Precision home switches

I am planning the home/limit switch configuration for my XBCW machine and I am considering what switches to purchase for the linear (X,W) axes. Both need to be high precision (~5um) and the only switches I have found which guarantee this repeatability are plunger or ball types which necessarily mean they must be combined limit/home switches. My question is: how much overshoot will I get during the homing sequence and can I control it to be less than the 1-2mm travel of the switch? Alternately, Is there a better kind of switch or configuration to provide this level of accuracy?
I have read several unofficial tests that indicate that Hall effect or Optical Interrupters can be this repeatable but the manufacturers don't stand behind it.

If you are using a two step home procedure, A simple flag on the ballscrew could increase an opto sensors resolution by say 15X to 50X

as far as overshoot that depends on many factors,
But if you look here
linuxcnc.org/docs/html/config/ini-homing.html
6.1. HOME_SEARCH_VEL
6.2. HOME_LATCH_VEL

depending on the speed above will have the most influence on the amount of overshoot.

Keeping latch vel set low/slow will have the most impact.

as far as switch type it depends on environment.

If you are cutting metals a magnet/hall based on may get build up and become less accurate.

An optical one that has the emitter and diode open to the elements will get dirty and could get triggered but cuttings.

There are opto mechanical switches that are a combo with a plunger that trips a light/laser beam.
These can have good accuracy while also providing a decent protection.

PCW:
A flag on the ballcrew kind of like a index?
so one would attach a disk the the end of the ballscrew with a small slit and an opto switch and use that to home?

One simple one step scheme would be to have a 1/2 disk on the ballscrew and an opto,
hall effect sensor or proximity sensor on the disk and logical and the output of the sensor with
the home switch. This would work if you could stop in 1/2 a turn.

You could also use an index like scheme with a narrow slot or flag but if this is a
step/dir system, homing to index is somewhat uncharted territory
(homing to index assumes encoders currently) Mesa products do support homing
to index on step/dir configuration s but that is also very new.

So, as I understand the homing sequence, the home switch has to remain activated until the motion is reversed. Then when the switch is opened the home position is set. This means an index like flag on the ball screw will not work unless there is a way to keep the switch/sensor activated until the motion returns to the home position - while the sensor is moving. Not impossible but a very tricky design problem.

I like the idea of an opto-mechanical switch which would be easy to keep clean but I haven't been able to find any other that the ones used for keyboards. any suggestions where to look?

I remember seeing some but my searching is not turning anything up.
I think if you went standard limit switch with some weather rating and kept the final approach speed slow you should get close to what you want.

This means an index like flag on the ball screw will not work unless there is a way to keep the switch/sensor activated until the motion returns to the home position - while the sensor is moving. Not impossible but a very tricky design problem.

This is why I suggested a 1/2 disk and stated that motion would need to stop in 1/2 turn. If this is possible
there is no issue with the home switch changing to off again at the home location.

Using an actual index avoids this issue (so the flag can be a narrow as the system can detect)

If the disk is fixed to the ball screw then the sensor would have to be attached to the moving part wouldn't it? In that case it seems like it would continue moving toward the disk for while after triggering the switch. In my case that is 2.5mm for a half turn. I can't picture how the disk would engage the switch.

No, the sensor is fixed and the disk is mounted on one end of the ballscrew
so the disk just turns in place. The disk signal is anded with the (coarse) home
switch signal so you get a 1/2 turn active home switch at the end of travel.
with disk_circumference/motion_per_turn improvement in sensing accuracy.

Ok, I understand what you are suggesting. Before I start designing a custom switch though, I want to continue searching for an off-the-shelf solution.