Solved Extruder jerk change in RRF3 vs RRF2 for delta

[kazolar]

Difference in M566 for extruder between RRF 2 and RRF3 ?

I had moved my delta over from RRF 2 to latest RRF3, and after doing input shaping and all other calibrations, I noticed an odd behavior with cornering and circles.
The previous instruction that I've seen was to set M566 for extruder high so it won't impact any operation, and that worked fine in RRF2 -- I checked my other duet printer and M566 is set to 1500 for each extruder and no issues.
On the delta, I've noticed gaps and missing lines -- and not due to skipping steps, but the head just flying around corners so fast that it doesn't even have time to lay down filament. My other M5666 settings are conservative at 300 for each tower stepper. So I'd expect corner not to be this aggressive, yet it is very aggressive, much different than the pre upgrade. Lowering this jerk setting for extruder from 1500, down to now 300, has made these corner and circle moves less jerky -- and more as to what I expect. My accelerations after input shaping are up to 6500, and that works fine, but something has changed in the jerk behavior, and I can't seem to figure out what. I replicated the same exact behavior where essentially XYZ tower jerk is ignored with different slicers, so it's definitely a change to jerk policy -- do xyz jerk on a delta do anything, is it all controlled by the extruder jerk?

[Phaedrux]

@kazolar said in Extruder jerk change in RRF3 vs RRF2 for delta:

My accelerations after input shaping are up to 6500, and that works fine, but something has changed in the jerk behavior, and I can't seem to figure out what.

I think it's more to do with your accelerations. What were they before?

[kazolar]

@phaedrux 3000, but I tried the same prints with lowered accelerations of 2000, 3000, and the cornering behavior of flying around circles persisted, the only way to get it to stop doing it was to lower extruder jerk

[kazolar]

@phaedrux what's more suspicious is that I had rebuilt this delta (it's big -- 500 mm diameter 1.5m tall) with new linear rails, SLS printed carriages and effector, 400 step nema 23s for the towers and I played around with trying some high speed prints to see what the new hardware could do before the move to RRF3, and I had set accelerations to 30k, and jerk to 1k in an speed benchy exercise, and i didn't see it act like it was acting after the move to RRF3 - the magnetic ball couplers can't handle G forces to push the jerk high enough to really print a super fast benchy, but 12 minute benchy that looks reasonably decent was possible. Clearly something is different with jerk behavior -- and I've been using P1 M566 policy with RRF2 as I am now.

[bberger]

@kazolar are you sure you're talking about E-Jerk?

I have stupid high E-Jerk (settings from official Orbiter v2 docs) and it has no influence whatsoever on my curves (XYZ@300). They're slow and "jerky" (for the lack of a better word) if my accelerations and speed is high enough.

On my 380mm Delta with my Moon's motors the highest acceleration it can handle at the outer edge of the bed is 10k@500mms before skipping steps, so that's my hard limit.

I print infill at 400mms@10k, 1st layer at 150mm@3.5k, inner perimeters 300mms@5k, outer perimeters 150mms@5k with mzv shaper@42Hz and XY jerk @ 300. And my circles sound like my printer is falling apart. I need to up the XY jerk to 600 or more to get stutter free circles, but then the effect of the input shaper gets noticably worse.

[kazolar]

@bberger I think this might be the fix. I increased my jerk on xyz and e to 1200, and so far that appears to be behaving much better. I am still doing tests, I also lowered pressure advance as it seems 0.02 was making more bulging corners than 0.01, which seems counter intuitive. Looks like there is something in the motion planner that results in essentially missed extrusions if the xyz jerk is not high enough when combined with input shaping -- that's my best guess. Still need to run more tests -- will sacrifice a couple of spools of cheap PLA to the testing to get this figured out.

Spoke too soon -- test just showed basically no extrusion as it's cornering -- lowered jerk, and it works fine again

[bberger]

@kazolar what's your acceleration on the first layer?

I had a similar "issue" yesterday. While my printer happily lays down 1st layers at 300mms@5k if it's straight lines and sharp corners it really struggled on a part that had a series of S-curves in them, looking like this:

|~~~~~~~~~~~~|

It happily laid down like the first 8 or so curves before skipping the last 4 and then instantly recovering on a 90 degree corner. This was repeatable independent from orientation and place on the build plate.

I then tried the same with a sharp cornered zig-zag pattern:

////////\

This was laid down without any problems.

The only thing that helped with the wavy pattern was to lower the acceleration to 1500, limit the speed to ~75mms and cut back pressure advance from 0.060 to 0.040.

I'm not sure if this is a bug in motion planning (I don't think so) or just an unfortunate combination of various parameters and a less-than-optimal shape.

[kazolar]

@bberger i think I found the issue. Part mechanical, part software. One of the upgrades I did was move the extruder really close to the hotend -- so the bowden tube is now maybe 25 mm tall, so my previous .02 pressure advance was basically messing up and not extruding where it should be. I tried .01 -- seems better, at least it's not - not extruding, where it should be. The other bit was my extruder sherpa mini which I rebuilt, started had a problem in the MJF glass nylon parts -- replaced with printed ABS body, and like night and day, not perfect, not yet -- I think the stepper got a bit damaged in the process, so needs to be swapped to, and I'll swap all the internal bmg gears to be safe, i think the nylon gear got a bit worn out, but looks like it there was an issue of binding maybe in repeated retractions, and the gears would slip, so the big nylon gear needs to go -- seems the mjf body is not stiff enough or some other property. Pretty sure after I replace all the extruder parts, it will be printing perfectly.