zillarob
commented
5 years ago Something to consider is that retract probably doesnt have much effect on the hot goo in the nozzle, beyond relieving pressure to 0. I dont think it can pull lower than atmosphere. I think the make-up air comes from the extruder end and through the bowden, not the nozzle. The solid filament pulls out just fine, but leaves the hot goo in the nozzle. Really not much you can do about this hot goo oozing, other than move faster than it can ooze.
I did draw up a gizmo to see if sealing the the filament in the bowden would help create the neg psi we want. It was noticeable on the spikey retract test part, but pretty minimal. I was using 1.75mm filament and suspect the small area of what would be the piston just didnt do a whole lot. Seemed like I would have to have an excessive amount of retract to create enough displacement to make a meaningful difference. Larger 3mm filament might have had more of an effect with its larger area/displacement.
richfelker
Bumpfish
thierryzoller
Description
A while back in https://github.com/Ultimaker/CuraEngine/issues/1055 for Cura, I requested a feature to perform retraction concurrently with travel. It was noted at the time that this would be difficult without being able to predict the printer firmware's incorporation of acceleration and jerk to motion.
Fast-forward to now: after playing around with Linear Advance in Marlin 2.0 quite a bit, including with very high K values for flex filaments with a bowden, it struck me that Marlin is already doing something very close to the sort of "retraction concurrent with travel" that I asked for in Cura - not for semantic retractions, but rather for negative adjustment of the extruder to relieve pressure when decelerating - and that the same idea could be extended to fully automatic retraction, probably largely reusing the existing math from Linear Advance. That is, the target pressure could just be held at some "negative" amount (not actually negative pressure, since air enters through the nozzle orifice; rather just an equivalent retraction) during non-extrusion/travel moves, instead of at 0.
This should yield significantly faster printing than having the slicer perform retraction, by doing in concurrently with motion, and should also improve print quality/reduce loss of material to oozing by being able to perform retractions in places where slicers don't want to (concurrent with "combing" over already printed lines or infill area, travel between infill lines, etc.) because it would be prohibitively slow to stop and retract.
Nominally, Linear Advance is supposed to reduce or eliminate the need for retraction anyway, but it doesn't, especially with flex materials where the model's accuracy to real material behavior is only so-so, and where the softness of the material has it leaking out even without any pressure.