Print destroys self

Basic Information:

Anycubic Mega Zero 2
ATmega1284p & R-pi4B V1.4 w/4GB
Klipper: v0.11.0-41-g9b60daf6
Octoprint 1.8.6
Python 3.7.3
Octopi 0.18.0
Filament: Overture PETG 1.75mm
klippy.log (1.3 MB)

Describe your issue:

Since recently adopting Klipper I’ve been wrestling with its configuration. I’m now at the point where test prints like the square tower used for PA analysis come out great. First layers = great. Calibration cube = great. Now try to print something a bit more interesting and larger and it starts off great but eventually turns into a scene from a B sci-fi horror flick. The darn thing starts to eat itself and it seems to be coincident with creating infill.

List of things that did not help during three print attempts:

  • Increased FADE parameter to16
  • Removed FADE parameters from BED_MESH
  • Re-ran all bed leveling procedures
  • Adjusted Z-offset during print when the horror scene started
  • Re-sliced model for slower print speed by 50%; 30mm/s

Out of ideas. Seeking more knowledge…

Your pressure advance value is extremely high. Try without and see if the issue persists

True, but that’s what the square tower results were. The square’s factor was 0.050 rather than 0.020. I am concerned about why the results were so high on this machine and know that it is probably indicative of some other issue that needs to be defined and investigated.

Well, the biggest influence is a bowden extruder. Bowden length, bowden quality, quality of the extruder in general and filament play a a role as well.

In any case:

  • PA is putting additional work / stress on the extrusion system
  • The higher the PA value, the more work / stress

Simply set PA to 0 and test again. If your problems go away then your extrusion system simply is not capable of dealing with such high PA values.

PA-Function is something like the “best guess” of a spring-force. “Spring” includes the length of the pressed filament, the elasticity of the bowden tube, actual temperature, material “pressing” strength and the process of “printing with a high dm/dt” (=change in matrial flow) as Sinoe already mentioned. Last point is very well adressable, if the “spring” behaviour is neglectable. Using a bowden extruder, you have that unfortunately. You may reduce it by using a 1.9mm Bowden tube instead of 2.0mm, cut the tube as shot as possible and so on (see list above).
The thing is, that PA works quite nicely, if your “spring” is short, else you have to guess and adapt. If I see your PA Tower: Yes, the upper corner looks nicest but “loading” the spring takes time and obviously during the infill, you do not have this time.

I would follow Sineos advice and use a PA of 0,025-0,03 (and reduce maybe the speed of the infill) and test if you see a change. If not, we are all on the wrong path…

We have to live with things like PA as long as someone integrates a pressure sensor in the heatblock. That day, we can change to a pressure-driven mathematic (mainly a PID for the pressure right before the nozzle and a viscosity and thixotropy model for the filament) which makes PA needless. But maybe this will never happen (price, space, complex maths, maybe our technology today is already good enough for printing 100mm/s with 0.6mm nozzle).

Turned off PA and did a series of little tests to get the regular old retraction dialed in again. Had lots of acne and blobbed/lifted corner issues too. Got those dialed in. The over-exuberant PA was indeed the root cause. Infill no longer tries to tear itself apart. Adding PA of 0.20 back in helped to relieve the last of the acne.

Re-ran the job that started this thread and about 90% of the way thru got a MCU disconnect error. Trouble is I was away so the printer timed out, cooled and the print disconnected from the bed. Augh, couldn’t save it. I hope my Rpi isn’t dying.