Vertical banding on COREXY machine
-
@zabana I do think you misunderstood my conjecture, although it may be wrong anyway. I wasn't suggesting it was overextrusion. I was suggesting that rather than a simple pulsing of the extruder, it might be the extruder and XY movements leapfrogging each other. That would cause alternating over- and under-extrusion, but little or no average difference in extrusion.
-
@donstauffer I'm seeing this identical issue on my Railcore II also. Test cubes and such show "waviness" identical to the images @DonStauffer posted. LDO 0.9 steppers and duet2 wifi (my Railcore is a filastruder kit version of recent vintage). Oddly it only seems to happen on X and not on Y. I've hooked up an accelerometer and captured some data with the 3.3 beta and note the difference between an identical simple move in the X and Y direction:
X
Y
These were something like a 100mm move at 80mm/s. I see basically the same thing at 40mm/s also as I recall.
After trying many different mechanical tweaks, and looking at some other prusa forum posts Don had mentioned, I was thinking it's possible it is stepper based and will be trying some moons steppers to see if that smooths and improves anything. It certainly could be how the steppers are being driven by the duet2 as well, though, of course, something I hadn't considered.
A bit of a mystery is why X seems more affected.. both steppers are moving, just in different directions, for a pure X and pure Y move, right? That's why I was inclined originally to think it was a mechanical vibration issue with my X carriage or X rails or something, but all sorts of tweaks and changes to that didn't yield any real notable changes to the accelerometer data.. even being able to make it worse would be an indication that it was mechanical, but I haven't been able to really change things much with mechanical tweaks. I could be just not doing the right thing, of course... I'm a computer engineer so real-world physical hardware is not my forte
Fixing or at least improving this would make a big impact on my print quality, so quite interested in any possible solutions. I will do some experiments with higher speeds as well, that is interesting.
-
@skrotz In corexy, moving Y moves only 4 of the pulleys on the belts. Moving X moves all the pulleys. It is a larger mechanical load on the motors to move the extruder in X than it is to move in Y- try pushing it manually with the motors off and you'll see.
-
Well, this afternoon did a few more test, you know just for the pleasure of havi g the same result always hehehe
I printed at 30mm/s all test and the changes was:
- updated fw to the latest one
- printed 20 , 40 and 80mm cubes
- changed nozzle to a phaetus one
- tried slicing with slic3r instead of simplify
- printed in vase mode
- printed at 210 instead 200 degrees
So I don't know what rlse to try but I am more than convinced that it is related on how on the duet wifi are managed slow moving as it happens on the extruder at such speed.
I also extruded on the air with the printed stopped 50mm of filament at 6 and 1 mm/s and the extruded filament has the pattern I see on the printed objects so it is like the stepper movement at low speeds is not smooth at all or something weird happens with back pressure or something like that...
-
@zabana said in Vertical banding on COREXY machine:
I also extruded on the air with the printed stopped 50mm of filament at 6 and 1 mm/s and the extruded filament has the pattern I see on the printed objects so it is like the stepper movement at low speeds is not smooth at all or something weird happens with back pressure or something like that...
Can you clarify exactly what it was you did for this test and perhaps post some pictures of the result? What happens if you extrude at the rate that is used with your "good" prints, does the extruded output then look smooth?
-
@gloomyandy it is difficult to take a pic of it but here it is. And the test is simple, just did an air extrusion lije when you want to prepare the hotend or simillar, just send thw G1 F6 E50 command on the console and that is all.
-
@skrotz I'm really happy to see the difference in 3.3beta. What RRF version is the top trace? The amplitude seems to be about half and the frequency quadrupled or so. Between the two I'd expect the artifact to be fairly obscure as a result.
What jerk & acceleration were these traces at? And typically how long before the released version of 3.3? I think I saw the RC announced recently.
-
@zabana I'm slightly confused. If the defect is caused by XY moves, why would it show in a midair extrusion, when theoretically only the extruder would be in play? Some kind of interplay between extrusion and XY moves?
-
@donstauffer here are the relevant settings:
M906 X1400 Y1400 Z1000 E800 I60 ; Set motor currents (mA)
M201 X3000 Y3000 Z100 E1500 ; Accelerations (mm/s^2)
M203 X24000 Y24000 Z900 E3600 ; Maximum speeds (mm/min)
M566 X1000 Y1000 Z100 E1500 ; Maximum jerk speeds mm/minuteI played with much lower jerk values but didn't see much change in the accelerometer data. Much lower accelerations seemed like it might have helped but on a test print of a calibration cube I still saw the exact same waviness as at higher acc.
Not sure what you mean by the difference in 3.3beta? I think you misunderstand, both of those images are from 3.3RC3, top one is an X movement and bottom is a Y movement.. I posted both to note the difference I see in X versus Y, and on my machine I don't see the waviness on Y really, just X, and looking at those two samples it kind of makes sense.
-
@donstauffer Maybe I did not explain properly but after all tests I did my only conclusion is that the artifacts are not related to XY moves but extruder moves. Of course there is a little bit ringing but also in the ringing due acceleration and deceleration I can see the periodic patron that gives the extruder. I don´t know why because I am not an expert in the firm or the internal management that the duet firmware do but as far as I can see and understand there is something different in my duet wifi between extruding slow at 30mm/s and 60mm/s that makes not smooth the extrusion at low speeds in straight extruded lines.
Today I ordered a LDO 8 pinion stepper so maybe having more steps per mm (55 insted 44 at full steps) and different inductance and such will help. It will arrive in about two weeks.
-
Seems to me at this point we need to verify if it's a mechanical extruder issue or firmware issue. So we'd either need to change the extruder, or change the firmware.
-
@phaedrux what can I do to change fw? Because the latest test was done with the rc3 and the previous one with the rc2.
Regarding the extruder change it is difficult because it involves a total redesign on the X carriage, maybe I have enough time to adapt it for a bmg again or do you sugest another thing to do and test? -
@zabana said in Vertical banding on COREXY machine:
@phaedrux what can I do to change fw? Because the latest test was done with the rc3 and the previous one with the rc2.
Regarding the extruder change it is difficult because it involves a total redesign on the X carriage, maybe I have enough time to adapt it for a bmg again or do you sugest another thing to do and test?Can you play a bit with the pinch-gear tension while you extrude in free air?
My gut feeling tells me, that the whole BMG-gear pinch geometry is wrong. (on my clones at least)
I pulled filament by hand through the gears of the sherpa mini and noticed, that any increase of pinch ends in a rough movement of filament. The dual gears get pushed together, before I get reasonable clamping on the filament side. The bracket shaft is parallel to the gears, so we can't avoid pushing both sides at once.If I had to re-construct it, I'd turn the idler bracket shaft 90°. That way the pivot point would be along the dual gears and the pinch force would only apply to the filament path.
-
@zabana
Since you are using a sherpa mini too, I think I will reconstruct the front bearing holder with a slightly larger distance between the main shaft and the hole that takes the pinch bracket shaft. The dual gears must not mesh too hard before the filament pinch force is right.
A bit of play between the gears is easier to accept than 'clenched teeth' -
@o_lampe I think I get what you say but can you post a pic or drawing bout what you say exactly?
-
from my experience this issue is a combination of things like geared extruder , short filament path , stiff filament (pla) and tiny melt zone .
try to change one of those and see if you have any improvement . easiest is to change filament to something less stiff like nylon .
first time i saw this type of artifact was when i shortened my filament path by half , all else was the same . -
@zabana Just to add to what the others have said about the drive gears, there are many reasons why a pair of meshed gears won't produce a perfect motion (even ignoring anything to do with how the hobbed gear bites into the filament). I design gearboxes for electric cars so this is pretty much my day job!
First reason is any manufacturing defects. There can be runout errors (where the gear teeth arent concentric to the rotational axis) which will cause once-per-rev errors. There can be pitch errors (where two consecturive teeth might be spaced slightly too close/far apart). There can also be form errors in the involute shape of the gear teeth meaning they don't roll over each other properly. As well as these issues affecting the meshing of the gears, they can also cause other issues (like double-flank contact where the backlash becomes zero and there isn't enough room for the teeth to mesh properly) that might change the loading on the gears & housings (particularly if they're only lightweight printed ones). Obviously on top of this, any errors in how the gears are mounted/aligned in the housings will affect them too.
I would also add that if your gears have come from two separate sources (one from the motor manufacturer(?), one from Bontech/extruder supplier), the gear teeth might not be perfectly matched (may have slightly different module, profile shift, base pitch etc).
The other issue is transmission error. Even if you have perfectly manufactured gear teeth, the teeth themselves will still bend a bit under load (particularly if they're plastic!). It might only be small, but it means you get some once/twice per tooth-pass error in the output shaft rotation. This will vary with load, but it may also interract with any manufacturing errors (e.g. to dampen them out as the teeth flex to compensate for them), so you often find that gears run best with some load on them.
All of these things are also often worse with small pinion gears, as 1% of a tooth pass is much bigger proportion of the shaft rotation, plus some other things on the actual tooth involute form.
So what does this all mean for you....
First thing I'd say is that if your issue is actually with the extruder mechanicals, it may be speed or load dependent. If you found that increasing the extruder flow rate helped improve things, it could actually be the increased load on the gears (from the nozzle back pressure) dampens out the variation source (whether that be because of how the plastic flows in the melt chamber, or something to do with the drive gears). I'd have another go at changing the nozzle temps, but look at going lower so that you build up more back pressure. Maybe you could also try a different size nozzle?
Defintiely play around with the tension spring that pushes the two drive gears tgether as @o_lampe suggests. Having it set too tight will only exacerbate any issues in the gear train
Edit: actually finished writing the post before hitting submit...
-
@zabana said in Vertical banding on COREXY machine:
@o_lampe I think I get what you say but can you post a pic or drawing bout what you say exactly?
Like rotating the whole pinch roller bracket 90 degree clockwise. The new pivot point is close to the dual gears and vertical.
I have to figure out a different place for the thumbscrew. -
@phaedrux I don't know if this counts, but I have two Bondtech BMG extruders, and I tried both with the same results. The one piece which I didn't change is the stepper gear.
-
@zabana The upper few mm of this print illustrate the effect of increasing perimeter speed from 80mm/sec to 120mm/sec, which I accomplished by lowering jerk to 110mm/sec (maintains adhesion on concave curves, but makes them ugly; paradoxically, my corners now look wonderful):
