Part II:
The circuit works very well! In the pic below, you can see the breadboard riding on my X-axis. A wafer neodymium magnet is mounted in a TDI mount.
Experiments were run with three different sensors, a Micronas HAL506UA, a DN6852A, and a neat Motorola 2ss52M magnetoresistive sensor, very sensitive. Of the three, the 6852 was least sensitive, the HAL506 in the middle. All three showed excellent repeatability, with the 2SS52M the best, consistently re-homing to 0.0005" or better, MUCH better than a microswitch. A combination including a more sensitive hall device (2SS52M) and a weaker magnet showed the best accuracy and repeatability. The air gap when the hall triggered ranged from 0.2" to 1" or more. Further improvement came when I turned a steel round into a point (like a pencil), stuck the magnet on the fat end, and lined up the point with the device.
The benefits of this circuit are many. I wouldn't use a mill with other than Normally Closed (NC) switches, either mechanical or like this, solid state. Any failure of a component will typically open the loop, causing the system to stop. No moving parts = better reliability. It is also more accurate than a microswitch. With a good setup, it should be VERY easy to adjust by simply moving some tiny magnets on a slim steel bracket.
Drawbacks are that the setup must be powered, but most controllers have 5V available for external digital logic. This circuit is very crude at this point, and I'll need to brainstorm failure modes a bit to ensure that the circuit does, in fact, open with ANY component failure.
Anyway, this was fun to mess with today. Maybe someone wants to do a bit more with it. I'm going to make an axis with this when the circuit is refined and quadruple checked. These components are very cheap. There's no reason why one couldn't have dual switches in seriesat either end of an axis, virtually guaranteeing a correct stop of an axis when a limit has been breached.
Swede