I don't know much about electronics, but I'm using LDT RM-88 Opto variants. They are connected to the 3rd rail/red instead of the ground/brown like the standard S88s. Maybe you can do the same and use the digital current?

Optocouplers are used to connect two circuits with different supply and ground.
Here you have common ground so it's useless.
As you are connecting to outside world through wifi you could get the 5 V supply from rectified digital signal, but I'm afraid that spikes and so could give problems.

Regards

assuming an optocoupler is of little help here (the Digital pin in actually 3.3V and the source power 5V) what do people suggest? Do I dare run the D3 pin directly to the isolated track and hence to GND?

Almost correct, it's like a transistor on the output side that opens.

I would either just use the digital current to power the input side of the optocoupler, since the current draw is very small,
and I guess only few of them will be powered at the same time. ( Larger R1 and a diode to prevent reverse current to the optocoupler )

Otherwise any cheap power supply could do the trick.



No current is used; but a small voltage drop.

Think of it like this: On ALL wires going to center rails, both detection zones and "not detection" zones, you insert 2 ( *or 4* )
diodes in anti-parallel.

Your detection is then to measure if there's a voltage on the diodes, which means that there's current passing through them.

Your optocouplers then measures if there's a voltage drop or not.

( *or 4*) is depending on type of optocoupler and if you want a diode in series with the optocoupler.

By mounting the anti-parallel diodes on both detection zones and "not detection" zones you'll have the same voltage drop on all tracks everywhere on your layout, and you'll not be able to see any difference - and even better if you can raise the track voltage the same as the voltage drop on the one or two diodes, then everything will be as is has always been.

Per.

Thanks for all the replies, let me see if I can make sense of it all....

So yes, as I expected, the diagram I showed does not really achieve anything with the optocoupler.....

If the goal is to use an optocoupler, yes, I could add an additional power supply to the small processor, (such as a 120V AC to 5V USB charger) but my mind suggests to me that in fact everything in my house ultimately has a common ground.. the charger and the power supply for the layout would be connected to the same power supply. (Even with AC, the neutral is bonded to ground at the breaker box.)

My entire layout system; Intellibox, two boosters, signals, s88 modules, track, 12V and 5V DC, circuits all use a common ground, and that ground is also tied to the ground of the house and the PC connected to the Intellibox. This allows all the accessory consumers of current to have a simple wiring system that ties back into a ground bus around the layout. Does it make sense to now introduce a quasi-separate set of power supplies?

Given that the goal is to know when the rail is no longer isolated, if I tied the GND of the ESP8266 to the layout ground, what are the possible bad side effects?

Possible dangers:

1. Digital current could find its way to the isolated rail (by a derailment or metal object bridging the center rail to the isolated rail), and only to that rail and not to ground, allowing a large current to run into ESP8288 digital pin and let the smoke out.
   
2. Noise?
   
3. Static discharge?
   

Is the solution perhaps to simply prevent those dangers from being realised?

Clearly I could add a diode to the circuit between the rail and the ESP8266 pin to prevent current running into the ESP8266.
I could also add a resistor to that circuit, (the 3.3V pins I will be using have pull-up resistors already but not sure if that is the same thing)
The pins can sink to 0V or supply 3.3 V with a maximum current of just 12mA.

Something like this:


(I added orange digital center rail signal - most likely source of stray current)

Does that sort out the danger of stray current?

For noise, what sort of capacitor is suggested and where? (I struggle to understand electrical filtering.)

I do expect some detection tracks to be bridged for extended periods of time. When a train stops on one, it will be bridged continuously.
I also expect many of these to be bridged simultaneously.

Per, If there is a voltage drop, current must be flowing, no?
You lost me completely on all the diodes stuff... I understand what anti-parallel is, but don't understand what you are suggesting.

I am also perfectly happy to have the sensors normally 'high' and go low when activated. I can flip the logic trivially in the software.

Mario, are you then encoding all the sensor data serially? Are you creating s88 bus signals?

well, I just did a test on my actual layout power system, with this configuration: (R1 = 0 Ohms)



and it works.

I thought I had built in debounce functionality but I need to look into that as I am getting a slew of sensor reports each time. (Just a software issue)

I already handle all the mutiple axles and chatter in my layout software so this is just fine tuning really, I increased the time from 50ms to 90ms in the debounce code.

so I have spent some time reading up on stray voltage protection, and here is my current idea:



C1 0.01uF
D1 1N5817
D2 1N5817
R1 1k Ohm
R2 10 Ohm