The test print I showed in the pictures is here: https://drive.google.com/open?id=1doV6srMX5rTykufwo9jgSxIBOiGrGZas

Printer config can be found here: https://pastebin.com/Zh8gesXR

I took a video of part of the test print with both firmware versions. The outer curves of the print are where the issue is most noticeable, like in the pictures I posted earlier. In the post-1.19.2 firmware video, you can see the printer move more slowly on those curves, with louder extruder jitter, compared to the 1.19.2 firmware.

1.19.2: https://youtu.be/vjKV19ek8iM

Post-1.19.2: https://youtu.be/Q5ceiCHCYeA

The test prints in the videos turned out identically to the earlier pictures in terms of surface finish.

Edit: Is there any other info I can post that might be helpful?

Issue is still present with the latest firmware, 1.20.1RC2. I have to switch back to 1.19.2 to print curves with any quality.

This totally doesn't help you now, but you should turn your stepper current down until they have just enough torque to do their job, and not enough to break things. I avoided similar mishaps on my printer by doing that.

@deckingman:

That's interesting. Can you confirm that those two parts printed with pressure advance and if so, what value you used?

Pressure advance was 0.5; I have a pretty long bowden cable.

I reverted to an earlier firmware version and did some comparisons using the same gcode and config files.

Left side is 1.19.2 and right is 1.20RC4

Left side is 1.20RC4, and right is 1.19.2

In 1.20RC4, the extruder chatters at about the same volume as 1.19.2, but if you look closely at the nozzle you can see it moving unsteadily around the curves. 1.19.2 seems to move at about the same speed on the curves as on the straights, but 1.20RC4's speed drops down significantly.

In the firmware changelog the only related thing I can see is in 1.20b2, which is "The start and end speeds of short segmented moves in accelerating or decalerating sequences are more closely matched where possible, which should give smoother motion". Is that referring to arcs or to SCARA movement?

I'm seeing similar issues, but not just for the skirt/brim.

https://www.duet3d.com/forum/thread.php?id=3745

I tested with an E jerk value of up to 1,000,000. It didn't seem to have any effect either.

I was under the impression that the goal with arcs was to have the resolution and the jerk high enough that the tool doesn't have to slow down when completing a segment, meaning that pressure advance shouldn't take effect at all. Is this right?

I checked the M122 report for buffer underruns, because deckingman's thread mentioned it, but I didn't see any. I'm only using one extruder anyway.

My config file is here: config. Its a simple CoreXY setup.

I read through a couple related threads about this issue before posting, mainly these ones:

Pressure Advance idea for arcs
More strange pressure advance behaviour

I haven't been able to fix my issue, which is that at higher speeds on arcs with pressure advance enabled, the extruder chatters enough to skip steps or ruin the finish.

The model I'm primarily using for testing is exported with 72 segments per circle. Firmware is 1.20RC2.

Movement settings:

M201 X2000 Y2000 Z100 E10000 ; Accel
M566 X1200 Y1200 Z30 E2000 ; Jerk
M572 D0 S0.5 ; Pressure advance

From looking around, it seems that the following things should reduce chatter:

• Lowering speed

• Increasing arc segments

• Increasing XY jerk settings

Lowering speed definitely works, but I have to reduce speeds from 60 to 80 to below 20 mm/s to get the chattering to stop.

Increasing the XY jerk settings doesn't seem to have any effect at all. I've tested from 1200 to 1,000,000, and accel up to 1,000,000 also. There's gotta be a maximum below that? The upside to this is that I found that my printer is way faster than I thought.

I haven't been able to test increasing segments. I can export the model with an arbitrary number, and the STL file size increases as expected, but Cura seems to do some downsizing because the resulting gcode size actually decreases. The gcode files Slic3r outputs are actually shorter than Cura's with the same settings, so it seems that's a no-go too.

Lowering pressure advance helps, of course, but the relationship seems linear. The chattering doesn't disappear like it should with adjusted jerk settings, it just gets quieter.

@dc42:

Pressure advance isn't perfect, especially when the Bowden tube is very long as I suspect yours is.

If you are using E3D or other quick-release Bowden tube fittings, are you using collet clips at each end to make sure that the collets can't move forward ans back 1 or 2mm at each end?

That test print isn't very realistic because fast-slow transitions like that don't generally occur in normal prints (hence that bug hadn't been spotted before). Another test you could try is printing 3 sides of a square, repeated with different amounts of pressure advance. Pick the pressure advance setting that gives the correct amount of extrusion in the corners.

Hey DC,

Even if it's not perfect, it still helps out a lot!

I printed some collet clips and applied them, but it didn't seem to reduce the amount of pressure advance needed. I think the amount of movement there was pretty small compared to the filament springiness; I'd estimate it was ~0.2mm, maybe.

I made a new script that prints the squares you suggested next to the speed transition line test. I found that both tests show best results at the same pressure advance setting, so maybe the speed transition lines are a good test for the real world. They are simple to read, too.

Thanks for the help and the firmware fix.

Hey DC,

I moved the M400 command as suggested.

I updated to 1.20RC2 and did some more testing. The extruder seems to be moving correctly now, but I'm seeing a bit of blobbing in the same spot that seems to be unaffected by changes in the pressure advance setting.

See this picture: https://imgur.com/a/FBhZ4

The line with S0.55 has a seamless slow-fast transition, but the fast-slow transition still has a bit of over-extrusion at the start. What do you think of that? Did you see that in your testing?

Also, I'm questioning the validity of this type of test. I'm not seeing it translate into better print quality. Any suggestions?