SOLVED Kossel linear plus: tuning a poor mesh grid

[gtrevi]

Hello all,

I'm struggling in trying to fine tune my Kossel linear plus printer.
Prints are coming out fairly good, but I'm getting out this "not so good" heatmap, with a corner bent down 0.216mm:

0_1550002324969_heightmap.csv

The whole printer is maniacally squared and leveled against the all metal CNCed corners (even carriages & head are metal).
The levelling probe is an original BLTouch.
The bed is a 240x4mm borosilicate glass disk, directly laying on the 3 lower-top 2020 aluminum extrusions, so absolutely flat and therefore I cannot understand why I’m getting this convexplane... which is btw pretty much compatible with some warping of the prints in that "blu" area.

Can anyone suggest what can I possibly look after, as I totally ran out of options here?

Thanks for any help!

[Veti]

http://boim.com/DeltaUtil/CalDoc/Calibration.html

looks a bit like Tower Placement Symmetry

[gtrevi]

@Veti thanks for that resource, really interesting article.

Not sure though it's a symmetry problem, as mentioned the whole printer is maniacally squared and leveled against the all metal CNCed corners & 2020 profiles... moreover, everything is assembled with proper chuncky T-nuts (not those useless square bolts).

Could it be perhaps be a slight rod/radius length difference (looking at the article you pointed out)?

[Veti]

What is the number of calibration factors that you used?

[dc42]

My guess is that you have a geometric error that is causing effector tilt when the effector is is the blue area. Mount a circular spirit level on the effector to check this.

[gtrevi]

@Veti I use 6 calibration factors (13pts), here is my bed.g: 0_1550510388221_bed.g

@dc42 The circular spirit level on top of the effector gives me of course the reference to the ground, but the positioning of the bubble is pretty centered and is consistent moving it across the bed. Given I placed it above the metal casing of the fans (40mm above the rod axes), the error should be even amplified, so should be ok...

That said, I’d expect that given that the glass bed is flat (measured with a precision square), the bed compensation should do the on-the fly corrections as the print is being executed, according to the poor mesh… instead it looks like it’s ineffective since it looks like the mesh just keeps the Zmax deviation (0.3mm) as the Z=0.
I've proven this to myself by manually jogging to Z=0, measuring -0.3mm of bed distance and then adding it to the M665 height along with disabling M501 in config.g, but keeping M375.

In lack of ideas, I even tried rotating the bed… but got roughly the same results, so one might say it could be the top frame extrusions on which the bed is directly laying (even though the whole frame is truly squared and measured with precision workshop tools), but again, if I put a flat surface on top of the frame extrusions, and given the bed is not being bent by any clamp (as it’s only firmly kept on the sides by plastic mounts), I'd expect the bed compensation should dynamically correct the print… or am I missing something?

Here is also my config-override.g, being loaded by M501 at the end of config.g:

0_1550511164818_config-override.g

Issue is I just can't get that perfect 1st layer... and it's driving me nuts

Thanks for any help/suggestion.

[dc42]

@gtrevi said in Kossel linear plus: tuning a poor mesh grid:

The circular spirit level on top of the effector gives me of course the reference to the ground, but the positioning of the bubble is pretty centered and is consistent moving it across the bed. Given I placed it above the metal casing of the fans (40mm above the rod axes), the error should be even amplified, so should be ok...

Is it definitely still in the same position when the nozzle is over the blue area of the height map?

Offset probes such as BLTouch are extremely sensitive to effector tilt. For example, if your BLTouch is offset from the nozzle by 25mm, then a change in effector tilt of just one degree will cause the relative heights of the BLTouch and the nozzle to change by 0.4mm, which will cause an apparent height error of 0.4mm.

[gtrevi]

@dc42 said in Kossel linear plus: tuning a poor mesh grid:

Offset probes such as BLTouch are extremely sensitive to effector tilt. For example, if your BLTouch is offset from the nozzle by 25mm, then a chan

Yes, it is definitively still.
My probe is a BLTouch, and being a delta effector, it's forcibly Y-offseted from the nozzle by exactly 40mm: adding a picture just to show the effector setup:

[Wyvern]

Yeah, that's far away from true center.
That magnetic probe switch your printer came with that sits right under the nozzle is probably a way better bet, I will be putting one on my FT-5, just because I like the simplicity of a mechanical switch.

Smart effector, that switch that I described or the IR sensor may be in order.

[gtrevi]

@wyvern yes agree w/ having a simple mechanical switch (at the time I got it it wasn't included... but anyways the only anycubic stuff left are the 2020 extrusions, rods and effector :)).

Just the tilt, is in any case unexplained to myself.

[gtrevi]

@dc42 @Wyvern guess the problem is solved: I just switched to a switch (ha) anycubic probe sitting right under the nozzle, and here is the result... which at least confirms the fairly good "trueness" of the frame (for which I ran out of any possible options):

Thanks once again for pointing me towards the right direction!!