Bed Leveling Issues- Adaptive Mesh Not Compensating Correctly

General Discussion
1

klippy.zip (994.7 KB)

Basic Information:

Printer Model: Ender3v3SE
MCU / Printerboard: 4.2.7
Host / SBC BTT Pad 7

Describe your issue:

No matter how level my bed is or what range my bed height map is from .08-.4 adaptive meshing always corrects to the same thing. however, the print head always to far from the bed in the middle causing gaps, and to close to the bed in the back left corner, casing scaring. If I lift the middle of the bed with say tape or paper, adaptive mesh still corrects to the same thing, to far from the bed in the middle and gaps.

Steps I’ve preformed-

  • Axis Twist Comp, manual and auto from 3 counts to 15 no noticeable changes
  • Verified x,y offset before twist comp
  • Verified frame is straight, tighten the frame brace to keep it straight
  • Verified the print head mount it level and at a 90 degree
  • Replaced the CR-Touch to be sure
  • Tried with out adaptive meshing with mixed results, depending on overall level of bed
  • tried both meshing algorithms, cause why not.
  • tried probe points from 5,5 to 15,15 and added mesh_pps 4,4

Not sure what else to do, as I thought I could just manual lift the bed, it took my longer than I care to admit before I realized i have a hill in the middle of my bed but adaptive mesh is still just compensating to the same distance, and still causing gaps

// probe accuracy results: maximum 0.094179, minimum 0.071047, range 0.023132, average 0.083030, median 0.086305, standard deviation 0.007642

IMG_1373
IMG_13731920Ă—2560 896 KB

Printed half a second layer
IMG_1372
IMG_13721920Ă—2560 932 KB

2 
Git version: '3f58baa3'
Branch: master
Remote: origin
Tracked URL: https://github.com/0xD34D/klipper_ender3_v3_se.git

This does not appear to be a genuine Klipper version.

You may ask the one who forked it.

3

Just glanced at the config and this jumped out at me:

[gcode_macro Test_Accuracy]
gcode = 
	PRTOUCH_ACCURACY SAMPLES=10 PROBE_SPEED=1

...
[bltouch]
...
speed = 20

This is like that scene where they put sawdust in the transmission…

The prtouch code isn’t in klipper mainline. I grepped around on github and found some of it. Looks like they have an hx711 ADC chip. We know those are too slow to use for probing, but they keep shipping them out anyway.

If you like the results of Test_Accuracy and you expect to get those same results when bed meshing, you need to set the speed: 1 in [bltouch].

I know that’s very slow, but you have a very slow probe. The maximum sample rate is 80 samples per second. If the probe is moving at 20mm/s one sample every 0.25mm.

We generally tell users that a probe with a resolution greater than 0.025mm is too bad to be useful. So the absolute maximum speed you should ever probe at is 20/10 = 2mm/s.

There is a thread here where people have OK results at 2mm/s. The slower you go, the better it should get.

4

Thanks, Ill give that a try, I have never come across that before. I don’t care if it is slow necessarily if its accurate.

Appreciate it.

1 Like
5

Its a fairly common fork for this printer, and I have not seen other with this issue. Is there any where else that could be the problem, or a direction you could point me in, beside the fork?

6

I have the same issue, printer does not follow the bed mesh.

  • I have 0.007 backlash on z axis
  • 0.015 probe deviation/accuracy (IR probe)
  • I have 0.001 X axis twist

I have warped my bed just to check how accurately it follow the bed, here is a 3,3 mesh :

image
image826Ă—514 57 KB

3,3 just to give you an idea, the actual mesh is 25,25 (500mm bed)

i found that on the highest spot, the nozzle is at ~80% of what it should be, thus digging into the bed, the lowest point also at 80% of what it is, being too far from the bed (if case of this bed, it’s about 0.5mm above or below bed)

There is definitely something wrong with following the bed mesh