I’ve got the otherwise excellent Bondtech LGX Lite, which seems to have a considerable amount of backlash, probably because of the many gears it has. I’m having a bit of problem getting it to play nice with PA. I was wondering whether the backlash could cause some errors. It would seem like it should be possible to compensate for backlash in software, but does Klipper do that already?
tl;dr: if PA is causing issues, use less of it.
No, Klipper is not doing any backlash compensation. PA is only a linear correction for spring-like effects of the filament.
While continuously extruding, the filament should remain compressed and the gears contacting on one side of the teeth, even during cornering. So in theory, backlash should only occur during retractions.
However if you overdo PA, the pressure might drop enough during cornering such that backlash comes into play. This will happen during tuning with a tower, producing a more or less large range of PA values where you cannot observe any change in bulging. You’ll have to keep this in mind and choose a conservative PA value slightly below the low end of that range.
Friction in the filament path or gears might exacerbate the issue. If you have some and cannot eliminate it, you’ll have to live with a lower than ideal PA value, lower print accelerations and higher smooth_time. (although the default 40 ms should be plenty for a direct drive system). Given the hardware limitations, it is more of a compromise than exact science.
Ok, thanks for the clarification! So PA should never be turning the gears backwards enough to pull the filament back, i.e. the extruder should never exert negative force on the filament. Somehow I was not expecting that, but good to know, and it makes sense if the assumption is that the filament is a linear spring with no mass. I will need to figure out what else is then causing the filament to not start extruding fast enough after a move with a retraction.My tuning towers have 3 cornes behave one way and the fourth is way different, and prone to tearing. Adding “UNRETRACT_EXTRA_LENGTH” seemed to help, but somehow I feel that is a bit dangerous, as I believe the extra lengths could add up and could lead to overextrusion in areas where there are lots of retractions. (Btw, lowering the smooth time to ridiculous values (4ms) also helped… My LGX and LGX lite extruders seem quite happy with those smoothing times, although it could quite well be that I just I am yet to notice the downsides with such low smoothing times.) I will re-build my filament path with new components, it could well be that I have a mechanical issue here. Thanks for the tip!