Bed leveling thoughts and concerns.
During a discussion with a friend that has a Cone One+ calibration and bed leveling. My experience with my MK3s+ and issues with the first layer, early in my print days, I did the silicone mod and it improved the accuracy of my prints. I just upgraded to 3.5s and with the expansion joints, I didn't think I needed the mod. I was back to first layer and z-level calibration issues. A print I made, part of the first layer separated from the rest of the print. Just did the silicone mod on the 3.5s and now things are working as expected and as experienced. There are two aspects to bed leveling. Full bed in relation to the Z-axis and surface flatness. I was also reading about how bed leveling has improved prints for people with a whole range of printers. In some cases, stopping large prints from failing. In the installation of the silicone mod, I found that the default bed had a variance of 0.446mm. 0.446mm is much more than the default layers I am using. With the mod, depending on how long the printer has heated, I am under 0.05mm in all the tests with the best variance of 0.021mm.
Reading through comments about bed leveling the Core One series, I started to think about how bed leveling with the Nextruder actually works in conjunction to a full print. This also fits for the Pinda leveling printers. In my case, the Pinda shows that there is a variance 0.446mm when doing the bed measurement and thus the printer is supposed to compensate for this error.
This brings up the question for discussion. If the printer is compensating for the first layer, does the compensation continue throughout the full print? Why is this a concern? Accuracy of prints. Using my default case, if the compensation is carried throughout the full print, the top layer is not going to be flat. This is not my experience.
Looking at this further, with comments about the Core One and the suggestion of using set screws to adjust the stop point for the Z-axis motor during homing, to help level the full bed, not just compensate for the expansion of the hot plate. Also, using shims to flatten the build plate. Lets say that the Z-axis carriage has a drop on one corner by 0.4mm giving the whole build plate a tilt. Nextruder pressure sensor should measure and the firmware should compensate for that tilt. Since the build plate is tilted, now it isn't square with the Z-axis and skewed with the X and Y axes. So, when printing, to remain square to the build plate, the rest of the print needs to be compensated on both X and Y angles for the printed object to remain square at all sides. Not an issue for small object but an issue for larger objects as the 0.4mm needs to be compensated for.
I visualize this as a box with a washer under one corner and how the top is not parallel to the table top anymore and the sides are not at 90° with respect to the table anymore.
I will agree, that for most people, this won't be an issue. As most of my prints are functional parts and in many cases have to mesh into other items, accuracy of prints is important with minimal post processing. If something isn't square, it may mean that the print is garbage.
RE:
Z is - IMHO - really one of the Achilles heels of the Core One architecture (arguably, it's got a few) and watching my Voron's bootstrap quad-gantry-level procedure (with four individually controlled Z motors) always makes me smile a little, even though waiting is annoying. This one needs to be accurate - Z endstop calibration switch at back, tool docks at front - any misalignment would cause crash on toolchange.
My understanding is that no 3D printer corrects the angular error between XY plane and Z axis. Otherwise it would need a correction margin unknown to the slicer, edge positions at high Z couldn't be printed reliably anymore. So the print gets skewed. That's why I calibrate Core One's printbed against the nozzle manually (I'd argue I can do better with a feeler gauge between nozzle and printbed than the firmware running the bed into mechanically only loosely defined endstops, shims or no shims...).
Print bed mesh carrying through to the top is another issue. My understanding (please do correct if I'm wrong) is, the XY transformation from bed mesh persists across Z. So if I need a level top surface, I'd better start with a level print sheet.
RE: Bed leveling thoughts and concerns.
Whoo Hoo. mnentwig I am in 100% agreement with you. Mechanical zeros, then electronic zeros from the start. From my experiments today, aluminium and thermal expansion for 3D printing is a problem. From short term hot build plate to long term hot build plate changed things by 0.015mm. I am thinking of making a build drawing of the MK3X series Y plate and getting one machined out of 6mm steel. Why. I don't know, because I can? Something to test. I am also testing polymer bearings right now, and other than needing to fine adjust them, I love them.
You just made me think. No CNC would use software offsets to correct for milling other than for engraving. It would be mechanical from the start. Most documentation for home made CNC's I have read about, mill the surface once the machine is built.
Feeler gauges for alignment of the Z Axis on the Core is a good suggestion.
For large prints, I didn't think about the limits of the build area and any skew.
Still doesn't answer the question about how the firmware compensates after the probe of the build plate.
Crash at tool change. An new issue for INDX mod? Something to think about.
RE: Bed leveling thoughts and concerns.
Ha! It's a software world. I have a CNC machinist friend who makes some really nice parts. I'm an amateur who makes just OK parts on manual machines. I was photographing one of his parts and, just for fun, did some measurements on it. It was darn near perfect, 0.0001" about everywhere. I asked him how he got the machine set up that well for taper and such. He said the machine itself is nowhere near that good- he programs out the error in the code after making the first couple parts.
I do wish the three Z screws were always matched and that the motors could be driven separately, but we have what we have. There is no mechanical fix for that, other than a decision of what altitude to set parallelism with the print head, if one is going to go that far.