Time for another of my "out of left field" questions. LoL
I've just started to get into 3D printing, and as I learn about them, I've found that some of them have what they call "auto-bed leveling". Of course it's not really that, but just adjusting the Z height during printing based on a survey done prior to starting the print.
I suspect that high end CNC machines have some similar process as well, so I'm wondering if anyone has thought of adapting it to a Tormch mill?
Every once-in-a-while someone on the forum talks about having to tram the mill's spindle column, or table sag. It seems the compensation that the 3D printers do would work on a mill as well.
If I understand correctly, what you suggest is often used for milling PCBs. There are lots of videos/websites if you Google using keywords such as "autolevel gcode probe" (without the quotation marks). The plan is usually to probe at a few (or many) points to generate an approximation of the variations in Z-axis across a slightly warped PCB. Then the gcode is adjusted using the interpolated values for the Z-axis adjustment.
It appears that there is a LinuxCNC builtin for Z-axis correction.
See LinuxCNC Documentation Wiki: ProbeKins
I'm unsure how much work it would take to activate this capability in PathPilot.
This discussion should probably move to the Pathpilot section...
I think what you said about the PCB milling sounds to be the same thing the 3D printer folks are doing.
If there is a Linux solution, I would bet that one of the Linux guru users here on the forum could figure out how to implement it in Pathpilot. The Probekins thing sounds like exactly what I was thinking of.
As I would envision it, it would not have to be run for every part like the 3D prints or the PCB's, maybe just once, or occasionally, to map the table of the mill. Seems like it would be a neat way to improve accuracy.
I'm confused; I assumed that you were thinking of solving the problem of an unflat PCB, engraving thin stock secured by double-sided tape or something similar. I mapped my table some time ago and, as I recall, the variation was only about +/-0.001 inches. That sort of variation is not a major contributor to errors on a PCNC. I don't believe that tram errors can be corrected by Z-axis compensation. What are your expectations?
No, i wasn't thinking of working with a particular part, but of compensating for machine errors. Several forum members in the past have mentioned having significant Z errors as they move the table in X and Y. If the table could be mapped in Z while moving in X and Y, then those errors could be compensated.
I realize it would be harder if it is an out of square column, but possibly that could be handled by doing the mapping at different Z heights.
I don't know, maybe just a pipe dream, but worth considering?
Maybe I was lucky to get a good mill! Z-axis compensation would be a great help if your table is warped or otherwise not flat.
Tram issues are usually only a problem because the leading/trailing edges of the cutter are not at the same distance from the table. This is mainly a problem with a fly cutter or other large diameter cutter. I don't think that Z-axis compensation can help with this problem.
Mapping the table won’t do much for a solid piece of stock in a vice or, in many cases, even clamped to the table. The work won’t flex to follow the table. PC board cutters map the work (PCB) surface. It would do no good to map your raw stock. Also small compensating Z changes would probably give a lousy top finish compared to even a poorly trammed machine. You would see the Z “fixes”.
More generally it can be thought of as volumetric compensation . Use a laser to measure the error with regard to the location of the spindle in 3D space and map the errors in the controller. That's the theory but the implementation is difficult. Error can come from cutter bending, thermal dimensional change, direction of motion, weight of a work piece and so on. All solvable but not easy. For the table in question do you measure the error with the table close to the column or far away, with a centered vise or a work piece overhanging one side, with the gibs snug or a little tight/loose?
Sounds like it might be too much work for a small possible gain. Oh well, wouldn't be my first idea that'd impractical. LoL
It might be something that could be done one time at the factory, with specialized gear, such as the laser you mentioned. The data could be shipped with the mill, maybe in firmware.