PA tuning Disparencies

petriheino

Re: PA tuning
Re: Enhancing pressure advance

Hi,

After countless trials with configurations, I can also confirm having similar difficulties in PA tuning and retraction tuning as other users on this forum.

By tuning extuder jerk, retraction distance and micro stepping resolution (16x to 256x) Ive been able to make the start point from missing material to having too much material (dot) with direct drive extruder.

The next segments are also effected - if a dot is formed, it measures 0.5mm, and the next lines measure 0.3mm, after it settles to 0.4mm after few millimeters of printing.

Another forum user suggested dynamic unretraction constant. But using firmware from mdealer fork didn’t seem to help.

Similar thing happens when attempting PA calibration method of Marlin PA1.5 so I tried injecting M572 D0 P0 commands to the gcode at G10 and M572 D0 P0.04 at G11. This helped in line test but not in round shape.

Also, when using low retraction amounts, they work nicely in calibration, but in real print start randomly stringing, leaking alot when retracting and moving (when printing larger areas). Large retraction amounts make gaps. Tested with both Simplify3D code and Cura code with absolute, relative and firmware retraction modes.

Could we make some changes to firmware where extruder or filament amount is reset at the event of retraction / unretraction?

Other possible solution would be if the DDA would detect that this is a retraction move, and all corrections should reset when the retraction event is noticed. What would firmware developers suggest?

petriheino

By doing more testing with Retraction distance 0.6mm and retraction speed 1800mm/min, the gaps increase if adjusting pressure advance from 0.04 to 0.07.

Could this issue be related to extrusionPending inside dda.cpp?

Because otherwise print quality is very good but starting point always fails by either missing material or having too much.

Would it make sense to add anything like this to end or beginning of DDA::Prepare etc?

if ( moveBuffer.hasPositiveExtrusion && axesMentionedExceptZ.IsNonEmpty() )
{
extrusionPending = 0
}

petriheino

Hi,

@dc42 could you please comment on this?

I was wondering if this problem could be related to other known bugs such as this one:

Github BugList.txt: Don't do pressure advance during accel/decel of sequences of short segments [LINK]

Could other kinematics (such as corexy vs cartesian) be more or less affected?

bot

I think the Duet team is in the midst of investigating and working on improvements to PA. It might be best to hold tight and wait for the announcements. IIRC RRF 3.4 is the earliest we'll expect to see anything.

o_lampe

@petriheino said in PA tuning Disparencies:

a dot is formed

I've noticed the dot, too. After printing several skirt lines, it retracts and builds a dot at the landingspot.
I blamed my Volcano hotend, because I print the first layer really slow and hot. The Volcano doesn't like that.

petriheino

Now that the RRF 3.4 is coming out is there any developments on this that would have discussed by the developers?

I suspect that this forum post may be related Small intermediate retractions detected with stepper analyzer.

dc42

@petriheino we have a researcher working on investigating the effect of PA and whether switching to a nonlinear (speed-dependent) PA or making other changes would be more effective. We probably won't have results in time fore the 3.4 release.

What type of hot end are you using, and what amount of PA are you using? If you are using a Bowden extruder, how long is the Bowden tube? What type of extruder drive are you using?

petriheino

@dc42

Hi, reference style testing has been made with D1.75 PLA filament, E3D nozzle D0.4mm and Brass and Stainless steel Hobb-goblin gears, all metal extruder with a gap limiter, spring force approx 2kg adjusted to set gap of 1.3mm with CNC machined Aluminium all parts and a steel ball bearing idler. Inlet side fully a PTFE tube, very low pulling force. Distance to nozzle approx 40mm from gear. Cold-end side measuring 36-42°C and hotend 205°C PLA. PA range tested 0 - 0.16, and for PLA it works best between 0.04 and 0.06 only (still remaining dimensionally accurate enough).

What makes it worrying is that if looking at motor movements with slow-motion camera, it appears as if motor is not moving at start after un-retraction moves. But this will surely need to be checked with a encoder still. Please note that in middle of print, performance of PA seems to be very good in corners, and perimeters, but all problems seem to be related to retraction points and infill start-end points.