Matter of fact, the symptoms are identical. But there’s a more than significant number of users experiencing them regardless of the hardware used (sometimes whit an arguably near-perfect motion system setup, like for @Murdock).
Also, why does it not translate to Marlin?
In the end, I don’t think that it’s not a mechanical issue (in many cases it’s only that). But software can’t be excluded (yet).
My primary goal with this branch is to make some minor improvements to bed_mesh and to provide a method for validating the adjustments made by bed_mesh. If there is a bug in the adjustment the mesh_analysis should find it rather quickly. With that data I should be able to reproduce the fault in a synthetic test, and if I can reproduce a bug I can fix it. Otherwise, I hope to put to rest the claims that bed mesh is responsible.
I’m experiencing this since I’m using Klipper. Actually I have never spend a big thought on it, took the calibrated z-offset as starting point and modified it until I was satisfied with the first layer.
Actually I have to modify it according to the material and temperature I’m printing. 255°C head / 105°C bed ASA is significantly different to 200°C head / 20°C bed PLA on my printer, i.e. thermal expansion etc.
This is purely anecdotal based on my own experiences with my heavily modified CR-10S Pro using stock & my own designed tool head with genuine BL-Touch 3.1.
I do not seem to be having any major issues with bed mesh, although some areas sometimes exhibit slightly more first layer squish than others. I have printed numerous large items up to 220 mm in size with virtually perfect first layer squish everywhere. I am however an engineer and am extremely detail oriented, keeping the printer tuned extremely well.
On the other hand, I do see evidence of some BL Touch “drift”, where at times the amount of first layer squish differs between prints. I have never measured it, but purely based on visuals I would say it’s on the order of 0.02 to 0.05 perhaps. This drift is not dependent on the amount of thermal stabilization. In addition, my probing speeds are quite low with the first approach being at 6 mm/s and the second at 2 mm/s. This is mainly because of my previous experience with slowly responding capacitive probes that needed approach speeds well below 1 mm/s. I fundamentally don’t consider slower mesh probing as an issue when compared to the length of time prints take. The mesh on my printer also remains valid for months across multiple prints and power cycles with no need to re-run.
To add yet another bit or information, running PROBE_ACCURACY on my installation at a single point returns extremely tight results:
// PROBE_ACCURACY at X:180.000 Y:150.000 Z:10.000 (samples=25 retract=3.000 speed=2.0 lift_speed=2.0)
// probe accuracy results: maximum 2.045781, minimum 2.042187, range 0.003594, average 2.043537,
// median 2.043437, standard deviation 0.000848
So I think that, all in all, BL Touch is definitely one of the potential contributors. I just don’t understand why, and I don’t really have the time to engage in another empirical characterization experiment right now.
In the rail-core community this is a hot topic, because a large percentage of the community have magnetic beds. Depending on who manufactured the bed, and what version it is you can have major issue or almost no issue.
The underlying issue seems to be that the beds using large powerful magnets generate enough emi that it messes with the blt.
@Arksine, Sorry to pollute the thread with my answers. To other users. Besides, if it suits everyone, I propose to create a thread to be able to exchange ways without polluting launched by @Arksine .
I’ll just respond briefly.
Yes that’s what I’m currently doing. Having never needed to do this I was just surprised. But if it’s not a problem, I’m fine.
Yes that makes sense.
Yes, I had thought of that option as well. Track to dig.
I think about it more and more. I will tinker with another sensor less sensitive to magnetic and thermal disturbances.
On this I suggest to those who want to discuss creating a dedicated thread or going through a collective dm.
Was there ever a follow up on this topic which I can’t find right now?
Basically I can say that BED_MESH or the compensation at Klipper has never worked for me, no matter which printer or which sensor. I have had several sensors over the last two years.
These include:
However, I had never really questioned it in the past, but assumed it was just me. In the meantime I read posts here which correspond exactly to my observations.
Of course I always used magnetic beds.
And it is definitely the case that there are no problems with non-magnetic beds?
EDIT: It would be great to know if somebody was able to narrow it down to specific probes or found probes that are not affected by this.
No, the issue is exactly the same regardless of the type of bed I use - default Build Tak, glass or magnetic PEI.
@Arksine, did you ever get a chance to merge these changes into the main branch? I’m having a world of problems getting my mesh to work well in areas that are outside the probe’s reach (the front of the bed in my case), and I’d like to see if the mesh-edge changes in this test branch might help (I think the bed continues to drop off at the front but the probe doesn’t know it, and the mesh is clamped to the last probe position, making the first ~40mm of my bed unusable.
UPDATE: I went ahead and forked the latest version and manually made the changes that affect the edges of the mesh (the extra control points). I couldn’t really tell that it helped my problem (it still seems the nozzle is too high at the front of the bed beyond roughly the point where the probe can’t really reach). I do think that this is the proper approach to the problem, in general. But in my case it looks like I’ll have to look for another solution. A flatter bed surface would be ideal but this is no consumer printer I’m using so I’ll have to fabricate something and come up with a new heater.
I’d need to look into trying this. I’ve got 3 different probes on my printers and they all work fine until the bed decided to move weirdly on my trident.
My SV04 has a BLtouch, its probe is bent but it still works
My Z9V5 has a inductive probe, not configured yet though.
And my trident has klicky installed, I probably should look into the dockable probe branch a bit for that one.
I think the BLtouch has issues with higher temps though.
There’s also an ADXL as a probe branch I’m pretty sure that I might use for my reborn 2
EDIT: scratch the ADXL probe, I tried to port over dymbuyn’s (can’t remember that username for the life of me) code for a ADXL as a probe but one file had a refactor between his fork and now