[3.5b1+] Inaccurate print dimensions

Philippe71

I have exactly the same problem with a Duet3-6HC + exp3HC + 3.5b1+.
Same problem as on the pictures.
I went back to version 3.4.5, everything works again (same GCode, same configuration).
Currently my machine is printing, I will switch back to 3.5b1+ and send an M98 P "config.g" and M201, M203, M566 and M906 when it is finished.

dc42

@Diamondback it looks to me that at some layers Z has not moved as far as it should, hence the extrusion bulges. Can you tell if it is all three Z motors that have not moved, or just one of them? If you can't tell, are you able to print e.g. a large hollow cuboid so thay you can tell?

@Philippe71 do you have a single Z motor, or more than one?

Diamondback

@dc42 The bulged lines go around the whole print at the same heights, so I would guess that it's all 3 motors? Do you want the cube in vase mode or normal? What wall thickness?

dc42

Hi @Diamondback, vase mode is different (Z changes continuously) so it might not show the same effect.

Diamondback

@dc42 So I printed a 75x75x50 empty cuboid on 3.4.5 (came out perfectly as expected) and I am now printing the exact same gcode on 3.5b1+.
Interestingly there does not appear to be any issue whatsoever this time around. The print looks exactly the same.
Things to note:

Wall thickness of 1.2mm/3 walls @0.4mm
There's no retractions going on (the previous prints with issues did have retractions)
I do not use Z-hop with retractions, Z only ever changes with completed layers and the active mesh leveling

Diamondback

@dc42 Two more images. The first shows the difference between a "well" printed (3.4.5) MFM housing variation and a faulty (3.5b1+) one, note the stark difference in the height of the posts, this is not a result of the viewing angle. Also note the bulging extrusion again. (one is rotated 180°, so don't be confused by the different sides being visible)

This one is the faulty one alone, note how clearly the hole in the front is not round at all, again this not an illusion from the photo.

(Sorry for the weird colors in the photos, looks like my phone is doing some over the top HDR stuff or something...)

Philippe71

@dc42
There are 2 Z motors

MaxGyver

@dc42

I can confirm the problem. With 3.5b1+ layers seem periodically “squished” in Z direction, almost like Z-Wobble.

My setup:

Duet36HC with 4 Z-Leadscrews.
Bed tramming is performed before every print, and mesh compensation is active.
I also redid the mesh probing after updating, just to be sure.

Diamondback

@dc42 anything else we can do to help you diagnose/replicate the issue? Looks like so far, everyone here has more than one Z stepper and is using active bed tramming?

oliof

@Diamondback maybe try printing without bed compensation to see if that triggers the same issue. Use a raft if required to ensure bed is not causing issues.

dc42

@oliof said in [3.5b1+] Inaccurate print dimensions:

@Diamondback maybe try printing without bed compensation to see if that triggers the same issue. Use a raft if required to ensure bed is not causing issues.

An excellent idea! Right now I don't know what is causing this issue. Knowing whether bed compensation is implicated would help me to narrow it down.

Diamondback

Ok cool, I can disable both mesh compensation and active leveling. (the leveling part before each print is not really needed anymore anyway since I upgraded my spindles to finder pitch)

oliof

@Diamondback if you disable two things, please make three runs: One without both, and one where either or is disabled

@dc42 the reason I thought of that is that elsewhere I recently helped debug an issue that looked like bad segmentation, but it was low Z jerk and acceleration combined with bed compensation that looked like the bulging corner issue.

IndeX4D

My print dimensions are correct, but when start a new print the z axis is exactly 0.2mm to deep.
+
After pause, It seemed like the z offset has completely gone. Should I make a new thread about this?

-4 z-axis
-Axis compensation active

dc42

@IndeX4D said in [3.5b1+] Inaccurate print dimensions:

My print dimensions are correct, but after every print, but every new print hast the Z axis 0.2 to deep.
+
After pause, It seemed like the z offset has completely gone. Should I make a new thread about this?

Yes please! Include your config file, and whether you are using babystepping or any other use of M290.

Diamondback

@dc42
Ok so, here are the results from 3.5b1+:

Active leveling before the print + mesh bed compensation active -> Broken
Only mesh bed compensation -> Broken
Only active leveling -> No issue

So it appears that mesh bed compensation is indeed related in some way.

dc42

@Diamondback thanks, that's very useful.

dc42

@Diamondback

Do you have any layer change script in those models, or anything else happening during the print (after initial heating etc.) other than G1 commands?
Do you have daemon.g running?

Diamondback

@dc42 Not exactly sure what you mean with the first point, I can share the gcode though.
AccuracyDebug.gcode

As for daemon, yes, at that point it was a regular daemon.g with NO loop in it that has this content:

if state.status == "idle" && global.deactivateToolAfterFilamentChange == true
	echo "Unloading tool after filament change"
	T-1

set global.deactivateToolAfterFilamentChange = false

if state.status == "processing" && state.gpOut[1].pwm > 0 && job.layer == 3
	M98 P"/macros/Misc/Lights/Off"

dc42

@Diamondback I tried to reproduce this on my tool changer, which uses a 6HC, but the print using the current RRF 3.5 build was the same height as the one using RRF 3.4.5 and doesn't show the layer artefacts. I had mesh bed compensation enabled.

Have you time to run some more tests? If so:

Please try the latest RRF build for the 6HC at https://www.dropbox.com/sh/5vxz29a7400gwcy/AAAPexxpGsP0LMo0jEFOOCqVa?dl=0
If it shows the same issue, try stopping the daemon task (rename daemon.g to something else)
If it still shows the same issue, try disabling bed compensation and instead enable segmentation using M669 S100 T10 (preferably, replace the 10 by the mesh spacing you use). This will tell me whether it is the bed compensation that provokes the issue, or the segmentation that occurs when bed compensation is enabled.