rtr12 posted:… And FWIW, I tried the circuit with the LM339 and was unable to make it work. I checked and re-checked all my connections and they seemed to be correct. Even tried a couple of things I thought might make a difference and they did not, still didn't work. I don't think it was due to a mistake on my part? I set it aside and am going to check it one more time in a day or two before giving up?
Perhaps moot now if you've moved on. But when you say it didn't work, in what way did it not work?
Did you have the 3-LED signal hooked up to the LM339 outputs? It's a longshot but the LM339 needs a load hooked up to it for the output voltage to change as intended. If no LEDs were hooked up and you just used a meter to see if the LM339 outputs changed, you would likely have read 0 or near 0V on all the outputs.
rtr12 posted:… I have to agree with him that the first circuit we came up with (R5.4 posted here) earlier in the thread seems to be the one we are leaning toward..
Not my circus, not my monkeys! 
But if trying to cover as many bases as possible with one circuit, I'd consider the PRR style signal head as I diagrammed above. You cannot use the voltage-starving technique depending on the Red-Green LED voltage difference. Though I think you've seen that the circuit difference is maybe a transistor instead of a diode and a resistor or whatever. In other words a few pennies to make the circuit independent of color.
Also, have you guys considered 2-aspect signals? The time-delay and inherent flicker suppression is a nice feature that few (if any?) off-the-shelf 2-aspect signals (e.g., dwarf) have. I haven't reviewed your circuits to see if this is a no-cost / just a jumper change.
