As an aside, does nobody use encoders for positional feedback? Every machine I see seems to rely on the axes being homed at the start and then not being stalled when in motion (which would throw the measurements out).
I think the problem with encoders is that as far as I know the open source software alternatives doesn't support them.
"Real" machines use servos, as you of course know, but yes, encoders would be a good thing, however what do you do once you missed a step on e.g. a curve? The part is potentially already ruined (although you could argue that a step "wrong & corrected" is better than just a step wrong, of course
Ah good point. The 7m beast of an inspection rig that we built uses steppers but I fitted encoders assuming our software guru/geek/git would be able to read them into the PLC and tell the axes where they'd gone wrong.
He couldn't so despite my best efforts to make a super-duper machine it is compromised by relying on the number of pulses sent to the steppers being correct.
Mind you, the reduction gearboxes do a bloody good job of making sure that nothing much stalls it
One of the guys at work is building a CNC in his garage for woodworking, trouble is that he's designed in so much adjustment to everything it's in danger of being more flexible than a Russian gymnast!
There's a couple of hardware solutions that support encoder feedback to close the loop. The KFLOP can do it and the CS Labs controller can do it. EMC (AKA LinuxCNC) can also do it in software.