Inductive Limit Switch problem

[Kitwn]

Embarrassing but I must own up: I realised during my tests that for all these years I've left the 36V supply for my stepper motors floating. Neither rail was grounded so the whole supply was bouncing up and down by 8 Volts at several KHz. It's a miracle the machine worked at all! Any way that's now fixed and some of the bits from RS have arrived, though not the screened cable for re-wiring the sensors yet.
-
Fortunately I have access to the highly sophisticated electronics test laboratory shown below and have been able to make some very revealing, noise-free measurements on the spare cheap Chinese breakout board. Earlier reported voltage measurements were made in the presence of noise and are not reliable.
-
With an input fed to the board header rather than the screw terminals, a falling voltage must drop below 0.6V to set the output of the board. A rising voltage must go above 1.7V to reset it. This gap is the hysteresis we expect from a Scmitt trigger input chip and provides some noise immunity.
-
When using normally closed (NC) microswitches as sensors the inputs are a good solid short circuit for most of the time and only open briefly during homing. Excellent noise immunity is obtained in this configuration as proved by the fact that my machine has worked reliably for a few years!
-
Using proximity sensors presents a problem, even with opto-isolators or other electronic interfacing, since the saturated collector-emitter voltage at the output of the proximity detector or from an isolator when triggered is about 0.6V. No guarantee of reliable triggering here even without noise. The use of a potential divider circuit as described earlier is an option but I've decided to use a more direct and low-impedance option... reed relays. These will be fed by an NPN emitter-follower transistor which serves to boost the current available from the proximity sensor.
-
Some will scoff no doubt, but it actually makes sense for a few reasons:
I know from past experience that a solid short circuit on the inputs works reliably, even in the presence of high noise.
Low-pass filtering is excellent and requires no additional components.
When wired as I plan to have them, with the 12V from the sensor holding the relay contacts closed as the normal state (this mimics the previously known-good NC microswitches), the circuit is proof against wiring faults as any breaks will trip the limit input.
-
Regarding positional accuracy, a key point of using proximity sensors is improved repeatability over microswitches. The reed relays I'm going to use have a specified response time of 0.5mS. Allowing for, say, 20% variation that gives a repeatability of within of 0.1mS. I can't remember exactly what my second approach speed is set at for homing in LinuxCNC but if it's even as high as 1m/min that equates to a variation of around 2 microns. Which equals nothing.

The 4 relays are $20 the lot from RS, the transistors and a wee bit of veroboard are existing stock worth much less. The whole circuit has been tested on the bench ( I have 2 relays in stock already) and works a treat. So I might just have a fully working machine again well before Mark McGowan let's me back in the pub!