Thanks for this. I think this circuit might work, but I worry about the opamp/comparitor (which I think might have to be an instrumentation type) being very high impedance and the possibility of noise picked up in that area from elsewhere. I'd rather have true isolation to be honest.
The original back-to-back optoisolator variant looks good, possibly with FET current limiters in place of the resistors and I may use an opamp configured as a schmidtt buffer with a fixed mid-point reference rather than a gate. In any case I'll try to do some LTSpice simulation if I can get the models or create them.
Mark.