley nudging the belt sideways should effectively shift resonance frequencies upwards by shortening the free wavelength. The theory would be that longer free running belts would cause more resonance at a lower frequency adding support points should both act as dampers and move the wavelength down, adjusting distances between supports could be used to manipulate resonance frequencies. But all this may come with new issues such as how to tune belt tension, for sure plucking belts for an app would need new reference numbers.
/Anders
RE: VFA Artifacts on X+Y Straight Edges
Along that thought I have been toying with the idea to simply add more support points on the longer free stretches of the belts, a simple pulley nudging the belt sideways should effectively shift resonance frequencies upwards by shortening the free wavelength. The theory would be that longer free running belts would cause more resonance at a lower frequency adding support points should both act as dampers and move the wavelength down, adjusting distances between supports could be used to manipulate resonance frequencies. But all this may come with new issues such as how to tune belt tension, for sure plucking belts for an app would need new reference numbers.
Would it really do that though? The belt is quite long, true, but its path is already divided up into several shorter sections by the existing idlers. The distance between the print head and the first idler in the path varies quite rapidly as the print head moves along the x-axis, and there's no indication that the VFAs are dependent on the x-position - rather they're dependent on the speed.
It seems more likely that the entire belt length has to be considered regarding resonant frequencies, since the vibrations along the belt length will be transmitted around the idlers, assuming they're effectively friction-free. Adding another idler isn't going to change that I think. Where an extra idler may help is in allowing a different design that adds some form of energy damping, e.g. friction, in a more controllable way.
RE: VFA Artifacts on X+Y Straight Edges
RE:
Josef has posted a third VFA update post on Reddit:
https://www.reddit.com/r/prusa3d/s/JrXCfVlfdk
Seems like they have some promising things in the works. Very glad that those changes/fixes will apply to the XL also.
RE: VFA Artifacts on X+Y Straight Edges
Along that thought I have been toying with the idea to simply add more support points on the longer free stretches of the belts, a simple pulley nudging the belt sideways should effectively shift resonance frequencies upwards by shortening the free wavelength.
I have stared at the belts for far too long while they are at VFA frequencies, I don't believe the resonance has much to do with with transverse waves. The belts don't seem to actually be vibrating. I have tried lightly holding the belts as they pass through my fingers and they don't feel very different at the problematic speeds. I think that it's more the belt stretching and contracting along it's length that creates the resonances. With my compliant idlers I can feel them moving in their carriers when it's at a resonant frequency so something is oscillating.
I think that your idea would work if you create some tension by having an idler that presses on the belt, displacing it's normal position, at the rear near the idlers that feed into the motor pulleys, that way tension changes will push against the idler and some energy could be absorbed. I have thought of trying this however I would prefer to go to what I believe the root of the problem is and make pulleys that don't cause resonance.
RE: VFA Artifacts on X+Y Straight Edges
Anyone ever notice the wear on the belts where they also rub on something?
RE: VFA Artifacts on X+Y Straight Edges
Along that thought I have been toying with the idea to simply add more support points on the longer free stretches of the belts, a simple pulley nudging the belt sideways should effectively shift resonance frequencies upwards by shortening the free wavelength. The theory would be that longer free running belts would cause more resonance at a lower frequency adding support points should both act as dampers and move the wavelength down, adjusting distances between supports could be used to manipulate resonance frequencies. But all this may come with new issues such as how to tune belt tension, for sure plucking belts for an app would need new reference numbers.
Would it really do that though? The belt is quite long, true, but its path is already divided up into several shorter sections by the existing idlers. The distance between the print head and the first idler in the path varies quite rapidly as the print head moves along the x-axis, and there's no indication that the VFAs are dependent on the x-position - rather they're dependent on the speed.
It seems more likely that the entire belt length has to be considered regarding resonant frequencies, since the vibrations along the belt length will be transmitted around the idlers, assuming they're effectively friction-free. Adding another idler isn't going to change that I think. Where an extra idler may help is in allowing a different design that adds some form of energy damping, e.g. friction, in a more controllable way.
I am not sure, I am just toying with ideas how to affect belt resonance. Like pushing down a guitar string changes the tune pushing a belt should as well. The belt on two sides of an idler is connected so yes they affect each other and that way the full length of the belt is connected but does that mean that it plays the same tune if all idlers where removed? I don’t think so intuitively but I am not sure. But in either case I agree taking energy out of the system by dampening is likely more effective.
/Anders
RE: VFA Artifacts on X+Y Straight Edges
Would there be value in getting a high-speed video (i.e. to slow down the motion) of the print head when VFA is occurring? Would uneven motion of the nozzle be visible?
Prusa MK4 since Jan 2024, Printables: @MikeB_1505898