Z_tilt vs mesh vs those fighting a flat mesh grid

..wasn't sure where to put this so figured it might help some people getting 'potato chips' as bed meshes. Yes...oddball basis here...playing with a brand new 300 bed, but mounted in a 500 frame. (IDEX style bed support)
I'm getting GREAT repeatability on my mesh. Sub 0.1mm no problem. I ran into something quite interesting, that is also a carry over from past Q's I've posted here trying to understand some aspects of Klipper in general. The image below, is a screen cap of a cluster of screen caps. Two z_tilts and what they were set to, and two 3x3 bed_mesh calibrations You'll note the two meshes are within 0.009mm of each other. (goes to the repeatability of the probe used itself as well) On previous 'play times' I'd experimented with reversing the 3 z_tilt points...EG: instead of front left->rear center->front right, I swapped the z_tilt probe points to left rear->front center->right rear. In this experiment I was trying to understand the final relationships created by the post z_tilt automated adjustments, on the mesh grid resulting from those z_tilt settings. what is interesting, is there is some math anomalies that must exist somewhere here (and this is my Q here ... is what am I missing?) The z_tilt adjustments made between the two sets of point configurations, does indeed result in a bed level opposite one another. Ok... that makes sense. But the math part that intrigues me.... the z_tilt adjustments made are 1/10th the mesh grid variation, and barely approach the difference between the two meshes (~0.009 mesh diff z_tilt variations being less in both cases) Yes, the system was re-homed after printer.cfg edits to change the z_tilt probe points. This sequence was repeat 6 times today. What is pictured, is the worst I could get.
8 Replies
ptegler
ptegler16mo ago
No description
ptegler
ptegler16mo ago
I guess the pondery here is why wouldn't reversing the z_tilt points show the same variation of adjustment needed to level the bed, as the measurable variations of the 'potato chip' of bed?
sherbs
sherbs16mo ago
Are you saying you'd expect the probed points range to be the same in both cases? I don't think there are any math anomalies here, just sampling error... But I might be missing what you are trying to get at
ptegler
ptegler16mo ago
sample 'error' is observable as 0.009mm z_tilt one one = .007 mesh @ 0.099 reversed points, tilt =0.002 but mess is greater at 0.108 mesh calcs varied 009 while tilts only varied 0.005 in either case 'repeatability' is better than diff between the two meshes
sherbs
sherbs16mo ago
I feel like the measurements that you are referring to are so small that they are beyond what the probe can measure? If you run the same tests again, do you get consistent results? I get different ztilt adjusts every time I run ztilt I've always assumed that was just due to variation in probe measurement
ptegler
ptegler16mo ago
I can only assume my wording could be better.... repeatability of measurement is a smaller portion of variations here. (repeatability is much better than measurement variations between the two scenarios)
sherbs
sherbs16mo ago
Yeah, perhaps. I do see how small variation in ztilt measurement could lead to larger variation in mesh measurement. But, on average, I think there wouldn't be much difference between your two scenarios. I think, rather than comparing one run of each scenario, you should compare several runs of each scenario and average the results together. If the bed ends up with more or less tilt due to z-tilt sampling locations, that could skew your mesh also
ptegler
ptegler16mo ago
...statements were based on several recent runs, as well as discussing this 'reversal' eons back here. When you have a low corner, I often suggest to people to swap the z_tilt points and re-run their mesh just to validate it's the bed, and not some other mechanical/electrical/magnetic anomaly effecting their mesh
Want results from more Discord servers?
Add your server