Mesh Bed Levelling Issue on Klipper, incorrectly working bed mesh compensation.
Вставка
- Опубліковано 11 жов 2024
- I have checked the probe accuracy and it seem to be fine. I will be happy to read your comments - maybe you have any idea how to fix the issue with incorrectly working mesh compensation.
The bottom right corner is always printed too low, and the opposite top left corner is always printed too high.
Unfortunately, I also have the same problem. I always thought it was a problem with my machine, but after seeing your video, I realized that other people had the same problem.
Hi again,
so after more investigation i am pretty sure that it is a system issue which cannot cope with a bed mesh leveling properly.
What i have done so far:
1. New probe - SuperPINDA form Prusa Research
2. Made the bed as flat as possible using bloody adhesive tape to rise the depressions (i have repeatedly done mesh calibration and stick a bit of tape in the depression spots to raise them to as close as possible to the ZERO level)
3. I made sure the probe is not affected by the temperature
4. I made sure that all the z lead screws are operating smooth, even and with ease
After all of that i have run a test print and while the bed is now much better leveled - the RatOS does not cope with the correct level compensation.
I will do more investigation with compensation settings and update here.
Same, I have no idea what BLTtouch is actually supposed to do, it sure as hell doesn’t compensate the bed mesh irregularities.
I have the exact same issue. Even narrowing down mesh variance to ~0.05mm or so shows the same thing : it's as the mesh was "flipped", yielding a wrong compensation in Z.
In my case, right side of the bed is ok, left side is a bit too low, so it's like having a "partial" 1st layer elephant foot on 3 out of 4 sides of a square shape. That thing didn't happen with Marlin, for instance.
If everything is ok on the hardware side and properly set up, it's logically some issue on the software side (imho).
New issues regarding that wrong compensation Z are often being opened on Klipper's Github, though the same answer is generally given ("bed mesh compensation works as intended").
Though, if your bed mesh uses the "bicubic" algorithm, maybe the stock "bicubic_tension" value is too high, which could explain the overcompensation. Yet, I tried both algorithms and ... even using "bicubic" while reducing bicubic tension was to no avail. Both algorithms show the same results.
So maybe it's really a cryptic Klipper issue.
Despite all the aconwladgment from the Klipper + Ratrig devs - i still belive they are missing some very tiny bug and this is happening to us who actually care about perfect 1st layer.
Yep. Definitely. There's something odd or wrong with the way Z compensation is applied with Klipper's bed mesh.
I may take some time at some point to read the source code of the bed mesh compensation and see if I can find any oddity somewhere.
I think this is also a very deep seeded issue I have specifically left Marlin due to this issue, thinking Klipper would be better. My mesh looks like a meteor hit the middle of the bed. It has a variant of .6. I'm going to get a credit card or something really flat and try to rub the magnetic sheet more flat for my PEI bed. See if that changes anything. Try to lower the variance.
Id suggest setting up KAMP now, instead of making a mesh of your whole print bed it only makes a mesh of the area that the print is going to be printed on. Faster meshing and much more consistent first layers. Its pretty easy to setup.
The algorithm used by bed mesh is screwing you over. It is fine for a flat bed on an angle but if you have a bed with warp, the interpolated points will be exaggerated so that it creates a smooth curve between them. Take a look at the difference between the probed points view and the mesh view. For example because I had a low point before two flat points, the algorithm invented a hill between the two flat points to fit a smooth curve.
i had the same issue, first make sure to include in your start gcode to use the mesh. second i had to take apart my print head and make shure to adjust the prope sensor to its correct position.
I'm really confused by this as well. I have a kingroon KPS 3 pro that doesn't seem to have any major issues, but my sidewinder x1 looks similar to this one. I have none of these issues when Marlin was installed with custom firmware.
I am experiencing exactly the same but with worse results due to the bed being a bit more warped. It s like klipper way overcompensates, so the low points on the bed are so much over done that the nozzle crashes on the surface, and the other way around on the otherside...
Wow, this is really bad then. What are your probe and bed mesh settings?
you can try mine, it works not bad, but not perfect too:
[probe]
pin: ^probe_pin # For NPN NC probes such as the Super Pinda / Vinda / SupCR / Decoprobe probes.
x_offset: -29.8
y_offset: -13.5
speed: 5 #predkosc podnoszenia stołu podczas mierzenia sonda
samples: 1
sample_retract_dist: 2
lift_speed: 15.0
samples_result: median
samples_tolerance: 0.02
samples_tolerance_retries: 5
#z_offset = 1.76 # poprzednia wartosc 1.77, mniejsza wartosc oznacza wiekszy odstep od stolu
[bed_mesh]
probe_count: 18,18
speed: 200
horizontal_move_z: 5 # wysokosc Z podczas przejazdow
mesh_min: 24,30
mesh_max:274,279
# Zanikanie trzymania sie siatki
fade_start: 0.4 # wysokosc w mm od jakiej zaczac zanikanie
fade_end: 2 # wysokosc w mm kiedy stol przestaje calkowicie kompensowac siatke
mesh_pps: 0,0 # interpolacja pomiedzy pomiarami (raczej przeszkadza zamiast pomagac)
algorithm: bicubic # algorytm interpolacji siatki
bicubic_tension: .2 # musze doczytac
# ustawienia czulosci reakacji na siatke z tego co zauwazylem to wiele nie daje
move_check_distance: 20 # poprzednia wartosc: 10
split_delta_z: 0.05 # poprzednia wartosc: 0.01
[z_tilt]
speed: 300
z_positions: # describing the location of each bed "pivot point" (referencja wzgledem rzeczywstej pozycji dyszy)
53,42
179,283 # 151.5,308
301,42
points: # trzeba uwzglednic offset probe
53,42
180,283 # 151.5,308
301,42
horizontal_move_z: 5
retries: 5
retry_tolerance: 0.02
Show us the probe info in the config file. Did you set the probe offset correctly? Show us how you calculated the offset.
IMHO, your bed level is excellent, +/- 0.05 is great result, and your test is, OK, maybe not excellent, but pretty good.
Possible reason for such unlevelation could be dislocation of printed circles from the spots where the probe measures the level. This is OK for common people, but you are perfectionist - so, adjust yout printed circles' locations.
Also check the order of your start gcode commands - is the home command (G28) AFTER nozzle and bed temperature setting WITH WAITING TO BE REACHED (M190 and M109)?
My bed leveling doesn't seem to make a difference after I level it. I did max points and nothing changes after I level it. Did you ever figure out how to resolve this?
Hi yes i disabled the mesh bed leveling and manually leveled the bed by means of sanding off and adding material whenre necessary - worked like a charm. I have not used bed leveling since a year.
Hi Delan3d. Did you ever find the cause of this? And fix ? I am having the same issue with my voron trident running klipper. Would be very keen to hear if you resolved the issue. Thanks.
Sorry Delan3d, I just watched your other videos explaining how you corrected the issue. Thanks!
@@evanmackenzie5811 it's impossible that a firmware can't level a bed at a such a low range what he did was make a flat bed that didn't needs klipper to work , i'm having the same problem and i saw that if you decrease your bicubic tension to 0.1 it seems to work i'm printing right know so i can't test but when i finish i will give it a try instead of losing days over a bed or just disable bed mesh and use only screw tilt calibration
@@diogocoelho496let us know the results, i am still simply printing with mesh bed leveling completely disabled and it works like a charm
@@delan3d422 when i put tension at 0.1 it inverted the compesation so i got to 0.15 and it seems to work just fine for now cause at klipper page they say the higher the tension the more is going to compensate. i say give it a try but if it works for you good also if is not consistent maybe i go just with screw tilt calibrate and leave bed mesh alone
@@diogocoelho496 Oh my... I have been fighting this issue for a while now. I've rebuilt my machine several times trying to convince myself it was a mechanical issue. I have no clue when I added the tension parameter but it sure enough mine was set to 0.2. Dropped that down to 0.1 and my first layers are perfect now... Thanks for the suggestion!!
Your gantry is either twisted or you have play in your x rail
did you try to delete the loaded bed Mesh and then do a new bed mesh without having the first Bed Mesh loaded?
Just seen your video. G28 delete the mesh, you have to load it again or do a new one for the print, I would do a temporary new one for every print , after G28 type g29.
Chris, you are a the MVP. I searched for the solution for a bit over a week now. This solved it for me. I simply had it the wrong way around in the start macro. Thanks a lot!
Your the best !!! looking for my problem sinse several week ... just forget to load bed mesh ..... Thanks
I have the same issue with quick draw, n had with bltouch too.
Manual leveling works much better than probe.
Probe bed mesh shows wrong
same problem here after two day to test any setup i forgive and use only bl touch for indicate me to turn each screw or not in manual leveling with knobs... :/
Solution:
Make your mesh_pps values 0,0.
Klipper compensation algorithms are useless. When you make them 0,0 Klipper compansates differences linearly. Bicubic compensation creates unnecessarily high and low imaginary point. I was going to crazy for this. Then I remebered the marlin algorithm.
Klippes is good and full of stuff but marlin is much much and MUCH more mature.
Jesus, this finally resolved the issue for me after months of investigation. Thank you so much!
I have the exact same issue with an Ender S1 Pro running Sonic Pad (Klipper). Drove me nuts the last few days! Do you have a solution yet?
Finally it works - I can‘t believe it! I‘ve mounted a cr-touch holder with Y offset 0 and changed the probing area to be rectangular ..
Strange that the y offset causing this. Not the first time i heard it.
@@magicmaui77 can you eli5 please. I’m having the same issue and I’m on day two and I can’t get my bed level for nothing.
@@magicmaui77 I have the same problem. I use Sonic Pad with an Ender 3 S1 Pro, I don't understand what the solution was, could you help?
@@GuLuppi Yes - you need to relocate the CR touch to a position, where it’s in line with the nozzle (Y offset = 0). There are some mounts available on the net. Config has to be adjusted (bltouch offset, mesh bed max min positions..). Here‘s the mount I use (5015 blower): ua-cam.com/users/shortsE4fWTpFdvzo?feature=share
I don't use klipper but I have similar problems with marlin.
Guys, proble was in [probe]
z offset just manual set it
Can you explain in more detail what the problem was?
The Z offset is set by Klipper itself after the calibration?
same happening to me
@Nicholas it turned out to be a warped bed and inaccurate BL-Touch due to the temperatures generated by the bed. I have sanded the surface of the bed to make it less warped, then i have used SuperPINDA probe. I thought that i can relay on the calibrated mesh, but the system does not really follow it all the way. Mesh calibration helps only a little. The biggest issue is always with a warped bed.
In my case was the z offset that varies between regions. I dont know If the carriage is tilt or not. I have an ender 5. For me a manual bed mesh should fix because It calibrates the nozzle distance not a probes distance.
@@delan3d422 but the whole point of using a BL touch is to account for a bed being warped.. thats literally why we buy these things
Has anyone solved this problem?
Unfortunately the only person who can solve this problem is the DEV TEAM - but they think their software is perfect and there is nothing wrong with it. So tell me - why my Prusa use mesh bed leveling and have no issues whatsoever ?
I’m having the same issue! So there is no fix for it?
@@joeybarnickle250 the only fix for it on Ratrig with Klipper is to change the whole system or make your bed dead flat - which i did and it works for me. Check my latest video
Yes, I solved it - mount your leveling sensor to Y Offset = 0 and it works even with a fairly bent bed ☝🏻
@@magicmaui77 what do you mean by saying "mount" to Y offset? Cna you elaborate on that please?
Check you enstop position
Just try making quickdraw probe or klicky probe. Works better than genuine bltouch
I had this issue to to fix it you have to go to printer config and go down to something about bed mesh and change it from 7x7 to 9x9 you can do this on moon sale, which is the web application thing