GFCI and an Open Ground

road-runner wrote:
The downstream neutral-ground fault detection does not use the current imbalance detector, but rather a different mechanism. An example can bee seen here: http://www.onsemi.com/pub/Collateral/FAN4147-D.pdf

Diagram on page 2, description starts near the bottom left of page 5. Older designs did use 120 Hz. signal injection, but still used a separate transformer for detection.

The "different mechanism" described in that document still uses the current imbalance detector; there is only one detector transformer. What I had found was the 120Hz style you mention but it too only had one detector transformer. The one you linked isn't really fundamentally different in that it still relies on inducing current for neutral/ground short detection. Instead of inducing a constant signal, it instead injects an amplified version of what it measures at the imbalance detector. This provides the positive feedback they describe where a small mismatch in current amplifies back and induces a larger one, which is in turn re-detected. The reason for this design is also stated: It's very sensitive to even small (high resistance) connections between neutral and ground because the current mismatch is amplified.

In either design the injection transformer surrounds the neutral and hot wires. With nothing plugged in, no current can be induced in those wires since the circuit is open. However if there is a neutral to ground connection downstream there is now a complete loop formed because the neutral and ground are also connected in the main service panel upstream. The transformer can now cause current to flow around this neutral/ground loop. Since the detection transformer sees this current in the neutral with nothing in the hot to cancel it, it will trip if the current is high enough.

To be honest, I'd never really looked at how GFCI's work until this thread and am thankful for the discussion as it helped me to learn a little in the process.