Limit switches and repeatability

You (and I, because I have the same issue at the minute) need precision microswitches. Metrol is one of the the main brands (spun out of... Toyota? I think?), and they make some pretty expensive switches and some less expensive switches. You may be able to find alternative brands too, but it's another order-of-magnitude thing. For precise location to 0.01mm, your resolution has to be 0.001mm/a micron. So I don't know how many options there may be at, y'know, sub-$50-apiece.

But I also haven't looked into it super hard either, because I'm not reliant on high-precision G53 positioning just yet, so it's not a major problem.

 

Your guess is as good as mine on that one, to be honest. I suppose it's probably technically possible to use a non-contact sensor in a precision fashion, but you'd end up using a dedicated circuit to read the outputs with a high-precision ADC and calibrate each one to the location you want it to trigger. I'm not sure how linear their response is, what their noise susceptibility is like, etc. Harder to produce something like that than to engineer a highly repeatable mechanical solution, would be my guess.

Honeywell and Omron also make some, I don't know how "precision" they are but they seem to be in the same general price range. The Metrols tend to be much more compact and easy to mount though, from what I've seen. I did buy a Metrol switch one time, but with no overtravel protection and at $150 a pop, I'm not sure how enthusiastic I am to jump on that ship again. The ~$50 switches at +/-0.005mm may be usable for your requirements, though, it just depends how precisely you're trying to locate each time. Automation Overstock has a decent selection.

 

Looks like the performance curve graph is the best you're gonna get there. It is just a sensor, not a switch, so I guess that makes sense. From the looks of the graph, I wouldn't naturally expect better than 0.1-0.25mm tolerances. It might be super repeatable, no real way of knowing without getting one and testing it.

 

If the machine is accurate to 0.01mm, and you need the magnet to repeat to 0.001mm to accomplish that, how can you use the machine to position it?

I think we're getting at the heart of the micro-switch problem. A mechanical solution can circumvent or augment the machine's lack of precision. A passive solution by design requires a higher precision machine than it has in order to set the machine it does have. It's a paradox.

 

I would hope, in theory, that there's nothing inherent to the frame or drive system that should be causing deflection or imprecision for homing itself- a negligible force, low speed operation. Nor should there be an impediment to its applicability to a G53 location- for toolchanging or modular workholding or whatever. How that translates to actual repeatability while cutting obviously depends on the machine, though a lot of of imprecision could be corrected out in software, assuming cutting forces are consistent.

A thought I just had, and not sure why I only just remembered this; on commercial CNC machines, high-precision home locations are done with encoder index locations. Either a linear scale index, typically 50mm or 100mm spacings, or the rotation index pulse on the servo encoder- which with tight, well-calibrated screws is fine to use and more common on VMCs. The home location frequently doesn't specifically correspond to any particular end of the axes, they're just set up to have specific amounts of travel in each direction from the encoder index.

So the homing sequence would go seek->home switch actuation-> index seek->index pulse-> homed. The control would know which direction it was going from the switch actuation. That way you don't need to use very expensive switches, just a normal Omron +/-0.2mm or whatever industrial switch. The problem with this method on hobbyist type machines is that they don't have ground ball screws, glass scales or servo encoders.

That said, the encoders on my closed-loop steppers should have an index pulse, hopefully, so maybe I just solved my own problem.

 

Assuming the steppers have an index pulse- ABZ, not just AB- on the encoder (which isn't necessarily a safe bet!), once per motor revolution that index pulse comes up. If you seek a ~0.2mm microswitch, you're then within ~3mm (assuming 1605/1204 screws) of that index pulse coming up in either direction, right? And that index pulse is going to be extremely repeatable, likely to within 0.005mm or thereabouts as long as you're always homing in the same direction. Depending on how many pulses per rev, etc.

So homing is still a two-step process, but instead of a slow re-home on the same switch, it simply uses the switch to find the right index pulse, and then homes on the pulse. The switch/pulse doesn't necessarily have to be in any particular location on the machine, you just define the travels in each direction from where you have it.

 

That, I'm not sure about. If it's six wires, that's likely differential A/B plus power. If it's 5 or 7, there may be an index signal that the driver may or may not use. Might be able to use a multimeter to test if you can rotate the motor slow enough. It's more of an oscilloscope project though, really (as are most things in CNC, I've found).

Of course, it's also possible that the encoder has the channel, but it's just not broken out to the serial cable because it's considered unnecessary for operation. Then you'd have to pull the cover off the motor extension to actually see what's inside.